%' Advanced Combustion Analysis The development of an advanced combustion analysis tool Automotive Egineering Project Report Aijaz Ahmed, Oskar Edholm, Ajanish Lutimath, Avanish Raj, Filip Tagesson Department of Mechanics and Maritime Sciences CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2021 Automotive engineering project report 2021:06 Advanced Combustion Analysis The development of an advanced combustion analysis tool Aijaz Ahmed Oskar Edholm Ajanish Lutimath Avanish Raj Filip Tagesson %' Department of Mechanics and Maritime Sciences Division of Combustion and Propulsion Systems Chalmers University of Technology Gothenburg, Sweden 2021 Advanced Combustion Analysis The development of an advanced combustion analysis tool © Aijaz Ahmed, 2021. © Oskar Edholm, 2021. © Ajanish Lutimath, 2021. © Avanish Raj, 2021. © Filip Tagesson, 2021. Automotive engineering project report 2021:06 Supervisors: Jonas Sjöblom , Lucien Koopmans & Ahfaz Ahmed Examiner: Jonas Sjöblom Automotive Engineering Project Report 2021 Department of Mechanics and Maritime Sciences Division of Combustion and Propulsion Systems Chalmers University of Technology SE-412 96 Gothenburg Cover: Engine cylinder. Gothenburg, Sweden 2021 iv Advanced Combustion Analysis The development of an advanced combustion analysis tool Aijaz Ahmed Oskar Edholm Ajanish Lutimath Avanish Raj Filip Tagesson Department of Mechanics and Maritime Sciences Chalmers University of Technology Abstract The main purpose of this project is to study and investigate advanced combustion analysis in compression-ignited (CI) and spark-ignited (SI) engines. A MATLAB code will be developed as the major deliverable, which will calculate the (apparent and real) rate of heat release with the provision of di�erent, e�ective intake valve closing, heat transfer models, blow-by loss and gas properties with automatic poly- tropic coe�cient detection. This code will also be used to compare and validate the di�erent zero level corrections with commercial data acquisition software by Dewe- soft (stakeholder for this project). For the study a 6-cylinder CI engine and am SI engine at Chalmers is used to record the in-cylinder pressure. Measurements of the in-cylinder pressure accurately plays an important role in pressure trace analysis. In this project, the experimental data like raw cylinder pressure and crank angle degree is recorded with the Dewesoft data acquisition system. Along with the recorded data, data from other test cells were also collected which was further analyzed and processed through the MATLAB code developed. After the investigation, the common assumption of constant values of certain pa- rameters, such as IVC e�ective, constant polytropic coe�cient, temperature at IVC, wall temperature, heat transfer loss and blow-by loss was found to skew the data significantly, thus giving incorrect results. Further, in-depth analysis for each of the above parameters is done and function scripts were developed and integrated into the main MATLAB program. The major findings are that the commercial combustion analyzer software’s usu- ally considers a constant polytropic coe�cient and also performs an apparent rate of heat release, thus neglecting the heat transfer and blow-by losses. But in reality, the polytropic coe�cient varies and the heat transfer and blow-by loss do contribute and a�ect the amount of heat energy liberated during combustion. The results from the analysis of the data’s influence on the heat release is presented in the report. Keywords: Combustion, Analysis, Heat release. v Acknowledgements We would like to thank Jonas Sjöblom for the help and guiding through the project, and for giving us his previously created combustion analysis code. This code has been very helpful both for general understanding and as a base skeleton for the newly produced code. We would also like to thank Ahfaz Ahmed for advice and general help throughout the project. Especially when it comes to gas properties your help has been very appreciated. Lucien Koopmans we would like to thank for your advice and help with the project. Sergej Abyzov thank you for helping us with the understanding of the combustion tool from Dewesoft. Finally we want to thank all the people helping us with data collection and running the engines in the test cells. Thank you! Aijaz, Oskar, Ajanish, Avanish, Filip vii Contents Nomenclature xi List of Figures xiii List of Tables xv 1 Introduction 1 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.4 Specifications of issues under investigation . . . . . . . . . . . . . . . 2 1.5 Deliverables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2 Theory 3 2.1 Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 Zero Level Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2.1 Fixed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.2.2 Measured . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.2.3 Thermodynamic 2-point . . . . . . . . . . . . . . . . . . . . . 8 2.2.4 Least Square Method . . . . . . . . . . . . . . . . . . . . . . . 9 2.3 Rate of Heat Release . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.4 Gamma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.4.1 Logarithmic regression . . . . . . . . . . . . . . . . . . . . . . 13 2.4.2 NIST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.4.2.1 Gas composition . . . . . . . . . . . . . . . . . . . . 15 2.4.3 Ganippa proposal . . . . . . . . . . . . . . . . . . . . . . . . . 16 2.5 Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 2.5.1 Woschni . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 2.5.2 Hohenberg correlation . . . . . . . . . . . . . . . . . . . . . . 18 2.5.3 Wall Temperature . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.6 E�ective Intake Valve Closing . . . . . . . . . . . . . . . . . . . . . . 19 2.7 Blow-by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.8 Compression Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3 Method 23 3.1 Test engines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.2 Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 ix 3.3 Zero level correction . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.4 Gamma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.4.1 Logarithmic regression . . . . . . . . . . . . . . . . . . . . . . 25 3.4.2 NIST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.5 E�ective Intake Valve Closing . . . . . . . . . . . . . . . . . . . . . . 25 3.6 Compression ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.7 Heat transfer models . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.7.1 Wall Temperature . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.8 Blow-by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.8.1 Blow-by code validation by using experimental values . . . . . 26 3.9 Rate of heat release . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.10 Determination of pressure derivative . . . . . . . . . . . . . . . . . . 28 3.10.1 Signal conditioning by current to voltage converter . . . . . . 28 4 Results and Conclusion 29 4.1 Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.2 Interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 4.3 Zero level correction . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 4.4 Gamma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 4.4.1 Logarithmic regression . . . . . . . . . . . . . . . . . . . . . . 32 4.4.2 NIST-Polynomial . . . . . . . . . . . . . . . . . . . . . . . . . 33 4.4.3 Ganippa proposal . . . . . . . . . . . . . . . . . . . . . . . . . 33 4.5 IVC E�ective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 4.6 Compression ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 4.7 Heat transfer models . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 4.8 Rate of heat release . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 4.8.1 Validation of RoHR analysis . . . . . . . . . . . . . . . . . . . 41 4.8.2 Wall Temperature . . . . . . . . . . . . . . . . . . . . . . . . . 42 4.9 Input data validation . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 4.10 Current to Voltage converter . . . . . . . . . . . . . . . . . . . . . . . 43 5 Discussion 45 5.1 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 5.2 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 A Suggested parameters 53 B User interface 55 x Nomenclature W Mechanical work J “ Specific heat ratio, gamma, cp/cv ≠ fl Lubricating oil type ≠ · Cycle type ≠ ◊ Crank angle ° A Area m 2 a Crank radius m B Bore m cp Specific heat constant pressure JK ≠1 mol ≠1 cv Specific heat constant volume JK ≠1 mol ≠1 dz Shape and clearance piston joint ≠ Eloss Wall heat and blowby losses J f Molecular degrees of freedom ≠ hc Heat transfer coe�cient W/m 2 ú K i Number of cylinders ≠ k Sealing ring type ≠ L Stroke m l Connecting rod length m li Number of piston rings n m Mass Kg N Engine speed RPM n Compound amount ≠ p Pressure Pa Q Heat transfer J Qch Chemical energy released by combustion J Qht Heat transfer J R The gas constant JK ≠1 mol ≠1 rc Compression ratio ≠ Sp Mean piston speed m/s T Temperature, mean Temperature K Us Internal energy J V Volume m 3 W Average gas velocity m/s wBB Blow-by gas flow velocity m/s z Wear parameter ≠ QBB Mass loss (Blow-by) J xi xii List of Figures 2.1 Piston-cylinder Model . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2 Ideal engine cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.3 Energy balance for an ICE . . . . . . . . . . . . . . . . . . . . . . . . 4 2.4 Experimental setup for measuring in-cylinder pressure and pressure derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.5 Operation of the piezoelectric transducer . . . . . . . . . . . . . . . . 5 2.6 Typical signal processing for a pressure transducer . . . . . . . . . . . 6 2.7 Savitzky-Golay filtering polynolial fit for n=3,8 is shown. . . . . . . . 7 2.8 Combustion inside the engine cylinder . . . . . . . . . . . . . . . . . 10 2.9 Dry air, mole percentage of composition gases. . . . . . . . . . . . . . 15 2.10 Diesel exhaust, mole percentage of composition gases. . . . . . . . . . 15 2.11 Left: logaritmic pv-plot. Rhight: in-cylinder-pressure vs CAD. . . . . 19 2.12 Blow-by During compression and expansion stroke with Crankcase ventilation system (PCV) . . . . . . . . . . . . . . . . . . . . . . . . 20 2.13 (a)Piston and rings assembly, (b) Flow through orifice . . . . . . . . . 21 3.1 The interaction of various function in the main code . . . . . . . . . 23 3.2 (a) Blow by measured values-for di�erent engine speed and Torque,(b) Blow by at engine speed 1200 [RPM] and di�erent Torques . . . . . . 27 3.3 Transducer signal conditioning through current to voltage converter . 28 4.1 The visual e�ect of the filtering process. . . . . . . . . . . . . . . . . 29 4.2 The visual e�ect of the interpolation process. . . . . . . . . . . . . . . 30 4.3 Comparison of di�erent Zero level correction methods . . . . . . . . . 31 4.4 Comparison of di�erent Zero level correction methods in log-log scale 31 4.5 Gamma output from Logarithmic regression and NIST. . . . . . . . . 32 4.6 Comparison of Heat transfer co-e�cient for di�erent heat transfer models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 4.7 Comparison of Heat transfer for di�erent heat transfer models . . . . 35 4.8 Rate of heat release curves of: chemical energy, internal energy, me- chanical work and heat transfer rate . . . . . . . . . . . . . . . . . . . 36 4.9 Cumulative mass fraction burned curve . . . . . . . . . . . . . . . . 37 4.10 Rate of Heat Release . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 4.11 Rate of heat release curves of: chemical energy, internal energy, me- chanical work and heat transfer . . . . . . . . . . . . . . . . . . . . . 39 4.12 Cumulative mass fraction burned . . . . . . . . . . . . . . . . . . . . 39 4.13 RoHR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 xiii 4.14 Motoring RoHR curves of CI engine . . . . . . . . . . . . . . . . . . . 41 4.15 Motoring RoHR curves of SI engine . . . . . . . . . . . . . . . . . . . 41 4.16 Di�erent datasets, out-of-bound and inside-bound . . . . . . . . . . . 42 4.17 Di�erent datasets, not following, and following the NIST-slope . . . . 43 5.1 Parameter sensitiveness towards RoHR, CI engine . . . . . . . . . . . 47 5.2 Parameter sensitiveness towards RoHR, SI engine . . . . . . . . . . . 47 B.1 UI, Mainpage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 B.2 UI, Setup page 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 B.3 UI, Setup page 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 B.4 UI, Setup page 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 B.5 UI, Run page . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 xiv List of Tables 2.1 Woschni coe�cients . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4.1 Engine, operating conditions, and RoHR methods . . . . . . . . . . . 36 4.2 Results of CI engine at 1200 RPM and 600 Nm heat release analysis . 38 4.3 Engine operating conditions, and RoHR methods . . . . . . . . . . . 39 4.4 Results of CI engine at 1200 RPM and 600 Nm heat release analysis . 40 5.1 Sensitivity Analysis dependencies . . . . . . . . . . . . . . . . . . . . 45 A.1 Recommended options for getting best RoHR results for CI and SI engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 xv xvi 1 Introduction Climate change is one of the most serious issues and ongoing that the world needs to address. The e�ects of this is not only on environment, but also on wildlife as well as the well being and life of humans. Hence it is our responsibility to address this seriously issue that threatens the planet today. One of the ways of doing this is by controlling and lower the emissions from transportation on a global scale. This can be done by imposing new, more stringent emissions regulations. These ever demanding emission regulation has forced automobile manufacturers to study the in cylinder processes in detail. The e�ciency, durability, fuel consumption, performance and much more depend on the process that takes place inside the combustion chamber. Hence it is very important to analyze the combustion in depth and understand it thoroughly. This project aims to improve combustion analysis and allow for e�cient engine optimization. A MATLAB based models will be developed, calibrated and verified against experimentally collected data. The project will be carried out by five students of the Automotive Engineering Master’s programme at Chalmers University and will be supervised by Jonas Sjöblom, Ahfaz Ahmed and Lucien Koopmans from the Combustion and Propulsion System (CaPS) division. The main stakeholder is Dewesoft. The developed tool will also be integrated into ongoing research projects at the division. 1.1 Background Dewesoft is a company dealing with data acquisition and analysis. They have a suite specifically designed for combustion analysis called "Combustion engine Analyser" (CEA). This analyzer takes the data collected by their data acquisition module and presents a combustion analysis to the user. The accuracy of an combustion analysis is of prime importance as this plays a major role in an engine calibration. Inaccurate calibration may lead to higher emissions and lower e�ciency of the engine. Therefore the accuracy of the models is of utmost importance. Dewesoft CEA provides various options and models for processing input pressure data from engine including three modes for zero level correction, suggestion of polytropic coe�cient, thermodynamic calculation for heat release etc. Their acquisition devices and software are used globally for engine development. Academia also has a large interest in developing combustion analysis tools that can help them in further understand the combustion process and the heat release. For academia it is important to have full transparency of the workings of the code used to truly understand the analysis and the results. 1 1.2 Aim The project aim is to do an in depth analysis of the combustion in a heavy duty diesel engine and a light duty SI engine by developing several model in MATLAB which will be correlated with the experimentally collected data. A new complete code for pressure trace analysis will be developed in MATLAB. This code will be used as point of reference to compare with the commercial software’s. Di�erent zero level correction methods, polytropic coe�cient determination methods and more, are investigated and if required the models are further optimized. The code written in MATLAB will also contain di�erent models for calculating heat transfer , blow-by, filtering and averaging. A method for input data validation will be created based in the automatic polytropic coe�cient determination and optimized. 1.3 Limitations 1. The experimental data is from the di�erent Chalmers test rigs. 2. Only the currently present sensors will be utilized. 3. The project is limited to two study periods. 4. The code is completed in MATLAB. 1.4 Specifications of issues under investigation 1. Di�erent zero level correction models will be investigated. The di�erences are evaluated and the best method is recommended 2. Improving the accuracy of thermodynamic zero level correction 3. Di�erent methods for determination of polytropic coe�cient are evaluated 4. Improve RoHR analysis 5. A model for blow-by is to be implemented and examined. 1.5 Deliverables Throughout this project a goal is set. This goal is to at the end of the project have created certain results. These products are presented here in the form of project deliverables. The main deliverable is the software that is to be created. The software will be a MATLAB code system that can perform certain tasks. These tasks include the use of di�erent zero level corrections, di�erent ways of calculating and using polytropic coe�cient, allow for di�erent models when calculating heat transfer, blow-by, loss angles and data smoothing/filtering and finally visualize the combustion using pressure trace diagrams, P-V diagrams and di�erent heat release diagrams. 2 2 Theory An internal combustion engine is a machine which is used for converting chemical energy stored in fuel into mechanical work. The fuel is mixed in appropriate amount with air and burned inside the combustion chamber. This converts the chemical energy of fuel into thermal energy. The thermal energy increases the pressure and temperature of the entrapped gas, which exerts a force on the piston causing it to move. The piston is connected to the crankshaft through a connecting rod which converts the applied force into rotational torque. The basic structure of an ICE is shown in Figure 2.1. Figure 2.1: Piston-cylinder Model [1] Modern day engines usually use a four stroke engine cycle in which power is delivered once every two rotations of the crankshaft. The four strokes of the engine are the intake stroke, compression stroke, expansion stroke and the exhaust stroke. Each stroke is equal to 180 deg rotation of the crankshaft and causes the piston to move between top dead centre (TDC) and bottom dead centre (BDC) positions. Based on the type of ignition mechanism used, engines are classified as Spark Ignited (SI) or Compression Ignited (CI) engines. In common SI engines a mixture of air and fuel is injected during the intake stroke (TDC to BDC), which is then compressed (BDC to TDC) and ignited using a spark plug. Combustion energy causes the gases to expand and pushes the piston down. Finally during exhaust stroke the burned charge is expelled out of the chamber. In a CI engine only air is initially inducted and compressed to high pressure and temperature. Ignition is then controlled by injecting highly pressurized fuel into the high pressure, high temperature environment which causes the fuel to ignite. 3 Ideal four stroke cycle for both SI and CI combustion are shown in figure 2.2. Figure 2.2: (a)Ideal Otto cycle (b)Ideal Diesel cycle [2] Engine e�ciency is dependent on a lot of factors including: the volumetric e�ciency of the engine (the actual mass of air trapped inside combustion chamber as compared to theoretical mass), the quality of fuel air mixture, the combustion characteristics including the speed of combustion, the temperature produced during combustion, combustion phasing etc. The energy produced during combustion is divided into heat energy dissipated through chamber walls, the exhaust enthalpy flux, blow-by and the useful mechanical work. Figure 2.3 shows the energy balance of an engine. Figure 2.3: Energy balance for an ICE 4 Monitoring of the in-cylinder combustion and heat release is extremely important in order to increase engine e�ciency. In-cylinder pressure is an useful indicator of fuel conversion e�ciency and can be used along with crank angle to analyse the combustion [3]. Combustion analysis from pressure data is used to calculate the work, heat transfer, the release of chemical energy, the distribution of energy, etc. In most cases, Mean E�ective Pressure (MEP) which is the work done normalised by volume, is used to compare di�erent engines. Gross Mean E�ective Pressure (GMEP) is used when work done is considered only for the compression and expan- sion strokes. Pumping Mean E�ective Pressure (PMEP) is used when work done during intake and exhaust stroke is considered. Indicated Mean E�ective Pressure is for the whole cycle i.e. the sum of GMEP and PMEP. Pressure data can be also used to estimate parameters like the TDC o�set, gamma, wall temperature, blow- by, compression ratio and e�ective intake valve closing [3]. Experimental setup for measuring in-cylinder data is shown in figure 2.4. Figure 2.4: Experimental setup for measuring in-cylinder pressure and pressure derivative The in cylinder pressure is measured using a piezoelectric transducer. The struc- ture of the piezoelectric transducer is shown in figure 2.5. The pressure variation rate experienced by the diaphragm of the pressure transducer is transmitted to the piezoelectric crystal which generates a current proportional to its deformation [4]. Figure 2.5: Operation of the piezoelectric transducer[4] 5 The charge produced by the transducer is very weak and thus needs to be amplified by an high impedance charge amplifier. The amplifier outputs a voltage proportional to the input charge, which is the time integral of the input current [4]. Pressure is related to voltage by following Equation 2.1 Pout ≠ Pref = Vch ú Ga Gs (2.1) where Vch is the voltage output from the amplifier, Ga is the amplifier gain,and Gs is the pressure transducer gain. Figure 2.6 shows the typical signal conditioning applied for a pressure transducer. The accuracy of the transducer is a�ected by the thermal strain on the sensor diaphragm caused by variable heat flux inside the chamber. This is called thermal drift and and is usually divided in short term drift caused by heat flux change during a cycle and long term drift or load change drift caused by changes in transducer temperature due to engine operating points [4]. Figure 2.6: Typical signal processing for a pressure transducer [4] The low current value of the piezoelectric transducer makes the signal prone to noise from the environment. Signal amplification can also act as a source for noise [5]. Filtering of input data is required to remove the high frequency noise content. The relative nature of the piezoelectric sensor along with thermal drift requires the need for correcting the baseline or zero level of the data. This process is called pegging or Zero Level Correction (ZLC). Crank angle encoders are used to measure the angle data of the crankshaft.The resolution of the encoder is dependent on the applica- tion and typically combustion analysis require a resolution of 0.1 degrees. Incorrect phasing of cylinder pressure and encoder data can cause errors in combustion anal- ysis especially in IMEP calculation [3]. Piezoelectric pressure transducer and crank angle encoder are the main sensors used for performing combustion analysis. Intake and exhaust pressure sensors, air, fuel mass flow senors can be used alongside for accurate gas exchange measurements. Fuel line pressure sensor, needle lift sensor, temperature sensor, valve lift sensor, ignition current sensor, blow-by meter and lambda sensors are some of the other sensors which can be used [3]. This project will mainly use data from in-cylinder pressure sensor and crank angle encoder. In- take manifold pressure, temperature sensors, volume flow meter and blow-by meter are also used in some analysis cases. 6 2.1 Filtering Most of the data collected in an engine is noisy due to large fluctuations in pressure and temperature. This noise would interfere with calculations and o�set the results due to a incorrect signal. This means that the signal will need filtering. The filter type Savitzky-Golay was recommended and used. In short, the Savitzky-Golay filter is based on a series of local low-degree polynomial fits fitted with a least square method. A user selected window is moved across the signal and within the window a least square low degree polynomial curve fit is performed. In the center of the fitted curve the now filtered data is extracted. In Figure 2.7 the process is shown. The original signal in blue is curve fitted by a 2nd order polynomial within the window, here the width is 5, and the mid-point of the polynomial is the new filtered signal. The new values for n=3-8 are shown. The interested reader is referred to: [6, 7]. Alternative filters that could be used are low-pass filters but would require a lot of tuning and thus not chosen. Figure 2.7: Savitzky-Golay filtering polynolial fit for n=3,8 is shown. 2.2 Zero Level Correction Piezoelectric transducers can only measure the relative change in pressure due to their inherent characteristics. In order to get the absolute pressure value, it needs to be referenced to a known pressure at some point. The charge amplifier used to convert charge to voltage is prone to thermal drifts and thus needs continuous resetting. Thus raw pressure data from the sensor amplifier combination needs to be corrected for baseline. The pressure value is referenced to a known absolute pressure value at a reference crank angle. The accuracy of combustion analysis is highly dependent on the selection of correct o�set method and the inputs used like the reference crank angle. There are various methods proposed in literature to 7 estimate this pressure o�set and usually falls within two categories: - Referencing relative pressure to a measured or known pressure at a suitable crank angle. Fixed and Measured correction methods fall under this category. - Estimation of pressure o�set using thermodynamic properties. Thermodynamic 2-point and Least Square method (LSM) are under this category. 2.2.1 Fixed This is the simplest way of performing pressure correction. Using this method the pressure value is set to a defined (static) value at a specified crank angle position [5]. This method can be used when there is not much pressure drop in the intake side or a good estimate of in-cylinder pressure is already present. It is a simple method which is not always correct. 2.2.2 Measured In this method the cylinder pressure is referenced to the intake manifold absolute pressure at a certain point usually when the flow can be considered to be balanced. This point is generally taken as intake BDC but can be di�erent for highly tuned engines or engines operating at high speed [3]. This method has the advantage of being simple and fast but require the presence of an additional sensor. The location of the intake manifold sensor, reference timing and sensor noise are the main sources of error for this method [3]. 2.2.3 Thermodynamic 2-point In this method the pressure o�set is calculated by measured cylinder pressure data at two points. The points are considered after intake valve closing and before start of combustion so that the compression process can be considered polytropic [3] i.e. p · V n = constant (2.2) Let P1 and P2 be the two points. Then p2 = p1 · (V1 V2 )n (2.3) and pmeasured = pactual + poffset (2.4) �p = p2,measured ≠ p1,measured = p2,actual ≠ p1,actual (2.5) From equations 2.3, 2.4 and 2.5 poffset can be calculated as poffset = p1,measured ≠ p2,measured ≠ p1,measured (V1 V2 )n ≠ 1 (2.6) This method assumes the coe�cient gamma to be constant which is required as an input for this method. Incorrect estimated of gamma can lead to errors in pressure o�set. 8 2.2.4 Least Square Method This method uses multiple points between intake valve closing and start of com- bustion to fit a linear regression curve and estimate pressure o�set. Using a linear regression reduces the impact of signal noise and error in individual measurement. The measured pressure is a sum of actual pressure and a pressure o�set. The pres- sure o�set remains constant for di�erent measurements. pmeasured = pactual + poffset (2.7) Considering a reference point, having parameters pref and Vref , then it can be said that pmeasured = 5 Vref Vmeasured 6 “ ú pref + poffset (2.8) For a single point it can be written in a matrix form as shown below p = 1 1 c 2 · A poffset pref B (2.9) For multiple points the above equation can be extended as Q ccca p1 p2 p3 pm R dddb = Q ccca 1 c1 1 c2 1 c3 1 cm R dddb · A poffset pref B (2.10) which is a set of linear equation in the form of Y = X · w (2.11) where Y= Q ccca p1 p2 p3 pm R dddb , X= Q ccca 1 c1 1 c2 1 c3 1 cm R dddb and w= A poffset pref B . w and thus Poffset can be calculated by the following equation w = (X Õ X)≠1 X Õ Y (2.12) This method assumes that gamma is constant and is known beforehand. Also matrix inversion for multiple points can be computationally expensive. Around 15 points is a good trade-o� between reducing signal noise and computational speed. 9 2.3 Rate of Heat Release Combustion of the air-fuel mixture inside the engine cylinder is the main process that controls engine e�ciency, power, and emissions. As shown in Figure 2.8 below combustion of the air-fuel mixture is an exothermic reaction for which the outputs are high pressure and temperature inside the engine cylinder. The chemical energy of the fuel is the sum of internal energy(dUs), mechanical work(”W ), heat transfer rate(dQht), and energy loss through mass flow(dQ BB ) [8]. For an accurate combus- tion analysis each of the above processes must be studied. Figure 2.8: Combustion inside the engine cylinder Rate of Heat Release [8] is a thermodynamic approach based on the first law of thermodynamics. The major advantage of using this approach is the change in pressure can be directly correlated to fuel energy released by combustion. Heat release as a function of crank angle degree can be derived by applying the first law of thermodynamics to a controlled volume and assuming energy loss through mass flow is zero (dQ BB = 0). Then heat energy released by the combustion (dQch) is given by: dQch = dUs + dW + dQ ht (2.13) This can be represented as: dUs = mCvdT (2.14) dW = pdV (2.15) pV = mRT =∆ mT = pV R (2.16) Di�erentiating equation 2.16 gives: mdT = d(pV R ) (2.17) 10 Further: mcvdT = cv R (pdV + V dp) (2.18) substituting in the equation 2.13 gives: dQch = cv R (pdV + V dp) + pdV (2.19) Further, expression for Apparent rate of heat release as function of crank angle de- gree is given by, =∆ dQch d◊ = cv R · V · dp d◊ + (cv R + 1) · p · dV d◊ (2.20) Now, energy loss by heat transfer is given by the equation, dQht dT = Ahc(T ≠ Tw) (2.21) Energy loss due to mass transfer i.e. crevice and blow-by losses which is assumed zero due to the fact that it is very small, but in reality cannot be zero and given by, dQ BB = ( “ “ ≠ 1) · p · V · (dm m ) (2.22) Hence, expression for real heat release rate as function of crank angle degree is given by, =∆ dQch d◊ = cv R · V · dp d◊ + (cv R + 1) · p · dV d◊ + A · hc · (T ≠ Tw) + dQ BB (2.23) Rewriting Equation (2.23) in terms of ploytropic coe�cient: cp = cv + R (2.24) and “ = cp/cv (2.25) =∆ dQch d◊ = 1 “ ≠ 1 · V · dp d◊ + ( “ “ ≠ 1) · p · dV d◊ + A · hc · (T ≠ Tw) + dQ BB (2.26) From Figure 2.1 the change in volume with respect to crank angle degree can be calculated by the following relation [1, 8]: V = Vd rc ≠ 1 + (Vd 2 ) · [RL + 1 ≠ cos◊ ≠ (RL 2 ≠ sin 2 ◊) 1 2 ] (2.27) 11 From the geometrical parameters of the engine, the displacement volume (Vd) can be determined by the equation: Vd = fi 4 ú B 2 ú L (2.28) RL is the ratio of connecting rod length to crank radius and calculated as: RL = l a (2.29) Di�erentiating the equation(2.24) with respect to crank angle gives: dV d◊ = Vd 2 · sin◊[1 + cos◊ ≠ (RL 2 ≠ sin 2 ◊) ≠1 2 ] (2.30) Further, determination of change in pressure with respect to crank angle degree can be determined by two di�erent methods: (1)Finite di�erence method, and dp d◊j = pj+n ≠ pj≠n 2 ú �◊ (2.31) Where,�◊ is sampling step between two crank angle degrees (2) Savitzky-Golay filtering Calculating (dP d◊ ) by finite di�erence method does not give accurate values due to uncertainties i.e. as the pressure transducer is induced to thermal stress the recorded signal is very noisy. But, with Savitzky-Golay filter the pressure signal can be fil- tered and noise can be reduced to a large extent and proves to be more suitable for this application than the finite di�erence method is. To determine the rate of heat release accurately, an in-depth analysis was required for the parameters which have a major e�ect on the rate of heat release. These parameters are: 1) Gas properties 2) E�ective Intake valve closing (IVC) and Exhaust valve opening (EVO) 3) Heat transfer rate 4)Wall temperature 5) Blow-by All parameters are investigated and explained in detail below. 12 2.4 Gamma In several of the equations governing the operating of an internal combustion engine the polytropic coe�cient appears. This coe�cient has di�erent names, sometimes being refered to as gamma, kappa, polytropic coe�cient or polytropic index. For ease of reading this will from now on be called gamma or “. This coe�cient is a result of the ideal gas law acting in an isentropic process[9]: p ú V “ = constant (2.32) or: p1 ú V “ 1 = p2 ú V “ 2 (2.33) The coe�cient can be calculated in di�erent ways. In this report logarithmic re- gression and NIST polynomial will be explained in detail. The value of gamma is constantly changing due to several factors, including but not limited to temperature, gas composition and pressure[10]. Thus the common method of using a fix gamma of 1.35 or 1.4 [8, 11] is not optimal and a method that can take these parameters into account is beneficial. A gamma of 1.4 is often used, this is derived from Equation 2.34. As air is mostly a diatomic gas f=5 [11]. “ = f + 2 f = 5 + 2 5 = 1.4 (2.34) 2.4.1 Logarithmic regression The logarithmic regression method uses thermodynamics to analytically calculate the instantaneous gamma. The regression is based on Equation 2.33 above, and can be rearranged as follows: ln(p1) ú ln(V “ 1 ) = ln(p2) ú ln(V “ 2 ) (2.35) ln(p1) ú “ ú ln(V1) = ln(p2) ú “ ú ln(V2) (2.36) ln(p1) ≠ ln(p2) = “ ú ln(V2) ≠ “ ú ln(V1) (2.37) “ = ln(p1) ≠ ln(p2) ln(V2) ≠ ln(V1) (2.38) This equation should give the instantaneous gamma value between two sample points. From the data acquisition all of the input data is known. This equation can be used during the compression and expansion strokes when valves are closed and combustion is not occurring, as the composition changes during combustion. 13 2.4.2 NIST A model for finding the cp of a mixture of compounds as a function of temperature was first introduced by C. Howard Shomate in 1944 [12]. This method is based on experimental measurements of heat change in compounds and defining several coe�cients. With these ceo�cients (A-E) the function describing cp is: cp = A + B ú T + C ú T 2 + D ú T 3 + E/T 2 (2.39) The coe�cients are availible in large quanteties for the majority of simpler and more common compounds. In this report the tables made public by The National Institute of Standards and Technology, U.S. Department of Commerce[13] were used. Equation 2.39 determines the cp of a singular substance [14]. These coe�cients can then be added to create a combined cp for a mixture of compounds as stated in Equation 2.40. cp,tot = nÿ i=1 moli moltot ú cp,i (2.40) The “ can then be calculated by: cv,tot = cp,tot ≠ R Mm (2.41) Where Mm is the weighted averaged molar mass. And: “ = cp cv (2.42) Equation 2.39 requires the instantaneous temperature for accurate calculation. From [9]: pV = nRT (2.43) If constant mass is assumed in the system [9]: pV T = nR = prefVref Tref (2.44) Equation 2.44 rearranges to: T = Tref ú p ú v pref ú vref (2.45) 14 2.4.2.1 Gas composition When calculating the NIST polynomial (sometimes referred to as Shomate polyno- mial) knowing the correct gas-composition is necessary as this is the basis of the calculation. The exact composition of the in cylinder mixture is very hard to model as the introduction of the fuel and its combustion results in a lot of intermediate compounds [8]. Thus, it needs to be simplified. Two modes are taken into ac- count, dry air and exhaust gas. The composition of gas can then be controlled by a weighting factor between the two compositions. These compositions are presented in Figure 2.9 and 2.10 [15, 16]. Due to the complexity and small percentage of the unburned hydrocarbons it is assumed to consist of the most common: C2H4 [17]. Figure 2.9: Dry air, mole percentage of composition gases. Figure 2.10: Diesel exhaust, mole percentage of composition gases. 15 2.4.3 Ganippa proposal To explore alternative methods of estimating gamma a search of literature was con- ducted. A proposed method presented by G. Abbaszadehmosayebi and L. Ganippa[18] was explored. Based on: dQ d◊ = “ “ ≠ 1p dV d◊ + 1 “ ≠ 1V dp d◊ (2.46) the Ahrr50 (apperent heat release rate at CA50) can be expressed as: Ahrr50 = “ “ ≠ 1 A p dV d◊ B 50 + 1 “ ≠ 1 A V dp d◊ B 50 (2.47) Gamma can then be extracted to create: “ = Ahrr50 + 1 V dp d◊ 2 50 Ahrr50 ≠ 1 p dV d◊ 2 50 (2.48) Substituting “ from 2.48 in 2.46, Ahrr can be expressed as: Ahrr = Ahrr50 ú K1 + K2 (2.49) Where: K1 = V dp d◊ + p dV d◊1 V dp d◊ 2 50 + 1 p dV d◊ 2 50 (2.50) and K2 = 1 V dp d◊ 2 50 ú p dV d◊ ≠ 1 p dV d◊ 2 50 ú V dp d◊1 V dp d◊ 2 50 + 1 p dV d◊ 2 50 (2.51) This equation can be rewritten to create the Residual Function RF: RF = A “ “ ≠ 1p dV d◊ + 1 “ ≠ 1V dp d◊ B ≠ (Ahrr50 ú K1 + K2) (2.52) Where, according to the authors, the correct gamma is found when RF is minimized. This method was pursued but did not work as presented. For further discussion see section 4.4.3. 16 2.5 Heat Transfer During the combustion phase inside the cylinder, the temperature of the burned gasses reach high temperatures[8]. Consequently the di�erences in heat are large and must be taken into account. The high heat flux that is generated during the combustion phases plays an important role because these high temperatures dete- riorates the lubricating oil film, overheating of the exhaust valves etc. The heat transfer plays an important role in the engine’s performance, e�ciency and emis- sions. The e�ect of heat transfer is also significant for the rate of heat release study, because neglecting heat transfer predicts that combustion will takes place faster than it would with it taken into account. In order to predict these heat fluxes in the cylinder various models have been presented over the years. In this project some of these models will be explored to see which model among the proposed models is best suited for CI/SI engines [8]. 2.5.1 Woschni The heat transfer between the charge and the wall takes place by convection and radiation. Since the heat transfer is dominated by the convection, the radiation e�ect is neglected because the calculation is complex and the additional influence is small [19, 8]. According to Newton’s law of cooling the convective heat transfer can be expressed in the following equation as [20]: dQht dt = hc ú A(T ≠ Tw) (2.53) Where hc is the convective heat transfer coe�cient, A is the area, T is cylinder gas temperature and Tw is the wall temperature. In Equation 2.53 the total area of the cylinder wall at any crank angle is calculated using the following equation: A = fi 2 ú B 2 + � ú B ú L 2 [l + 1 ≠ cos� + (l2 ≠ sin�) 1 2 ] (2.54) The Woschni heat transfer coe�cient, b, is taken as the characteristic length and the characteristic velocity is the mean gas velocity in the cylinder [8, 21]. hc = 3.26 ú B ≠0.2 ú p 0.8 ú T ≠0.55(2.28 ú W )0.8 (2.55) In Equation 2.55, W is the average gas velocity and is calculated using: W = [C1 ú Sp + C2 Vd ú Tr pr ú Vr (p ≠ pm)] (2.56) In the above equation C1 and C2 are constants that vary depending on the combus- tion period. The values are shown below in Table 2.1 [8, 21]. Table 2.1: Woschni coe�cients Phase C1 C2 Intake-Exhaust 6.18 0 Compression 2.28 0 Combustion-Expansion 2.28 0.00324 17 The motoring pressure in Equation 2.56 is calculated using Equation 2.57. pm = pr(Vr/V )n (2.57) 2.5.2 Hohenberg correlation Hoehenberg suggested a correlation in 1979. He examined Woschni formula and made few changes. There are quite a few di�erences with Woschni heat transfer coe�cient equation and Hohenberg heat transfer coe�cient equation. First of which is that, for characteristic length instantaneous cylinder volume is used instead of bore, and the temperature exponent is changed from -0.55 to -0.4. The Hohenberg equation is stated below hc = 3.26 ú V 0.06 ú P 0.8 ú T e≠0.4(Sp + 1.4)0.8 (2.58) 2.5.3 Wall Temperature The wall temperature is defined as the temperature of the cylinder liner towards the combustion chamber. This is important when a heat analysis is to be carried out, and the analysis will be more accurate if this value is as close to reality as possible. The first and simplest method is to use a fixed value. This is a fast method and can work if the value is known and close to reality. The problem is that this value often is a guess, or a more or less good estimation based on earlier knowledge. Another problem is that the wall temperature most likely changes depending on load and speed. Therefore, a value that is calculated for the actual load point is more accu- rate. To do this the cooling water temperatures and flow is used. The temperature of the inlet and outlet water needs to be known as well as the cooling water flow and cp value for the fluid. Thereafter the energy taken away from the cylinders can be calculated by Equation 2.59. Q̇ = ṁcp�T = ṁcp(Tout ≠ Tin) (2.59) Twall = ( Q̇ hA ) + Tout (2.60) The outgoing water can also be assumed to have approximately the same temper- ature as the water surrounding the cylinders. The simplest way now is to use this value. The next step for an even better assumption is to use the material data of the cylinder liner. This method works best for engine blocks that have an open deck design. This is because most of the cylinder liner is in contact with the water, compared to closed deck design. Now the heat flow can be calculated through the cylinder liner. Since the material is known as well as the outside temperature, cylin- der geometry and heat carried away with the cooling water the inside temperature can be calculated. This is done by the help of equation 2.60, and gives the wall temperature. All these calculations are made for steady state cases to make sure the temperatures have stabilized. 18 2.6 E�ective Intake Valve Closing The intake valve closing is often given from the manufacturer. This IVC however is not always the actual closing. The given value can be a geometrical value, often based on the cam geometry. If this value is used for heat release calculation the answer will most likely di�er from reality. The intake valve closing determines how much charge that is trapped in the cylinder. Therefore, this value is vital for know- ing how much charge actually is inside of the cylinder during combustion. The value for the IVC that can be used instead is called IVC e�ective, and is calculated by the help of a logarithmic regression analysis of the in cylinder pressure and volume. To calculate the IVC e�ective a logarithmic regression analysis is made. How the logarithmic regression is done can be seen in section 2.4.1 The in cylinder pressure and cylinder volume with associated CAD is used. First the CAD for the given geo- metrical IVC is used to know when to start looking for the e�ective IVC. Based on this a value of how many CAD:s after this the analysis should include is estimated, and a maximum limit. Thereafter the logarithm of the volume and pressure is used to make a linear regression, where the pressure trace deviates more than a specified limit is the e�ective IVC. With other words where the plot makes a “bend”. This is illustrated in figure 2.11. Figure 2.11: Left: logaritmic pv-plot. Rhight: in-cylinder-pressure vs CAD. 19 2.7 Blow-by Despite the accurate construction of piston and cylinder assembly, the engine com- bustion chamber will have small volumes/ gaps which are known as crevices. The largest crevices are between piston, piston rings, and the cylinder wall. During the engine operating cycles, due to pressure di�erences, some amount of gas mixture will escape from these volumes into the crankcase and this phenomenon is known as blow-by [8][3]. Most of the blow-by occurs during compression and expansion stroke as shown in the Figure 2.12 below. Figure 2.12: Blow-by During compression and expansion stroke with Crankcase ventilation system (PCV) Even though its not optimal, earlier, the blow-by gas mixture used to be vented in the atmosphere, but now due to more stringent emission regulations, it has to be stopped. Because the blow-by gas consists of unburned hydrocarbons this type of emission needed to be addressed. In addition to this due to blow-by, the power generated and e�ciency of the engine is reduced as some amount of air-fuel mix- ture escapes, also, the fuel vapors escaping the cylinder can mix with the oil in the crankcase, thus compromising the lubrication property of the oil, which causes more friction and frictional wear of engine components. To avoid venting the gas mixture in the environment a system know as a posi- tive crank case ventilation system (PCV) is introduced as shown in above picture 2.12. The gas mixture accumulating in the crankcase is passed through the PCV, in which the oil and other impurities are separated (i.e.filtered ). And then the filtered gas is recirculated back to the intake manifold which contributes to the improvement of engine e�ciency. The disadvantage of PCV is the gas mixture filtered is not 100 percent clean, It forms a smudge layer in the intercooler and the intake manifold thus hampering the engine parts[22]. 20 Over the last years a lot of research have been carried out, but there is no established mathematical equation defined from which blow-by mass can be determined accu- rately, because blow-by mass generated depends on several factors such as engine speed, load and BMEP. It is therefore di�cult to generalise for all engine types. It also depends on engine wear properties which is very di�cult to predict and it can generally be said that the older the engine the higher the blow by rate [23][24]. In order to determine blow-by accurately, the only way is to install blow by me- ter and measure the values for that particular engine as blow-by intensity will be di�erent for every engine. A way to simplify the blow-by phenomenon it can be assumed as a choked flow through a small orifice [25]. Figure 2.13: (a)Piston and rings assembly, (b) Flow through orifice [8] The blow-by flow velocity for non-chocked flow can be determined by[25]: wBB = Û 2 ú “ “ ≠ 1 ú R M ú T [1 ≠ (p0 p )( “≠1 1 )] (2.61) and blow-by flow velocity for chocked flow can be determined by: wBB = Û 2 ú “ “ ≠ 1 ú R M ú T (2.62) The maximum mass flow occurs when the flow velocity at the minimum area (Agap) reaches the speed of sound, and this condition is known as choked flow. which is calculated by introducing the Mach number (M)[8]. M = V a ; where, a = Ò “ ú R ú T (2.63) The real mass flow rate can then be calculated by: dm = wBB ú Agap ú d◊ ú N ≠1 (2.64) 21 2.8 Compression Ratio The compression ratio is defined as the geometrical volume ratio between the volume at BDC and TDC. This is however not always the actual compression ratio. This is due to manufacturing tolerances, piston and valve sealing as well as porosity of the materials. The compression ratio is vital to get a correct heat analysis. Therefor the e�ective compression ratio is needed. The e�ective compression ratio is calculated by using the in-cylinder pressure. The data needs to be collected during motoring of the engine, which means fuel cut and no combustion occurring in the cylinder. This can be made in two ways. Either by letting a motor, for example the starter motor, power the engine, or by running the engine and then do a fuel cut (engine breaking). The benefits by doing it with engine breaking is that the compression ratio can be calculated for di�erent speeds and with di�erent heat levels. This means that for example the piston ring sealing is taken to account, which can vary. The drawback is that the engine is slowing down while the data is collected, which means more unstable data. Equation 2.65 shows the formula for calculating the compression ratio. The displace- ment volume is already known from the geometry of the crank. Whats unknown is the clearance volume. To calculate this, equation 2.66 and 2.67, must first be anal- ysed. Since there is no combustion occurring the losses are only dependent on losses to the wall and blow-by, hence the equations. When this is known the clearance volume can be calculated with the use of equation 2.68 [26]. “ for air is used at the given temperature. The walltemperature and blowby are already calculated and are therefore not calculated again. rc = Vmax Vmin = Vd + Vc Vc (2.65) dQn = “ “ ≠ 1pdV + 1 “ ≠ 1V dp (2.66) dEloss = dQwall + dHblowby (2.67) Vc = ≠“pdV ≠ V dp + (“ ≠ 1)dEloss dp (2.68) 22 3 Method Figure 3.1: The interaction of various function in the main code The code is composed into multiple functions which can be divided into three main sections. 1. Data loading - interface between the data acquisition software and MATLAB code. Included function: • loadComb 2. Data processing - Process the input data into a form suitable for analysis. Included functions: • filterComb • selectCycles • interpolate_time_2_CAD • zero_level_correction • avgComb 3. Data Analysis - Performs the rate of heat release analysis while taking inputs from multiple functions to improve its accuracy. Included functions: • RoHr • gasProp • IVCe� • heatTransfer • blowby 23 The code gives the user flexibility in choosing di�erent approaches and methods for data processing and analysis purposes. The data coming from Dewesoft is sampled in time domain. The code allows the user to continue the analysis in the time domain or convert it into CAD domain of specific resolution. The user also has the flexibility of choosing to run the analysis for multiple cycles or using an average cycle. Figure 3.1 shows the interaction of the various functions. 3.1 Test engines Data was collected from three di�erent engines in this project, to develop and val- idate the code. All those engines had di�erent parameters and sizes, two were compression ignited engines and one spark ignited. The most important data for those engines are stated below. • A 12.8 liter, in-line, 6 cylinder turbo diesel engine with common rail fuel injection, over head camshaft and 4 valves per cylinder. • A 2 liter, in-line, 4 cylinder, twin-turbo charged SI engine with direct injection, dual over head camshafts and 4 valves per cylinder. • A 2 liter, single cylinder, turbo diesel engine with common rail fuel injection, over head camshaft and 4 valves per cylinder. These engines were then run at di�erent operating points as data was collected. The reason for using three di�erent engines was to validate that the code was working, and to test it for both SI and CI engines. 3.2 Filtering The filtering of the data is done using a Savitzky-Golay filter. This is implemented in Matlab by using the function sgolay to calculate the convolution coe�cients and the function conv is used to produce the filtered signal and its derivatives. 3.3 Zero level correction The code zero_level_correction takes structs of cylinder pressure and the engine data as inputs along with the correction method and returns a struct containing the o�seted pressure data. Pressure values are o�setted for each individual cycle ensuring that the e�ects of long term drift are minimised. The code o�ers four dif- ferent correction methods. The user is asked to enter specific inputs for the selected method. For the fixed method user needs to input the reference CAD and the static pressure value. The measured value methods requires reference CAD at which the intake manifold pressure will be referenced to cylinder pressure. Thermodynamic 2-point asks the user for two reference points and the polytropic coe�cient. The function calculates the pressure o�set based on equation 2.6. Least square method asks the user the reference CAD and the gamma value. 24 3.4 Gamma The code gasProp was created do to give the user the ability to select between three di�erent types of Gamma calculation. The first method lets the user decide values for bot cp and cv to calculate gamma. The second and third applies the methods outlined in Section 2.4.1 and 2.4.2. 3.4.1 Logarithmic regression The logarithmic regression method applies Equation 2.38 at di�erent indices and calculates gamma. As the pressure and volume of the system is known the equation can be used to calculate mean values for compression by using IVC and SOC as indices and EOC and EVO as indices. The equation can also be applied in a time resolved manner and by a moving window calculate the instantaneous gamma by shifting the indices and evaluating gamma over short change in pressure and volume. 3.4.2 NIST Prior to calculating cv for any substance the instantaneous temperature is required. This is calculated using Equation 2.45 where the reference is set to the IVC as the pressure and temperature is known from measurements in the intake manifold. The gas composition at IVC can either be assumed air or a mixture of air and exhaust gas to simulate EGR. The transformation of the gas composition from air/air-EGR mix to pure exhaust gas is best approximated with MFB and 1-MFB respectively. This model does not take in-combustion gases into account. 3.5 E�ective Intake Valve Closing To determine the E�ective IVC the code "IVC E�ective" was created. This code needs the following inputs: Pressure, volume and CAD. The format this should be inputted in is that the vectors of those data sets needs to be of the same size, and all three needs to be from the same datapoint. Meaning that for every CAD there is corresponding volume and pressure values. When this data is inputted a logarithmic regression analysis can be made to determine the e�ective IVC, as explained in section 2.6. 3.6 Compression ratio The code "E�ective compression ratio" was created. This code needs motoring data as input.The input from this data is the pressure and volume along with the CAD. Both the actual volume and pressure as well as the deltas for those are needed. Fur- thermore gamma for air, from the "gamma code" are needed, the wall temperature losses and blowby losses from each accordingly code are also needed. From those inputs the actual compression ratio can be determined, by the help of equation 2.68 & 2.65. 25 3.7 Heat transfer models A MATLAB code was created in which, the provision was given to select the heat transfer model among the two models. The input data necessary for calculation of heat flux for both Hohenberg and Woschni were taken from the generated engine data file. For the Woschni model the heat transfer coe�cient is calculated using Equation 2.55, where the Woschni correlation constants C1 and C2 takes up di�erent values for di�erent phases of the combustion. 3.7.1 Wall Temperature To calculate the wall temperature the code "Wall Temp" was created. This code needs the geometry of the cylinder liner (bore & stroke). And material data for this, the heat transfer coe�cient value are taken from the Woshni calculations. This is to not build in more uncertainty. It also needs the mass flow of the cooling water, as well as the outlet and inlet temperature to the engine. The last input needed is the cp value for the cooling liquid. From this the wall temperature can be calculated by the help of equation 2.59 & 2.60 3.8 Blow-by To determine the blow-by of the engine a code is created based on simplified com- pressible flow model by using the equations 2.62 and 2.64. The inputs for the code are polytropic coe�cient, gas constant, temperature, crank angle degree, piston ring gap and engine speed. For which the resulting output is mass loss due to blow-by, this mass loss is added to the main heat release equation 2.26 as dQBB because the pressure data recorded is after blow-by losses [3, 25]. 3.8.1 Blow-by code validation by using experimental values Measured blow-by data through a blow-by meter collected from the HD diesel en- gine. From the data a 3D graph is plotted as shown in Figure 3.2. 26 Figure 3.2: (a) Blow by measured values-for di�erent engine speed and Torque,(b) Blow by at engine speed 1200 [RPM] and di�erent Torques From the above graph it can be interpreted that blow-by intensity changes with change in engine speed and Torque. Blow-by is highest at maximum rpm and torque. Data recorded with Dewesoft acquisition system on a HD diesel engine is at 1200rpm and T=600[N/m] and this can be correlated to the measured data blow-by of 1.13[Ltr/sec]. The blow-by percentage can be calculated [22] for the engine with 4-stroke(n=2), Vt=12.8 litres, at N= 1200 [rpm] [22]. V = (N ú Vt)/n = 128[Litres/sec]. (3.1) Blow ≠ by in percentage = 1.13/128 ú 100 = 0.882%. (3.2) The Blow-by estimation model is tuned and validated with respect to the experi- mental values. 3.9 Rate of heat release The filtered pressure, crank angle degree, and engine data is used as inputs and performs the rate of heat release analysis as described in the section 2.3. The rate of heat release and cumulative mass fraction burned curves are the main outputs from the code and this code will be the base and reference for comparison of the analysis. 27 3.10 Determination of pressure derivative In cylinder pressure accuracy plays an important role in performing combustion analysis. A pressure transducer is installed on cylinder-2 of the engine and the experimental set up is as shown in Figure 2.4. As described earlier the pressure transducer is sensitive to outside e�ects and requires signal conditioning. Signals of pressure transducers can be conditioned by: • Signal conditioning by charge amplifier • Signal conditioning by current to voltage converter Most of the data acquisition systems use signal conditioning by a charge amplifier, which has a major disadvantage of instability and drift [4]. As the data acquisition system previously used by the local researchers was charge amplifier method, signal conditioning by current to voltage converter is investigated further. 3.10.1 Signal conditioning by current to voltage converter In order to measure the in cylinder pressure derivative accurately the output current from the transducer should be converted to voltage signals[4]. The values will be more precise if this is performed with a low impedance device. Current-to-voltage converter is a low impedance device which eliminates the signal distortions caused by transducer self capacitance. The construction of a current to voltage converter is simple as it only consists of two operational amplifiers and a resistor [4]. The circuit diagram is as shown in the below Figure 3.3: Figure 3.3: Current to voltage converter [4] The measured pressure variation through the unit converter is given by: dP dt = Vcv Gs ú Ra (3.3) Where Vcv = Output voltage, Gs = Transducer gain and Ra =Resistance value. Further, pressure data can be determined by fourth order Runge-Kutta numerical integration method [4]. The current to voltage converter device was built in house calibrated and tested for recording the in-cylinder pressure data. 28 4 Results and Conclusion In this chapter results and conclusions for each separate sub-code will be presented. The results will be chosen in a way to show a representative measure of the e�ect of the process. All measures will be presented in the Appendix for further reference. In conjunction with with the results a short specific conclusion will be presented. For a overall discussion, see Chapter 5. 4.1 Filtering The filtering of the pressure signal by Savitzky-Golay is carried out using 5th order filter and a window size of 301.This is presented below in Figure 4.1. Figure 4.1: The visual e�ect of the filtering process. To compare the e�ectiveness of the filter the di�erence in IMEP was calculated between unfiltered and filtered pressure data, and the total mean di�erence of ≠3.8147 ú 10≠6 Pa, which represents a percentage di�erence of 4.8776 ú 10≠5%. 29 The e�ect of the filtering process is advantageous, the signal is smoothed and any high derivative is smoothed down to reasonable levels. When comparing the di�er- ence between an unfiltered and filtered IMEP only a small di�erence is seen. This indicates that the analysis following the filtering isn’t a�ected by the small di�er- ence between the signals due to filtering. The di�erence of 4.8776 ú 10≠5% is seen as negligible. 4.2 Interpolation IMEP was used to justify the e�ectiveness of interpolation as was done with filtering. The mean IMEP di�erence with an interpolation of 0.1 deg crank angle came out to be 1.77 ú 10≠4, which represents a percentage di�erence of 0.0015. Figure 4.2: The visual e�ect of the interpolation process. 30 4.3 Zero level correction The raw pressure obtained from Dewesoft was corrected for zero level by three di�erent methods. Figure 4.3: Comparison of di�erent Zero level correction methods Figure 4.4: Comparison of di�erent Zero level correction methods in log-log scale The di�erent zero level correction give di�erent o�sets. This is more evident in figure 4.4 which shows the logP-logV diagram for di�erent methods. This can lead to di�erences in gamma, max pressure and combustion identification parameters like CA50. 31 The choice of best method to use depends on use case, availability and confidence of the available data. If accurate pressure values inside cylinder is known then fixed method should be used. Engines operating at low speeds without having a heavily tuned intake (low ram e�ect) can use measured value method. Pressure readings for this method is highly influenced by location of intake manifold pressure sensor. Thermodynamic 2pt and Least square methods can be used interchangeably and don’t require additional data or sensor. LSM method provides more isolation towards sensor noise at a cost of high computational time. 4.4 Gamma 4.4.1 Logarithmic regression The method of logarithmic regression is highly dependant on the input data and a general result is di�cult to produce. Presented in Figure 4.5 is the resulting output from a data-set acquired from an SI-engine. Figure 4.5: Gamma output from Logarithmic regression and NIST. The method of logarithmic regression has its beginning in a universal thermody- namic truth. This should by logic give it the true value of gamma. The problem with this assumption is that the measurements taken aren’t perfect. If the pressure trace collected is incorrect the gamma calculated will be incorrect. The advantage of the logarithmic regression method is its simplicity. As the pressure and volume of the cylinder should be known the gamma can be calculated with only these vari- ables. There is therefore no need to introduce more assumptions, ex. compositions, and the model takes care of the change between SI and CI independently. The 32 weakness of the method is that it can and will give gamma values that are outside of the expected range if the input data is wrong. The gamma range should be be- tween 1.25-1.4, but this method can give gamma values well outside of this range if any of the inputs are incorrect. Another disadvantage of the logarithmic regression method is the need of adiabatic process, this results in an inability to calculate the coe�cient during the actual combustion. During the project, it was found that using the logarithmic regression method when the pressure sensor and amplifier gain is set incorrectly will result in correct outputs as logarithmic regression seemingly corrects the error of incorrect gains. This has been attributed to that the calculated gamma is made from the same data it is later used to correct, which will correct any inconsistency in data acquisition. 4.4.2 NIST-Polynomial The output from the NIST-Polynomial is compared to the output from the Loga- rithmic regression in Figure 4.5. The method of using the NIST-polynomial gives some apparent advantages, mainly that it is a tested method that gives a gamma output that is within the physical range of the coe�cient. The second advantage is the academic agreement that the method is correct enough to use for further analy- sis. The third advantage of the method is that the coe�cient can be approximated throughout the combustion which some other methods cannot do. The disadvantage of the method is the need for a proper mixture composition or at least an approxi- mation of the composition which will introduce errors. The use of the NIST method on combustion will always be approximate as there are a large amount of burnt and unburned species reacting throughout the combustion. This was in the code ap- proximated with a pre-ignition and a post-combustion mixture and the composition inbetween these two points approximated as a transfer between the two states with a MFB-curve. It can be seen in Figure 4.5 that the approximation of gamma by NIST-Polynomial gives a gamma that follows the expected curve and at all points fall within the expected range of the coe�cient. 4.4.3 Ganippa proposal The method of calculating gamma presented by G. Abbaszadehmosayebi and L. Ganippa [18] was pursued but could not be made to work. In Equation 2.47 the ap- parent heat release rate is calculated at CA50, what is not included in this equation is the index for gamma “50 and instead “ is used. Due to the fact that gamma will vary throughout the compression and expansion this index is important. Later in Equation 2.52 the two gamma are assumed the same. This is not correct and the residual function will give an incorrect estimation. 33 4.5 IVC E�ective The IVC e�ective di�ered from the stated value by the manufacture in most of the cases. This was in the range of around 0-7 CAD and in all cases later closing than stated. For some engines tested no values for IVC was stated, and therefore no conclusion on how much it di�ered can be made. However, by looking at the log-log P-V diagrams it could clearly be seen that the compression started where the code set IVC e�ective. There was also one engine which the code had problem calculating the actual closing, and it was also hard to determine by looking at the graphs. This was due to fluctuations/ vibrations a�ecting the pressure. Whether that was due to valve closing or not is di�cult to determine. Except from this single case the code seems to deliver accurate values. The IVC e�ective will a�ect a lot of parameters and are in that sense important to get accurate, and to minimise false marginals going further in the code. The IVC e�ective is used for determining how much charge that is trapped in the engine, and how big the “explosion” will be. But as seen in the sensitivity analysis it does not a�ect the overall RoHR that much, compared to other factors. 4.6 Compression ratio The compression ratio di�ered from the stated from the manufacture by around 0.01-0.1 depending on the tested engines. This seems reasonable and can as said depend on tolerances in the manufacturing process and porosity, as well as valve closing o�set. This gives a small di�erent in the overall RoHR, but are far from the most a�ecting factor, as seen in the sensitivity analysis. 4.7 Heat transfer models Comparison graphs for heat transfer coe�cient and heat transfer for both Woschni and Hohenberg is presented below in Figure 4.6 and 4.7 respectively. From the figures it can be seen that the maximum HTC (heat transfer coe�cient) observed for Woschni is around 1250 W/m2-K and for Hohenberg it is approximately 2400 W/m2- K. There is a significant di�erence between the models for heat transfer curves as well. The maximum heat transfer observed for Woschni is 4 J/deg, whereas the maximum heat transfer observed for Hohenberg is 8 J/deg. The di�erence in these values is due to the di�erence in the characteristic velocity and the temperature exponent. The Hohenberg model has the highest heat transfer coe�cient compared to the Woschni because of the change in the temperature exponent. Hohenberg assumes constant characteristic velocity across all crank angles, whereas the Woschni equation calculates a di�erent characteristic velocity across all crank angles. It can be seen from the Figure 4.7 that, as the piston is in downward motion, and during the intake process the heat flux is minimum. The heat flux increases when the piston is compressing the gas mixture. 34 Figure 4.6: Comparison of Heat transfer co-e�cient for di�erent heat transfer models Figure 4.7: Comparison of Heat transfer for di�erent heat transfer models 35 4.8 Rate of heat release The major project delivery,the modular rate of heat release analysis software is completed. This software have model based sub functions that can be used with dif- ferent options for di�erent RoHR analysis. The MATLAB based software is capable of performing both apparent and real rate of heat release calculations. Below are the results of apparent and real rate of heat release of the six cylinder CI engine. Table 4.1: Engine, operating conditions, and RoHR methods Engine type CI Inline Six Turbocharged Operating condition RPM: 1200 Torque: 600Nm ZLC method Thermodynamic 2pt -100 -65 1.37 Gamma method Logarithmic Regression 2,3 IVC method IVC e�ective Rohr type Apparent/Real Wall temp method User input Heat Tansfer method Hohenburg dP/dCad method Savitsky golay In Table 4.1 the details of the engine and the details of di�erent options used to perform the rate of heat release analysis are presented. Figure 4.8: Rate of heat release curves of: chemical energy, internal energy, mechanical work and heat transfer rate 36 Based on the behavior of rate of heat release curves, as seen in figure 4.8 and figure 4.11, several conclusions can be drawn. The internal energy is a function of temperature, mass and calorific value of the fuel. Thus at the end of compression where temperature and pressure are at their maximum, also internal energy will reach its maximum value close to TDC. As the expansion progresses the internal energy will reduce as pressure and temperature decreases. The mechanical work curve will initially be negative as work is required to compress the air-fuel mixture. After TDC and during the expansion phase the work will be positive as pressure from the combustion will power the piston downwards. The heat transfer can be seen to have a small e�ect on the heat release but can be seen during the compression and expansion strokes as heat transfer will take place to piston and cylinder walls during these phases when the temperature is high [8]. Figure 4.9: Cumulative mass fraction burned curve The mass fraction burned curves of apparent and real heat release of both CI and SI engine is as shown in the figures 4.9 and 4.12. From the mass fraction burn curve the di�erence can be seen, especially at the end of the combustion, where the left over mass burns cannot be easily quantified [8]. When considering real rate of heat release, blowby and heat transfer are also added to pressure data obtained from in-cylinder measurement. This increases the cumulative energy released as compared to apparent case. During normalisation, both the curves are normalized by their respective cumulative sum, and this causes the real mass fraction burned curve to appear lower than apparent curve. If instead of the respective cumulative sum, the fuel energy content i.e. mf ú QLHV (mass of fuel multiplied by its lower heating value) is used as the normalising factor then the results would be di�erent and apparent will show incomplete combustion and unexplained loss of energy. This means the real heat release analysis reflects the actual combustion inside the engine cylinder and there is room for improving the e�ciency. 37 Figure 4.10: Rate of Heat Release Table 4.2: Results of CI engine at 1200 RPM and 600 Nm heat release analysis RoHR method Apparent Real IMEPnet 8.39 8.39 IMEPgross 8.12 8.12 PMEP -0.27 -0.27 Pmax 88.36 88.36 Gamma 1.37 1.37 hcmax 0.00 2384.58 dQhtmax 0.00 8.46 Tmax 1433.02 1433.02 SOC -4.70 -6.20 CA50 6.30 6.70 CA90 19.50 22.80 The result of CI engine at 1200 RPM and 600 Nm after rate of heat release analysis is presented in table 4.2. Similarly Apparent and real rate of heat release analysis done for SI engine and following are the results shown below: SI engine at 1500 RPM and BMEP 9 Bar. 38 Table 4.3: Engine operating conditions, and RoHR methods Engine type SI Inline Four Turbocharged Operating condition RPM: 1500 BMEP: 9bar ZLC method Thermodynamic 2pt -100 -65 1.37 Gamma method Logarithmic Regression 2,3 IVC method IVC e�ective RoHR type Apparent/Real Wall temp method User input Heat Tansfer method Woschni dP/dCad method Savitsky golay The above table 4.3 gives the details about the engine and the details of di�erent options used to perform the rate of heat release analysis. Figure 4.11: Rate of heat release curves of: chemical energy, internal energy, mechanical work and heat transfer Figure 4.12: Cumulative mass fraction burned 39 Figure 4.13: RoHR Table 4.4: Results of CI engine at 1200 RPM and 600 Nm heat release analysis RoHR method Apparent Real IMEPnet 9.32 9.32 IMEPgross 8.82 8.82 PMEP -0.50 -0.50 Pmax 48.53 48.53 Gamma 1.37 1.37 hcmax 0.00 471.36 dQhtmax 0.00 1.12 Tmax 2406.77 2406.77 SOC 2.20 2.10 CA50 11.20 11.50 CA90 18.80 20.30 Hence the results of heat release analysis of SI engine at 1500 RPM engine (Speed) and 9Bar (BMEP) are presented in table 4.4. 40 4.8.1 Validation of RoHR analysis In motoring case there is no chemical energy involved from fuel and thus its rate of heat release curve should be a flat line. This fact can be used to validate the di�erent RoHR methods and can be used to justify the selection of various param- eter sub-models. From the figures 4.14 and 4.15,the chemical heat release curve is approximating a flat behaviour. The values are not exactly zero, but are in range of +/- 2 J/deg. This can be considered negligible when it is compared against the values obtained during combustion. This curve can be further tuned by perform- ing data quality checks and ensuring that incoming data is free from small defects like charge leakage etc. The presence of a relatively flat profile can be seen as a validation for the code. Figure 4.14: Motoring RoHR curves of CI engine Figure 4.15: Motoring RoHR curves of SI engine 41 4.8.2 Wall Temperature The wall temperature gives values of around 80°-95° Celsius. The problem is to validate those numbers, since no actual measurements have been made. However, earlier assumptions on the same engine tested have been 80°C, compered to this it seems to be in the right region. As seen in the sensitivity analysis it does not a�ect the overall RoHR in that great manner compared to the other factors. This is also a hard parameter to determine since di�erent engine blocks will have di�erent geometries and material properties. The physics should at least give a good as- sumption/starting point, but to get it more accurate than this actual measurements needs to be made. 4.9 Input data validation The analysis of in cylinder pressure is as shown sensitive to pressure irregularities and errors. Therefore, it is valuable, to early in the process, be able to identify these problems and thus not rely on incorrect data acquisitions. Inconcistency in the input data can be seen in two ways. Both are based on the logarithmic regression model for calculating gamma, and compared to the known bound NIST polynomial. Therefore if the calculated data is inconsequential in several regards compared to the NIST polynomial the data might be incorrect. First, it can be identified by comparing the computed gamma to the known physical range of gamma. This is shown in Figure 4.16. The left image with "incorrect" input data will have values well above and below the expected margin of gamma, while the right data is more of what should be expected from correctly acquired data. Figure 4.16: Di�erent datasets, out-of-bound and inside-bound Second it can be checked by the mean gamma slope within the compression and expansion stroke, this can be seen in Figure 4.17. For the left figure it is easy to see the di�erence in the slope between the NIST polynomial and the logarithmic regression, while in the right figure the mean slope from both -110°- -75° and -75°- -25° follows the mean slope of the NIST polynomial quite well. 42 Figure 4.17: Di�erent datasets, not following, and following the NIST-slope This validation can be done to indicate that there are obvious errors in the data acquired and some further action in cleaning or collecting must be made. The calculated gamma curve and the NIST polynomial results can not be expected to completely match as this would make one of the two models superfluous but the general trend of the two should be close. 4.10 Current to Voltage converter The current to voltage converter that was built behaved non-linearly during testing and calibration. This uncertainty might be attributed to the soldering as the op-amp is sensitive to temperature, the non-linearity can also be due to current saturation of the op-amp. 43 44 5 Discussion 5.1 Sensitivity Analysis Rate of heat release curves are a�ected by various parameters such as trigger o�set, zero level correction, gamma, heat transfer, compression ratio, IVC, temperature at IVC, wall temperature and RoHR method namely apparent or real. Inclusion of sub models like blowby can also a�ect the results. A sensitivity analysis was performed to quantify the e�ect of all these parameters and to establish the greatest contributors. Figure 5.1 represents the inter dependencies of various parameters. Table 5.1: Sensitivity Analysis dependencies Trigger o�set Trigger o�set is used to synchronise the crank angle encoder data with the pressure data. Change in trigger o�set changes IMEP, gamma, heat transfer, temperature and rate of heat release. Pressure o�set Pressure o�set or zero level correction a�ects IMEP gross and PMEP calculation but IMEP net remains constant. Maximum pressure, gamma, heat transfer, temperature and rate of heat release parameters are changed. Gamma Gamma a�ects rate of heat release as shown in equation 2.26. It does not a�ect IMEP, pressure, temperature and heat transfer properties. 45 Compression Ratio Compression ratio changes volume which in turn changes gamma, temperature, heat transfer and rate of heat release. IVC e�ective IVC e�ective is the reference point chosen to calculate in cylinder temperature. Temperature and heat transfer are a�ected which leads to a change in rate of heat release parameters. Temperature at IVC Temperature at IVC a�ects temperature calculation which in turn a�ects heat trans- fer and rate of heat release calculations. Wall temperature Wall temperature a�ects heat transfer which causes slight change in rate of heat release parameters. Sensitivity comparison Figures 5.1 and 5.2 show the change in CA90 in crank angle degrees with a delta change in di�erent parameters. Since parameters have di�erent units and compar- ison with unity change is not possible (for example-unity change of intake valve temperature is a small quantity, but that of gamma is a very huge quantity) e�ort has been made to keep the delta change reasonable for each parameter. Comparing the parameters at this scale shows that rate of heat release (as evident by CA90) is most a�ected by pressure o�set and gamma, followed by trigger o�set and compres- sion ratio. IVC does not have a significant change in RoHR analysis. 46 Figure 5.1: Parameter sensitiveness towards RoHR, CI engine Figure 5.2: Parameter sensitiveness towards RoHR, SI engine 47 5.2 Concluding remarks The code developed for calculating combustion analysis parameters investigates dif- ferent parameters which can a�ect the heat release calculations. It provides a mod- ular platform to calculate RoHR with di�erent levels of complexity incorporating multiple models to be used according to availability of data and engine type. No model can be completely perfect, but the code tries to give the theoretical best approximation for the combustion parameters. This is important as it will help analyse the split of loss from the engine indicating the e�ciency of the engine and major loss areas. Some of the parameters gives a small change in the overall RoHR but all together might add up to a noticeable di�erence. Also, correct estimation of parameters like combustion duration is important as it is a validation point used for finalising cylinder head design and configurations aimed at providing faster com- bustion. True estimates will allow better correlations with simulation results and faster and e�cient development. The created code gives the user freedom to choose what models to include and which to omit. It is based on modularity and allows the user to create their own models to use alongside this code. The code is also open for the user to read and there are no embellishments of the results and the user is only given the truth. 48 49 50 Bibliography [1] Dr.Allan T. Kirkpatrick. Slider crank model. https://www.engr.colostate.edu/~allan/thermo/page1/page1f.html. [Online accessed 18-Jan-2021]. [2] Wikipedia contributors. Internal combustion engine — Wikipedia, the free encyclopedia. https://en.wikipedia.org/w/index.php?title=Internal_ combustion_engine&oldid=996699589, 2020. [Online; accessed 17-January-2021]. [3] Rakesh Kumar Maurya. Reciprocating engine combustion diagnostics. Springer, Switzerland, 2019. [4] Luiz Fernando Milanez Andre Valente Bueno, José Antonio Velásquez. A new engine indicating measurement procedure for combustion heat release analysis. Applied Thermal Engineering, 29(6):1657–1675, 2009. [5] DEWESoft. Combustion Analyzer, 1.1.0 edition. [6] Abraham. Savitzky and M. J. E. Golay. Smoothing and di�erentiation of data by simplified least squares procedures. Analytical Chemistry, 36(8):1627–1639, 1964. [7] R. W. Schafer. What is a savitzky-golay filter? [lecture notes]. IEEE Signal Processing Magazine, 28(4):111–117, 2011. [8] John B. Heywood. Internal Combustion Engine Fundamentals. McGraw-Hill, New York, USA, 1988. [9] Frank M.White. Fluid mechanics, Seventh edition in SI units. McGraw-Hill, NY 10020, New York, USA, 2011. [10] Engineering Tool Box. Air - specific heat at constant temperature and varying pressure, 2009. [2020-11-09]. [11] Francis W. Sears and Gerhard L. Salinger. Thermodynamics, Kinetic Theory, and Statistical Thermodynamics. Addison-Wesley, Reading, Massachusetts, 1975. [12] C.H. Shomate. High-temperature heat contents of magnesium nitrate, calcium nitrate and barium nitrate1. Journal of the American Chemical Society, 66(6):928–929, 1944. [13] The NationalInstitute of Standards and Technology. Standard reference database number 69. Electronic, https://doi.org/10.18434/T4D303, 2018. [14] Amyn S. Teja. Simple method for the calculation of heat capacities of liquid mixtures. Journal of Chemical & Engineering Data, 28(1):83–85, 1983. [15] Konrad Reif. Diesel Engine Managementn. Springer Vieweg, Wiesbaden, Germany, 2014. 51 https://www.engr.colostate.edu/~allan/thermo/page1/page1f.html https://en.wikipedia.org/w/index.php?title=Internal_combustion_engine&oldid=996699589 https://en.wikipedia.org/w/index.php?title=Internal_combustion_engine&oldid=996699589 [16] Engineering Tool Box. Air - composition and molecular weight, 2003. [2020-10-23]. [17] Grzegorz Kosza≥ka and Jacek Hunicz. Detailed speciation of emissions from a diesel engine fuelled with canola methyl ester. In International Scientific Conference on Aeronautics, Automotive and Railway Engineering and Technologies, 2018. [18] G. Abbaszadehmosayebi and L. Ganippa. Determination of specific heat ratio and error analysis for engine heat release calculations. Applied Energy, 122:143–150, 2014. [19] Joop Somhorst. Reduction of heat transfer and heat load in internal combustion engines. International journal of engine research, 29:1–21, 2019. [20] G. Woschni. A universally applicable equation for the instantaneous heat transfer coe�cient in the internal combustion engine. SAGE journal, 1967. [21] Talal Yusaf A. Aziz Hairuddin, B. Andrew P. Wandel. E�ect of di�erent heat transfer models on a diesel homogenous charge compression ignition engine. IJAME, 8:1292–1304, 2013. [22] Digsy. Blow-by and breather systems. Designing Breather Systems, 2005. [23] Karol F. Abramek. An attempt at as analytical description of blow-by intensity to a crankcase. Automotive vehicles operation, (17):7–14, 2007. [24] Shardul Gargate, Rohan Jacob1 Rohan Aher, and Sunil Dambhare. Estimation of blow-by in diesel engine: Case study of a heavy duty diesel engine. International Journal of Emerging Engineering Research and Technology, (18):165–170, 2014. [25] Luca Marchitto Simona Silvia Merola Adrian Irimescu, Cinzia Tornatore. Compression ratio and blow-by rates estimation based on motored pressure trace analysis for an optical spark ignition engine. Applied Thermal Engineering, 61:101–109, 2013. [26] Joop Somhorst et.al. A method to evaluate the compression ratio in ic engines with porous thermal barrier coatings. SAE Technical Paper, 1(1778), 2018. 52 A Suggested parameters Table A.1: Recommended options for getting best RoHR results for CI and SI engine 53 54 B User interface To make the software usable by the average user a user interface was created to give potential users a simple input to the otherwise complex code and code structure. This UI was coded and designed using the built-in app creator in MATLAB. The UI is capable of loading data-sets and setting engine parameters and then allows the user to select what models that are to be used. The code can save previous setups if the same data or data with the same variable names is to be used. There is a hope to eventually complete a post analysis UI to visualize results in an understandable way but is of yet not completed. Figure B.1: UI, Mainpage 55 Figure B.2: UI, Setup page 1 Figure B.3: UI, Setup page 2 56 Figure B.4: UI, Setup page 3 Figure B.5: UI, Run page 57 Nomenclature List of Figures List of Tables Introduction Background Aim Limitations Specifications of issues under investigation Deliverables Theory Filtering Zero Level Correction Fixed Measured Thermodynamic 2-point Least Square Method Rate of Heat Release Gamma Logarithmic regression NIST Gas composition Ganippa proposal Heat Transfer Woschni Hohenberg correlation Wall Temperature Effective Intake Valve Closing Blow-by Compression Ratio Method Test engines Filtering Zero level correction Gamma Logarithmic regression NIST Effective Intake Valve Closing Compression ratio Heat transfer models Wall Temperature Blow-by Blow-by code validation by using experimental values Rate of heat release Determination of pressure derivative Signal conditioning by current to voltage converter Results and Conclusion Filtering Interpolation Zero level correction Gamma Logarithmic regression NIST-Polynomial Ganippa proposal IVC Effective Compression ratio Heat transfer models Rate of heat release Validation of RoHR analysis Wall Temperature Input data validation Current to Voltage converter Discussion Sensitivity Analysis Concluding remarks Suggested parameters User interface