Design of an Electric Steering Actuator for Outboard Engines Master’s Thesis in Systems, Controls, and Mechatronics BRIAN BONAFILIA Department of Signals and Systems Chalmers University of Technology Gothenburg, Sweden 2015 Master’s Thesis 2015 Abstract This thesis presents an approach to developing an electro-mechanical steering actuator for outboard engines in order to replace existing hydraulic and electro-hydraulic options. System requirements are analyzed and an overview of the solution is given. The relevant theory behind the components of the system, including brushless DC motors, three-phase bridges, and closed-loop feedback control using Internal Model Control, is discussed. A model-based approach is used in which a model of the electrical and mechanical elements of the steering system is developed with the MATLAB/Simulink package SimScape. The results of bench tests and sea trials with the steering actuator are presented, showing performance that is comparable with a commercially available electro-hydraulic steering actuator. The development results in a functional prototype, but fails to achieve all of the performance and functional requirements. Recommendations and considerations for future development are given. Keywords: motor control, steer-by-wire, actuators, brushless DC motor, field oriented control Preface This thesis was written in order to show fulfillment of the educational goals of the Systems, Controls, and Mechatronics Master’s Program at the Department of Signals and Systems at Chalmers University of Technology in Gothenburg, Sweden. The thesis work was carried out at CPAC Systems AB, a Volvo Group company located in Gothenburg, Sweden. Acknowledgements There are many who deserve acknowledgement for their assistance with this thesis work, as without their help I may not have been able to progress this far. Firstly, I would like to thank CPAC Systems and my supervisor during this project, Marco Monzani, for giving me the opportunity to do this work. It has been quite an experience and the knowledge that I have gained during this project is tremendous. At CPAC, I would also like to thank Victor Stensson for sharing his vast knowledge of power electronics and brushless motor control and for his help in developing the mo- tor driver, and Dan Olsson for all of his assistance with solving the problems posed by the steering actuator system. I would also like to acknowledge the hardware develop- ment team, including Peter Siljehov, Fredrik Karlsson, Bo Eriksson, Fredrik Karlemon, Michael Pettersson, and Stefan Lundgren, for their patience and assistance with all of my many electronics questions. With testing and installation, Johan Karlsson and Mathias Lindeborg have provided great assistance that is much appreciated. Also, I would like to thank Claes Segerfelt, Jesper Björnek, Angelica Baumann, and Mårten Elfving for their assistance with cabling, weatherproofing, assembly, and other mechanical issues. It was thanks to the efforts of the SKF team, including Jerry Dahlgren, Mathias An- dersson, Pontus Claesson, Leif Larsson, and Niklas Henningsson, that the final assembly of the actuator was realized. I would like to acknowledge Freescale engineers Mats Henriksson and Mikael Bohm, and Arrow engineer Magnus Grönqvist, for their support in getting the microcontroller up and running and in troubling-shooting the motor driver. I would also like to thank Melisa Milak for her generous support during this project. Finally, I would like to thank the Department of Signals and Systems at Chalmers, and my examiner Professor Jonas Sjöberg, for the chance to continue my education and gain the skills that were necessary for me to attempt this project. Brian Bonafilia, Gothenburg, Sweden 9/9/15 Contents 1 Introduction 1 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1.1 Industry Trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1.2 Electro-Mechanical, Hydraulic, and Electro-Hydraulic Actuators . 2 1.2 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.3 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.4 Development Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.4.1 Design Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.4.2 Model-Based Design . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.5 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2 Steering System Requirements 6 2.1 Performance Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.2 System Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3 Electrical Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.4 Functional Safety Requirements . . . . . . . . . . . . . . . . . . . . . . . . 9 2.4.1 Functional Safety Background . . . . . . . . . . . . . . . . . . . . . 9 2.4.2 Functional Safety Goals . . . . . . . . . . . . . . . . . . . . . . . . 10 3 System Architecture 12 3.1 Outboard Engine Configuration . . . . . . . . . . . . . . . . . . . . . . . . 12 3.2 Actuator Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 4 Motor Theory 15 4.1 Principles of Electromagnetism . . . . . . . . . . . . . . . . . . . . . . . . 15 4.1.1 Lorentz Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 4.1.2 Faraday’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4.2 DC Motor Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 4.3 Voltage Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 4.4 Brushless DC Motor Model . . . . . . . . . . . . . . . . . . . . . . . . . . 21 i CONTENTS 4.5 Commutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4.5.1 Six Step Commutation . . . . . . . . . . . . . . . . . . . . . . . . . 23 4.5.2 Field Oriented Control . . . . . . . . . . . . . . . . . . . . . . . . . 26 5 Motor Selection 33 5.1 36V Prototype Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 5.1.1 Prototype Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . 33 5.1.2 Prototype Performance . . . . . . . . . . . . . . . . . . . . . . . . 34 5.2 Motor Parameter Requirements . . . . . . . . . . . . . . . . . . . . . . . . 36 5.2.1 Parameter Verification . . . . . . . . . . . . . . . . . . . . . . . . . 36 5.3 Model of BLDC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 6 Control Electronics 39 6.1 Electronics Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 6.2 Controller Board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 6.2.1 Voltage Regulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 6.2.2 CAN Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 6.2.3 PWM Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 6.3 Driver Board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 6.3.1 Power Transistors . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 6.3.2 Sensing Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 6.3.3 Voltage Input and Filtering . . . . . . . . . . . . . . . . . . . . . . 45 6.4 Current Sensing Board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 6.5 Position Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 6.5.1 Absolute Position . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 6.5.2 Commutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 6.6 Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 7 Mechanics 49 7.1 Components and Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 7.1.1 Ball Screw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 7.1.2 Worm Drive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 7.1.3 Bevel Gearbox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 7.1.4 Planetary Gearbox . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 7.1.5 Electronics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 7.1.6 Housing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 7.1.7 Mounting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 7.2 Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 8 Control System 53 8.1 Control Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 8.1.1 Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 8.1.2 Control Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 8.1.3 CAN Communication . . . . . . . . . . . . . . . . . . . . . . . . . 54 ii CONTENTS 8.1.4 MC33937A Interrupts . . . . . . . . . . . . . . . . . . . . . . . . . 54 8.1.5 Current Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 8.1.6 Position and Speed Feedback . . . . . . . . . . . . . . . . . . . . . 57 8.1.7 Commutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 8.2 Closed Loop Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 8.2.1 Cascade Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 8.2.2 IMC Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 8.2.3 Feedforward Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 8.2.4 Position Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 8.2.5 Anti-Wind Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 8.2.6 Evaluating the Controller . . . . . . . . . . . . . . . . . . . . . . . 65 9 Testing and Results 71 9.1 Load Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 9.2 Test Rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 9.3 Boat Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 9.3.1 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 9.3.2 Tracking Performance . . . . . . . . . . . . . . . . . . . . . . . . . 78 9.3.3 Current Consumption . . . . . . . . . . . . . . . . . . . . . . . . . 78 9.3.4 Fault Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 10 Discussion and Conclusions 83 10.1 Performance Requirement Analysis . . . . . . . . . . . . . . . . . . . . . . 83 10.1.1 Actuator Power and Efficiency . . . . . . . . . . . . . . . . . . . . 83 10.1.2 Explanations for Failure . . . . . . . . . . . . . . . . . . . . . . . . 84 10.1.3 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 10.2 System Requirement Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 85 10.2.1 System Requirement Status . . . . . . . . . . . . . . . . . . . . . . 85 10.2.2 Failure Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 10.2.3 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 10.3 Electrical Requirement Analysis . . . . . . . . . . . . . . . . . . . . . . . . 87 10.3.1 Current Consumption Observations . . . . . . . . . . . . . . . . . 87 10.3.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 10.3.3 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 10.3.4 Comparison to Teleflex . . . . . . . . . . . . . . . . . . . . . . . . 89 10.4 Safety Requirement Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 90 10.4.1 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 10.4.2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 10.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 Bibliography 96 iii CONTENTS A SimScape Models 97 A.1 BLDC Motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 A.2 Control Electronics Models . . . . . . . . . . . . . . . . . . . . . . . . . . 97 A.3 Control System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 B Additional Images 101 B.1 Assembly Photos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 iv 1 Introduction 1.1 Background Outboard engines are the most common type of propulsion system in small boats [1]. Since 2012, companies such as Yamaha Motors have been marketing electronic steering systems for outboard engines [2]. Among the advantages of directional thrusters such as the the outboard engine is that they are able to generate a net propulsive force or turning moment in almost any direction. This allows for maneuvers such as rotation or sideways motion without needing bow or stern thrusters. When interfaced with a control system such as CPAC Systems’ Electronic Vessel Control (EVC), the steering system can be used to increase the maneuverability of the small boat and allow for steering via intuitive interfaces [3]. Currently, the leaders in the electronic outboard steering market offer their steer- ing systems with hydraulic or electro-hydraulic steering actuation [4] [5] [3]. The next technological step is to reduce the complexity of the system and increase the reliability and energy efficiency by replacing the electro-hydraulic steering system with an electro- mechanical system. A fully electric system eliminates all hydraulic connections, valves, and pressure gauges, as well as the power and signal connections to these components, and replaces them with a single mechanical connection between an electric motor and the outboard engine. Maintenance and installation are simplified and the reliability of the system is improved by reducing the number of components which can potentially suffer a failure or malfunction during operation. 1.1.1 Industry Trends The move from hydraulic actuators to electric motors is present in many industries. A market review report from “Research and Markets” indicates that “electric actuators are expected to replace hydraulic and pneumatic actuators used in the medical industry due to their reliability and operational effectiveness” [6]. 1 1.1. BACKGROUND The change has been especially pronounced in automotive power-assisted steering systems. According to “Car and Driver”, the percentage of cars produced with electric power steering in the North America, the EU, Korea, and Japan, has increased from 25.8% in 2005 to 58.2% in 2011 [7]. Japanese manufacturer NSK expects an increase to over 80% by 2018 [8]. Among the most often cited reasons for this increase are improvements in fuel efficiency (1-3 mpg more) and reduced complexity [7]. This increase in popularity is based on the advancements in the electronic control systems based around the electrical motors, which have improved the controlability of electro-mechanical systems [9]. With microcontroller more powerful, more compact, and less expensive than before, many features can be added to these electrical power steering systems to improve vehicle stability and optimize power consumption. A similar trend is occurring in the naval world. The Italian Navy is funded several research projects to move away from hydraulic actuators on their vessels. A study had found that the hydraulic systems required 60% of all maintenance of transmission and steering systems [10]. To that end, a number of electrical actuators are being developed and tested to perform some of the tasks traditionally done by hydraulic actuators on large vessels, such as rudder control. 1.1.2 Electro-Mechanical, Hydraulic, and Electro-Hydraulic Actuators In an electro-mechanica actuator, the motive force comes from an electric motor which uses the interaction between the motor’s magnetic field and currents in the motor’s windings to generate a force. The motors generally are connected to some gearing or transmission mechanisms to deliver the required force or torque in the proper direction at the necessary speed. Rack-and-pinion systems and lead screws can be used to transform rotary motion into linear motion. A hydraulic system uses pressurized fluid to exert the force[11]. A pump, generally driven by an electric motor, is run constantly to keep the fluid at the required operating pressure. A reservoir of the pressurized fluid is necessary to allow expansion and con- traction of the actuators. When movement is desired, a valve is opened and the fluid is allowed to flow. When the valve is closed, the fluid is not permitted to leave its enclosure and the actuator holds its position [11]. In the third system type, electro-hydraulic actuators, the valves and pump are re- placed by a reversible motor-driven pump which maintains pressure in the fluid, elimi- nating the need for the valves but not the hydraulic couplings [12]. In the hydraulic and electro-hydraulic system, extremely large forces can be delivered by the hydraulic system, limited only amount of pressure the hoses and valves can with- stand. The power delivered can be an order of magnitude greater than electro-mechanical actuators of the same size, which suffer power density limitations from saturation limits of the magnetic field and current carrying limits of the wires [11]. In a hydraulic sys- tem, heat generated at the load can also be carried away by the moving fluid to reduce thermal problems relative to the electro-mechanical actuators [11]. These systems also have the benefit of being able to hold their position without requiring additional energy input (although pure hydraulic systems must run a pump to keep the fluid pressurized). 2 1.2. CONTRIBUTION CYLINDER CYLINDER ACTUATOR CONTROL UNIT MOTOR CONTROL CONTROL CONTROL UNIT VALVES INPUT UNIT RELIEF INPUT INPUT RESERVOIR PUMP MOTOR RESERVOIR PUMP MOTOR HYDRAULIC ELECTRO-HYDRAULIC ELECTRO-MECHANICAL Figure 1.1: Schematic of Motor-Driven Systems showing reduced complexity of electro- mechanical systems The hydraulic coupling allows distribution of the system; a pump can be placed some distance away from the actuator cylinder. Hydraulic systems can also be quickly disen- gaged by releasing a fluid pressure valve. Hydraulic systems must be maintained against leaks and hose wear, or else effectiveness will drop as pressure is lost and environmental damage can occur. Additional power and signal connections are required to monitor pressure and control valves. The electro-mechanical system requires fewer components, needing only a motor and gearing. With no hydraulic fluid or hoses, electrical systems can have lower maintenance and installation costs [13]. In many applications, electrical actuators promise additional efficiency over hydraulics [14] as hydraulics also lose power in pumps and valves through fluid friction and leakage [15]. Electro-mechanical actuators do not have the power density of hydraulics, and require additional work to mechanically disengage from the parts they are moving. These two factors are key concerns in developing an electro- mechanical steering system for outboard engines. 1.2 Contribution The contribution of this thesis work is the development and production of a functional prototype electric actuator. The actuator features custom motor control electronics developed as part of this thesis work and is capable of steering two 250 HP outboard engines on a 30-foot boat using the engines’ 12 V driveline batteries. The thesis results also include recommendations on what is necessary to bring the prototype actuator closer to a commercially viable product. 1.3 Previous Work The Yamaha Helm Master system, which uses a electro-hydraulic actuators for outboard engine steering, is referenced and used as a benchmark for the electrical actuator system designed in this thesis work. The electro-hydraulic actuator in this system is provided by Teleflex, and is referred to as the ”Teleflex Actuator” in this text. 3 1.4. DEVELOPMENT METHOD Figure 1.2: Flowchart of design method At the beginning of this thesis work, CPAC Systems had provided an earlier prototype actuator. This actuator used stock power electronics that were contained in a separate enclosure. Three 12 V batteries in series were used to supply the actuator with a 36 V source. The 36 V source, separately enclosed electronics, and sensor choices were issues that were to be solved in this thesis work. To limit the scope of this thesis, the ball screw mechanism from the 36 V actuator was reused. The term ”36V Actuator” will be used to refer to this earlier prototype, and ”12V Actuator” will be used to refer to the actuator developed in the thesis project when the actuators are to be compared in any way. 1.4 Development Method 1.4.1 Design Process Figure 1.2 illustrates the method used to solve the problems faced in this thesis work. Section 1.5 shows where the details for each step can be found. The performance and power input requirements of the electrical actuator were gath- ered from industry standards and an analysis of the 36V actuator. Electrical and mechan- ical components that were capable of delivering upon those requirements were selected. Electronics were developed to control the actuator’s electric motor to achieve the oper- ating points found to be necessary based on the analysis of the 36V actuator. Control logic was written using a model-based approach, and this control logic was programmed into the microcontroller inside the actuator in C. Testing and verification of the design was performed on a weighted test rig as well as in a twin-engine installation on a boat on open water. 1.4.2 Model-Based Design To develop a system to meet a target outcome, it is generally necessary to develop a model of the system which links the required inputs to the desired outputs. Due to the large number of non-linearities in the actuator system, a state-space model would not have the level of accuracy desired. Instead, a simulation package for physical systems was 4 1.5. THESIS OUTLINE used. SimScape is a package for MATLAB/Simulink that contains a number of electronic and power system components. The physical components of the actuator were modeled in SimScape’s ”Sim Power Systems”, while the logical components and controller were modeled using the basic Simulink elements. 1.5 Thesis Outline The structure of the thesis is arranged as follows: Section 1 Introduction to the thesis with background information, industry trends, and motivation for pursuing an electric steering actuator. Section 2 Requirements for the actuator system taken from industry standards and CPAC internal documentation. Section 3 Overview of the actuator design. Section 4 Background theory on electric motors is given. Section 5 Details of the motor selection process. Section 6 Details of electronics development process. Section 7 Details of mechanical component selection and mechanical configuration. Section 8 Details of control system design and simulation results. Section 9 Results of actuator testing. Comparison to electro-hydraulic actuator. Section 10 Discussion of results and presentation of conclusions. Appendices Additional details on MATLAB/Simulink model are given. 5 2 Steering System Requirements This chapter gives the requirements for actuator system. The requirements comes from ISO Standards, the American Boat and Yacht Council (ABYC) recommendations for electronic steering systems, internal CPAC documentation, and Volvo Penta require- ments for marine systems. Performance, electrical, safety, and system requirements were needed in order to know what targets the actuator had to be designed for and to give standards against which the finished actuator could be evaluated. 2.1 Performance Requirements The relevant performance requirements are set by internal documentation and have been designed through cooperation between CPAC Systems and a client company. The performance requirements that are relevant to this thesis work are: 1.1 The response speed of the steering actuator should be sufficient for the driver to perceive that they have direct mechanical control of the rudder. The rudder control response targets for various requested steering speeds are shown in Table 2.1 and defined in Figure 2.1. The maximum requested speed for the actuator is 20 deg/s. The timing variables correspond to the acceptable delay between a requested change of 1◦ to a rudder movement of 1◦ (T1), the acceptable delay between a change of direction request and a change of rudder movement direction (T2), the acceptable delay between a requested change of 1◦ after a change of direction and a rudder movement of 1◦ (T3), and the acceptable delay between a request of 0.5◦ from the final target and the rudder reaching 0.5◦ off the final target (T4). 1.2 The position control loop should be stable enough to ensure that the response to any ramp does not show oscillatory behavior. 6 2.1. PERFORMANCE REQUIREMENTS Figure 2.1: Performance target definitions Table 2.1: Nominal time delays at different angular request slew rates Requested Speed (deg/s) T1 (ms) T2 (ms) T3 (ms) T4 (ms) A1 (deg) A2 (deg) 2.5 260 460 260 185 -0.1 -0.1 5 195 280 195 120 0.03 0.03 10 165 190 165 95 0.26 0.26 20 125 165 185 95 0.8 0.37 1.3 To provide a suitable range of motion and to allow the vessel to maneuver appro- priately, the steering system should be able to steer the outboard engine to ±30◦ from the straight forward position. 1.4 ”The steering system shall be fully operational within one second after becoming energized at nominal voltage” [ABYC P-17 27.5.1] 1.5 ”The steering system shall be energized whenever the propulsion engine(s) are running” [ABYC P-17 27.5.2] 1.6 ”The steering system shall complete its full range of travel at a maximum of 3.2 s against a minimum 340 kg linear force or 1016Nm torque opposing.” [ABYC P-17 27.5.7.1] 1.7 ”The steering system shall respond to the input command within 0.2 s.” [ABYC P-17 27.5.7.2] 7 2.2. SYSTEM REQUIREMENTS 1.8 ”Steering helms ... shall be oriented in such a manner that clockwise rotation of the steering input device will result in the boat turning to starboard and counter- clockwise rotation of the steering input device will result in the boat turning to port.” [ABYC P-17 7.5.6.1] 1.9 ”Upon propulsion engine startup, the steering system shall default rudder position to straight ahead or provide a visual indication of rudder position at the command station.” [ABYC P-17 27.5.6.3] 1.10 The steering unit position shall be accurate to within 1 mm in the final direction of travel after linearization. 2.2 System Requirements The following system requirements apply to the functional requirements of the electric actuator in general and may not be traced back to a specific standard. 2.1 The actuator should be able to installed on single, twin, triple, or quadruple engine installations. 2.2 The actuator should not interfere with the tilting of the propeller of the outboard engine out of the water. 2.3 ”Provisions shall be made to permit manual operation of the steering systems.” [ABYC P-17 27.7.4] In case of emergency of malfunction, the steering system should be able to be disengaged and steering performed manually. 2.4 The actuator should be interfaced with an electronic steering system. 2.3 Electrical Requirements To operate the system on Yamaha and Volvo Penta engines, the system should follow the electrical standards set out by Volvo Penta and all other relevant electrical standards. 3.1 The actuator for each engine should use only the 12 V battery connected to that engine for power. 3.2 The maximum steady-state current drawn must be less than 70A. 3.3 ”The system shall be tested and capable of operation over a voltage range at the power supply terminal from 80% to 120% of system voltage.” [ABYC P-17 27.6.5.1] 3.4 The device must function normally after being exposed to a reverse voltage at the power supply leads for 60 s. (Not required to function during reverse voltage test). Due to the limited scope of this thesis work, additional electrical requirements, such as EMC immunity, were not considered. 8 2.4. FUNCTIONAL SAFETY REQUIREMENTS 2.4 Functional Safety Requirements 2.4.1 Functional Safety Background A steering system must be considered from a safety perspective as well. Steering system failure was a contributing factor to 70 reportable accidents in recreational boating in the United States in 2012-2013, according to the US Coast Guard. These accidents led to 54 injuries and 7 deaths[16] [17]. The potential for loss of life due to failure of this device warrants spending time on a safety analysis. What follows is a brief hazard analysis which seeks to establish some guidelines for development of the electrical steering actuator beyond the prototype level. ISO 26262 is a standard framework which provides a method for designing safety- related electrical/electronic systems in road vehicles [18]. One of the concepts of ISO 26262 is the ”Automotive Safety Integrity Level”, or ASIL. This is a classification which determines what safety requirements are needed for a process. As this is not an auto- motive product, a more general ”SIL” could be used. (A)SIL D is the most stringent level of safety requirement, and (A)SIL A the least. Below the (A)SIL spectrum is the designation QM for ”quality management”, where the manufacturer’s own quality control process should be used. Any evaluation of risk must take into account that risk can never be avoided com- pletely, it can only be mitigated to acceptable levels. The (A)SIL determines the steps that must be taken to properly mitigate that risk. To determine the (A)SIL of the item, an analysis is made on three dimensions, shown in detail in Table 2.2. The first dimen- sion is severity, a measurement of consequences of malfunction in terms of the threat to the health of people involved. The second dimension is probability of exposure, which is a measure of the probability of the system being in a state in which failure would cause the specific harm. The third dimension is controllability, or how likely it is that, should the system fail, a human operator would be able to gain enough control over the system to avoid the specific harm. The (A)SIL for all combinations is shown in Table 2.3. For the electrical steering actuator, there is only one associated malfunction that it is capable of: unintended steering. This includes both an ignored request to change rudder angle, as well as an unexpected change in rudder angle when no change is requested. This malfunction can occur due to a number of failures of individual elements within the system, such as failure of the voltage regulator to maintain power to the unit, sensor failures, or failure of communication with the EVC. Functional safety guidelines give methods that, if followed, can mitigate risk of failure of these elements, and successfully following the guidelines to mitigate the risk eliminates the need for redundant sensors or interfaces. The SIL of the steering actuator is given in Table 2.4. From the analysis, the worst case scenarios are during high speed operation, where collisions are expected to cause life threatening injuries. Less severe injuries are expected to occur due to sudden movements in open water, however the vessel will be spending much of its time in this situation. Severity, exposure, and controllability were based on estimates. 9 2.4. FUNCTIONAL SAFETY REQUIREMENTS Table 2.2: ASIL Determination Criteria [18] Severity S0 Not Severe No Injuries S1 Low Severity Light and moderate injuries S2 Moderate Severity Several injuries with probably survival S3 High Severity Life threatening injuries, fatal injuries Exposure E0 Incredible Unlikely E1 Very Low Probability Once a Year E2 Low Probability Few Times a Year or 1% of operating time E3 Medium Probability Once a month or <10% of operating time E4 High Probability Every Use or >10% of operating time Controllability C0 Controllable in General C1 Simply Controllable 99% of human operators can safety navigate risk C2 Normally Controllable 90% of human operators can safety navigate risk C3 Difficult to Control <90% of human operators can safety navigate risk Table 2.3: ASIL Determination [18] C1 C2 C3 S1 E1 QM QM QM E2 QM QM QM E3 QM QM ASIL A E4 QM ASIL A ASIL B S2 E1 QM QM QM E2 QM QM ASIL A E3 QM ASIL A ASIL B E4 ASIL A ASIL B ASIL C S3 E1 QM QM ASIL A E2 QM ASIL A ASIL B E3 ASIL A ASIL B ASIL C E4 ASIL B ASLI C ASIL D 2.4.2 Functional Safety Goals The recommended Safety Goal for all cases is ”Unintended steering is to be prohibited”. With SIL B comes a number of recommendations for software and hardware design. As this is a prototype not meant for commercial production, only a few critical to 10 2.4. FUNCTIONAL SAFETY REQUIREMENTS Table 2.4: Hazard Analysis of Electric Steering Actuator for Unintended Steering Mal- function Situation Hazard S E C SIL Docking Impact (Stationary Objects) 0 3 1 QM Low Speed in Narrow Waterway Impact / Running Aground 1 3 1 QM High Speed in Narrow Waterway Impact / Running Aground 3 2 3 SIL B Low Speed Around Other Vessels Collision (Vessel) 1 3 1 QM High Speed Around Other Vessels Collision (Vessel) 3 2 3 SIL B Low Speed Around Humans in Water Collision (Swimmer) 3 1 1 QM High Speed Around Humans in Water Collision (Swimmer) 3 1 3 SIL A High Speed in Open Water Unstable Movement 1 4 3 SIL B the design of the actuator are considered in this work. One key implication is that a microcontroller rated for SIL B or better should be used. With the emergence of ISO 26262, many semiconductor companies have begun to launch products with ASIL D certification, which also covers ASIL C, B, and A [19]. With each SIL comes a recommend hardware failure rate to design for[20]. At SIL B, the target rate is 10−7 h−1 [18]. Based on this metric, component selection prioritized components with a long mean time to failure (MTTF) to reduce the chance that com- ponent failure could occur, leading to faults which would cause an unintended steering malfunction. Guidelines for this SIL recommend certain safety strategies, such as the use of parity bits in communication to prevent communication errors, sensor redundancy, and simu- lation of dynamic components, which were followed throughout this development. The overall functional safety requirements for the actuator system are: 4.1 Unintended steering is to be prohibited. 4.2 ASIL B or better rated microcontroller to be used. 4.3 High reliability and MTTF to be prioritized for component selection. 4.4 Whenever possible, components should be automotive grade or have diagnostic functions to allow detection of a fault. 11 3 System Architecture This chapter gives an overview of the 12V actuator’s structure and how it achieve steering of the outboard engine without going into details of component parameters. Justifica- tions for the actuator’s configuration and types of components are given. 3.1 Outboard Engine Configuration The outboard engines that this actuator is designed to be used with are configured as in Figure 3.1. The rudder angle is controlled by linear displacement of the actuator attachment point. The actuator is fixed to the body of the boat at is ends and the engine at the actuator attachment point. A twin installation was used for this development process. Rudder Angle Engine Pivot Actuator Attachment Figure 3.1: Outboard engine layout with twin installation. Linear motion at actuator attachment point results in engine pivot and changes to the rudder angle. 12 3.2. ACTUATOR DESIGN SPI Hall Effect Pos. Sensor Current Sensor Abs. Pos. Sensor SPI CAN MCU PWM Gearing PWM EVC Power Stage BLDC (PCU) OutboardSCU Engine SPI CAN MCU PWM Figure 3.2: Schematic overview of the actuator system developed as part of this thesis work 3.2 Actuator Design The system overview is given in Figure 3.2. Locations of sections which give details on specific components of the actuator are given in Table 3.1. Table 3.1: Design Details Component Section BLDC Motor 5.2 MCU (Physical) 6.2 MCU (Logical) 8.1 Driver and Power Stage 6.3 Current Sensor 6.4 Position Sensors 6.5 Gearing 7.1 System Model A To provide a solution to Requirement 2.1, it was determined that the motor and electronics should be placed together within the same housing to make a self-contained unit. A single unit should be able to be attached to each outboard motor, regardless of the number of engines, with only a power and signal connection required for control. A separate microcontroller would be enclosed within each actuator in a self-contained Steering Control Unit (SCU). This microcontroller would handle communication and process the various sensor inputs and send the control signal to the power stage to drive 13 CAN Driver Driver 3.2. ACTUATOR DESIGN the motor. From Requirement 2.4, the actuator was to be integrated into Volvo Penta’s EVC (Electronic Vessel Control) system, or other equivalent electronic control system. The SCU’s communication to the EVC is through the Powertrain Control Unit (PCU) node of the EVC via CAN (Controller Area Network) bus, which is standard in many automotive systems. The EVC would provide system instructions and a target reference to the actuator. For an installation with multiple actuators, each actuator would be connected to the others via a CAN bus to observe the positions of the other actuators and adjust their own response if necessary in order to prevent engine collisions. This also gives redundancy in communication to avoid a loss of contact with the EVC, helping to satisfy Requirement 4.1. To meet the reliability and performance requirements of this application, specifically Requirements 4.3 and 1.6, it was determined that a permanent magnet brushless di- rect current (BLDC) motor should be used. Using permanent magnets in motors allows for greater power density than in motors where the magnetic field is generated by a separately supplied electric current [21], which permits a smaller motor and makes Re- quirements 2.2 and 1.3 easier to achieve. Brushless motors have longer life and quieter operation than brushed motors because there are no mechanical contacts between the stator and the rotor [22]. In order to drive a brushless motor, a three-phase power stage needed to be designed around the operating conditions for the motor. Two separate position sensors were necessary–a relative position sensor which could provide accurate commutation information to drive the motor, and a a single turn sensor which would be able to provide the rudder angle via the position of the motor along the screw axis. This was necessary so that the system would always start in a known state, even if powered down completely. Due to Requirement 1.9, short-range proximity sensors could not be used to provide information about the end points in order to calibrate the system for use with a single multi-turn position sensor, as no calibration sweep would be allowed. Some form of gearing was necessary to turn multiple turns of the motor into one turn of the absolute position sensor. More detail on the specifics of this design are given in Chapter 7. All position sensors were to be contact-less sensors for the purpose of Requirement 4.3. A contact-less sensor has theoretically unlimited mechanical life under standard operation, which suits the functional safety goals of the project. A ball screw was to be used to transform the rotational motion of the motor into linear motion for the outboard motor to provide the actuator as shown in Figure 3.1. This approach had been used successfully in the 36V actuator. The ball screw would mount in the same way as the hydraulic cylinder in the electro-hydraulic system so that no changes would need to be made to the mounting arrangement of the outboard engine. 14 4 Motor Theory This chapter details the principles of operation of direct current (DC) motors and brush- less DC (BLDC) motors that are essential to understand in order to understand the design process of this thesis work. The material in this chapter can be found in most textbooks on electric machines but is included here due to its importance in understand- ing much of the rest of this report. 4.1 Principles of Electromagnetism This section includes information on the basics of electromagnetism as they apply to operation of electric motors, as well as the necessary background for understanding the principles of commutation for brushless permanent magnet motors. The electric motor is a device which converts energy in the form of electricity into kinetic energy. (Throughout this report, the term ”motor” will be used exclusively to refer to devices of this nature, and ”engine” will be used to refer to devices, such as the boat’s outboard engine, which convert convert chemical energy into kinetic energy via combustion.) Electromagnetism is one of the four fundamental forces of nature [23]. For under- standing the operation of electric motors, especially the need for electronic commutation of brushless permanent magnet motors, it is necessary to have a basic understanding of some of the principles of electromagnetism. The most important concept is the magnetic field. The magnetic field is a vector field which is created by either a permanent magnet or the flow of an electric current through a conductor [23]. (It is not necessary to understand how the field is created, only that it is.) Indicated by the symbol B, the magnetic field has the unit of the tesla, T . The magnetic flux φ is the surface integral of the normal component of the magnetic field B through a given surface and it is expressed in the unit of the weber, Wb. B, the 15 4.1. PRINCIPLES OF ELECTROMAGNETISM F N I S Commutators F Brushes Figure 4.1: A simple DC motor strength of the magnetic field, is also called the flux density, as it is the amount of flux per unit area. In a uniform magnetic field through a surface of area S: dφ = B · dS (4.1) 4.1.1 Lorentz Force The Lorentz force is the force experienced by a charged particle as it moves through an electric and magnetic field. The force is generally expressed as in (4.2). [24] F = q(E + v×B) (4.2) Here, the force F is a function of the charge q of the particle, the velocity v and the flux density of the magnetic field B and of the electric field E. In the case of an electric motor, the contribution from the electric field is negligible [25]. The current in a conducting wire is the measure of the movement of the electrons through the material in terms of the rate of electric charge moved per second. Each of these electrons experiences the Lorentz force, resulting in a net force on the wire itself [23]. This force, also called the Laplace force, can be determined by substituting the definition for an electric current into (4.2) and omitting the electric field term . F = I`×B (4.3) The force acting upon the conducting wire is proportional to the current through the wire as well as the length ` of the wire and the strength of the magnetic field. Figure 4.1 shows how this force becomes a torque in a very simple example of a single-coil electric motor. In a brushed motor, the type shown here, physical contact is made between the armature wires and the voltage supply via brushes and a round commutator. As the coil rotates, the commutators switch between the positive and negative supply leads so that the current always flows the same direction relative to the magnetic field and thus generates a torque in the same direction. [26] Functional motors have many more coils than this, so the force calculated in (4.3) should be applied to each turn of the armature coil. Because the length of the motor, the 16 4.1. PRINCIPLES OF ELECTROMAGNETISM strength of the magnetic field, the number of coils, and the effective radius of the rotor are constant in permanent magnet motors, (4.3) leads to the expression of the motor’s output torque in terms of a constant kt with the units Nm/A, called the motor’s torque constant or simply the ”motor constant”. T = ktI (4.4) 4.1.2 Faraday’s Law Faraday’s Law of Magnetic Induction describes the way in which a voltage is induced in an electric circuit which experiences a changing magnetic field. For the purpose of electric motors, Faraday’s Law can be expressed as in (4.5).[24] E − dφ= N (4.5) dt The electromotive force E is referred to as the EMF, or in the case of electric motors, the back EMF, and has the units of V . The flux φ is determined from the flux density of the magnetic field and the surface for which one loop of the conducting wire forms a boundary. In an electric motor, the armature windings are normally present in coils of approximately the same size, which experience approximately the same change in flux. For this reason, the equation includes the term N for the number of turns in each coil of the winding. The EMF is the amount of electromagnetic work done per unit of charge on a charge carrier in a coil as it travels once around the loop. It is a manifestation of the same phenomenon which causes the Lorentz force [23]. Lenz’s law underscores the importance of this effect to electric motors. Lenz’s law states that an induced EMF gives rise to a current whose magnetic field opposes the original change in magnetic flux, hence the negative sign in (4.5) [11]. It is seen that the EMF induced according to Faraday’s Law opposes the voltage which would be driving the excitation current in the simple motor of Figure 4.1. The result of this is that electric motors are inherently self-limiting and stable sys- tems. The dφdt term in (4.5) shows that the back EMF increases proportionally to the motor speed. Because this back EMF opposes the voltage driving the motor, the faster the motor spins, the less voltage there is available to drive a current that would generate a torque. At some speed, the back EMF will be equal to the supply voltage (minus losses) and thus there would be no current in the armature, and therefore no torque to further accelerate the motor. Like the calculation for the motor’s torque, the relationship between speed and back EMF can be simplified by using defining a back EMF constant, or velocity constant, kv for the motor. This term has the units of V s/rad. The units V s/rad and Nm/A are dimensionally equivalent, and kt and kv are equivalent terms [25]. E = kvω (4.6) Knowing this, it is possible to develop a mathematical model of a DC motor. 17 4.2. DC MOTOR MODEL R L I + U ω - Figure 4.2: Schematic of motor armature circuit 4.2 DC Motor Model For the sake of simplicity, we will limit our study to permanent magnet motors or separately excited motors where the strength of the magnetic field generated by the field current is constant. The windings of the motor armature can be expressed as a simple electric circuit as shown in Figure 4.2. From standard equations for an RL circuit and the back EMF term from (4.6), the current can be calculated as in (4.7). dI U = RI + L + kvω (4.7) dt The back EMF and the armature current depend upon the speed of the motor, ω. This speed can be found by considering a simple mechanical model of the system considering an inertia J , viscous damping term b, and torque from (4.4). dω J + bω = ktI (4.8) dt Taking the Laplace transform of (4.7) and (4.8) yields a simple system model. Ω(s) kt = (4.9) I(s) (Js+ b) I(s) 1 = (4.10) U − kvΩ(s) (Ls+R) 4.3 Voltage Control From the simple model given in (4.7), it is apparent that the steady-state behavior of the DC motor is a linear relationship between the current and the speed. Because the torque is proportional to the current, there exists a linear relationship between the torque and speed of a DC motor, as shown in Figure 4.3. As the motor increases in speed, the back EMF reduces the voltage across the winding resistance until the current flowing generates only enough torque to balance the load applied to the motor. With a net torque of zero, there is no acceleration, and the motor reaches steady-state. 18 4.3. VOLTAGE CONTROL Continuous Operation V = Vs Fan/Pump Load V = 0.67 Vs Mech. Load V = 0.33 Vs No Load Speed Figure 4.3: Torque vs. speed curve of a DC Motor If Figure 4.3, the operating point for the motor is found at the intersection of the torque vs. speed curve of the motor at the supply voltage and the torque vs. speed curve of the load. Even with no external load, a motor has internal friction that generates an opposing torque. In a permanent magnet DC motor, the only parameter in (4.10) that can be controlled is U . In a separately excited DC motor, in which the magnetic field is generated by a current passing through the stator, the variable kv is also adjustable by varying the stator current. To change the speed of the motor, it is necessary to adjust the applied voltage, V in the figure, so that the motor’s torque vs. speed curve intersects the load’s at the desired point, keeping in mind that the motor can only operate continuously up to a certain current level. Above that, damage to the motor may result due to prolonged operation. In Figure 4.3, to control the speed of the motor driving the load, V can be set to any level up to Vs, which is the motor’s supply voltage. For the steering actuator, the the electrical power that drives the motor system comes from the boat’s battery, which is charged by the outboard engines. The voltage of this battery, floating at 13.6V , is something that the control system is unable to directly alter. 19 Torque 4.3. VOLTAGE CONTROL 1/f T25 Carrier Signal T75 Vsup 25% Duty Cycle Vavg Vsup Vavg 75% Duty Cycle time Figure 4.4: Basic principle of PWM In order to vary the voltage that is being applied to the motor, pulse width modula- tion (PWM) is used. PWM provides an adjustable average voltage from a fixed voltage source by generating a square wave with the same magnitude as the source voltage. This wave is controlled to be high for a portion of time coinciding with the desired average voltage level [11]. Vavg DutyCycle = × 100% (4.11) Vsup PWM is achieved by generating a carrier signal with a desired frequency, and enabling or disabling a switch based on the value of this signal. Figures 4.4 and 4.5 show this in more detail. Many microcontroller have a module on them dedicated to generating a PWM signal, as it is very common in voltage control applications. Open Iavg Closed Figure 4.5: The current is sampled in the middle of the PWM duty cycle. Motor inductance maintains the current while the circuit is open. It is most common to sample the current in the middle of the duty cycle and to use that value as the average current [27]. 20 4.4. BRUSHLESS DC MOTOR MODEL 4.4 Brushless DC Motor Model The brushed motor previously discussed has several disadvantages. The brush makes physical contact with a fast moving commutator and therefore is subject to wear which can lead to failure. This contact also produces audible and electrical noise. The brushes are susceptible to corrosion which makes this model undesirable for many environments. Finally, there is a possibility of sparking when the commutator makes contact with the brush. BLDC motors operate according to the same principle as discussed above, but with- out brushes. Instead, the commutation is done electronically. Sensors detect when each pole is in position to provide maximum torque, and the flow of current through the coil is switched on and off as necessary. Figure 4.6 provides a schematic of the operation of a 3-phase BLDC motor. ω R L UA ω R L UB ω R L UC Figure 4.6: Schematic of 3-phase BLDC motor armature circuit The mechanical model is the same as for the brushed case, as there is still only one rotor and stator. The electrical model becomes slightly more complicated as it now includes three phases. [28]          Uas R 0 0Ia d L−M 0 0 IaU = 0 R 0 I + 0 L−M 0 I + eabs b b edt b (4.12) Ucs 0 0 R Ic 0 0 L−M Ic ec The term M is the mutual inductance between the phases. A current flowing in one phase also has some effect on the magnetic fields in the other phases. [29]. In (4.12), the back EMF terms ea, eb, and ec are functions of the rotor speed ω as well as the position θ. The magnetic fields in the motor are generated by permanent magnets on the motor axis, and as those magnets rotate, so does the magnetic field, B. 21 4.4. BRUSHLESS DC MOTOR MODEL The construction of the BLDC motor determines the nature of the back EMF voltage profile. The shape and position of the permanent magnets determine how the magnetic flux changes for the coils wrapped around each pole, and from (4.5), this governs the back EMF. Many motors are made with sinusoidal back EMF profiles, while others are made with trapezoidal back EMF profiles. In the case of a symmetric sinusoidal back EMF motor, the per-phase back EMF terms are: ea = kvω sin((θ) ) 2π eb = kvω sin(θ − ) (4.13)3 2π ec = kvω sin θ + 3 For a three phase BLDC motor, it is generally true that the phases are placed sym- metrically, i.e. 120◦ apart. This has implications for the torque generation as well. In the previous sections, it was mentioned that kv and kt are equivalent terms, and that remains true here as well. Bm 0° 90° 180° 270° A+ -Bm kv,a C- B- N 0° 90° 180° 270° S C+ -kv,a B+ kt,a A- 0° 90° 180° 270° -kt,a Figure 4.7: Motor constants as function of rotor position Figure 4.7 gives some insight into this. With the BLDC motor, the flux density is changing with the angle of rotation, designated α in the figure. The profile of B(α) at the airgap for the positive coil of A is shown. When the north pole of the magnet is pointing directly at the coil for Phase A, the flux density is greatest. Because of the sinusoidal distribution of the flux density around the rotor, at this point the change in flux density is zero. From (4.5), the back EMF constant kv,a is also zero. [25]. 22 4.5. COMMUTATION In (4.4), F is proportional to the magnetic field strength, B(θ). Being a measure of torque, kt,a for Phase A would then be greatest when it generated a force acting perpendicularly to the magnetic axis, i.e. when α = 90◦ or α = 270◦, depending on what direction you were interested in turning it. Powering Phase A at α = 0◦ would result in a force parallel to the magnetic axis and would generate no torque, thus kv = 0. It is clear that kv,a = kt,a is true still, hence the torque generated by each phase can be given as: Ta = ktIa sin((θ) ) 2π Tb = ktIb sin(θ − ) (4.14)3 2π Tc = ktIc sin θ + 3 In (4.14), it is shown that in order to generate torque in the appropriate direction, the currents must be controlled so that they are in phase with the back EMF at all times. [25] 4.5 Commutation The process of switching the current direction so that torque is generated continuously in the desired direction is called commutation. The work documented by this thesis looked at two common ways of approaching the problem of commutation: Six-Step Commutation and Field Oriented Control. 4.5.1 Six Step Commutation As seen in Figure 4.6, the flow of current through the BLDC motor is controlled using switches. These switches are power transistors, semiconductors which are turned on or off by electronic means. Generally, six power transistors are used which control flow in either direction in each phase. The simplest method of control for a three phase BLDC motor is to toggle through six commutation steps, each step having one switch open on the high side and one switch open on the low side. Thus, current flows through two phases at once, positive for one and negative for the other. This gives six combinations of current vector direction that can be switched to best track the rotating rotor flux. Figure 4.8 demonstrates the concept by showing two steps in a commutation cycle. In the top, shown in red, the high side of phase A is open and the low side of phase C is open. Current flows in the positive direction in phase A and in the negative direction of phase C. To the right, it is shown that this generates a magnetic flux vector, shown as a red arrow, which leads the magnetic axis of the rotor by 90◦ (along the axis labeled q-axis, for quadrature), thus generating the desired torque. As the rotor turns counter- clockwise, this angle is reduced. As the magnetic axis (labeled the d-axis, for direct) passes the dotted line between the positive A and negative B pole, a switch must occur 23 4.5. COMMUTATION Q axis A+ D axis B- B+ D axis A+ Q axis N S A- Figure 4.8: Illustration of current flow during two commutation steps so that the red flux arrow remains close to the q-axis. Current that generates a flux in the direction of the q-axis is called ”quadrature current”, and current that generates a flux in the direction of the d-axis is called ”direct axis current”. After the commutation step, the high side of phase B has been opened so current can continue to flow in the negative direction in phase C and now also in the positive direction in phase B. This deactivates the phase A coils and activates the phase B coils. Once again, in the image to the right, the red flux axis represents the direction of the force that is pulling on the north pole of the magnet, resulting in torque in the desired direction. Here, though, this axis does not line up directly with the q-axis, because the six step commutation can only generated a flux in the direction of one of the six dotted lines. This limitation in this commutation scheme results in considerable torque ripple for motors with a sinusoidal back EMF because the torque angle between the rotor flux vector and the stator quadrature current is not constant. The commutation scheme is shown in Figure 4.9. Simple commutation toggle only two switches, the high side switch of one phase, and the low side of the other. This is not ideal, because it results in a case where only a single switch is open. Because the motor has inductance, current in the motor cannot change instantaneously, so when the high side switch has closed, the current must continue to 24 N S 4.5. COMMUTATION V 0 -V I 0 -I open closed open closed open closed open closed open closed open closed 0 100 200 300 400 500 600 700 800 900 1000 Theta (degrees) Figure 4.9: Six Step Commutation Scheme for Positive Torque flow in the motor. It does this by passing up through the freewheeling diode in the low side switch. In the process, there is a voltage drop and a considerable power dissipation. To avoid over-heating, it is best to use complementary commutation, which calls for the low side switch to be opened during the part of the PWM phase that the high side switch is closed for. This allows the current to pass through the transistor instead of the diode, and the power dissipation is far lower. Figure 4.10 demonstrates this.[30] Current Waveforms Figure 4.9 shows the overall commutation scheme for six step commutation. In the second graph in the figure, the ideal current shape is shown. The ideal current for each phase is positive Imax for a 120 ◦ sector of the cycle, then zero for 60◦, then −Imax for 120◦, before being zero for another 60◦, and repeats. This is not physically achievable, as the current phases have finite inductance which requires time to build up the current. 25 Phase C Phase C Phase B Phase B Phase A Phase A Low Side High Side Low Side High Side Low Side High Side Target Back Current EMF 4.5. COMMUTATION Figure 4.10: Illustration of free wheeling current during complementary commutation At the same time, the back EMF is changing during the current cycle while the supply voltage is not changing, so the voltage left to drive a current over the motor’s resistance is not consistent over each commutation phase. These effects lead to current waveforms shown in Figure 4.11. It should be noted that the greater the inductance of the motor, the more the first effect is apparent, but the less the second is. [31] Low inductance motors are especially hard to control in this regard. The current fluctuation as a result of this effect is significant. In testing, the current ripple within a single commutation phase could be as large as 25A between the peak and the trough when the average current was 38A. Increasing the controller speed and switching speed helps to mitigate this effect, but there are many other considerations that make this not always a practical solution.[31] 4.5.2 Field Oriented Control In addition to six step commutation, another method of commutation was explored for the motor. In six step commutation of a motor with a sinusoidal back EMF waveform, after each commutation the stator current requires a finite amount of time to return to its steady-state current value. This effect causes a ripple in the motor torque as the 26 4.5. COMMUTATION 100 Measurement Simulation 50 0 -50 -100 30 20 10 0 -10 -20 -30 Figure 4.11: (top) Single phase current under load (bottom) Measurements of phase cur- rents current decays and rises again with each commutation step. Furthermore, even with a steady-state current, there is torque ripple due to the changing rotor flux position within each commutation step. A method to solve this problem is Field Oriented Control. The underlying principle of Field Oriented Control (FOC) is to express the current and rotor flux as vectors, and to control the current vector in reference to the rotor flux vector to achieve the desired performance by adjusting the individual current compo- nents. In essence, it is a method of keeping the current and the back EMF in phase at all times. Field Oriented Control is achieved by using Space Vector Modulation (SVM) as opposed to the PWM technique. This statement may be a bit misleading, as SVM also uses pulses of a constant voltage to control the phase voltages. The conceptual difference is that PWM adjusts each individual phase using modulation, while SVM treats the three-phase bridge circuit as a single unit to produce a pair of voltage vectors. This allows the current to flow in all phases, and the flux vector to be aligned with the magnetic axis regardless of the rotor angle, as in Figure 4.12. Clark and Park Transformation As previously discussed, the torque generated by the BLDC motor is proportional to the cross product of the current and the back EMF, that is the rotor flux. To simplify 27 Current (A) Current (A) 4.5. COMMUTATION D axis A+ Q axis N S A- Figure 4.12: Snapshot of Space Vector Modulation analysis of a three-phase circuit, it is useful to use the Clark and Park Transforms to convert the stator current vector from a time-varying three-phase non-rotating frame of a reference to a time-invariant 2-coordinate rotating frame of reference. [32] b ia a j2 /3 e ib is j4 /3 e ic c Figure 4.13: Complex space representation of current vector In general, the stator current of a three phase BLDC in complex space is given by: i = i + e2jπ/3i + e4jπ/3s a b ic (4.15) Instead of this expression, the three-phase current can be expressed as in two sta- tionary coordinates using the Clark Transform [32]: [ ] [ ] ia iα 2/3 −1/3 −1/3  = √ √ i b (4.16) iβ 0 1/ 3 −1/ 3 ic 28 4.5. COMMUTATION b i  , a i is  c Figure 4.14: Reconstruction of three phase current vector in two dimensions With this expression for α and β, the current is still dependent upon the position of the rotor flux. A second transform, the Park Transform, is necessary to convert this stationary coordinate system into one which rotates with the rotor flux so that the stator current can be expressed in terms of a torque-generating component, q, which is at 90◦ to the rotor flux [33], and a non-torque-generating component, d, which is line with the rotor flux and can be controlled to adjust the flux linkage, which can be used to modify the values of kt and kv. [ ] [ ][ ] iq cos(θ) sin(θ) i= α (4.17) id − sin(θ) cos(θ) iβ These transformations yield the current in terms of a component which can be directly related to torque. A PID controller can now be used to provide closed loop control for a reference for both iq and id. This is typically done using two separate PID controllers, each of which returning a reference voltage, vq and vd. [30] One references voltages for vq ref and vd ref are found, the inverse of the Park Trans- form is used to return references vα ref and vβ ref Vector Modulation The basic principal behind SVM is to use the entire three-phase bridge as a single entity to produce a voltage vector. In the Figure 4.16, each of six possible vectors is displayed (in addition to this there are two null vectors which are not shown). In the figure, the value 1 corresponds to the top-side switch being open, and the value 2 corresponds to the bottom side switch being open. In the case of what is labeled v4, the top-side switch for phase A is open, and the bottom side switches for B and C are open. From (4.15), we know that an equal voltage for phase B and phase C would negate each other and result in vβ = 0, while vα = va−vb/3−vc/3. Hence, this vector points along the positive α axis. The other vectors have similar meanings. 29 4.5. COMMUTATION q d id e i  i is  iq Figure 4.15: Reconstruction of current vectors in rotating reference frame The numbering of the vectors and sections is arbitrary. What is shown here is just the internal numbering in the method used in the electric steering actuator’s SVPWM mode, but no single convention exists. What matters is that a consistent algorithm is used to determine what sector a reference voltage vref falls into. Trigonometric calculations can be expensive computationally in microcontrollers, especially when they must be done several times in one controller cycle, so other methods can be used. V β2 V6 (010) (110) I V t V III1 6 Vref V4 V (100)3 V7 V0 (011) t2V4 α IV II VI V V51 (101) (001) Figure 4.16: Reference in terms of voltage vectors To create a voltage vector in the direction desired, first the sector is determined based on the α and β components of the voltage vector which came as the results of the 30 4.5. COMMUTATION controller in the previous section. Once the sector is known, the two voltages vectors which define that sector are used to create the reference vector. [34] This is done in the same was as PWM is used to set a voltage, only instead of a scalar value set by two states the SVPWM scheme modulates between four states to set a vector value. The value vref is a vector sum of the form: vref = t1v6 + t2v4 + (ts − t1 − t2)v0 (4.18) If the magnitude of vref is less than the maximum magnitude of the reference voltage vector, shown by the circle in Figure 4.16, then t1 + t2 will be less than ts, and the system must be off for a period of time, t0. The time t0 is divided between v0(000) and v7(111). The result is a control over the switches in the bridge looking as in Figure 4.17. t0 t2 t1 t0 t1 t2 t0 4 2 2 2 2 2 4 Phase A Phase B Phase C Figure 4.17: Three phase bridge control for reference voltage in Figure 4.16. Duty cycle contains portions when all low side switches are open, as well as portions when all high side switches are open. There are numerous algorithms one can follow to derive the final values for t1 and t2. One example is that method used in this application, and an example is given for Sector 3. Once the sector is known, a matrix representing the two voltage vectors which define that sector is constructed: [ ] M = Vs V1 V2 (4.19) The voltage reference vector is constructe[d of]the form: V V αref = (4.20) Vβ The weighting matrix for ea[ch]sector is a function M : h ( )−1 = MTM MTVref (4.21) k 31 4.5. COMMUTATION The values h and k now give t1 and t2 in terms of what percentage of the time that V4 and V6 should be active. Based on this, duty cycles for each phase can be set: 1− h− k dutyA = h+ k + 2 1− h− k dutyB = k + (4.22) 2 1− h− k dutyC = (4.23) 2 (4.24) The final term in (4.22) sets the time that all high side switches should be open. By using center-aligned PWM, the timings in Figure 4.17 are respected. Voltage and Current Waveforms In theory, the results of the FOC method produce current and voltage waveforms as shown in Figure 4.18. Note the constant value for Iq. Vs 0 Is -Is Iq Id Is -Is 0 100 200 300 400 500 600 700 800 900 1000 Theta (degrees) Figure 4.18: Resulting waveforms for Field Oriented Control using SVPWM. (Top) Voltage waveform and (Bottom) current waveform are kept in phase. Iq in the third row is constant. Iα and Iβ are shown in the second row. 32 Current Current Current Voltage 5 Motor Selection This chapter gives information on how the BLDC motor for the 12V actuator was chosen. The process of determining the required speed and torque of the motor is shown, and the method used to find the motor parameters from this information is explained. 5.1 36V Prototype Analysis To limit the scope of the thesis, the ball screw from the 36V actuator was reused. The 36V actuator was subjected to load testing and disassembly to determine the required motor speed and torque needed for the performance requirements. This information could be used to define parameters of a motor which would allow Requirement 1.6 to be met based on voltage and current limitations set by Requirements 3.1 and 3.2. 5.1.1 Prototype Mechanics The mechanical actuator used in the 36V actuator is a ball screw jack. This is a type of lead screw which uses recirculating ball bearings to reduce friction. If the screw is fixed in place and the nut spins, the nut will move down the length of the screw. The spinning of the nut is achieved using a worm drive with the nut mounted inside of the worm gear and the worm attached to the motor shaft. The dimensions of the screw and mounting are given in Figure 5.1. The pitch (or lead) of the screw is 5.0mm. The length of the screw and nut housing are shown in Figure 5.1. The actuator has a stroke of 21 cm from end to end. The worm has 4 starts and the worm gear has 27 teeth for a ratio of 6.75 : 1 between motor turns and turns of the screw. Each motor turn advances the screw by 0.74mm. The attachment point of the outboard engine to the actuator is 20 cm from the pivot point of the outboard engine itself. For a 21 cm stroke, the maximum rudder angle that can be achieved lower than specified in Requirement 1.3: 33 5.1. 36V PROTOTYPE ANALYSIS Engine Pivot 20 cm 26 cm 47 cm Figure 5.1: Outboard engine pivot scheme with ball screw dimensions ( ) θ = tan−1 10.5 cm = 27.7◦rudder (5.1) 20 cm Both the Teleflex actuator and the 36V actuator use a linear approximation of the rudder angle with 3.38 mm of linear motion for each 1◦. For a maximum angle of 30◦ from center in either direction, this approximation of the rudder angle around the zero point yields errors of 2.3◦ at a requested angle of 30◦. This same approximation was used in the 12V actuator design. The required motor speed is found in (5.2) using this relationship and the maximum slew rate of 20◦/s in Requirement 1.1. (20 deg/s) (3.38mm/deg) (6.75motor turns/screw turns) (60 s/min) Nt = = 5475RPM (5.0mm/screw turn) (5.2) 5.1.2 Prototype Performance The 36V actuator was subjected to a load test to determine the efficiency of the ball screw and the torque needed for Requirement 1.6. The actuator was mounted vertically with a cable attached to a lever below the actuator, as in Figure 5.2. Weights up to 100kg were suspended from the end of the lever to give effective axial loads up to 530kg. The driving torque Td of the screw based on the load F is: FPh Td = (5.3) 2πη1 The efficiency of the ball screw and worm drive, η1 and η2, are found by analysis of the motor current at different loads. The FL57BL03 motor used in the 36V prototype 34 5.1. 36V PROTOTYPE ANALYSIS 150 cm 41 cm mg Figure 5.2: Load test setup for the prototype actuator 6000 Measurements Target for ± 30° 5000 Target for ± 25° Target for ± 20° 4000 3000 2000 0 0.5 1 1.5 Torque (Nm) 30 25 20 15 10 Measurements 5 Calculations 0 0 0.5 1 1.5 Torque (Nm) Figure 5.3: Speed vs. Torque of 36V actuator with markers showing rudder angle stroke in 3.2 s at 340 kg of load 35 Speed (RPM) Current (A) 5.2. MOTOR PARAMETER REQUIREMENTS has a torque constant, kt, of 0.051 Nm/A and draws a no-load current, INL, of 3.8 A (the motor without the screw draws only 0.8Nm). The gear ratio, g, is the worm drive reduction ratio of 6.75. Based on (5.4), the current vs. load profile in Figure 5.3 shows a combined efficiency of 60%. This is reasonable for worm drives. [35] FPh η = η1η2 = (5.4) 2π(I − INL)ktg The 36V actuator did not satisfy Requirement 1.6, as shown in Figure 5.3. It could only manage 40◦ of motion in that time. To move 340 kg of load at that speed, the new motor would need to provide 0.8 Nm at 5133 RPM . The actuator should be able to operate continuously at this operation point, with intermittent operation of twice this level allowed as a safety factor. 5.2 Motor Parameter Requirements Based on the assessment documented in Section 5.1, a new BLDC motor was chosen using estimates that ignored switching and commutation losses. In order to reach the target speed of 5475RPM off of a nominally 12 V supply, the motor should have a speed constant less than: U −RINL 12 V − 0 V kv = = = 0.021 V s/rad (5.5) ω 573.3 rad/s To generate the required amount of torque with this torque constant at the operating point of 0.8Nm at 5133RPM , the motor would require: τ 0.8Nm I = = = 38A (5.6) kt .021 V s/rad The motor resistance necessary to provide this current at the operating point is: U − kv ω 12 V − (0.021 V s/rad) (537.5 rad/s) R = = = 18.7mΩ (5.7) I 38A The closest match that could be found in a reasonable time frame was the BY80BL300 from Boyang Motor. This has a 80mm frame and is rated for 6000RPM and 30A with a nominal voltage of 12V . (Previous test data of the 36V actuator shows that the 340kg load test case was an extreme example and that it would be safe to operate above 30A for 3.2 s, and that normal operation would rarely exceed 30A, with 72A peak current.) 5.2.1 Parameter Verification To build a model of the motor for controller development, the motor parameters must be verified. Winding resistance and inductance were found using an LCR meter. The back EMF constant was found by spinning the motor shaft using a drill. The line-to-line voltage amplitude, in Figure 5.4, at 91.6 rad/s was 2.03V , for a kv of 0.02V rad/s. This figure also shows verification of the back EMF constant used in the simulation. 36 5.2. MOTOR PARAMETER REQUIREMENTS 4 2 0 -2 -4 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 Time (s) 2 Simulation Output 1 0 -1 -2 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 Time (s) Figure 5.4: (Top) BY80BL300 Back EMF constant determination at 91.6rad/s. (Bottom) Verification of simulation results at 80.9 rad/s against phase current measurements. The motor inertia was found using acceleration of the motor and the supply current during a voltage step test with a current-limiting power supply. From the supply current in Figure 5.5 and the relationship in (4.4), the torque applied in the test can be calculated. The range of interest is from t = .04 s to t = .11 s where the current supply is limited to 10.4A. ∆ω (352 rad/s− 130 rad/s) αmean = = = 3171 rad/s 2 (5.8) ∆t .07 s The inertia is found using Newton’s Laws for rotational motion: τ 0.020Nm/A · 10.4A J = = = 6.6× 10−5 kgm2 (5.9) α 3171 rad/s2 The inertia could also be estimated using the BY80BL300’s frame size of 80 mm and the frame size and known inertia of the 57 mm FL57BL03 motor. The inertia of a cylinder is proportional to the mass and the square of the radius. With a uniform density, the mass of the rotor is also proportional to the square of the radius. Treating the rotor as a cylinder with a radius proportional to the frame size gives an inertia for BY80BL300 that is c(lose )to the calculated value using (5.9) and (5.8). 80 4 · 1.73× 10−5 kgm2 = 6.7× 10−5 kgm2 (5.10) 57 37 Volts (V) Volts (V) 5.3. MODEL OF BLDC Motor Acceleration 600 400 200 0 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 30 20 10 0 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 Time (s) Figure 5.5: Speed and Current Response for BY80BL300 A comparison of the motors in the 36V and 12V actuators is given in Table 5.1. Table 5.1: Motor parameters FL57BL03 BY80BL300 Poles 8 4 Nominal Voltage 24 V 12 V Torque Constant 0.051Nm/A .020Nm/A Line-to-Line Resistance 0.21 Ω 0.032 Ω Line-to-Line Inductance 750 uH 42.6 uH No Load Current 0.8A 2.2A Rated Speed 3000RPM 6000RPM Rated Torque 0.32Nm 0.48Nm Inertia 1.73 × 10−5 kgm2 6.6 × 10−5 kgm2 5.3 Model of BLDC The SimScape model setup of the BLDC motor is given in Appendix A. 38 Current (A) Speed (rad/s) 6 Control Electronics This chapter discusses the development of the control electronics which was a significant part of this thesis work. Final component selection is given along with an explanation for why the components were chosen. Other design concerns which influenced the layout of the printed circuit boards are discussed. 6.1 Electronics Configuration The electronics that drive the electric steering actuator consist of three printed circuit boards. One board is used for the power stage and motor driver, another for the micro- controller and communication peripherals, and a third for current sensing. This setup minimizes the package of the electronics. Multiple boards of equal size can stacked on top of each other, offset by only the height of the components. To place these side by side on a single board would require a much larger area, and would have increased the width or length of the final actuator packaging. Additionally, the power electronics generally need to be scaled to fit the motor control application, but the microcontroller does not. Using two different boards for this allowed development on the microcontroller to be done separately from the power electronics so that as much of the solution as possible could be reused if the motor driver requirements changed, for example due to a change in the gearing or selection of a different motor. The current sensing board was attached as a third board to keep the size of the power electronics board more compact, and because not all motor control approaches would require three-phase current sensing. 6.2 Controller Board The microcontroller that was chosen was a 180 MHz microcontroller from Freescale called the MPC5744P. This microcontroller was chosen due to its peripherals and ASIL 39 6.2. CONTROLLER BOARD D certification in accordance with Requirement 4.2 [36]. It has the necessary number of CAN modules for communication with the EVC and the steering sync, as per Require- ment 2.4 as well as ample analog-to-digital conversion modules (ADC), PWM modules, and other communication interfaces. These are described briefly in this section. Figure 6.1: Finished Controller Board PCB with MPC5744P 6.2.1 Voltage Regulation The MPC5744P microcontroller requires well-regulated voltage inputs to function. It requires 1.25 V for core power, and 3.3 V and 5.0 V for digital and analog signals. It must also provide 5.0V to the motor driver for over-current comparison reference. From electrical Requirement 3.3, it is specified that the system should be operational within ±20% of the nominal voltage, so a constant voltage source can not be depended upon. A voltage regulator circuit was used in order to guarantee these voltage levels in the operating environment at varying input voltage. The key components are a PWM-based closed-loop regulator for the 5.0 V supplies and two linear regulators for the 3.3 V and 1.25 V supplies. The large capacitors seen in the center of the board in Figure 6.1 are to reduce the ripple in the 5.0 V that comes as a result of the switching regulator. To minimize electromagnetic interference with the signal lines, the switching compo- nents of the voltage regulator were laid out as far as possible from the sensitive signal areas. An unbroken ground layer on the board provides a return path for currents in the voltage regulator that do not cross paths for signal currents elsewhere in the board. 6.2.2 CAN Module The main form of communication between the electric steering actuator and the EVC system was through a CAN bus line. CAN uses a differential pair of wires to carry a 40 6.2. CONTROLLER BOARD digital signal to different nodes. Figure 6.3 shows the relationship between the voltage on the wires and the bit being sent. A logical 0 is a dominant bit, and in the dominant state the voltages on the wires are driven apart by the transceiver of a node on the bus line. If no dominant bits are transmitted, the line is pulled to the recessive state, a logical 1, by two 120 Ω termination resistances. In the case of the controller board, a CAN transceiver IC with a split voltage pin was used, illustrated in Figure 6.2. [37] The CAN system uses bit-wise arbitration at each node. If a node attempts to send a recessive bit at the same time as another node is attempting to send a dominant, the voltage on the bus line will show the dominant bit. This will be seen by all nodes, including the one attempting to send the recessive bit, which will stop transmitting due to the conflict and reattempt at another time. Each message begins with a message ID, and the lowest ID has the first dominant bit and therefore priority on the line. The most critical messages are given the lowest identifier to avoid delays in communication. [37] Two CAN circuits like the one shown in 6.2 were built into the controller board. One CAN bus was shared between the actuator and the EVC system, and the other was shared between both actuators and a computer used for debugging. TX PIN TX CAN HI CAN HI 60 SPLIT 120 60 RX PIN RX CAN LOW CAN LOW CAN TRANSCEIVER EVC SYSTEM ACTUATOR CAN CIRCUIT Figure 6.2: CAN Communication Circuit Dominant CAN HI Recessive CAN LO 0 1 0 1 0 Figure 6.3: Dominant and recessive bits of a CAN signal 41 6.3. DRIVER BOARD 6.2.3 PWM Module The MPC5744P has a built-in module which produces PWM signals. Due to the low inductance of the motor and a desire to improve current control, the switching frequency used was 24 KHz. Switching transistors on and off requires a finite amount of power, so higher switching frequencies lead to lower efficiency. The faster the switching time, the more closely the applied voltage resembles the desired Vavg, so the less current and torque ripple during each PWM cycle. Because most adult humans can hear sounds with frequencies up to 16 − 20 KHz[38], the PWM fundamental frequency should be kept high enough to not cause audible noise. 6.3 Driver Board The driver board uses the MC33937A three-phase gate driver from Freescale to control the switching of six transistors in a three-phase bridge circuit in order to control the voltage on each phase of the BLDC motor. This board directly controls the power going to the motor, based upon the PWM commands from the controller board. Figure 6.4: Finished Driver Board PCB for the MC33937A 6.3.1 Power Transistors Six N-channel metal-oxide-semiconductor field-effect transistors (MOSFETs) were used to construct the power stage, as in Figure 6.7. For power electronics, MOSFETs are preferable to other transistor types because they are voltage controlled, which requires less power to hold them on and allows them to be made to turn on faster [26]. Turn-on time is a concern because until the MOSFET is fully open, it has a higher resistance and thus heat losses are greater. N-channel MOSFETs are used instead of P-channel MOSFETs because N-channel MOSFETS generally have a lower on-resistance. [39] 42 6.3. DRIVER BOARD The transistors were selected based upon the current requirements (72 A peak) and power dissipation needs. The IPB180N04S4-H0 MOSFETs that were selected are rated for 180A and have a drain-to-source resistance, RDS(on) of 1.1mΩ when fully open [40]. The expected power loss during full-on conditions at peak load is: P 2max = ImaxRDS(on) = 5.7W (6.1) At the continuous load of 30A, the power dissipation is: Pcont = I 2 contRDS(on) = 1.0W (6.2) Acceptable power loss in the MOSFET is found from the thermal resistance between the MOSFET and the PCB under recommended conditions and the allowable temper- ature rise[41]. With a 6 cm2 cooling area, the thermal resistance is RthJA = 40 K/W [40]. Assuming ambient temperature of 25◦, the amount of power that can be dissipated before reaching the maximum operating temperature of 175◦ is: 175◦ − 25◦ Pok = = 3.35W (6.3) 40K/W This allows continuous operation up to 55A, but only intermittent operation above. The above calculations are only valid when the MOSFET is fully open. A MOSFET begins to open and allow current to pass through when VGS , the voltage between the gate and the source, is greater than, Vth, which for this MOSFET is 4 V . RDS(on) is a function of VGS , with a value of 2.0mΩ at VGS = 5 V and 1.1mΩ at VGS = 10 V . VGS is increased by moving charge to the MOSFET gate, as in Figure 6.5. Values for this MOSFET are Qgs = 71 nC and Qg = 173 nC. VGS QGS QGD QG QG Figure 6.5: Gate to source voltage as a function of gate charge For a switching frequency of 24KHz, the PWM period is 41.67 µs. It was assumed that a 0.5% rise time was acceptable, meaning that the gate should be fully open within 208 ns. To fully open the gate the driver must put out a current of at least: 43 6.3. DRIVER BOARD 173 nC Idrive = = 0.83A (6.4) 208 ns The MC33937A was selected for its capability of driving 1.7A of gate current [42], and the fact that it was a single-chip solution to three phase drives with built-in diagnostics to help satisfy Requirement 4.4. A ”bootstrap circuit” is needed for the high side MOSFETs. Once the MOSFET is driven open, the source voltage and the drain voltage are approximately the same at VDC , so to maintain VGS the gate voltage must be lifted above VDC . The bootstrap capacitor, shown in Figure 6.6, is charged by the supply when the low side MOSFET is open. The voltage across the bootstrap capacitor, Vboot, is approximately equal to the supply voltage minus the voltage drop of the diode. When Vboot exceeds Vth, the high side MOSFET can be driven open. When the hide side MOSFET is open, the voltage at its source and the voltage at the negative side of the bootstrap capacitor will rise to VDC , pushing the positive side to VDC + Vboot. As long as that voltage exceeds VDC by an amount of more than Vth, VGS will be large enough to hold the MOSFET open. Vcc = 12 V Vcc = 12 V 11.3 V 23.3 V D D + + CBOOT c HS CBOOT o HS 11.3 V 0 V G closed 11.3 V 23.3 V G open - - So S c 0 V 12 V Motor Motor o LS c LS open closed c o CHARGING CHARGED Figure 6.6: Operation of the bootstrap circuit showing how VGS can exceed VDC , and how internal transistors in the driver are used to conrol the voltage at the high side MOSFET gate. The input capacitance, Cis to the MOSFET is 17.6 nF . The bootstrap capacitors were chosen to be 220nF to be ”10-20 times” higher than this in accordance with design guidelines [42] so that the bootstrap capacitors do not discharge themselves in the process of opening the MOSFET gate. With only 100 nF capacitors, the motor would fail to spin at high PWM duty cycles. Because the bootstrap capacitor is only charged when the low side MOSFET is open, complementary PWM switching must be used in this arrangement; the low side should open each time when the high side closes. 44 Driver Driver Driver Driver 6.3. DRIVER BOARD Figure 6.7: Current and DC Voltage sensing circuit 6.3.2 Sensing Circuits Other key elements of the driver board are the DC voltage sensor and the overcurrent comparator. Shown in Figure 6.7, VDC is scaled by a 0.2475 : 1 voltage divider so that the expected voltage levels are within the 0 − 5 V sampling range of the MPC5744P’s ADC module. This is necessary because the controller must know what the supply voltage is to choose the proper PWM duty cycle for the control input voltage. Overcurrent detection is achieved by comparing a Voc,th = 2.63 V reference to the output of the current sensing circuit. The voltage drop across the 1mΩ sensing resistor is amplified into a readable signal, Vamp,out, that is proportional to the current. The amplifier gain in (6.5) gives a maximum allowable current of 71A from (6.6). Vout 43 kΩ = + 1 = 36.83 (6.5) Vin 1.2 kΩ 2.63 V Ioc = = 71A (6.6) (36.83) (1mV/A) 6.3.3 Voltage Input and Filtering An ideal diode circuit was used at the power input to limit the current to one direction in order to fulfill Requirement 3.4 which states that the system should be able to handle reversed input voltage. Figure 6.8 shows the configuration of this input circuit. The ideal diode driver opens the MOSFET only when the battery voltage at the input, VBAT , is 45 6.3. DRIVER BOARD higher than the voltage at the output to the motor, VDC . During operation, it is common to have the capacitors charged over VBAT by the flyback current. At this point, the ideal diode closes and no current flows towards the battery. VBAT 2.2 H VDC 9400 F 9500 F IDEAL DIODE CONTROLLER Figure 6.8: Power input to the motor driver circuit The LC circuit also filters the noise on the input line from PWM chopping and reduces the DC voltage drop that comes with pulling large currents [43]. When the high side MOSFETs are closed, the current in the motor windings freewheels while it decays, but the current flowing into the motor is cut to zero. The current built up in the 2.2 µH inductor in Figure 6.8 instead flows into the right capacitance. When a high side MOSFET opens, the current still in the motor windings now draws the charge from the capacitance until the current in the inductor is built up again. As a result of this filtering process, it is observed that the current flowing from the battery can be smaller that the average current being measured in the motor windings. The large capacitors seen in Figure 6.4 have a total capacitance of 18.8 mF . It is hard to predict exactly how much capacitance is necessary on the input, but some ”back of the envelope” calculations can give a recommendation. If the maximum expected current is 72 A, and the motor inductance is 42 µH, then the total energy that would need to be stored in the motor inductors is: 1 E = LI2 = 0.1089 J (6.7) 2 This amount of energy should be available from the capacitors without causing a large voltage drop, so we assume that 0.5V is all we can accept from a 12V supply. For this energy to be stored in capacitors with that voltage drop, the capacitance must be: 1 E = CV 2 ⇒ 2 1 E = C((12 V )2 − (11.5 V )2)⇒ 2 2(0.108J) C = = 18.4mF 11.75 V 2 (6.8) 46 IN GATE GND OUT 6.4. CURRENT SENSING BOARD This analysis is quite simple and ignores the energy lost from the resistance of the capacitors and motor inductors while the inductors are being charged, which is not a negligible term. 6.4 Current Sensing Board Connector to Controller Board ACS759 Motor Driver Connection 61x56 mm Motor Connection Figure 6.9: Finished PCB for the ACS759 Sensors For more advanced motor control, it is necessary to be able to measure the motor current in each phase. Three ACS759 Hall effect current sensors from Allegro, shown in Figure 6.9, were mounted on a board above the driver board. These sensors were chosen based upon their current rating of ±100A. The output is an analog voltage read by the MPC5744P’s ADC. 6.5 Position Sensors 6.5.1 Absolute Position The absolute position sensor chosen was the RFC-4801 from Novoteknik. This was cho- sen due to high durability, high environmental resistance (IP67/IP69k), and its feature of digital feedback, which is considerably easier to work with than analog channels. In keeping with Requirement 4.3, this device has a MTTF of 91 years, and an unlimited mechanical life because it is a contact-less sensor. 47 6.6. MODELING Figure 6.10: (Left) Stock image of RFC4801 and magnet from Novoteknik (Right) Stock image of AM256 from Renishaw The communication with the device is via serial peripheral interface (SPI). The most common SPI setup is a 4-pin interface with chip select (CS), clock (CLK), data from master to slave (MOSI), and data from slave to master (MISO) [44]. The RFC-4801 used a 3-pin interface with only a single wire for data, so SPI ports on the MPC5744P were with the MOSI pin set to ”open drain” so that the MISO pin could accept incoming data without being pulled to the level of the MOSI pin. This device allows a position update rate of 1KHz. The feedback from the sensor is a 12-bit signal, which has 4096 counts and is shown by (6.9) to be sufficient resolution. The minimum resolution needed from the sensor is set by the TargetRudderAngle signal from the PCU, which gives the reference position in 0.1◦ increments, corresponding to 0.338mm of linear travel with the linearization used. Chapter 7 gives the gear reduction to this sensor as 343 : 1. 1 sensor turn (343motor turns/sensor turn)(5mm/screw turn) = 752 counts (6.9) 0.06mm 6.75 screw turns/motor turn 6.5.2 Commutation The AM256 from Renishaw is another Hall effect device which senses angular position. It is mounted on the motor shaft and used to detect the phase of the motor for commutation purposes. The output of this sensor is two analog voltages corresponding to the sine and cosine of the magnet’s angular displacement from an on-chip reference. More details about how this signal were read are given in Section 8.1.6. 6.6 Modeling The Simulink model of the electronics components was also done using SimScape in order to generate as realistic a representation of the electronics as possible. The SimScape models are given in Appendix A. 48 7 Mechanics The mechanics and assembly of the 12V actuator are covered in this chapter. Final component selection is given along with the process used to establish the requirements of those components. 7.1 Components and Layout The final assembly is shown in Figure 7.2, and key components are rendered in Figure 7.1. 7.1.1 Ball Screw The ball screw actuator from the 36V prototype was used as the basis for this actuator. The ball screw and the worm drive were kept ”as is”, and the motor was mounted in parallel to the screw axis to make the actuator shorter in an attempt to increase the amount of tilt that the outboard engine could achieve. The efficiency of the screw is unknown, although typical efficiencies for a ball screw are about 90%, while other forms of lead screw have efficiencies of 40%− 60% [45]. The ball screw was provided by SKF. The active thread area is 47 cm, and the screw pitch is 5mm. 7.1.2 Worm Drive The worm drive is enclosed in the same housing as the ball screw. The nut which contains the ball bearings in the screw assembly is connected to the worm’s gear for a ratio of 6.75 : 1 between motor turns and turns of the nut. The worm resists back-driving. In testing, it was found that a force of roughly 1400 N is necessary to start to turn the worm by applying a linear force to the screw. The efficiency of the worm drive and ball screw combination was estimated to be 60− 70%. 49 7.1. COMPONENTS AND LAYOUT RELATIVE POSITION SENSOR BEVEL GEARBOX BLDC MOTOR CURRENT SENSOR BOARD DRIVER BOARD WORM DRIVE PLANETARY CONTROLLER GEARBOX BOARD ABSOLUTE POSITION SENSOR BALL SCREW Figure 7.1: Rendering of mechanical assembly [Modeling courtesy of SKF] 7.1.3 Bevel Gearbox A 1 : 1 bevel gear transmission was used to connect the motor axis to the input of the worm drive. It was also concluded that a design using a bevel gearbox could most easily be modified to include a manual over-ride in the form of a removable crank that could attach to the bevel gearbox on the opposite side of the output shaft, although this idea was not explored any further. The bevel gearbox used is a Tandler HW-000. This was the smallest stock component that was readily available. It nominally 90% efficient at its rated torque of 10Nm, but this value is much larger than the torques required in this prototype, so the losses are significant. The bevel gearbox is a 1 : 1 ratio connection. The motor shaft and the planetary gearbox connect with a rigid component which passes through the hollow shaft of the bevel gearbox. A second shaft comes off of this main shaft at a 90◦ angle. The worm for the worm drive is mounted on this second shaft. The BY80BL300 motor requires approximately 2.2A no load current at 5000RPM . The motor turning only the Tandler HW-000 gearbox with no load attached requires 9.0 A to run at this same speed. The bevel gearbox represents the single largest source of losses in the design of the electric steering actuator. This will be discussed more in Chapter 10. 50 7.1. COMPONENTS AND LAYOUT Figure 7.2: Completed steering actuator 7.1.4 Planetary Gearbox It was necessary to have a sensor providing information about the position of the actuator along the length of the ball screw. One method of providing this information was to use a planetary gear reducer to convert the multiple turns of the motor required to travel the length of the screw to a single turn which could be read using another position sensor. The planetary gearbox is a PD040 from Planetroll. This gearbox has a 343 : 1 gear reduction. The target range of motion was 60◦. Using the output measurements from the Teleflex actuator and the 36V prototype, it was considered acceptable to use the same 3.38mm/deg measurement as before. So, the total gear reduction necessary was: (60 deg) (3.38mm/deg) (6.75motor turns/screw turns) R = = 273.38 (7.1) (5.0mm/screw turn) The PD040 was the smallest available stock planetary gearhead found that was ca- pable of this reduction. To verify that Requirement 1.10 can be satisfied, the backlash, φb must equate to less than 1mm of movement along the screw. From (7.2), the PD040’s backlash of 0.33◦ is acceptable. 51 7.2. MODELING (360◦/sensor turn) (1mm) (6.75motor turns/screw turns) φb = = 1.42 ◦ (7.2) (5.0mm/screw turn) (343motor turns/sensor turn) The moment of inertia of this component is listed as 3× 10−6 kgm2. 7.1.5 Electronics The electronics were mounted within the housing so that the finished actuator would be a complete, self-contained unit. The left images in Figure B.1 show how this mounting was done using a long ”L” bracket. 7.1.6 Housing The housing consisted of an aluminum case in several parts. The final fit and configu- ration was designed by SKF. The individual pieces fit together with silicon paste to fill the gaps and make the unit waterproof. The weather resistance was verified by leaving the actuators mounted on the boat in the water for approximately three weeks between installation and testing. (This test was not planned; scheduling issues made it happen that way). Both actuators were fully functional upon a return to the boat. 7.1.7 Mounting More details of the actuator mounting are given in Appendix B. Figure B.1 shows mount- ing and partial disassembly. This mounting is done to allow the pivot action discussed in Figure 5.1 in Section 3.1. 7.2 Modeling The mechanics were modeled by adding inertia units to the SimScape model of the BLDC Motor in A. The efficiency of the screw and worm drive was assumed to be 60% from the previous analysis. The inertia of the full system is estimated as in (5.8) in Section 5.2.1. Several tests give an average result of about 1.75×10−4kgm2. This is quite reasonable, as it is roughly two and a half times that of the motor alone. The fully assembled actuator, including the screw, draws about 13.4A at a speed of 5300 RPM off of a nominally 13.4 V supply. This no-load torque is a result of internal viscous friction and equates to a value of about 4.78× 10−4Nms/rad. These values are based on estimates and may differ slightly between actuators. They are also dependent upon temperature. The SimScape model of the system mechanics including these parameters is given in Appendix A. 52 8 Control System The chapter discusses the control structure of the actuator system. The closed-loop motor controller, as well as the program structure of the microcontroller are given in detail. The performance of the control loop is shown in simulations in the last section of the chapter, and the simulations are compared against measured data. 8.1 Control Flow 8.1.1 Initialization When a battery voltage is first applied, the MPC5744P and MC33937A take about 2 − 3 ms to reach full voltage and begin executing their programming. The initializa- tion phase of the program of the microcontroller consists of defining the system clocks and system interrupts, initializing the General Programmable Input and Output pins (GPIO), initializing the Analog-to-Digital Converter (ADC), calibrating the ADC in- puts, initializing the MC33937A, and initializing the control loop. Initialization of the 90MHz system clock, GPIOs, communication peripherals, and MC33937A via SPI are completed in a matter of microseconds. Because the high and low side switches must be toggled as part of the MC33937A initialization phase, the PWM output does not begin until after the driver is running [42]. The ADC calibration verifies the zero level of the current sensors by taking an average reading of the voltage output over 500 samples. The current detection is triggered by the midpoint of the PWM cycle, which was set at 24KHz, so the ADC calibration requires 21ms. Once this function has been cleared, the system begins to communicate with the EVC system. The longest message cycle time, the motor current feedback to the EVC, is 100 ms, so it is 100 ms before full communication is established and full function is permitted. The actuator requires less than 0.2s after receiving power before steering is functional. This is well below the one second standard set by Requirement 1.4. 53 8.1. CONTROL FLOW 8.1.2 Control Timing The microcontroller uses a fixed-priority preemptive scheduler. If an interrupt is de- tected, the microcontroller will only perform the associated function if it is a higher priority than any of the other waiting functions. The cycle times and relative priority are given in Table 8.1. Table 8.1: Microcontroller Function Timing Function Cycle Time Priority Control Loop Update 41.67 µs 14 CAN Communication 500 µs 12 Relative Position Sensor Read 25 µs 16 Absolute Position Sensor Read 100 µs 15 Current Sensor Read −− 18 MC33937A Interrupt −− 17 Idle 0 7 The timings of these functions was not tuned beyond what gave a functional response. 8.1.3 CAN Communication Every 500 µs, the CAN communication function is executed. This function iterates through a series of CAN messages to be sent over the course of 20 function calls, as well as reading received CAN messages received from the EVC. Generally, most messages were sent with 10ms cycle times. The messages include steering feedback communication to the EVC, as well as system state reporting for debugging and data logging. The messages associated with steering control and feedback were given the lowest identifiers, and those signals used for debug- ging were given the highest, so that the bitwise arbitration would grant higher priority to the safety-related messages. Upon receiving an updated TargetRudderAngle value from the PCU, the controller would convert this value into a motor position in radians relative to a pre-calibrated zero point, and use the new value as a reference for the feedback controller. A new value was received once every 10ms. 8.1.4 MC33937A Interrupts The control of the motor driver, the MC33937A, is critical to achieve controlled motion of the motor. The MC33937A motor driver does the important task of operating the internal transistors that are seen in the single-phase bootstrap circuit in Figure 6.6 in 54 8.1. CONTROL FLOW accordance with the PWM signal coming from the MPC5744P in order to drive open the MOSFETs. After an initialization sequence leaves the MC33937A in the ENABLE state, opera- tion proceeds as normal. The MPC5744P sends PWM signals to the MC33937A, which drives the MOSFETs open according to those PWM signals. The MC33937A interrupt is a function called in response to the interrupt pin on the MC33937A being pulled high as a result of a fault detected in the driver. As part of the initialization phase, the faults which are set to generate interrupts are desaturation (a high-side MOSFET gate is unable to be opened in the allotted time), under-voltage (the driver lacks sufficient voltage to charge the gate capacitance), over-current (the current through the 3-phase bridge is too high), and phase error (similar to desaturation, but only on the low side). All interrupts except for the over-current force a reset of the driver back into the RESET state, which requires it to then be re-initialized into the ENABLE state. Over-current causes a system counter to be incremented based on the detected current at the shunt resistor, and the interrupt is cleared. If the over-current counter exceeds a certain threhold, the motor driver is shut off. Often during startup, the motor can draw very large current values for tens of mi- croseconds. Too aggressive overcurrent protection would cause frequent shutdowns in the motor driver. Instead, the overcurrent limit was based conceptually on power dissi- pation. The square of the current, which was measured every 41.7µs, was integrated and checked against a value proportional to the product of the square of 72 A over 200 ms, which was assumed to be a safe period of time to operate at that current. Larger cur- rents would exceed the threshold value more quickly that the 200 ms. Once this value was reached, the motor driver would be shut off to protect the motor and electronics. It is possible that the interrupt pin could experience a fault. As part of the main control loop, a request is periodically sent from the MPC5744P to MC33937A for a status register update via the SPI. If this status register indicates a fault, then action is taken as it would be if an interrupt signal was received. If no response is received, the MPC5744P immediately shuts down its PWM signal to prevent damage to the driver circuit. This fault control and diagnosis is meant to reduce the chance of faults which can lead to the unintended steering condition that is forbidden in Requirement 4.1. 8.1.5 Current Feedback The current sensor is given the highest priority because the timing of the current detec- tion needs to be done at the same time each PWM cycle in order to ensure a meaningful and consistent current measurement [27]. A software trigger was programmed to cause an interrupt at the peak of the PWM carrier wave, so that sampling could be done in the middle of the ”open” period of the switching cycle regardless of duty cycle. The current used in the case of six-step commutation was the total DC current through the motor at that moment. This was measured from all three phase current sensors: 55 8.1. CONTROL FLOW 20 Supply Current Sampled Phase Currents Reported Current (CAN) Instantaneous Phase Current 15 10 5 0 -5 0.226 0.2265 0.227 0.2275 0.228 0.2285 0.229 0.2295 0.23 Time (s) Figure 8.1: Current sampling and real currents (Simulation) 1 Is = (|Ia|+ |Ib|+ |Ic|) (8.1) 2 The sign of Is was established based on the commutation period and which sensors reported a positive current and which reported negative. Attempts at FOC used the procedure for solving for Iq from the phase currents as discussed in Section 4.5.2. Several different currents can be discussed when talking about motor current. Figure 8.1 illustrates some of this. The black lines represent the instantaneous phase currents– the current flowing through the coils of wire in the stator. Each of these is sampled in the middle of the PWM pulse, producing the smoothed red line, which is used for current feedback control. Every 10 ms the microcontroller gives a state feedback over the CAN bus for debugging and datalogging purposes. The current reported via CAN, which is used in many cases in this report when the actuator’s current consumption is being discussed, is a filtered version of the current as sampled over that period of time. Finally, the blue line shows the supply current flowing from the battery into the actuator, which is lower than the current in the motor windings. The DC supply voltage to the motor, VDC is sampled during the current sampling phase. 56 Current (A) 8.1. CONTROL FLOW 8.1.6 Position and Speed Feedback The relative position sensor is read at a faster rate than the control loop update speed so that the reading can be filtered before the controller receives it. The timing of the control loop was set to the same rate as the PWM frequency so that the controller could update the PWM duty cycle with each PWM reload. One byte of data is exchanged with the absolute position sensor every 100 µs. To receive a position update, a 10 byte message must be exchanged with the RFC position sensor via SPI. The 12-bit position is contained in two bytes, and an inverse of the position is given in another two bytes for verification purpose. If there is a mismatch, the position data is discarded and the last good value is used. The update rate of the RFC position sensor is 1KHz. The relative position sensor was used for commutation purposes as well as position and speed feedback. The output of the AM256 rotary magnetic encoder is in the form of two analog sine waves, offset by π2 . During the control loop, the voltage levels of the last measurement are analyzed to determine the position of the rotor. The calculation of this is done by comparing the measured values Vsin and Vcos against calibration data of minimum and maximum for the position sensor: ( ) Vsin,max+Vsin,min φ = sin−1 Vsin − 2 s (8.2)( Vsin,max − Vsin,min ) − VV cos,max+Vcos,min φ = cos−1 cos 2 c (8.3) Vcos,max − Vcos,min φ = mφs + nφc (8.4) The value of m and n that were used depended upon the position in the rotation. The angle estimate for φ should be more heavily weighted to the sine wave voltage around 0◦ and to the cosine wave voltage around 90◦ to make the most precise measurements, and the same for other quadrants. A startup, the absolute position of the actuator, θa, was taken from the position of the absolute position sensor, θRFC . To keep the position update fast enough, and to account for potential drift between the accumulated position based on the rotational sensor and the position of the absolute sensor, a complimentary filter was used to resolve the final position based on both sensors. Testing showed that α = 0.88 was suitable. θ[k] = α (θ[k − 1] + φ[k]− φ[k − 1]) + (1− α) θRFC . (8.5) The speed was calculated from this during the CAN message cycle as part of the same function call. The speed was filtered to try to minimize noise as well as to reduce the effect of the real speed fluctuation due to th(e current ripple o)n the control loop. θa[k]− θa[k − 1] ω[k] = βω[k − 1] + (1− β) (8.6) ∆t 57 8.2. CLOSED LOOP CONTROL 8.1.7 Commutation Six-step commutation was used for the primary testing, due to the relative simplicity. When the final position was determined via the position sensor, the active phases were determined as in Table 8.2. The lowside MOSFET for the negative phase was held open by disabling the PWM on the output pin and setting the GPIO corresponding to the high side to logic 0 and the low side to logic 1. The positive phase is given the PWM duty cycle corresponding to the reference voltage desired: Vref duty = × 100% (8.7) VDC To rotate in the negative direction, the positive and negative phases were reversed. The table is based off of φe = 2φ because there are two pole pairs, so the commutation must happen twice per motor rotation. Table 8.2: Commutation Table φe Phase A Phase B Phase C 0◦ − 60◦ + − 0 60◦ − 120◦ + 0 − 120◦ − 180◦ 0 + − 180◦ − 240◦ − + 0 240◦ − 300◦ − 0 + 300◦ − 360◦ 0 − + 8.2 Closed Loop Control Closed-loop control of the motor position, speed, and current are used to quickly and accurately reach the targeted rudder angle without exceeding current limitations. 8.2.1 Cascade Controller The feedback controller used for the electric steering actuator is a cascade loop controller, based on a structure given in [46]. In a single loop controller, the output of the controller is a control variable which when given as an input to the plant will steer the primary system state variable in the desired way. A cascade loop controller is a controller in which the output from the controller of the primary control loop is fed into a second, faster control loop as the set point. The controller of the inner loop outputs the control variable necessary to steer a secondary state towards this set point, and in doing so assures that the system’s primary state variable is controlled. Figure 8.2 shows the configuration of this in this application. There are three cascade loops used for the electric steering actuator: the position loop, which is the primary 58 8.2. CLOSED LOOP CONTROL system state variable, the speed loop, and the current loop. One benefit to cascade loop control is setting of saturation limits for the secondary state variables. For example, if the output of the speed controller requests a value for the current set point, Ir, that is beyond the 72 A limit, the value is kept at 72 A before going into the current control loop. This will keep the current at safe levels even under high loading. b kv d d d r Position r Speed ir Current Vr V Actuator i k Actuatort Controller Controller Controller Electrical Mechanical Figure 8.2: Actuator cascade control loop with position, speed, and current controllers. For a cascade control loop to be used, there must be a relationship between the primary state variable and the secondary state variable(s). In the case of the BLDC motor, this relationship is clear in that the position of the motor is the integral of the motor’s speed, and the speed is the result of the acceleration caused by the motor’s torque, which is proportional to the current. The inner loop must also be at least 5− 10 times faster than the outer loop [47]. This is satisfied by the electrical speed vs. the mechanical speed of the motor. 8.2.2 IMC Scheme The position, speed, and current controllers in the cascade structure were designed using the Internal Mode Control scheme. IMC Scheme Theory Internal Model Control (IMC) is a control framework based around the concept that if a model of the system exists, the feedback controller should be used only to correct for new information that deviates from this model, such as disturbances or model errors [48]. The IMC scheme provides a fairly easy framework for controller design in which a limited number of parameters are required for tuning. Beginning with an open loop system in Figure 8.3 where the controller Gc(s) is used to control a process Gp(s), it follows that the system output Y (s) is equivalent to R(s)Gc(s)Gp(s). If Gp(s) can be modeled accurately as M(s), then perfect setpoint tracking is achievable via G (s) = M−1c (s) [49]. Feedback is unnecessary if there are no model inaccuracies or disturbances. An illustration of this is shown in Figure 8.3. There will always be model inaccuracies and disturbances, so closed loop control is still necessary. The IMC uses the feedback loop only to account for these disturbances and model inaccuracies, as illustrated in Figure 8.4. 59 8.2. CLOSED LOOP CONTROL R(s) Y(s) Gc(s) Gp(s) Figure 8.3: Open Loop IMC D(s) R(s) U(s) Y(s) Gc(s) Gp(s) M(s) Figure 8.4: Structure of IMC Scheme ∫ t d u(t) = Kpe(t) +Ki e(τ)dτ +Kd e(t) (8.8) 0 dt The objective is to find an equivalent PID controller in the form of (8.8) for the IMC controller Gc(s) [50]. This requires expressing the controller for a traditional PID structure, given in NEED EQ in terms of Gc(s) and M(s). It can be shown from Figure 8.5 that for the traditional PID case, the controller output U(s) is C(s)(R(s) − Y (s)). Taking the feedback loop in the IMC schematic in Figure 8.4 reveals that: Gc(s)(R(s)− Y (s)) U(s) = (8.9) 1−Gc(s)M(s) So it follows that: Gc(s) C(s) = (8.10) 1−Gc(s)M(s) D(s) R(s) U(s) Y(s) C(s) Gp(s) Figure 8.5: Final Closed Loop Form of IMC Scheme 60 8.2. CLOSED LOOP CONTROL The IMC algorithm calls for factorizing the plant model into minimum phase and non-minimum phase components [50]. Fortunately, both the mechanical and electrical systems involved in the BLDC motor are minimum phase, and time delays are considered to be negligible. The controller, Gc(s) in the above figure is then chosen as the inverse of the minimum phase component, M(s)−1. If this controller is not stable, causal, or proper, a filtering function is used: f(s) = (λs+ 1)n (8.11) The final step is to express this function in terms of the traditional PID controller, Kp + Ki s + Kds. In this case, we opt for the simpler PI controller, as it was found that the derivative term is unnecessary. K M−1i (s)f(s) C(s) = Kp + = − (8.12)s 1−M 1(s)f(s)M(s) Substituting in the transfer function for the electrical part of the system plant, (8.13), into (8.12) yields the structure of the controller, given in (8.14). 1 Gp(s) = (8.13) sL+R L R C(s) = + (8.14) λ λs Only the value λ is left as the tuning value to determine the time constant of the regulator [50]. Tuning is a trade-off between robustness and performance. Absent dis- turbances or model errors, the closed loop response as shown in Figure 8.5 using (8.13) and (8.14) is: Y (s) 1 = (8.15) R(s) (λs+ 1)n With IMC, the system will behave as a first-order system with a bandwidth of λ−1. The same procedure is done and the controller for the speed loop looks like: J b Cm(s) = + (8.16) λm λms Loop Parameters For the IMC scheme, the values used should be: L Kp,e = (8.17) λ R Ki,e = (8.18) λ 61 8.2. CLOSED LOOP CONTROL The choice of lambda is based on the response time of the system. The time constant of the RL circuit based on the resistance and inductance in the BY80BL300 is: L τ = = 1.33ms (8.19) R For the current loop, a value of λ = 166 µs was decided upon from tuning. From here, the value for Kp,e and Ki,e were calculated to be 0.25 and 197, respectively. The values for speed are the inertia and the viscous friction constant. These values were estimated at J = 1.75 × 10−4 kgm2 and b = 4.78 × 10−4 Nms/rad. The time constant for the mechanical system is: J τm = = 366ms (8.20) b It was decided that for the speed control loop to see the current as a constant value, the current controller should be much faster than the speed controller [47]. The speed controller should also be several times faster than the mechanical system. Without trying to make the speed loop too fast, a time constant of 90ms was chosen. Based on this, values of .02 Nms/rad and .0054 Nms/rad were chosen for propor- tional and integral terms. With the torque constant kt of 0.02 Nm/A, the final values for Kp,m and Ki,m for the speed loop were 0.8 and 0.27. The value for Kp,m is much larger than the calculation should yield, which is a value of around 0.1. Testing with these values revealed that improvements could be found with a bit of tuning. As with other PID tuning methods, the IMC scheme gives a theoretical framework from which a guideline can be derived, but to reach the performance goals of the system a further tuning is generally necessary. Microcontrollers do not deal with continuous states. All computations are done with discrete numeric values, so the continuous form of the transfer function in terms of the Laplace variable ”s” must be discretized in terms of the ”z” operator to be useful in code. The final controller, was not discretized as a single transfer function because it is necessary to maintain a separate counter for the integrator value for reasons discussed in the next section. Each term is trivial discretized in a fairly simple way: a proportional term re- mains proportional, and the integral term is divided by the controller update frequency, 24KHz, and multiplied by the accumulated integral error. 8.2.3 Feedforward Terms The idea behind the IMC scheme is to only use the feedback controller for handling unknown values. With the BLDC motor, we have two known values which the controller can use to improve the response of the system. The first is the Back EMF. The voltage that the current controller needs to apply at a given speed to maintain that speed is known as: V1 = 0.02ω (8.21) 62 8.2. CLOSED LOOP CONTROL Table 8.3: Controller Parameters Used Parameter Value Units Kp,e 0.25 V/A Ki,e 197 V/A Kp,m 0.8 As/rad Ki,m 0.27 As/rad K −1p,s 250 s Kemf 0.02 V s/rad Knlc 0.024 As/rad In addition to this, the no-load current which results from viscous friction is known as a function of the motor speed. At roughly 13.4A no load current at 5300RPM , the no load current per Amp is estimated at: If = 0.024ω (8.22) In reality, this is not a linear relationship, but no-load current consumption at a few speed lower than the maximum have shown that a linear approximation is close enough. 8.2.4 Position Control The position control loop uses a proportional controller only. The error between the target position and the actual position is amplified and used as the reference speed. The value of this gain is Kp,s = 250s −1, and it was tuned manually. The idea being that the controller should try to drive the actuator to the maximum speed as quickly as possible in all cases, unless the target position is very close by. The phenomenon of ”hunting”occurs when a controller is close to, but unable to settle exactly upon, the reference target. Within a small range, the controller will continuously try to move the system back and forth around the target point. This may be because of unaccounted for nonlinearities in the system (backlash, switching delay), or due to small errors in the position sensor. This needlessly consumes energy and generates both audible and electrical noise. To prevent the controller from hunting, a deadband is introduced into the system. The deadband tells the controller when it is ”close enough” so that hunting behavior is not seen. The control signal from the HCU is given to a precision of 0.1◦, so it is not necessary for the actuator to attempt to be any more accurate than within half this value. The deadband width, φdb, in motor radians is: 63 8.2. CLOSED LOOP CONTROL (0.05◦) (3.38mm/◦) (6.75motor turns/screw turns) (2π rad/motor turn) φdb = = 1.43rad (5.0mm/screw turn) (8.23) When the position error was less than this, the position error was set to zero and all integrator memory functions were zeroed out. In order to prevent any collisions between the actuator and the boat body, a hard stop limit was set for the position. If the signal from the HCU attempted to set the target position beyond this limit, the reference position would be set to the hard stop limit. 8.2.5 Anti-Wind Up No real world actuator behaves in a truly linear manner; all are subject to some restric- tions in their force, speed, or range of motion. Such a limitation in the system is called a saturation nonlinearity, and it leads to a phenomenon called ”wind up”. ”Wind up” is best described as in [51] as ”the degradation in performance that occurs when a saturation nonlinearity is inserted, at the plant input, in an otherwise linear feedback control loop.” The chief saturation nonlinearities that have to be dealt with in the electric steering actuator are the limited voltage of the boat’s battery, nominally 12.0V , but generally floating at 13.6V , and the current-handling capability of the motor and the power electronics, capped at 72 A for a short term. The actuator has a finite range of movement to consider as well. The performance degradation from a saturation nonlinearity is due primarily to the integrator term in the closed-loop feedback controller. With a saturation nonlinearity at the plant input, the integrator terms can ”wind up” to larger values than the controller is capable of outputting. When this effort is finally dissipated into the system, it will be at an undesirable time and can cause the system to behave with a sluggish, divergent, or oscillatory response [51]. Reference 1200 With Anti-Windup Without Anti-Windup 1000 800 600 400 200 0 0 0.5 1 1.5 2 2.5 3 Time (s) Figure 8.6: Simulation of system with and without anti-windup showing that anti-windup control prevents overshoot due to actuator saturation. 64 Position (radians) 8.2. CLOSED LOOP CONTROL Feedforward Terms Saturation Error kp u=0 u 0 ki 0 u Figure 8.7: Basic concept of clamping for anti-windup. When the output of the controller is different from the saturated output, integration is halted. The simplest approach to anti-windup is clamping. Clamping prevents integration when the controller output is already at saturation [52]. In Figure 8.7, a clamping controller is shown in block diagram form. In Appendix A, Simulink models of the speed and current controller are shown featuring anti-windup as in Figure 8.7. The results of simulation with and without this anti-windup adjustment are shown in Figure 8.6. Here the actuator’s saturation nonlinearity prevent it from tracking a ramp on the reference. Due to the ”winding up” of the integrator, the position overshoots the reference. By introducing the anti-windup control, this issue is prevented. 8.2.6 Evaluating the Controller Closed Loop Controller Response The current control step input in Figure 8.8 shows the degree to which the current can track the reference. There is good agreement between the model and the actuator in terms of the levels of current chatter. The use of a six-step commutation with a low inductance motor causes a large degree of current fluctuation that is difficult to control without creating a much faster control loop than was practical in this application. The target current rise time of τr = 4λ = 0.664 ms is much faster than the 10 ms update speed of the CAN feedback used to monitor the motor current, so the rise time appears to be instantaneous. The speed response in Figure 8.9 shows a fairly satisfactory behavior when tracking a speed input. Overall, the actuator tracks a 2000 RPM reference input with 27 ms, and the 5300 RPM reference with about 120 ms of rise time. It does not quite reach the 5300 RPM target, as the voltage is insufficient to achieve this speed with the large no-load current. The actuator saturation is obvious in the current tracking plot for this reference target. The current controller requests a current of 72 A to provide the actuator with the torque necessary to accelerate as desired, however, once the speed increases, the actuator is unable to draw that much current off of the 13.6 V source. In Figure 8.10, the actuator is not able to output the maximum speed desired by 65 8.2. CLOSED LOOP CONTROL Measurements 20 Simulation Output Reference Current 10 0 -10 0 1 2 3 4 5 6 Time (s) Figure 8.8: Current step response showing the degree of current tracking achieved and the agreement in current behavior between the simulation and the actuator. 6000 4000 2000 Measurements 0 Simulation Output Reference 0 0.05 0.1 0.15 0.2 0.25 0.3 Time (s) 100 50 0 -50 0 0.05 0.1 0.15 0.2 0.25 0.3 Time (s) Figure 8.9: Speed step response showing simulation accuracy and actuator saturation behavior in both speed and current. the controller, however it is able to drive nearly the full range of motion in roughly 2.5 seconds, though the model seems to disagree very slightly in terms of the maximum achievable speed, and thus the time required for the stroke. 66 Current (A) Speed (RPM) Current (A) 8.2. CLOSED LOOP CONTROL 1000 500 0 -500 Measurements Simulation Output Reference -1000 0 0.5 1 1.5 2 2.5 3 3.5 4 Time (s) 800 600 400 200 0 0 0.5 1 1.5 2 2.5 3 3.5 4 Time (s) Figure 8.10: Position step response showing simulation accuracy and maximum actuator speed. To determine if the controller were satisfactory or not, a simulation was done of the performance requirements given in Section 2.1. The results are given in Table 8.4. This simulation established that the performance requirements could be met by this controller for this system. Amplitude A2 for the 5◦/s slew rate and T4 for the 20◦/s slew rate are suspected failures, although the specifications do not give any indication that a negative value for A2 is unacceptable for the 5◦/s slew rate test. 67 Position (Radians) Speed (rad/s) 8.2. CLOSED LOOP CONTROL Position Output Position Command 3 3 2 2 1 1 0 0 0 1 2 0 0.5 1 1.5 Time (s) Time (s) 3 3 2 2 1 1 0 0 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 Time (s) Time (s) Figure 8.11: Performance Requirement Simulation. . Table 8.4: Simulation response at different angular request slew rates. Requested Speed (deg/s) T1 (ms) T2 (ms) T3 (ms) T4 (ms) A1 (deg) A2 (deg) 2.5 80/260 55/460 80/260 80/185 0/-0.1 -0.04/-0.1 5 80/195 55/280 80/195 80/120 0/0.03 -0.04/0.03 10 70/165 55/190 85/165 85/95 0/0.26 -0.04/0.26 20 90/125 55/165 35/185 95/95 0/0.8 -0.04/0.37 Field Oriented Controller Behavior The Field Oriented Control Loop was also tested, though not as rigorously as the Six Step Commutation controller was in Section 8.2.6. 68 Rudder Angle (Degrees) Rudder Angle (Degrees) Rudder Angle (Degrees) Rudder Angle (Degrees) 8.2. CLOSED LOOP CONTROL 15 10 5 0 Six-Step Commutation FOC -5 0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4 Time (s) 10 5 0 -5 0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4 Time (s) Figure 8.12: Torque-generating current and total supply current comparison (simulation) using FOC is more constant than with six-step commutation. Figure 8.12 shows that the FOC controller is able to maintain a smoother quadrature current, meaning a smoother torque, while drawing a more consistent current from the battery. Electrical and audible noise were noticeably reduced in practice. The reason for this shown in Figure 8.13. The current and back EMF should be in phase to maintain a steady torque level. In the top graph in this figure, the six step commutation controller produced the characteristic waveform in the phase A coil. This is not an ideal match to the sinusoidal back EMF current. In the bottom picture, the FOC controller is able to produce a sinusoidal phase current to match the back EMF. The FOC required more tuning and was far more sensitive to errors in the commu- tation position, so development with that controller ceased. Figure 8.14 shows that the phase currents in practice were not the desired sinusoidal shape. This is largely due to the fact that the low inductance and high speed of the motor make precise control of the current very difficult. The abrupt change in the phase currents in the top graph of the figure is due to the fact that the CAN feedback was not fast enough to give real-time phase current data (10 ms gives only 4 − 6 data points per commutation period at top speed), so the debugging program was set to record 32 ms of phase current data, and then play the record back over the course of 8 seconds. The change in current in this instance happened while the recorder was playing back data and not recording it. The image is meant to illustrate the phase current shape and relative current magnitude to demonstrate the limited amount of success with the FOC scheme. 69 Supply Current (A) Quadrature Current Iq (A) 8.2. CLOSED LOOP CONTROL 20 4 10 2 0 0 -10 -2 -20 -4 0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4 Time (s) 20 4 10 2 0 0 -10 -2 -20 -4 0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4 Time (s) Figure 8.13: Phase current and back EMF (simulation) showing FOC algorithm keeping current in phase with back EMF. 20 10 0 -10 -20 15 10 5 Measured Reference 0 0 5 10 15 Time (s) Figure 8.14: (Top) Phase currents with FOC. (Bottom) Current response of actuator. 70 Current (A) Current (A) FOC Phase Current (A) 6 Step Phase Current (A) Back EMF (V) Back EMF (V) 9 Testing and Results This chapter gives the results of testing with the actuator. Load tests were conducted to determine the speed and current consumption of the actuator against loads of varying size. Tests were performed to verify steering functionality when interfaced with the EVC system in a laboratory environment. The final test was a live test with the actuators installed on a twin-engine setup on a 30’ boat. 9.1 Load Tests The load tests consisted of running the actuator in a vertical position while supporting a weight of up to 350kg. The speed vs. load results are shown in Figure 9.1, where they are compared to the results from the load tests with the 36V actuator. The 12V actuator is only capable of a stroke of ±20◦ in the 3.2 seconds under 340kg of load as specified by Requirement 1.6. Mechanical efficiency is the primary concern of the actuator. Using the ball screw formula in (9.1) to determine the torque experienced by a linear load, it is estimated that the overall mechanical efficiency, including friction torque of the motor and bevel gearbox, is 72%. Based on the screw torque calculation repeated below, the current, including the no-load torque at the measured speeds, follows the expected value for a screw and worm configuration with this efficiency. This is higher than the previously estimated 60%, but these values are only estimates. FPh Td = (9.1) 2πη1 The current vs. speed plot in 9.2 shows that the motor behaves in a manner predicted by the model (average current values plotted). It is the high torque requirements for a given linear load that drives the current up and the speed down. 71 9.1. LOAD TESTS 60 40 12V Actuator 36V Actuator ° 20 ±30 25°± ±20° 0 0 500 1000 1500 2000 2500 3000 3500 40 30 20 12V Actuator 72% Screw Efficiency 10 0 500 1000 1500 2000 2500 3000 3500 4000 Load (N) Figure 9.1: Load Test Comparison. The 12V actuator falls short of the performance tar- gets, achieving about ±21◦ range of motion in the allotted time. Screw efficiency calculations estimate 72% efficiency for the screw drive. 4500 4000 3500 3000 2500 Simulation Results (11.9V / 11.0V) Measurements 2000 5 10 15 20 25 30 35 40 45 50 Current (A) Figure 9.2: Speed and Current in Simulation and in Testing. The test results agree with model simulations which account for the voltage drops observed in testing. 72 Speed (RPM) Current (A) Speed (mm/s) 9.2. TEST RIG 45 40 35 30 25 10 15 20 25 30 35 40 Current (A) Figure 9.3: MOSFET temperature rise during load test is moderate. Maximum operating temperature for the transistors is 175◦C. During the load test, it was observed that the DC voltage dropped from 11.9 V with an unloaded actuator (approximately 15 kg) to 11.0 V with 350 kg of load. This may be due to a weakened battery or to circuit breakers and cables rated for only 40 A. This voltage drop corresponds to roughly 23 mΩ of resistance. The plot in 9.2 shows the current vs. speed curve from the model for both 11.9 V and 11.0 V . The measured values appear within this range. In Figure 9.3 it is shown that the maximum temperature rise of the MOSFETs during the load test was lower than expected from Section 6.3.1. Using ample surface area, many thermal vias, and a large heatsink in the form of the mounting bracket helped to keep the temperature down below the data sheet values. Because the actuator was running at full duty cycle up, and then running down with a much lower current, it is best to consider that the power dissipation during these tests was half of what it would be in steady state. This would meant that the practical thermal resistance of the MOSFETs appears to be 24K/W , and that the MOSFETs may possibly handle the full 72A continuous. 9.2 Test Rig The electric steering actuator was tested using the actual Yamaha Helm Master hardware (Powertrain Control Unit (PCU), Helm Control Unit (HCU), displays, battery, and steering wheel) in a laboratory setting. The actuator was tested both independently as well as in parallel with a Teleflex electro-hydraulic steering actuator that is currently part of the Helm Master system. This test was to verify that the CAN communication between the actuator and the Helm Master PCU was fully functional in both directions, as well as to test the control system against the reference signals that would be expected in actual operation. Figure 9.4 gives an example of the behaviour of the electric steering actuator in response to a target rudder angle. The position readout from the electric steering actu- ator is done using the complimentary filter and the two rotational sensors. The Teleflex actuator uses a linear sensor. The validity of the comparison was established by mea- 73 Maximum Temperature (°C) 9.2. TEST RIG Figure 9.4: Reference tracking comparison. The 12V actuator tracks the reference within 120ms vs. the Teleflex actuator which has a 200ms delay. Target Rudder Angle (Starboard) Teleflex Rudder Angle (Starboard) 20 Electric Actuator Rudder Angle (Starboard) 10 0 -10 -20 47 48 49 50 51 52 53 54 Time(s) Figure 9.5: Reference tracking comparison. Electric steering actuator generally performs better than the Teleflex actuator. suring the displacement of the hydraulic cylinder against the linear travel of the screw in the electric steering actuator; both had the same displacement for the same value of the CurrentRudderAngle signal. In Figure 9.4, the position tracking delay is between 100− 120ms for the electric steering actuator, and roughly 200ms for the Teleflex ac- tuator. This delay was fairly consistent during the testing. The delay was considered acceptable based on the performance requirements given in Section 2.1. It might be remembered that during the simulation, the delay was roughly 80 ms. This was accurate for those control parameters, however the control parameters used with the boat installation and test rig were made less aggressive in order to more smoothly handle the nature of the input. Figure 9.5 gives another example of a more natural steering curve. The electric steering actuator once again maintains a 100-120 ms delay at turns of up to 13.5◦/s (It is fairly difficult to turn the wheel any faster than this.) and tracks the signal well. Figure 9.6 gives a final example showing the reference tracking over a longer period of time. 74 Rudder Angle (deg) 9.2. TEST RIG Target Rudder Angle (Starboard) 30 Electric Actuator Rudder Angle (Starboard) 20 10 0 -10 -20 -30 230 232 234 236 238 240 242 244 246 Time(s) Figure 9.6: Reference tracking. Electric steering actuator tracks reference target well. With a programmed sequence input, the steadiness of the input was less of an issue, and the more aggressive of the two controllers could be used without high levels of noise or current hunger. The performance of the actuator was evaluated against the requirements in Section 2.1. Figure 9.7 shows the response to the different slew rate inputs, at 2.5◦ − 20◦ slew rate. Table 9.1 shows the timing results as compared to the requirements. The actuator is able to achieve the results desired, although it is clear that it is unable to track the slew rate of 20◦/s. The response is fast enough that the 1◦ target can be hit before the lagging speed results in a much longer delay. This is obvious from the change in direction happening with the actuator at 2◦ instead of the desired 3◦. The documentation detailing the requirements gives no indication that this value should be measured in anything other than this manner. Because of the nature of the CAN protocol, the results of these tests should be considered valid within ±10 ms. It is entirely possible that the receipt of the target message could be delayed while other communication on the CAN bus is processed, likewise it is possible that the position responses could be nearly 10ms old by the time they are received by the monitoring software. For this reason, the narrow margin on T4 for 10◦/s and 20◦/s should not be considered as passing values. In further development, a more aggressive position control loop with integral weighting should be considered. The requirements do give some leeway in terms of overshoot. In all cases, the system response was well-within the 200ms specified by the ABYC for Requirement 1.7. Based upon the results of the tests on the simulator and the load tests in 9.1, it was determined that because the actuators could adequately track the reference signal and handle the expected loads, it was safe to test on the boat. 75 Rudder Angle (deg) 9.3. BOAT TESTS Position Output Position Command 3 3 2 2 1 1 0 0 0 1 2 3 0 0.5 1 1.5 2 Time (s) Time (s) 3 3 2 2 1 1 0 0 0 0.5 1 0 0.2 0.4 0.6 Time (s) Time (s) Figure 9.7: Results of the performance test. 2.5◦/s at top left, 5◦/s at top right, 10◦/s at bottom left, and 20◦/s at bottom right. Table 9.1: System response at different angular request slew rates Requested Speed (deg/s) T1 (ms) T2 (ms) T3 (ms) T4 (ms) A1 (deg) A2 (deg) 2.5 80/260 70/460 90/260 80/185 0/-0.1 <.01/-0.1 5 80/195 50/280 70/195 80/120 0/0.03 <.01/0.03 10 80/165 50/190 50/165 85/95 0/0.26 0.1/0.26 20 70/125 40/165 20/185 95/95 0/0.8 0.2/0.37 9.3 Boat Tests 9.3.1 Installation Sea trials were carried out with the electric steering actuators. The vessel used in the test was a 30’ boat propelled by two Yamaha F250 outboard engines in a twin installation. The boat used Yamaha’s Helm Master electronic control system. Both the port and starboard engines were fitted with one of the electric steering actuators as developed in this work. 76 Rudder Angle (Degrees) Rudder Angle (Degrees) Rudder Angle (Degrees) Rudder Angle (Degrees) 9.3. BOAT TESTS Figure 9.8: Mounting of Actuators on Boat Figure 9.8 shows how the sizing of the actuator impacted the installation and limited the range of motion to only ±24◦ from the original target of ±30.6◦ Figure 9.9: Mounting of Actuators on Boat 77 9.3. BOAT TESTS 9.3.2 Tracking Performance Figure 9.10 shows a sampling of the starboard engine’s rudder tracking the input ref- erence at a speed of 10-20 knt. The steering CAN bus did not include engine speed information, so speed estimates are from average readings off the boat’s speedometer. The bottom of figure 9.10 gives a close-up of an 11.5◦/s turn of the rudder. At this turning rate, the actuator is able to track the reference input with an approximately 120ms delay. This is well within the specification as given in Section 2.1. From the figure, it is observed that the position control loop could have been more aggressive to avoid slowing down when nearing a setpoint. Figure 9.11 shows tracking over a longer period of time. No problems were observed in tracking the reference on the water. It was not possible to turn the wheel quickly enough to request a turning speed faster than the actuator could achieve. 10 5 0 -5 -10 0 5 10 15 Time (s) 10 5 0 -5 Target Rudder Angle Rudder Angle (Starboard) -10 6 6.5 7 7.5 8 8.5 Time (s) Figure 9.10: (Top) Starboard engine on open water showing reference tracking. (Bottom) Detail. 9.3.3 Current Consumption Figure 9.11 shows the current consumption over a wider range of motion and longer time. Current consumption was well below the 70 A continuous maximum specified in Requirement 3.2. The noisy current, difficulty in repeating exact steering maneuvers, and difficulty in maintaining a constant boat speed made it difficult to replicate steer- 78 Rudder Angle (degrees) Rudder Angle (degrees) 9.3. BOAT TESTS ing maneuvers at different speeds to judge the impact that boat speed had on current consumption. Not enough data was gathered to attempt to develop a model relating boat speed and rudder angle to steering actuation force, although such a model could be extremely important for further design of a steering system. Other measurement tools aside from the motor current of the steering actuator should be used in this testing. 40 20 0 -20 Target Rudder Angle Rudder Angle (Port) -40 50 55 60 65 70 75 80 Time (s) 100 50 0 -50 -100 50 55 60 65 70 75 80 Time (s) Figure 9.11: Motor current at 5 knots Figure 9.12 and Figure 9.13 show a similar turn at 5 knots and 20 knots and the similar current consumption of both maneuvers. Both maneuvers averaged 9.5◦/s. At 5 knots, the motor current over the shown range was an average of −21.4A. The current is negative due to the direction the motor was turning. At 20 knots, the motor current averaged −22.6A. A 5% difference in motor current with the slight variation in steering maneuver does not given enough information to draw any conclusions about the effect of boat speed on steering forces. Due to limited testing time and weather conditions, the tests were conducted at 28 knots and below. At higher speeds, the effect may be more pronounced and some useful data could be gathered, but given the actuator’s already excessively high current consumption at these speeds in its present state, this data would not be useful to the next prototype development stage. 79 Motor Current (A) Rudder Angle (degrees) 9.3. BOAT TESTS 20 Target Rudder Angle 10 Rudder Angle (Port) 0 -10 -20 59.5 60 60.5 61 61.5 62 62.5 63 63.5 60 40 20 0 -20 -40 -60 59.5 60 60.5 61 61.5 62 62.5 63 63.5 Time (s) Figure 9.12: Motor current at 5 knots 20 Target Rudder Angle 10 Rudder Angle (Port) 0 -10 -20 215 215.5 216 216.5 217 217.5 218 218.5 219 60 40 20 0 -20 -40 -60 215 215.5 216 216.5 217 217.5 218 218.5 219 Time (s) Figure 9.13: Motor current at 20 knots 80 Motor Current (A) Rudder Angle (deg) Motor Current (A) Rudder Angle (deg) 9.3. BOAT TESTS Target Rudder Angle (Starboard) 15 Rudder Angle (Starboard) Rudder Angle (Port) 14.5 CAN Bus to EVC 14 13.5 13 12.5 88.5 88.6 88.7 88.8 88.9 89 89.1 89.2 89.3 89.4 89.5 Time (s) Figure 9.14: Steering Actuator Behavior with Loss of CAN Communication. When the starboard side loses CAN communication with the EVC, the target rudder angle used is the port side rudder angle. 9.3.4 Fault Detection Communication Loss The control system in the steering actuators is equipped with a basic fault detection sys- tem in case the steering unit loses contact with the EVC temporarily. This functionality was tested during the boat tests. In Figure 9.14, the rudder angles for port and starboard side are shown, as well as the target rudder angle for the starboard side. Between 88.9s and 89.0s, the CAN communication between the EVC and the starboard steering actuator is interrupted for a period of 1-2 seconds. After 4 consecutive missed CAN frames (0.04s), the actuator goes into a fault recovery state. During this time, the starboard actuator uses the port actuator’s position as the reference target so that normal steering can occur until communication is restored. The motivation for this system is from Requirement 4.1 in Section 2.4, which specifies that unintended steering can not occur, including also that steering inputs cannot be ignored. This function is disabled when the engines are pointing in different directions, such as in docking mode. In that case, both steering actuators are held in their current position until communication is restored to avoid an engine collision. (This functionality was not tested.) This is not an ideal solution, as it violates the previous statement about unintended steering, but testing could not be delayed until a solution was found. Due to the low speeds, docking mode was found to be below the SIL scale, so it was not considered a safety hazard. It was realized after the test that port and starboard reference targets can differ by several degrees due to the settings of the EVC system. The tracking should account for that, but did not. 81 Rudder Angle (degrees) 9.3. BOAT TESTS Overcurrent The overcurrent detection system has also been tested in the steering application. When detecting an impact or excessively high current draw, the actuator would successfully shut down the motor driver to avoid damaging the electronics or the motor. When a locked rotor condition was triggered in the actuator, during a voltage step, the current increased over the 72A limit that was allowed. After a short time, the motor driver was disabled and the current dropped to zero. Motor Current Voltage Step Duty Cycle Overcurrent Warning Overcurrent Threshold Motor Driver Enabled 150 100 50 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Time (s) Figure 9.15: Steering Actuator Behavior with Overcurrent Condition. Motor current drops to zero when motor driver is shut off in response to overcurrent condition. As shown in Figure 9.15, this method allows the motor to be stopped when locked rotor conditions or excessive loads are detected, but does not behave to aggressively that transient currents interrupt normal operation. 82 Current (A) / Duty Cycle (%) 10 Discussion and Conclusions This chapter contains discussion of the 12V actuator’s ability to pass the stated require- ments and offers recommendations for changes that should be made in order to meet requirements that were not met by this prototype. Additional notes on actuator per- formance are given. The conclusion is presented with all requirements listed as being passed or failed. 10.1 Performance Requirement Analysis 10.1.1 Actuator Power and Efficiency In Section 9.2, it is shown that the actuator passes the stated requirements for response to a ramp input. In Section 9.1, it is shown that the actuator does not pass Requirement 1.6, calling for the system to complete its full range of travel in 3.2 s or less against 340 kg load. The power required to lift the load at the necessary speed is calculated in (10.2). ( ) · (3.38mm/deg) (60 deg)P = F v = (340 kg) 9.81m/s2 = 211W (10.1) 3.2 s When lifting 340 kg, the actuator achieved a top speed of 43.1mm/s at 38.4 A and 11.02 V . The efficiency of the act(uator at th)is operating point is: Pout (340 kg) 9.81m/s 2 (43.1mm/s) 143W η = = = = 0.34 (10.2) Pin (11.02 V ) (38.4A) 423W From the results in Chapter 5 and Section 9.1, it was estimated that the efficiency of the ball screw and worm drive is 60 − 72% when converting load to torque, but the overall efficiency is much lower, as shown in Figure 10.1. 83 10.1. PERFORMANCE REQUIREMENT ANALYSIS 0.4 0.3 0.2 0.1 0 0 500 1000 1500 2000 2500 3000 3500 Load (N) Figure 10.1: Actuator efficiency during load test. 10.1.2 Explanations for Failure There are several possible explanations for the failure of the actuator to pass the load test requirement. The input power was higher than the output power necessary to reach the performance target, so power losses in the system must be discussed. In Section 9.1, Figure 9.1 shows the current drawn during the load test against the theoretical current drawn if the ball screw and worm were 72% efficiency. The current from the load being lifted is added onto the no-load current of the motor. The minimum current drawn with no load besides the actuator’s own weight during the load test is 12A. The no-load current was so large that it had to be figured into the controller as a feedforward term as discussed in Section 8.2.3. In Chapter 5, the no-load current of the motor is given as 2.2 A at rated speed. As mentioned in Section 7.1.3 The difference in the no-load current is due primarily to the Tandler HW-000 bevel gearbox which requires 9.0A worth of torque-generating current to turn at 5000 RPM . Assuming the friction behaves as viscous friction, this gearbox costs the system 97W at no load and 70W at 340kg. The actuator may have succeeded without this gearbox. Additional sources of loss are electrical. The motor was selected based upon a simple speed and load calculation used for DC motors which did not take into account other electrical losses. Figure 10.2 shows that what the DC motor model predicts as the behavior of the BY80BL300 was not actually realized in practice. Despite a back EMF constant of .02 V s/rad, the motor was not actually able to be driven to 600 rad/s off of 12V . The 32mΩ resistance of the BY80BL300 was also higher than the 18.7mΩ needed. The measured results agree with SimScape simulations done in MATLAB/Simulink. By more accurately modeling the switching and commutation of the power electronics, the real behavior of the motor is found. The discrepancy in behavior is most likely due to power losses in switching and commutation, which can be significant when driving a sinuisoidal back EMF motor with six-step commutation [53]. Energy built up in the motor inductor which must be dumped at each commutation phase switch is also a culprit for losses [54]. 84 Efficiency 10.2. SYSTEM REQUIREMENT ANALYSIS 600 400 200 Theory Observed 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Torque (Nm) Figure 10.2: Observation of motor performance falls short of theoretical performance. 10.1.3 Recommendations The bevel gearbox must be eliminated. Another solution to the absolute position sensor must be found so that the motor output does not pass through another lossy gear stage. The ball screw and worm drive actuator could be replaced with a more efficiency actuator. At 72% efficiency, this mechanism is a significant source of loss. Using more efficient gearing between the motor and the screw will reduce the self-locking capability of the drive and require a holding torque from the motor, but the overall power consumption may be less as a result. More research and development is needed in this direction. Rather than naively applying the DC calculations to determine the necessary motor specifications as in Section 5.2, the actuator simulation created in SimScape should be used to determine what motor parameters and configuration are needed to reach the results. The BY80BL300 motor has too much winding resistance, and possibly too high of a motor constant to achieve the desired operating point. A different motor should be selected using the knowledge gained from this work. Because losses from six-step commutation were expected to be significant, it is also recommended that FOC is used to improve the efficiency of the motor commutation[53]. 10.2 System Requirement Analysis 10.2.1 System Requirement Status Two of the four system requirements for the actuator were failed, and one performance requirement related to these is discussed here as well. Requirement 2.2 called for the actuator be small enough to allow the outboard engine’s propeller to be tilted completed out of the water. Requirement 2.3 demanded a manual override in case of steering system failure. Requirement 1.3 gave a actuator stroke of ±30◦. All three of these requirements were failed. 85 Speed (rad/s) 10.2. SYSTEM REQUIREMENT ANALYSIS 24.25 cm 0.15 cm 43.5 cm 47 cm 21 cm 11 cm Figure 10.3: Actuator installation schematic (top view) 10.2.2 Failure Analysis The lack of manual override is due to a lack of time in the thesis development. Several options were considered, but none implemented. The other requirements were failed because the actuator is too large. Figure 10.3 gives a schematic of the actuator installation. The overall length of the actuator body is 43.5 cm. From the engine pivot point, the actuator extends 24.25 cm in one direction, and 19.25 cm in the other direction. Between the pivot point and the side of the boat, there is 34.5 cm of free space. The linear approximation for rudder angle that assumes 3.38mm per degree of rudder angle. For a ±30◦ stroke, 10.1 cm of travel is needed. The clearance of 1.5mm was not sufficient for safety during testing, and on another boat installation the available space could be even smaller. 17.5 cm 8 cm Vertical Pivot Axis Screw Axis 22 cm Figure 10.4: Actuator installation schematic (side view) Figure 10.4 shows a schematic of the actuator side view. When the outboard engines are tilted out of the water in order to preserve the propeller, the engine and attached steering actuators pivot around the axis shown here. To be able to pivot fully out of the water, there must be sufficient clearance for the steering actuator to tilt down into the boat. This was not the case with this actuator. With a total distance of 25.5 cm from the pivot, the actuator could not tilt far enough down without striking the inside surface of the boat, which was 22 cm at the deepest point. 86 10.3. ELECTRICAL REQUIREMENT ANALYSIS 10.2.3 Recommendations The mechanical configuration of the actuator needs to be changed. Chapter 7 shows that the bevel gearbox and planetary gear reducer are responsible for much of the actuator body size. Finding another solution to the absolute position sensor that does not require these components would reduce the size to workable dimensions. A manual override solution should be incorporated into the new mechanical layout. 10.3 Electrical Requirement Analysis The actuator did not fail any of the electrical requirements. This section instead discusses potential improvements to the actuator to reduce current consumption. 10.3.1 Current Consumption Observations The only requirement for current consumption was Requirement 3.2 which requires less than 70 A of steady-state current. The current consumption during operation was well below this, but it is still higher than is could be. In Section 9.3, motor winding current is shown to regularly be between 20 − 50 A while turning on the water. The current is not smooth, even though many steering motions are. The roughness of the current is seen in Figure 10.5. 4000 2000 0 6 6.5 7 7.5 8 8.5 60 40 20 0 -20 6 6.5 7 7.5 8 8.5 10 0 Reference Actual (Starboard -10 6 6.5 7 7.5 8 8.5 Time (s) Figure 10.5: Performance of Internal Control Loops 87 Rudder Angle (degrees) Motor Speed (RPM) Motor Current (A) 10.3. ELECTRICAL REQUIREMENT ANALYSIS 4.4 4.2 4 3.8 3.6 3.4 100 102 104 106 108 110 Time (s) Reference Response 20 10 0 −10 −20 100 102 104 106 108 110 Time (s) Figure 10.6: Motor current during slow turn 10.3.2 Examples The actuator is very sensitive to small irregularities in movement, which will be common with a human helmsman. At the 6.5 ms mark in Figure 10.5, the TargetRudderAngle signal from the helm reaches a plateau and remains constant for several milliseconds. The actuator approaches this level, and the target speed drops to below 2000 RPM , dropping the target current down to very low levels. After the wheel is turned again, the actuator must accelerate to continue tracking the signal, driving the current up over 60 A in the process. The turn is a fairly smooth maneuver which could be reproduced with a constant current of about 20A. The TargetRudderAngle signal comes only in 0.1◦ increments. Figure 10.6 shows how even during a very slow turn, requiring only 30 RPM from the motor and less than an amp of constant current, the motor regularly spikes over 20 A of current to accelerate, and experiences very large oscillations in the current. 10.3.3 Recommendations It is clear that the controller should be improved to draw a smoother, more regular current. Some form of trajectory planning is recommended so that the actuator tracks a reference more intelligently. Something more advanced than a first-order filter on the input is needed to provide smoother tracking without increasing the tracking delay to 88 Current (A) Rudder Angle (deg) 10.3. ELECTRICAL REQUIREMENT ANALYSIS unacceptable levels. 10.3.4 Comparison to Teleflex The battery input current that should be considered when discussing power consumption is somewhat lower due to the nature of the three phase bridge, as explained in Section 6.3.3. Figure 10.7 shows the relationship between battery current and winding current at different duty cycles. Winding Current 20 30 Battery Current 15 20 10 10 5 0 0 0.1 0.2 0.3 0.4 0 0.05 0.1 0.15 0.2 Time (s) Time (s) 30 30 20 20 10 10 0 0 0.2 0.25 0.3 0.35 0.4 0.25 0.3 0.35 0.4 Time (s) Time (s) Figure 10.7: Supply and winding current for (Top Left) 25% Duty Cycle, (Top Right) 50% Duty Cycle, (Bottom Left) 75% Duty Cycle, (Bottom Right) 99% Duty Cycle. Supply current is lower than winding current. Figure 10.8 shows a comparison between the supply current to the Teleflex actuator and the 12V Actuator in an unloaded test. A back-and-forth sweep of 180◦ of the test rig steering wheel was done at the same rate for both actuators. During the test steering motions, the average current consumption of the electric steering actuator was 5.29 A, and the average current consumption of the Teleflex actuator was 8.69A. This provides only a single data point to compare, but it suggests that the electric actuator may be competitive against the electro-hydraulic actuator despite these problems. More testing is required to know how both perform under load. 89 Current (A) Current (A) Current (A) Current (A) 10.4. SAFETY REQUIREMENT ANALYSIS 40 30 20 10 0 0 0.5 1 1.5 2 2.5 Time (s) 40 30 20 10 0 0 0.5 1 1.5 2 2.5 Time (s) Figure 10.8: Battery supply current of Teleflex actuator vs. Electric Steering Actuator Prototype. 10.4 Safety Requirement Analysis 10.4.1 Observations No major problems were observed in the area of functional safety. No unexpected steering occurred during the boat test. The actuator did need to be restarted several times due to a communication fault. If such a fault occurred during turning, it is possible that unexpected steering could occur, however CPAC’s EVC system is expected to prevent unexpected steering events by warning the helmsman of the fault, disabling the steering system, and stopping the engines. 10.4.2 Recommendations The cause of the communication loss is unknown. For several seconds, CAN messages did not reach the port side actuator until the unit was restarted. That the problem was corrected with a restart suggests it is a software programming issue, or possibly a signal integrity issue. In either case, the CAN bus lines on the controller board should be reviewed for EMC/EMI problems, and the microcontroller software should be adjusted to use a proven standard software structure. 90 Electric Steering Current (A) Teleflex Current (A) 10.5. CONCLUSION 10.5 Conclusion Table 10.1: Requirement fulfillment Requirement Pass/Fail Section Notes 1.1 Performance Pass 9.2 Close to failure on some 1.2 No overshoot Pass 9.2 - 1.3 ±30◦ Fail 9.3.1 Actuator too large 1.4 Active in 1 s Pass 8.1.1 - 1.5 Power up with engines Pass 8.1.1 Powers up with battery 1.6 Speed vs. Load Fail 9.1 Not powerful enough 1.7 Response within 200ms Pass 9.2 - 1.8 Clockwise to starboard Pass - Achieved by design 1.9 Default rudder position Pass - Follows EVC command 1.10 Accuracy of sensors Pass 6.5.1 - 2.1 Twin/Triple/Quad Pass 9.3.1 One unit per engine solution 2.2 Tilting / Trim Fail 9.3.1 Actuator too large 2.3 Manual overrride Fail - Did not implement 2.4 Interface with EVC Pass 9.2 - 3.1 12 V power Pass 9.3.2 Successful boat test 3.2 Less than 70A current Pass 9.3 Current can be limited 3.3 Voltage range Pass 6.2.1 Successful tests 3.4 Reverse voltage Pass 6.3.3 Achieved by design 4.1 No unintended steering Pass 9.3.4 Only slight problems 4.2 ASIL B MCU Pass 6.2 - 4.3 Component Selection Pass - Achieved by design The actuator has performed its basic functional goals, but has not satisfied all of the requirements as shown in Table 10.1. The unit is under-powered and cannot pass the requirement for moving a full stroke under 340 kg in 3.2 s. Nor can it achieve a full stroke due to the size. The motor lacks the speed and acceleration to pass the performance requirement test with a suitable margin. The motor needs to draw less current in order to achieve the speed required. Improvement is necessary in this area to have a competitive efficiency. The closed loop controller requires better tuning and some method of planning a 91 10.5. CONCLUSION smoother trajectory. More time should have been spent addressing this problem. The motor passes the electrical specifications, as far as operating voltages and survival of reverse voltage, but the requirements that were omitted, including EMC immunity and voltage surge resistance, are not expected to be passed. This work has satisfied the goal of producing a functional electric steering actuator, and despite not fulfilling all of the requirement, has demonstrated the soundness of the concept while detailing many of the concerns involved in the development. 92 Bibliography [1] B. Sisson, New outboards, market share figures and good numbers from washington, Trade Only Today (June 2014). [2] Newly developed ’helm master’ joystick control system, Yamaha Motor Press Re- lease, http://global.yamaha-motor.com/news/2012/1019/helmmaster.html (Octo- ber 2012). [3] C. Plueddeman, The new magically maneuverable outboard, Power and Motoryacht. [4] Optimus electronic power steering, http://www.seastarsolutions.com/products/electronic- power-steering/ (2014). [5] G. Mitchell, Lauding the merits of the yamaha helm master integrated steering system, http://yamahaoutboards.com/owner-resources/blog/pro-anglers/lauding- merits-yamaha-helm-master-integrated-steering-system (2013). [6] Global electic power steering market 2012-2016, Research and Markets (March 2013). [7] D. Sherman, ’are we losing touch?’ a comprehensive comparison test of electri- cal and hydraulic assist steering, http://www.caranddriver.com/features/electric- vs-hydraulic-steering-a-comprehensive-comparison-test-feature, car and Driver Tech Feature (January 2012). [8] H. Greimel, NSK: Electric power steering saves fuel, Automotive News (August 2014). [9] T. S. Sayeed Mir, Mohammad S. Islam, Role of electronics and controls in steering system, in: Industrial Electronics Society, 2003. The 29th Annual Conference of the IEEE, Vol. 3 of 29th, IEEE, 2003, pp. 2859–2864. [10] C. Bruzzese, A high absolute thrust permanent magnet linear actuator for direct drive of ship’s steering gears: Concept and fem analysis, in: 2012 XXth International Conference on Electrical Machines, IEEE, Marseille, France, 2012, pp. 556–562. 93 BIBLIOGRAPHY BIBLIOGRAPHY [11] C. de Silva, Mechatronics: A Foundation Course, Taylor and Francis Group, Boca Raton, FL, 2010. [12] S. Frischemeier, Electrohydrostatic actuators for aircraft primary flight control, in: 5th Scandinavian International Conference on Fluid Power, Linköping, Sweden, 1997. [13] A. Garcia, J. Cusido, J. Rosero, J. Ortega, L. Romeral, Reliable electro-mechanical actuators in aircraft, IEEE A-and-E Systems Magazine (2008) 19–25. [14] A. Wroblaski, Debunking the myths of hydraulics to electric actuator conversion, Thomson Industries Whitepaper. [15] E. M. Corp., Hydraulic efficiency, Technical Topic Whitepaper (2009). [16] US Coast Guard, 2012 recreational boating statistics (2012). [17] US Coast Guard, 2013 recreational boating statistics (2013). [18] Road vehicles - functional safety, ISO 26262:2011, international Organization of Standards. [19] Safety-integrated hardware solutions to support asil-d applications, Freescale Semi- conductor Whitepaper (2013). [20] K.-J. Lee, Y.-H. Ki, H.-S. Ahn, Automotive ecu design with function safety for electro-mechanical actuator systems, World Academy of Science, Engineering, and Technology 7. [21] A. E. Mehrdad Ehsani, Yimin Gao, Modern Electric, Hybrid Electric, and Fuel Cell Vehicles: Fundamenetals, Taylor and Francis Group, 2009. [22] G. Prasad, N. S. Ramya, P. Prasad, G. T. R. Das, Modelling and simulation analysis of the brushless dc motor by using matlab, International Journal of Innovative Technology and Exploring Engineering (IJITEE) 1 (5) (2012) 27–31. [23] R. D. Knight, Physics for Scientists and Engineers, Vol. 4, Pearson Addison Wesley, 2004. [24] D. C. Hanselman, Brushless Permanent Magnet Motor Design, McGraw-Hill, Inc., 1994. [25] J. R. Mevey, Sensorless field oriented control of brushless permanent magnet syn- chronous motors, Master’s thesis, Kansas State University (2006). [26] A. Hughes, B. Drury, Electric Motors and Drives, fourth edition Edition, Elsevier Ltd., Oxford, UK, 2009. 94 BIBLIOGRAPHY BIBLIOGRAPHY [27] F. Briz, D. Diaz-Reigosa, M. Degner, P. Garcia, y. m. p. Juan booktitle=Power Electronics Conference (IPEC), 2010 International, title=Current sampling and measurement in PWM operated AC drives and power converters. [28] N. Muruganantham, S. Palani, State space modeling and simulation of sensorless permanent magnet bldc motor, International Journal of Engineering Science and Technology 2 (10) (2010) 5099–5106. [29] A. Tashakori, E. Ektesabi, N. Hosseinzadeh, Modeling of bldc motor with ideal back-emf for automotive applications, in: Proceedings of the World Congress on Engineering, Vol. II, WCE, London, UK, 2011. [30] R. Krishnan, Permanent Magnet Synchronous and Brushless DC Motor Drives, CRC Press Taylor and Francis Group, 2010. [31] J. Fang, X. Zhou, G. Liu, Instantaneous torque control of small inductance brushless dc motor, IEEE Transactions on Power Electronics 27 (12) (2012) 4952–4964. [32] D. Morgan, Field-oriented control, Embedded Systems Programming. [33] J. R. B. A. Monteiro, A. A. Oliveira, M. L. Aguiar, E. R. Sanagiotti, Electromag- netic torque ripple and copper losses reduction in permanent magnet synchronous machines, in: European Transactions on Electric Power, John Wiley and Sons, 2011. [34] P. Perruchoud, P. Pinewski, Power losses for space vector modulation techniques, in: Power Electronics in Transporation, 1996. IEEE, IEEE, 1996, pp. 167–173. [35] Mead Info, Comparison of gear efficiencies, http://www.meadinfo.org/2008/11/gear- efficiency-spur-helical-bevel-worm.html (October 2012). [36] Mpc5744p data sheet, Tech. rep., Freescale Semiconductor. [37] Robert Bosch GmbH, Stuttgart, CAN Specification 2.0 (September 1991). [38] H. Moller, C. Pedersen, Hearing at low and infrasonic frequencies, Noise and Health. [39] A. Kita, Metal oxide semiconductor field effect transistor and h-bridge driver circuit, Mechatronics Fact Sheet (2004). [40] Ipb180n04s4-h0 data sheet, Tech. rep., Infineon. [41] B. K. Bose, Power Electronics and Motor Drives: Advances and Trends, Academic Press, 2010. [42] Mc33937a data sheet, Tech. rep., Freescale Semiconductor. [43] A. C. Chow, D. J. Perreault, Active emi filters for automotive motor drives, in: Power Electronics in Transportation, 2002, 2002, pp. 127–134. [44] SPI Block Guide V03.06 (January 2000). 95 BIBLIOGRAPHY [45] SDP/SI, Ball and acme lead screw technical information, http://www.sdp- si.com/D810/PDFS/Ball [46] M. Morari, E. Zafiriou, Robust Process Control, Prentice-Hall, Inc., Englewood Cliffs, New Jersey, 1989. [47] T. Olsen, N. Ito, Implemented an effective loop tuning strategy, Chemical Engineer- ing Progress. [48] T. Glad, L. Ljung, Control Theory. Multivariable and Nonlinear Methods, Taylor and Francis Group, 2000. [49] S. Moussavi, M. Alasvandi, S. Javadi, A novel combination pid strategy of speed control for independent excited dc motor, in: Advances in Power and Energy Sys- tems: Proceedings of the 12th WSEAS International Conference on Electric Power Systems, High Voltages, Electric Machines. [50] Q. Gang, S. Le, H. Ling, A practical application of imc-pid controller in unmanned vehicle, Telkomnika. [51] A. R. Teel, L. Zaccarian, Modern Anti-Windup Synthesis: Control Augmentation for Actuator Saturation, Princeton University Press, 2011. [52] X. lan Li, J.-G. Park, H.-B. Shin, Comparison and evaluation of anti-windup pi controllers, Journal of Power Electronics 11 (1). [53] D. Y. Ohm, R. J. Oleksuk, Influence of pwm schemes and commutation methods for dc and brushless motors and drives, P. E. Technology Conference (2002). [54] The effects of inductance on motor performance, Ellipash LLC Technology Blog, http://www.ellipsah.com/blog1/?p=75 (April 2013). 96 A SimScape Models A.1 BLDC Motor Figure A.1: SimScape Model of the BLDC and Mechanical System A.2 Control Electronics Models Figure A.4 shows how different blocks representing the driver, current sensor, and BLDC motor are connected. Figure A.2 shows the ACS759 current sensors, as well as the on- chip low pass filter. In these circuits, C = 72 pF and R = 100 Ω. Figure A.3 shows the layout of the model of the driver board. The model does vary slightly from reality in that the MOSFETs used are ideal, i.e. they respond instantly to the command to open or close. It did not appear that this inaccuracy was significant in the final simulations. 97 A.3. CONTROL SYSTEM MODEL Figure A.2: SimScape Model of the Current Sensors Figure A.3: SimScape Model of the Motor Driver A.3 Control System Model Figure A.5 shows the Simulink model of the control logic. The position, current, and voltage signals are passed to the MPC5744, and the MPC5744 passes commands for high and low side MOSFETS to open or close based on feedback control that will be 98 A.3. CONTROL SYSTEM MODEL Figure A.4: SimScape Model of the Physical System discussed in the next section. Figure A.5: Control Flow 99 A.3. CONTROL SYSTEM MODEL Figure A.6: The cascade control loop Figure A.6 shows a Simulink diagram of the cascade controller as used in this actuator application. Figure A.7: The current control loop with anti-windup. Figure A.8: The speed control loop with anti-windup. 100 B Additional Images B.1 Assembly Photos Figure B.1: Additional Assembly Pictures showing (top left) unenclosed actuator, (top right) finished actuators with cabling, (bottom left) detail of electronics mounting below planetary gearbox, (bottom right) mounting of actuator to outboard engine 101