Assembly Instructions for the Swedish Manufacturing Industry of the Future Designing and comparing effective assembly instructions in line with digitalization Master’s thesis in Production Engineering NICLAS BUSCK and FREDRIK SVENSSON Department of Product and Production Development CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2017 Master’s thesis 2017 Assembly Instructions for the Swedish Manufacturing Industry of the Future Designing and comparing effective assembly instructions in line with digitalization. Niclas Busck and Fredrik Svensson Department of Product and Production Development Division of Production Systems Chalmers University of Technology Gothenburg, Sweden 2017 Assembly Instructions for the Swedish Manufacturing Industry of the Future Designing and comparing effective assembly instructions in line with digitalization. Niclas Busck and Fredrik Svensson © Niclas Busck and Fredrik Svensson, 2017. Supervisor: Dan Li, Department of Product and Production Development Examiner: Åsa Fast- Berglund, Department of Product and Production Develop- ment Master’s Thesis 2017 Department of Product and Production Development Division of Production Systems Chalmers University of Technology SE-412 96 Gothenburg Telephone +46 31 772 1000 Cover: Operator interface of a augmented reality-based assembly instruction. Typeset in LATEX Printed by Chalmers Reproservice Gothenburg, Sweden 2017 iv Assembly Instructions for the Swedish Manufacturing Industry of the Future Designing and comparing effective assembly instructions in line with digitalization. Niclas Busck and Fredrik Svensson Department of Product and Production Development Chalmers University of Technology Abstract Digitalization within the Swedish manufacturing industry is today realized in many different ways. It often involves high levels of automation, with robotics and internet of things. How digitalization can benefit the manual assembly work where simple paper assembly instructions are used for industries with fewer levels of automation is however less mentioned. This thesis work focuses on designing and comparing three assembly instructions for a manual assembly operation in line with digitalization that supports human cognitive processes. The three instructions types considered are text and picture (T&P), Video and augmented reality (AR) and they use dif- ferent technologies that could be connected to the digitalization concept. These instructions are then tested in experiments to find out differences in their individual performances regarding time to complete an assembly, achieved product quality and perceived acceptability by inexperienced operators. All three instructions are effec- tively designed, which entails a detailed study of both planning and presentation of instructions. The instructions have also been individually enhanced by the usage of instruction guidelines from available literature to reach each of the instruction types highest potential before compared in the experiments. The results of this thesis consists of instruction pre-work (planning), three designed instructions (presenta- tion) and an experimental study that gives industrial engineers directions on how to design assembly instructions for inexperienced operators and which type of tech- nology to employ. The Video instruction performed overall best in the experiments and is therefore recommended to be used considering inexperienced operators. The thesis concludes that big improvements in instructions design can be reached with familiar technologies that are in line with digitalization, which could have a large impact on Swedish companies’ short term production efficiency and long term global competitiveness. Keywords: assembly instructions, augmented reality, guidelines, product design, assembly sequence, comparison, experiments, video, text and picture. v Acknowledgements There are a lot of people who have helped us during our master thesis work and de- serve appreciation. First of all, we would like to thank Richard Hedman and Johan Bengtsson from GTC who made this thesis possible together with our examiner Åsa Fast-Berglund. We hope that our work have contributed to the Smarta Fabriker project and we wish them the best of luck in building the exhibition at Universeum later this year. Then, we would of course like to thank our supervisor Dan Li for his full commitment from the start and valuable feedback, which was very influential. Also, we would like to specifically thank Pär Magnusson from Stora Enso in Skene who helped and guided us in our cardboard design development work. The thesis would not be the same if not for all the participants in our workshop and experiments, who gave valuable qualitative feedback and quantitative data. So, a very big thanks to all of the involved researchers from Chalmers University of Technology and high school students from GTG. Lastly, we would like to thank all the other thesis workers connected to Smarta Fabriker who contributed to the interesting weekly meetings at Visual Arena, Lind- holmen science park, Gothenburg. Niclas Busck and Fredrik Svensson, Gothenburg, June 2017 vii Contents List of Figures xiii List of Tables xvii 1 Introduction 1 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1.1 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.1.2 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . 4 1.1.3 Delimitations and Scope . . . . . . . . . . . . . . . . . . . . . 4 1.2 Introduction to thesis methodology . . . . . . . . . . . . . . . . . . . 5 1.2.1 The VR-product: Initial design . . . . . . . . . . . . . . . . . 6 1.3 Thesis overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2 Methods 7 2.1 Pre-work: Setting the foundation for designing effective instructions . 7 2.1.1 Improving the product design . . . . . . . . . . . . . . . . . . 8 2.1.2 Finding the most appropriate assembly sequence . . . . . . . . 8 2.2 Assembly instructions: designing, improving and experimenting . . . 9 2.2.1 Method for designing the T&P instruction . . . . . . . . . . . 9 2.2.2 Method for designing the Video instruction . . . . . . . . . . . 9 2.2.3 Method for designing the AR instruction . . . . . . . . . . . . 10 2.2.4 Improvement workshop: Improving the assembly instruction designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.2.5 Experiments: Evaluating assembly instructions . . . . . . . . 11 3 Theory 13 3.1 Pre-work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.1.1 Design for Assembly . . . . . . . . . . . . . . . . . . . . . . . 13 3.1.2 Assembly sequence planning and MTM-SAM . . . . . . . . . . 13 3.2 Assembly instructions . . . . . . . . . . . . . . . . . . . . . . . . . . 14 3.2.1 In general . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 3.2.2 T&P instructions . . . . . . . . . . . . . . . . . . . . . . . . . 16 3.2.3 Video instructions . . . . . . . . . . . . . . . . . . . . . . . . 16 3.2.4 AR instructions . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4 Results 21 4.1 Pre-work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 ix Contents 4.1.1 The improved product design . . . . . . . . . . . . . . . . . . 21 4.1.2 The selected assembly sequence . . . . . . . . . . . . . . . . . 22 4.2 Assembly instructions . . . . . . . . . . . . . . . . . . . . . . . . . . 22 4.2.1 T&P instruction . . . . . . . . . . . . . . . . . . . . . . . . . 24 4.2.1.1 Guidelines used for the T&P instruction . . . . . . . 24 4.2.2 Video instruction . . . . . . . . . . . . . . . . . . . . . . . . . 25 4.2.2.1 Guidelines used for the Video instruction . . . . . . . 25 4.2.3 AR instruction . . . . . . . . . . . . . . . . . . . . . . . . . . 25 4.2.3.1 Guidelines used for the AR instruction . . . . . . . . 28 4.2.4 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 4.2.4.1 Assembly time and quality measurements . . . . . . 28 4.2.4.2 Survey responses . . . . . . . . . . . . . . . . . . . . 30 5 Discussion 35 5.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 5.2 Pre-work implications . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 5.3 Assembly instructions . . . . . . . . . . . . . . . . . . . . . . . . . . 36 5.4 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 5.4.1 Assembly time . . . . . . . . . . . . . . . . . . . . . . . . . . 38 5.4.2 Quality errors . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 5.4.3 Survey results . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 5.5 Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 6 Conclusion 43 Bibliography 45 A Fixture design process and results I A.1 Theory introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . I A.2 Design methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . II A.3 Design results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II A.3.1 Removing waste material . . . . . . . . . . . . . . . . . . . . . II A.3.1.1 "The Box" . . . . . . . . . . . . . . . . . . . . . . . . III A.3.2 Removing sheet frame . . . . . . . . . . . . . . . . . . . . . . III A.3.2.1 "Frame a’la nose" . . . . . . . . . . . . . . . . . . . . IV A.3.3 Combined solutions . . . . . . . . . . . . . . . . . . . . . . . . V B General Idea Generation for the Assembly Work IX B.1 QR-Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX B.2 Guidance by light and vision control . . . . . . . . . . . . . . . . . . X B.3 Gamification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X B.4 HMI with YuMi robot . . . . . . . . . . . . . . . . . . . . . . . . . . XI B.5 Continuous Improvement . . . . . . . . . . . . . . . . . . . . . . . . . XI C Requirement specification XIII D Product design improvement suggestions XV D.1 Lock mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XV x Contents D.1.1 Lock solution A . . . . . . . . . . . . . . . . . . . . . . . . . . XVI D.1.2 Lock solution B . . . . . . . . . . . . . . . . . . . . . . . . . . XVI D.1.3 Lock solution C . . . . . . . . . . . . . . . . . . . . . . . . . . XVI D.2 Lens- and support- section . . . . . . . . . . . . . . . . . . . . . . . . XVII D.2.1 Support Parts . . . . . . . . . . . . . . . . . . . . . . . . . . . XVII D.2.2 Middle Section . . . . . . . . . . . . . . . . . . . . . . . . . . XVIII D.3 Front Section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIX D.3.1 The Pocket Size . . . . . . . . . . . . . . . . . . . . . . . . . . XIX D.3.2 The hole for the phone’s camera . . . . . . . . . . . . . . . . . XIX D.3.3 Unnecessary Cardboard . . . . . . . . . . . . . . . . . . . . . XIX E MTM-SAM calculations XXI F Improvements suggestions: Workshop XXIII xi Contents xii List of Figures 1.1 The VR-reality goggles in an unfolded cardboard sheet state, together with waste material and frame (left), and in a folded state (right), which is the desired end-state of the manual assembly operations. . . 2 1.2 Flowchart of all manual assembly operations considered. The assem- bly instructions will only be constructed towards the final assembly operation and fixtures will only be designed for the first two opera- tions; removal of waste material and frame. . . . . . . . . . . . . . . . 3 1.3 Start and finish-positions of the final assembly, which is the consid- ered operation for the assembly instructions. To the left, the VR- cardboard goggles without waste material and frame together with lenses and the right image shows the desired end-state. . . . . . . . . 3 1.4 Main components of the preparation process to create assembly in- structions [14]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.5 The initial VR-product design on a 600 x 400 mm cardboard sheet. . 6 2.1 The overall thesis methodology inspired by Delin [14], with added fixture design, workshop and experiments process steps, showing the portions of the two main thesis parts; Pre-work and Assembly in- structions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 3.1 Precedence diagram of the initial product design. . . . . . . . . . . . 15 4.1 Left; Initial product design on the cardboard sheet. Right; Final product design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 4.2 An HTA tree of the final assembly sequence. The product part- names included in the HTA is related to the part-names in Figure 4.1. . . . 23 4.3 The layout of the T&P instruction. . . . . . . . . . . . . . . . . . . . 24 4.4 Explains the interface of the Video instruction. . . . . . . . . . . . . . 26 4.5 The Augmented Reality instructions viewed in Unity3D and corre- sponding trackers. Top Left; Control panel of the animation. Top Right; 3D model of the VR-product with embedded animation. Bot- tom Left; tracker for the control panel. Bottom Right; tracker for the VR-product. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 4.6 The reticle (blue circle) interacts with the control panel. . . . . . . . 27 xiii List of Figures 4.7 Left; An operator using the augmented animation as guidance for the assembly work, from the operator’s viewpoint. Right; A QR- code that links to a YouTube video where an operator assembles the VR-product using the AR instructions. . . . . . . . . . . . . . . . . . 28 4.8 Three box-plots of the assembly time results with incorporated median- time (with a line), mean-time (with an cross), interquartile range and max/min values for all considered instruction types. The dot above the AR plot is an outlier. . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.9 A bar-chart of the achieved product quality. The number of qual- ity errors per assembly is shown in relation to its frequency for all considered instruction types . . . . . . . . . . . . . . . . . . . . . . . 30 4.10 Results to the question regarding how well the participants under- stood the instruction type after the LEGO tutorial (Q1). . . . . . . . 31 4.11 Results to the question regarding how easy the participants thought the instruction was to use during the VR-product assembly (Q2). . . 32 4.12 The results on how much the participants would like to use the in- struction types for assembly tasks in the future (Q3). . . . . . . . . . 32 4.13 The results when the participants were asked how amused they were during the assembly task (Q4). . . . . . . . . . . . . . . . . . . . . . 33 4.14 The results from the participants when they were asked how stressful the assembly task was (Q5). . . . . . . . . . . . . . . . . . . . . . . . 34 A.1 Overview of the waste material (in red color) . . . . . . . . . . . . . . III A.2 "The Box" suggestion. From upper to lower picture; showing the front view, the indentation and the space underneath. . . . . . . . . . . . . IV A.3 The "Frame a’la nose" suggestion. From upper to lower picture; show- ing the whole fixture on a workbench and a focused image of the nose V A.4 A combined solution to the waste material- and frame- removal steps called "Combined 1". . . . . . . . . . . . . . . . . . . . . . . . . . . . VI A.5 A combined solution to the waste material- and frame- removal steps called "Combined 2". From upper to lower picture; showing a focused image on the fixture and a image of two fixtures attached to a sheet- metal framework. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII B.1 Flowchart of the entire manual assembly work. . . . . . . . . . . . . . IX D.1 Initial product design with part-names. . . . . . . . . . . . . . . . . . XV D.2 Lock solution A. Left to right shows how to close the lock . . . . . . . XVI D.3 Lock solution B. The left image shows the open state and the right shows the locked state. . . . . . . . . . . . . . . . . . . . . . . . . . . XVII D.4 Lock solution C. Left image shows the open state, and image shows the locked state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XVII D.5 Shows how the part "Middle Support" is assembled, to the left in- correctly, and middle correctly. The right image shows how the red "Middle Support" assembled incorrectly while the green "Middle Sup- port" is assembled with a smaller misalignment by moving the middle hole closer to the part "Bottom". . . . . . . . . . . . . . . . . . . . . XVIII xiv List of Figures E.1 MTM-SAM calculation of initial product design. . . . . . . . . . . . . XXI E.2 MTM-SAM calculation of improved product design. . . . . . . . . . . XXII xv List of Figures xvi List of Tables 3.1 Guidelines for making T&P instructions [13, 21]. . . . . . . . . . . . . 17 3.2 Guidelines for video instructions [22]. . . . . . . . . . . . . . . . . . . 18 3.3 Guidelines for making AR instructions. . . . . . . . . . . . . . . . . . 19 4.1 Assembly times statistics (minutes:seconds). . . . . . . . . . . . . . . 29 4.2 Comparison of the instruction types in regards to mode and median of the Likert scales for each survey question. The Likert scales are from 1 to 7, where 1 indicates a negative response and 7 a positive response. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 A.1 Stages to design a fixture [7, 8] . . . . . . . . . . . . . . . . . . . . . I A.2 A fixture verification system [9] . . . . . . . . . . . . . . . . . . . . . II C.1 Specification of requirements for the product, assembly work and fix- tures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIII F.1 Generated improvement suggestions from the workshop. . . . . . . . . XXIV xvii List of Tables xviii 1 Introduction In the introduction, a background to the origin of the thesis is presented with related problems and opportunities that the manufacturing industry are currently facing. Then, a more narrowed definition of the focus areas in the thesis is presented fol- lowed by a description of the purpose, which narrows down into three research questions. Lastly, an introduction to the thesis methodology, delimitations and a thesis overview is presented. 1.1 Background This thesis is part of a project named Smarta Fabriker, run by Göteborgs Tekniska Collage (GTC) [1], which in turn is based on Sweden’s new industrialisation strat- egy; Smart Industry [2]. The aim of the new industrialisation strategy is to boost Swedish manufacturing companies’ capacity for change and global competitiveness by focusing on utilizing the potential of industrial digitalization, enhancing sustain- able production and increasing the general knowledge and skills about digitalization, which hopefully will make Sweden world leader in research and development within this field in the upcoming years. Digitalization is paving the way for new ways of working and major opportunities will come as a result. It is therefore of most im- portance that Swedish companies are able to see the benefits of digitalization and also have the right skills to best utilize the future potential. The project, Smarta Fabriker, is aimed to focus on increasing industries and academia’s knowledge about modern industrial production and digitalization. To achieve this, engineering stu- dents from Chalmers University of Technology and high school students from Göte- borgs Tekniska Gymnasium (GTG), together with several companies, are building a modern miniature factory, partly founded by the government, which will display the future manufacturing potential and hopefully attract new people into becoming interested in industrial digitalization. The factory will be exhibited at Universeum in Gothenburg later this year and people who come to visit the factory will be able to order and assemble a pair of cardboard goggles suitable for virtual reality (VR)- ap- plications. The cardboard goggles, shown to the left in Figure 1.1, will be produced in the factory by an fully automatic process part, than later assembled manually together with plastic lenses. The factory will therefore utilize a combination of mod- ern automation and manual operations to show the suitability of digitalization for different manufacturing companies with different levels of automation. 1 1. Introduction Figure 1.1: The VR-reality goggles in an unfolded cardboard sheet state, together with waste material and frame (left), and in a folded state (right), which is the desired end-state of the manual assembly operations. This thesis will treat the manual assembly operations of the miniature factory. This entails creating solutions to how the product should be assembled in an efficient way without generating quality defects or jeopardizing exhibition visitors’ safety, connected to industrial digitalization within the manual assembly field. The exhi- bition visitors will have no previous experience of assembling the product and will therefore need cognitive aids through effectively designed assembly instructions and fixtures. The exhibition visitor will from now on be named operator in this the- sis to ease reading. The operator will be able to select which kind of instruction technology to use when doing the assembly work, this to try out all and experience the differences between each. Assembly instructions in line with digitalization of- ten involve the usage of digital pictures, 2D drawings and 3D models, presented on screens, tablets or through Augmented reality (AR) [3]. The assembly instructions designed in this thesis include text and picture based (T&P) on paper, Video based on a touch screen and AR based on a mobile phone enclosed in a headset. These three instruction technologies will be as thoroughly and effectively designed as pos- sible and later tested to find out differences in generated assembly time, product quality and operator acceptability. How digitalization can benefit the manual assembly work for industries with fewer levels of automation in general is less mentioned and the focus is more towards fully automated processes [3]. Regarding instructions for manual assembly opera- tions, industries today often use mentors to teach or instruct new assembly tasks [4], which is very resource demanding, and other information about the assembly work are often text heavy non-updated paper instructions [5] in a binder that is located far away from the operator. Digitalization offers new ways of working con- sidering instructions, e.g. learning in virtual environments, increasing connectivity that enhances communication abilities, which keeps instructions updated for flexi- ble production, and keeping instructions close to the operator presented on smart tables, computer screens or in viral space [3]. The increasing demand of factory resource efficiency, mass-customization with globalization [6] will make flexible and digital instructions an important asset for any industry with manual assembly work. 2 1. Introduction The manual assembly operations at the miniature factory that are being considered in this thesis work can be summarized into Figure 1.2. The manual work starts when the operator has collected a sheet of cardboard, see left image in Figure 1.1, at the end of the automated process part. The first operation to be executed is to remove waste material and frame from the cardboard sheet so that the goggles can be assembled. The next operation is to get a pair of lenses, and the final operation is to assemble the goggles and lenses together, see Figure 1.3. The designed assembly instructions does only consider the final assembly operation and fixtures are only designed for the first two operations, see Figure 1.2. The focus of this reports’ research questions is on the designed assembly instructions for the final assembly and not on the fixtures, though plenty of work have been put into them. The fixtures have been designed according to Haschemi and Kang [7, 8] and validated with Kang [9]. The fixture design process, relevant theory [10, 11, 12] and results can be found in Appendix A. Also, a better overview of all manual assembly operations at the miniature factory together with interesting design suggestions for the assembly work can be found in Appendix B. Figure 1.2: Flowchart of all manual assembly operations considered. The assem- bly instructions will only be constructed towards the final assembly operation and fixtures will only be designed for the first two operations; removal of waste material and frame. Figure 1.3: Start and finish-positions of the final assembly, which is the considered operation for the assembly instructions. To the left, the VR-cardboard goggles without waste material and frame together with lenses and the right image shows the desired end-state. 3 1. Introduction 1.1.1 Purpose The purpose of the thesis is to explore how digitalization can be applied in a manual assembly context at large, focusing towards the assembly instruction field, and for- mulate recommendations to the Swedish manufacturing industry regarding how to design instructions and which type of technology to employ considering the specific circumstances at Smarta Fabriker. 1.1.2 Research Questions The first research question (RQ1) is related to finding out how to design instructions in the best way considering the specific environment. RQ1: How can instructions be designed for an inexperienced operator in an assem- bly context? The second and third research questions (RQ2 and RQ3) investigates how the three instruction designs differ in regards to important quantitative parameters, which will have an impact on factory efficiency, such as productivity. RQ2: How do assembly instructions differ in performance regarding achieved prod- uct quality and assembly time when being used by inexperienced operators in an assembly context? RQ3: How do assembly instruction technologies differ in perception by the inexpe- rienced operators in an assembly context? Perception in this research will consider the following parameters; understanding, usability, future preference, amusement and stress level. RQ2 and RQ3 will be answered through conducting experiments. The uniqueness of RQ2 and RQ3 from previous research are the inexperienced operators, that only have one try to complete the assembly, and that all assembly instructions used in the experiments have been individually enhanced from available assembly instruction theory, so that all used technologies show their greatest potential. 1.1.3 Delimitations and Scope The first, second and third operation, see Figure 1.2, to remove waste, remove frame and get lenses, will be aided with the help of physical fixtures. It was decided to not develop and compare assembly instruction technologies regarding these opera- tion steps. The fourth and final operation, the final assembly, will be guided with properly designed assembly instructions based on the three different technologies; text and pictures (T&P), Video and augmented reality (AR). The instructions designed in this thesis are intended to be used without any other help of e.g. a human instructor or mentor and since the instructions will later be 4 1. Introduction used in an exhibition environment, any form of incorporated sound is not considered in the instruction design work. Thus, no solution will include the use of headphones, speakers or microphones, recording or playing sound. The work will not be executed primarily to Smarta Fabriker, the solutions will be more general and perhaps more technically complex than what is possible to have in large crowd exhibitions. Smarta Fabriker will later be able to pick concepts they see fit in their exhibition. There will be no final manufacturing of components or fixtures from the concepts that are generated in the report. Though some prototypes may be created to facilitate concept validation. 1.2 Introduction to thesis methodology To effectively design assembly instructions, one must according to Agrawala [6] si- multaneously consider both planning and presentation of instructions. The planning of instructions regards developing and selecting the best assembly sequence that is easy for operators to understand and follow. The presentation of instructions is about conveying the selected assembly sequence in an appropriate way [6, 13]. A case study from Volvo trucks [14] confirms the view that making assembly in- structions requires a lot of preparation work, in e.g. the form of gathering initial requirements, looking at the product design to improve assembly and conducting time analyses (see Figure 1.4). The thesis work have therefore been divided into two main parts inspired by Agrawala [6], pre-work and assembly instructions. One part elaborating instruction planning including relevant pre-work according to Delin [14] and one about the construction and presentation of assembly instructions. Figure 1.4: Main components of the preparation process to create assembly in- structions [14]. 5 1. Introduction 1.2.1 The VR-product: Initial design The VR-product design considered in this thesis work can be found in Figure 1.5 on a 600 x 400 mm cardboard sheet. The product has been stamped on the cardboard sheet. Before the final assembly, some waste material need to be removed together with the outer frame. The product will be assembled with a pair of plastic lenses to achieve the desired VR-effect. The product design in Figure 1.5 will be the starting point in the planning phase for the assembly instruction designs. Figure 1.5: The initial VR-product design on a 600 x 400 mm cardboard sheet. 1.3 Thesis overview The thesis is divided into six chapters, excluding appendix. First, a method chapter will be presented that elaborates the thesis methodology in a detailed way. Then, a theory chapter that present relevant theory about concepts brought up to the method. Thereafter, results, discussion and conclusion chapters follow. Answers to the thesis’ research questions are indirectly incorporated in the broad discussion chapter and more directly formulated in the conclusions chapter. The appendix consists of interesting parts connected to all manual assembly operations that are not included in the results. The thesis is characterized by pre-work and assembly instructions, connected to effective instructions [6], and the division of both pre- work and assembly instructions will be throughout every thesis chapter, excluding the last conclusion chapter. 6 2 Methods This section describes the method used to fulfill the thesis purpose and answer the research questions. The work is split up in the main parts; Pre-work and Assembly instructions. Figure 2.1 shows the method inspired by Delin [14] that have been used during the thesis work. The method shows a linear and sequential process, but the actual work in both Pre-work and Assembly instructions has been iterative, i.e. going back and forth between the tasks, because the process steps influence each other and changes in the later process steps effects the earlier. Since this thesis also has a lot of different stakeholders, which includes supervisors, companies and other thesis workers connected to Smarta fabriker, changes and late adjustments have been inevitable. This methods chapter will though be ordered according to the linear process of Figure 2.1. The fixture design process has been conducted in parallel to the ordinary flow and its method together with its results can be found in Appendix A. Figure 2.1: The overall thesis methodology inspired by Delin [14], with added fixture design, workshop and experiments process steps, showing the portions of the two main thesis parts; Pre-work and Assembly instructions. 2.1 Pre-work: Setting the foundation for design- ing effective instructions The Pre-work part of the thesis is necessary in order to make effective assembly instructions [6]. This part implied making a requirement specification, summarizing all different requirements from various stakeholders regarding the product design and workplace design, with assembly sequence and fixture design taken into consid- eration. The specification acted as a guiding document in the development process 7 2. Methods and in the decision making process. The specification can be found in Appendix C. Then, the product design was analyzed with the purpose of making it more suit- able for assembly, and thereafter finding possible assembly sequences considering the improved design and requirements. The sequences were later evaluated to find the most suitable sequence regarding the preconditions at the workstation. The Pre-work will lay the foundation for the later thesis part; Assembly instructions. Interviews have been conducted (semi-structured and unstructured) to guide and validate the pre-work. The interviewees in this thesis are anonymous and the inter- views were recorded only if permitted by the interviewee. The recordings have only been used within the project group during the thesis work. 2.1.1 Improving the product design The starting point of improving the product design was considering the demands and requests of the requirement specification. The specification focused on reducing hard cognitive tasks, improving the product quality, by reducing the number of assembly errors, and reducing the general time of assembly. The chosen method based on the requirement specification was Design-for-Assembly (DFA) [15, 16], in line with Delin [14]. After looking at the general guidelines in the DFA literature, an idea generating session was conducted with the purpose of finding possible product improvements that would simplify the assembly work. After generating a lot of improvement ideas, a semi-structured interview was conducted together with a senior cardboard designer at Stora Enso Packaging AB in Skene (Sweden), evaluating and validating each of the improvements found in the idea generating session. The improvement suggestions can be found in Appendix D. 2.1.2 Finding the most appropriate assembly sequence In order to find the most appropriate assembly sequence, according to the require- ment specification, all possible assembly sequences for the final assembly in Figure 1.2 needed to be studied. A precedence diagram of the improved product design was thereby constructed. Several possible sequences were eliminated because of their lack of use, which would, if used, result in unnecessary work for the oper- ator. For example, some tasks could be executed before others but it would not add any additional value or reduce the operator’s cognitive load. The generated sequences that implied extra work got screened out from the final evaluation step. The final evaluation step was executed using MTM-SAM [17, 18] looking in more detail at which sequence resulted in fewest body movements and thereby would give the fastest assembly time. The selection of using MTM-SAM was based on an un- structured interview with a production analysis researcher at Chalmers University of Technology and is also supported by Zha [19]. When the assembly sequence was determined, a Hierarchical task analysis (HTA) tree [20] was constructed to visualize and explain the inherent steps and used as a basis for the instruction layouts [6, 13]. 8 2. Methods 2.2 Assembly instructions: designing, improving and experimenting All designed assembly instructions (T&P, Video and AR) were founded on the result of the pre-work section, i.e. on the improved product design and selected assembly sequence, together with assembly instruction guidelines gathered from relevant the- ory [13, 21, 22, 23]. Therefore, a review of the relevant literature was conducted. All three assembly instructions were designed with the intention of reaching the highest potential of the instruction type. The generated assembly instructions were there- after validated through an improvement workshop before tested in experiments, which were aimed to quantitatively find out differences between the instructions regarding time of assembly, achieved product quality and operator perception. 2.2.1 Method for designing the T&P instruction To make the T&P based instructions, a camera1 was used to take realistic assembly pictures, showing each step of completing the assembly. The product was positioned on a brown table with a brown background as well. The idea was to only show details in the pictures that are relevant to the operator. The pictures were then put together in the software Microsoft Word2 to generate the instruction layout. Adobe Photoshop CS53 was used to crop and remove backgrounds in some of the pictures. It was decided to fit all of the necessary instructions in one page only, so the operator can get a better overview of all the assembly steps. An alternative would have been to use more pages to describe the sequences in more detail, but it might have also made the instructions as a whole more complex to comprehend. The page with the instructions was designed to be around A3-size. The pictures of the assembly steps have sufficient size to fit the A3 paper format. Several alternatives of instructions have been tested, to have the paper horizontal or vertical, using numbered pictures instead of sequences, having the instructions structure oriented in a number of ways etc. 2.2.2 Method for designing the Video instruction When the Video instruction was made, the same camera was used as in the T&P instructions to capture the studied assembly sequence in video. The video of the whole assembly sequence was thereafter divided into six parts, between 4 and 12 seconds, in Videopad Video Editor4 and converted into .gif-pictures. These pictures were then imported into Microsoft Powerpoint5, one gif-picture per slide, to increase usability that ensures better operator control and prevents having to review the entire video sequence if one task is unseen [24, 25]. Snapshots of start and finish 1https://www.dpreview.com/reviews/canoneos500d/ 2https://office.live.com/start/Word.aspx 3https://helpx.adobe.com/creative-suite/kb/cs5-product-downloads.html 4http://www.nchsoftware.com/videopad/ 5https://office.live.com/start/PowerPoint.aspx 9 https://www.dpreview.com/reviews/canoneos500d/ https://office.live.com/start/Word.aspx https://helpx.adobe.com/creative-suite/kb/cs5-product-downloads.html http://www.nchsoftware.com/videopad/ https://office.live.com/start/PowerPoint.aspx 2. Methods positions were thereby added as a complement to the gif-pictures on each slide, with the purpose of further minimizing non value added waiting time. 2.2.3 Method for designing the AR instruction At first, an idea generating session was held to find plausible effective solutions that could instruct how to assembly the product. It was early established that a good solution would be to have an virtual 3D-model of the cardboard that was animated accordingly to the assembly sequence. The idea was that the operator should be able to control the animation to some extent, so that the instruction was presented according to the preferences of individual operators. This was the vision of the instruction. The next step was to find a solution on how this could be realized. Unity3D6 was initially a game engine but is today also a common engine to develop Augmented Reality, Virtual Reality or Mixed Reality with software development kits (SDK). SDK’s can be used for many purposes, for example to build projects to different platforms, like Android, Playstation 4 or Samsung SMART TV. Unity3D was used together with Android SDK to be able to test the project on an Android phone with Virtual Reality glasses. This enabled a simple way to test solutions, without the need for additional hardware, like Microsoft’s Hololens7. When Stora Enso were finished making the final product design, based on our im- provement suggestions, it resulted in a 2D PDF drawing. This was imported into Google Sketchup8 to make a 3D-CAD model of the product. The 3D model was therafter imported into Blender9, an open-source 3D creation software, to animate the assembly sequence and lastly imported into Unity3D. To make Unity3D function as intended, additional SDK was necessary to build augmented reality projects. Because of the research team’s lack of experience with Augmented reality or Unity3D, all of the following SDK’s might have been up for the task, thus had to be tested. It should be noted that the team did not have any experience with any of the software mentioned in this section, except to a small extent Google Sketchup. First, Vuforia SDK10 was tested, then Google’s Cardboard SDK11 and they were tested combined. Finally, Vuforia SDK and Vuforia’s AR/VR sample12 was used to successfully build a solution that was aligned with the vision. Trackers to the instructions was designed in Adobe Photoshop CS5. 6https://unity3d.com/ 7https://www.microsoft.com/en-us/hololens 8https://www.sketchup.com/ 9https://www.blender.org/ 10https://developer.vuforia.com/downloads/sdk 11https://developers.google.com/vr/unity/ 12https://developer.vuforia.com/downloads/samples 10 https://unity3d.com/ https://www.microsoft.com/en-us/hololens https://www.sketchup.com/ https://www.blender.org/ https://developer.vuforia.com/downloads/sdk https://developers.google.com/vr/unity/ https://developer.vuforia.com/downloads/samples 2. Methods 2.2.4 Improvement workshop: Improving the assembly in- struction designs After the draft versions of the assembly instructions were completed, a workshop was made with the purpose of finding instruction improvements and set the final instruction design before conducting later experiments. The workshop participants consisted of six researchers from Chalmers University of Technology within different fields related to production (e.g. cognitive/physical ergonomics, human-machine- interaction and productivity), to reach a wide range of interrelated perspectives. The participants were divided into three separate groups and each group was as- signed a specific instruction. The task was then to assemble the product with the help of the specific instruction and thereafter give feedback (positive and/or nega- tive) on perceived instruction effectiveness and suggest future instruction improve- ments. Then, each participant got the chance to try all the other instructions and compare those with the first one in a joint discussion, to see if they agree with the first feedback round and/or comes up with other improvement suggestions. The first round of feedback is very valuable, because the participants have then only assembled the product one time, which will generate as realistic assembly conditions compared to the exhibition setting as possible. When the participants try the other instruction-types in the second round, they will have cognitively remembered motion patterns and work sequences on how to perform the assembly, which will reduce operators need of using the instructions thoroughly and will increase the risk of not receiving as thorough feedback as possible. 2.2.5 Experiments: Evaluating assembly instructions The experiments were conducted to answer research questions two and three by testing how each designed assembly instruction perform regarding assembly time, achieved product quality (RQ2) and perceived usability by operators (RQ3). The participants in the experiments consisted of students and teachers from GTG in Gothenburg, ranging from 15-55 in age, who had no previous experience of assem- bling the VR-cardboard product and they were in total 30 people. The participants were equally divided, so that 10 participants tested each assembly instruction. The T&P instructions was printed on a A3 paper and the Video instruction was showed on a laptop screen. The AR instructions was run on an Android OS with an Sony Xperia Z513 and used with an HMD called Homido Virtual Reality Headset V214. The experiment procedure was designed according to the following agenda; three participants at a time were placed in a prepared room (LAB-environment) and assigned an instruction each at random. All participants are anonymous in the the- sis and they were thoroughly informed of their anonymity before the experiments started. The participants were guided to separate stations containing the instruc- tion, which were shielded from each other to prevent seeing other participants. They 13https://www.sonymobile.com/global-en/products/phones/xperia-z5/ 14http://www.homido.com/en/shop/products/homido-hmd-v2 11 https://www.sonymobile.com/global-en/products/phones/xperia-z5/ http://www.homido.com/en/shop/products/homido-hmd-v2 2. Methods were also not allowed to talk or leave the station during the experiments. After be- ing placed at a station they got to try out the specific instruction type by doing a tutorial consisting of a small LEGO assembly before doing the VR-cardboard assem- bly, where the functionality of the instruction types were explained continuously by an instructor. The purpose of the tutorial was to level out the participants’ differ- ent previous experiences with the instruction type. Some participants may be very familiar with e.g. augmented reality related technology and other might not, and the research group wanted to reduce the effect of those individual preferences [23]. Thereafter, they were given the VR-cardboard product as the next assembling task and measurements of time to complete the assembly and related quality errors were taken and summarized in Microsoft Excel15. Lastly, they were given a questionnaire made in Google forms16 with related questions regarding how they perceived the assembly task in general and the specific instruction. This experiment procedure was repeated 10 times until all 30 participants had assembled the VR-cardboard product once. During the assembly of the VR-cardboard product, the experiment leaders used stopwatches to measure the time of assembly. All participants started at the same time and when they were done assembling they raised a hand to signal the exper- iment leaders to stop measuring the time. To asses the achieved product quality, each finished assembled product was examined between every experiment procedure, looking for assembly errors. An assembly error could be e.g. misplacement of lenses, folding the cardboard the wrong way and thereby damaging the product function- ality or poorly folding cardboard parts resulting in unused interstice function. All types of assembly errors were equally weighted in the later analysis. The questionnaire consisted of six questions, one yes/no question regarding their previous experience of assembling products in an industry context and five likert scale [26] questions with a seven-point-scale about perceived amusement, stress, if the instruction type was simple to understand and use and if the participants would like to assemble products with the related instruction type in the future. The anal- ysis of the questionnaire answers consisted of calculating mode (the most frequent answer), median and studying the variation of answers in histogram diagrams, which are common practices when doing analyzes of likert scale data [27]. 15https://products.office.com/sv-se/excel 16https://www.google.se/intl/sv/forms/about/ 12 https://products.office.com/sv-se/excel https://www.google.se/intl/sv/forms/about/ 3 Theory This chapter presents the theory that has been used during the thesis work, which has been brought up in the thesis methodology. The chapter is divided according to theory related to pre-work and theory related to assembly instructions, excluding the workshop and experiments, see Figure 2.1. The Theory chapter is built to first give a small introduction in each section regarding the specific subject and gradually go into more details related to the thesis topic. 3.1 Pre-work This section describes the necessary theory related to pre-work. This includes Design for assembly (DFA), related to improving the product design, and assembly sequence planning together with MTM, related to selecting the most appropriate assembly sequence for the final assembly operation. 3.1.1 Design for Assembly Design for assembly (DFA) is a method used when designing products with the purpose of facilitating the products assembly work [28], aiming to reduce assembly time and quality errors resulting in lowering the total manufacturing costs. The DFA analysis is often made manually but can be incorporated with computer al- gorithms [29]. There are general design guidelines to be followed when conducting a DFA analysis considering manual assembly [15, 16]. The most relevant of these guidelines are listed below with relation to the VR-product in parentheses: • Reduce the number of parts (reduce the amount of cardboard pieces). • Get parts to fit more easily (easy folding cardboard construction reduces as- sembly time). • Design parts with self-location features (prevent operators making quality er- rors). • Minimize reorientation of parts during assembly (reduce unnecessary body movements). 3.1.2 Assembly sequence planning and MTM-SAM Assembly sequence planning is a method to find the optimal assembly sequence [30] regarding e.g. cost, assembly time or other parameters. It can be conducted manu- ally or with help of computer algorithms and programs, depending on product part 13 3. Theory amount and part complexity [29, 31]. In general assembly sequence planning, the first step is to determine which parts that are included in the assembly [29, 30]. Thereafter, an analysis of the connections between parts is made, which will be the input to graphs and diagrams such as AND/OR graphs and precedence diagram. The precedence illustrate the constraints between part connections and will show all possible assembly sequences, see Figure 3.1 for a precedence of the old VR-product design. Though, it needs to be analyzed further with the help of the initial param- eters (e.g. reducing assembly time) in order to find the most appropriate sequence for the assembly work [31]. Looking at the parameter; reducing assembly time, a calculated MTM- time (from e.g. MTM-SAM) could be used as a quantitative fac- tor when determining the most optimal sequence out of all possible [19]. MTM is an abbreviation of "Methods-Time Measurements" which constitute of sev- eral methods (MTM-1, MTM-2, MTM-SAM) [17]. All methods are used to ob- jectively calculate the time a specific operation (e.g. assemble the goggles) in an manual assembly should take, called norm time [18]. The norm time is calculated in the unit time factor, where 5.6 factors equal to 1 second. Each manual work operation is divided into the basic movements required to perform the work oper- ation. Each of these basic movements is assigned a certain time value, which is determined by the way the human body moves and the conditions under which the movements are performed [17]. The difference between MTM-1 and MTM-SAM is that an MTM-SAM analysis is not as detailed as an MTM-1 analysis. Thus, the MTM-SAM analysis takes shorter time to perform than MTM-1, with enough precision needed to correctly analyze assembly methods in production settings [32]. 3.2 Assembly instructions This section describes relevant theory for designing assembly instructions, from a presentation of instructions point of view [6]. The chapter is divided into sections re- lated to assembly instructions theory in general and to specific assembly instruction theory (T&P, Video, AR). At the end of each section, guidelines will be summarized, which were considered in the assembly instructions design process. 3.2.1 In general In this research an assembly instruction is defined as a set of standalone procedural instructions that structurally shows how parts in certain predefined positions are as- sembled. Standalone in this case means that the instruction technology by its own, and not with the help of i.e. an instructor, is needed. When constructing assembly instructions it is important to arrange the information in a way that suits human cognitive processes [33]. The cognitive processes include perception, through vision and hearing, memory and attention [34]. When using the cognitive processes, the operator experience differences in cognitive load (intrinsic, germane and extraneous load) that effects assembly performance [21], depending on how the work instruction is created [35]. Intrinsic and germane cognitive load is affected by the complexity of the assembly task and depending if the task is new or have been learned before 14 3. Theory Figure 3.1: Precedence diagram of the initial product design. from previous experiences, whereas extraneous load is unnecessary cognitive load that have no positive effect on the assembly performance [13]. Some general guidelines to help assembly instruction designers to effectively support operators are the following [21]: • Support active cognitive processes (i.e. not too much information, focus on the most important and consider operator experience) • Support operators mental models (how a person perceives a situation affects his/hers behaviour) • Support cognitive abilities and limitations (memory and attention is limited, 15 3. Theory thus take away redundant information) • Support individual preferences (humans might want different information) • Support perception through correct placement of information and the usage of pictures. Instructions can be presented in many different ways. It could be in a descriptive form, e.g. text-based, or depictive form [21], e.g. pictures and video, animation. It can also be a combination of both descriptive and depictive, e.g picture- and text- based (T&P). The type of technology used (information carrier) also differ between instructions and will highly impact operators’ active cognitive processes [21]. 3.2.2 T&P instructions Text instructions, often combined with pictures, are the most common type of as- sembly instruction in a manufacturing setting [21]. One big issue with only having text is the delimitation regarding communication. Production operators often speak different languages and have limited knowledge in other languages, thus pictures can be used instead as a compliment [36]. Combining both text and pictures makes the assembly information more easy to understand [34]. It has been shown in several studies that mixed information formats increases performance [37]. Several other case studies show that depictive work instructions (e.g. picture and video) outper- form descriptive instructions (e.g. text only) when it comes to achieved product quality and assembly time [21]. When designing T&P instructions, one could consider guidelines from Söderberg [13] and Mattson [21], see Table 3.1. The guidelines are divided into the subcategories; structure, layout, and T&P, whereas structure refers to how to plan the instruction and layout and T&P are about presentation [6]. 3.2.3 Video instructions Video instructions are procedural instructions that are displayed on a screen. The video assembly instruction is often played from the beginning to the end, without pauses, but the operator have the chance to reverse and play back sections of the video. Previous research has been focusing on comparing video instructions and print instructions [24], and/or augmented reality (AR) instructions with remote guidance [25]. The benefits of using video instructions over print instructions are that animations in videos are useful for tasks that involve complex assembly actions or procedures, specifically those that are difficult for users to imagine [24]. Although print instruc- tion often combines pictures and text, video instructions utilize sound, animation, images, and text to make meaning. Research have shown that operators perform better when animation is combined with sound or words [38]. Although video in- structions have shown to improve assembly efficiency, problems have arisen regarding its usability and acceptance among operators. There seem to be a thin line between 16 3. Theory Table 3.1: Guidelines for making T&P instructions [13, 21]. Guideline Description Structure The structure should be based on a planned procedure of assembly, for example by the use of HTA. Support the instructions by adding separate presentations with pictures of the finished product. Depending on the space available in the instruction layout, the separate presentation can be placed either in the same information presenter or on a separate presenter. A separate presentation can also be added with pictures of high complex parts. Layout The layout should make it easy to find information and be consistent throughout the instructions. The instruction steps should include headings that are clear and concise, intuitive and informative (support the under- standing of the task). Text and pictures The instructions should have a high focus on pictures and text should only be used when pictures are not sufficient. All pictures should be realistic, photographs are to prefer when possible. In order to be clear the pictures should be big, have high contrast and reduced shadows. Text and pictures should only include relevant information. Eliminate unnecessary details in pictures. Highlight the most important information and use e.g. dif- ferent colors or arrows to direct attention. the video playing too fast, which means that it has to be reviewed, or the video play- ing too slow so that the operator has to wait for the next sequence to show,[24, 25] which generates frustration among workers. There is a need of making the operator feel more in control when using the video instruction by making it easier for opera- tors to find specific assembly sequences or tasks [24]. Plaisant and Shneiderman [22] have also gathered some guidelines to consider when creating video instructions, see Table 3.2. 3.2.4 AR instructions Augmented Reality (AR) is a technology that can be used in assembly work to guide and instruct the operator with the help of combining virtual graphics (animation, arrows, text) and physical objects. The virtual graphics can be overlaid on the phys- ical objects to enhance operator’s perception of reality [23]. This is sometimes called "Mixed reality" instead and there are currently debates as to which term should be 17 3. Theory Table 3.2: Guidelines for video instructions [22]. Guidelines Description In general Coordinate demonstrations with text documentation Synchronize spoken narration and animation carefully Be faithful to the actual user interface Use highlighting to guide attention Ensure user control Keep file sizes small used [39]. When using AR, a head-mounted display (HMD) with a camera is often used to capture the physical world and depict the combined reality, e.g. Oculus rift1 or Microsoft Hololens2, with one difference that Hololens do not completely block your line of sight, which allows for a different AR experience. Using AR with HMD, the assembly instructions becomes part of the actual assembly work and will also aid the operators by hands-free interaction [40]. Another way to use the AR technology is to employ a hand-held PC, which have been proven to be more ben- eficial for learning situations than the usage of HMD:s [41]. Though, the usage of hand-held PC in an assembly context is not yet fully studied since it requires at least one operator’s hand to hold and steer the PC. A lot of previous studies have been made comparing the effectiveness of different assembly instructions media. AR-based instructions and paper-based instructions (e.g. T&P) are often compared looking at objective quantitative differences, e.g. in assembly completion time and error rate, and subjective qualitative/quantitative differences, e.g. operator acceptance or usability [42]. The studies often conclude that the use of AR technology, when guiding operators in assembly work, outper- forms paper-based instructions, achieving lower assembly time and fewer assembly errors [43, 44, 45, 46]. Operator acceptance and usability of AR technology regard- ing assembly instructions have though historically not been better than paper- based instructions [23]. Aspects summarized by Syberfeldt [23] that have a proven nega- tive affect on operator’s perceived usability regarding AR technology with HMD for assembly work consists of: • Improper operator training of the AR functionality before usage. • Time lag experienced. • To low complexity of the considered product in assembly. In Table 3.3, there is a summery of aspects to consider when designing AR-based instructions regarding improving assembly work efficiency (e.g. assembly time). 1https://www.oculus.com/ 2https://www.microsoft.com/en-us/hololens 18 https://www.oculus.com/ https://www.microsoft.com/en-us/hololens 3. Theory Table 3.3: Guidelines for making AR instructions. Guideline Description Efficiency Graphical information and visual features does not need to be lo- cated in the task area to be useful. But aim for no visible misalign- ment between the graphics (animation) and the physical object to achieve best performance [40] The type of visual features used ( e.g. arrows, 2D sketches, text, 3D animations) should be adopted to the assembly operations relative difficulty level, since simpler visual features are easier to understand and thus faster recognized [47] 19 3. Theory 20 4 Results This chapter will present the results generated in this thesis work, divided into the sections; Pre-work and Assembly instructions. 4.1 Pre-work This section presents the results of the pre-work. The pre-work result consists of a presentation of the improved VR-product design and the selected final assembly sequence that was used as a foundation for all designed instructions. 4.1.1 The improved product design A comparison of the new and old design of the cardboard sheet with the product design on it can be found in Figure 4.1. Appendix D discusses several feasible product improvements, including some of which have been implemented. Figure 4.1: Left; Initial product design on the cardboard sheet. Right; Final product design. The product has been through several different improvements. All demands on the product from the requirement specification (Appendix C) are being met. The product require less cardboard in the construction to fulfill its functions, without any noticeable decrease in the construction’s overall strength. The cardboard sheet (around the VR-product) has the same measurements, except that one corner has been cut off, as is shown to the right in Figure 4.1. The corner is a poke-yoke1 1http://leanmanufacturingtools.org/494/poka-yoke/ 21 http://leanmanufacturingtools.org/494/poka-yoke/ 4. Results solution that makes it impossible to place the cardboard sheet flat in the fixture incorrectly before the waste material are to be removed. The cardboard sheet mea- surements have not been reduced since the sheets will be delivered on EU pallets which is 1200 mm x 800 mm and will precisely fit four sheets per layer. The product design could not be reduced in size enough to make one pallet able to contain more sheets with products per layer. Since the VR-product require less cardboard and the cardboard sheet have roughly the same measurements it will initially result in more cardboard waste, though the reduced VR-product measurements will result in possibilities to lower the cardboard sheet dimensions. One big change is that the whole middle-section has been optimized and moved ("M1" and "M2" in the new design). Initially there were three similar parts, but during development it was concluded that one of the parts was not necessary to fulfill the VR-product functions. The whole middle-section was moved to achieve a more natural assembly sequence. Now the product does not need to be reoriented during assembly. Before, parts of the product needed to be folded several times in order to be assembled. Now, the "Bottom" part act as a base during assembly, the other parts are folded against it. The component "M1" have also gotten a cutout according to a new lens design for the glasses. The "lock", see Figure 4.1, has been changed as well into a simple piece of cardboard without any hooks on it. It is held in place with the help of friction and is easily assembled and disassembled. The product part "FL", see Figure 4.1, includes a small cutout, this is so the telephone can be connected to headphones when the goggles are being used. In order to be compatible with a large quantity of telephones, the product part "Front" does contain an additional hole for the camera. This way the telephone can be oriented so the headphone jack fits the cutout while the telephone’s camera is in any of the camera holes. 4.1.2 The selected assembly sequence An HTA of the selected sequence for the final assembly can be seen in Figure 4.2 and it is the sequence that resulted in the lowest assembly time when doing MTM- SAM analyses. The MTM-SAM time for the assembly sequence is 77 factors, which is 13,75 seconds. Comparing that MTM-SAM assembly time to the MTM-SAM assembly time of the old product design, which also had another sequence, it is a reduction of 20,6%, see Appendix E. The reduction in assembly time is, as said above, achieved due to both improved product design and new sequence, i.e. one aspect can not achieve the reached reduction without changing the other. The achieved reduction in assembly time of 20,6% is a significant improvement, which is solely based on the result from the pre-work. 4.2 Assembly instructions This section will bring up the results of the three instruction designs and exper- iments. All instructions are based on the assembly sequence in Figure 4.2. The 22 4. Results Figure 4.2: An HTA tree of the final assembly sequence. The product part- names included in the HTA is related to the part-names in Figure 4.1. 23 4. Results instructions have been designed with guidelines, presented in the theory chapter, taken into consideration and a sub-section is therefore dedicated for each instruc- tion type, explaining which of the guidelines that were used. The instructions have also been improved based on what was brought up during the workshop. A summary of the improvement suggestions from the workshop can be found in Appendix F. 4.2.1 T&P instruction The layout of the T&P instruction can be seen in Figure 4.3. Figure 4.3: The layout of the T&P instruction. The instruction is ordered according to eight sub-sequences which explains all as- sembly steps. Each sub-sequence begins and ends with an initial and final position, indicating how the product should look before and after the sub-sequence. The end positions are clearly highlighted so the operator could continuously check if the sub- sequence was carried out correctly. Between the initial and final position there are several pictures with the purpose of explaining how to carry out the sub-sequence. The images are complemented with arrows and additional highlights in the form of green circles to make details even more clear to understand. When appropriate, text has been used to indicate sound that the parts make when assembled together. 4.2.1.1 Guidelines used for the T&P instruction Considering the guidelines from the theory chapter, the end-picture of sub-sequence eight have been highlighted with a black border, since this shows the completely assembled product. The blue border at the top contain the start position of the 24 4. Results goggles as well as a fully assembled pair. The product itself does not contain any complex parts, it was considered one instruction for everything was enough. To minimize the extraneous cognitive load and since all steps is simply a matter of folding the product in the correct way, sub-sequences have no headings. The instruction is based almost only on pictures. Text are used to describe how the instruction work (for example, "Step" and "End of Step"). The pictures was taken with only brown background to reduce unnecessary details in the pictures and keep a high contrast between the product and the background. The lightning was adjusted during the photo shoot to reduce shadows in each of the pictures. The pictures have been cropped and re-sized to be large and easy to understand. Some pictures only shows parts of the product, this is to reduce unnecessary details, save layout space and to have the shown parts larger instead. Arrows, colored circles and text which indicates sound have been used to highlight important details. 4.2.2 Video instruction The interface of the Video instruction is depicted in Figure 4.4. The general concept of the instruction is that a video of the whole final assembly sequence is divided into six sub-sequences, see the grey boxes in Figure 4.2, and converted into gif-pictures, which are placed in the lower left corner, see Figure 4.4. The function of the gif- picture is that it enables the sub-sequence to be looped automatically when ended. In the instructions upper layout part, two snapshots of start and finish positions are located. These pictures are a complement to the gif-pictures, showing how it should look when a sub-sequence is completed. To switch between sub-sequences, buttons to reach previous and next sequences are placed in the lower right corner, green button for next and red for previous. The instruction is created so that it can be presented on a touch screen with interactive touch buttons. 4.2.2.1 Guidelines used for the Video instruction In the instruction design phase, a lot of focus has been towards achieving high operator control since it has been largely documented to be one of the biggest issues towards operator acceptability. The division of the sequence into parts that are looped through gif-pictures should ensure that the operator experience less stress about missing a step or having to unnecessary wait for the next sequence. The instruction interface is built with the operator’s cognitive abilities in mind, e.g. different colored touch buttons, placed in the bottom right corner and pictures of start and finish position. As for T&P instructions, the aim of the Video instruction has also been to use realistic captured video-shots, use minimum amount of text in the interface and reduce unnecessary information (extraneous load). 4.2.3 AR instruction The final result of the AR instruction can be seen in Figure 4.5 in Unity3D’s environ- ment. The result is an app (.apk file) which can be installed on an Android device. The app works by putting the device in a HMD; in this case a pair of VR-glasses. 25 4. Results Figure 4.4: Explains the interface of the Video instruction. The object to the top right in Figure 4.5 is a 3D model of the VR-product (blue) and lenses (red). There is a pre-programmed animation attached to the objects accord- ing to the assembly sequence. The animation is controlled by the control panel, seen to the top left in Figure 4.5. The control panel can play and rewind the animation with the green and red buttons, respectively. The 3D model of the VR-product has approximately the same size as the physical one, to mimic the physical assembly as much as possible. The instructions also need trackers so that the control panel and instructions can be displayed in the physical world. The tracker for the control panel can be seen in the bottom left and for the VR-product in the bottom right in Figure 4.5. The trackers is in the form of pictures (.PNG) that have been printed on A4 papers. The control panel is operated with a reticle, which can be seen in in Figure 4.6. The reticle is the small blue circle that are to the left in each of the three images in Figure 4.6. The reticle is stuck in the operators view at all times, i.e it will be in the same position in relation to the HMD- screen when the operator looks around. The animation is in a paused mode when the reticle is not hovering over any of the buttons, as in the picture to the left in Figure 4.6. The animation will play when the reticle hovers over the green play button, as shown in the middle picture in Figure 4.6. The picture to the right in the same Figure shows how to rewind the animation, simply by hovering the reticle over the red rewind button instead. When any of the buttons are activated the button’s image will turn blue and the arrow becomes a pause symbol, to visually indicate that the button has been pressed. Figure 4.6 also shows certain part of the product as green. These parts are the active parts in the current animation sequence; when any of the buttons are pressed the green components will move according to the assembly sequence. This is to visually guide the operator to focus on the current parts that are next to be assembled. 26 4. Results Figure 4.5: The Augmented Reality instructions viewed in Unity3D and corre- sponding trackers. Top Left; Control panel of the animation. Top Right; 3D model of the VR-product with embedded animation. Bottom Left; tracker for the control panel. Bottom Right; tracker for the VR-product. Figure 4.6: The reticle (blue circle) interacts with the control panel. To the left in Figure 4.7 there’s an operator using the animation to assemble a physical product. The idea is to have the control panel and the virtual product in front of the physical product, as the Figure shows. The operator starts by playing a comfortable length of the animation and assemble accordingly. If necessary the animation can be re-winded and played again until the physical product has been assembled accordingly. This is simply repeated with the rest of the animation un- til the product has been successfully assembled. The trackers displays the virtual objects at a fixed position, meaning that the trackers can be moved or rotated and still display the virtual objects. This enables the possibility to get closer to the VR- product, look at details or critical movements from different angles or simply turn the VR-product around; it is very alike a physical VR-product in terms of position and movement. There are also an YouTube video, which the QR-code to the right in Figure 4.7 links to, that shows an operator using the AR instructions’ functions to assemble the VR-product. 27 4. Results https://youtu.be/WTdKCr1l-I0 Figure 4.7: Left; An operator using the augmented animation as guidance for the assembly work, from the operator’s viewpoint. Right; A QR-code that links to a YouTube video where an operator assembles the VR-product using the AR instructions. 4.2.3.1 Guidelines used for the AR instruction The virtual product is a replica of the physical product. This means that the op- erator can continuously check and identify that all parts are assembled correctly. The functionality of the AR instruction is quite simple; just play, pause and rewind the animation. Since it is combined with simple visual features, for example com- ponents becoming green, it should be easy for operators to understand and use the instructions effectively. Because of the ability to play as long as preferable of the animation, the operator can adjust the instructions according to preferences. 4.2.4 Experiments This section presents the results from the experiments. The section is divided into two parts; Time and quality measurements, which were measured during the exper- iments, and survey responses, from the survey that the participants filled in after the experiments. 4.2.4.1 Assembly time and quality measurements Figure 4.8 shows box-plots of the measured assembly time. There are three plots, one for each instruction type. The plots shows the interquartile range, mean and median values as well as the maximum and minimum assembly times. The AR instruction- plot has one sample that is an outlier, i.e. it is located above 1.5 multiplied by the interquartile range of the third quartile. This sample can be seen as a black dot above the AR box-plot. Detailed statistics of all the box-plots can be found in Table 4.1. 28 https://youtu.be/WTdKCr1l-I0 4. Results Figure 4.8: Three box-plots of the assembly time results with incorporated median- time (with a line), mean-time (with an cross), interquartile range and max/min values for all considered instruction types. The dot above the AR plot is an outlier. Table 4.1: Assembly times statistics (minutes:seconds). T&P Video AR Lowest 00:58 01:14 02:21 Quartile 1 01:52 01:42 02:52 Median 02:34 01:59 03:17 Mean 02:47 01:59 04:08 Quartile 3 03:46 02:12 04:59 Highest 05:52 02:37 09:03 It should be noted that the T&P instructions have the overall lowest assembly time at 00:58. The highest T&P time was 05:52 and the median time was 02:34. The AR box-plot stretches from 02:21 to 05:23 or to 09:03 with the outlier included, have a median time of 03:17 and a mean of 04:08. The Video box-plot has the shortest range between min and max- time. The mean and median is also the lowest, both at 01:59. Since the mean and median times in the T&P and AR differ, their individual distributions are skewed. The Video instruction on the other hand have the median equal to the mean, and therefore it does not have any skewness. Figure 4.9 shows a bar-chart of the measured quality errors. The x-axis displays the number of errors per assembly for each instruction type. The y-axis shows how many products that yielded the specific amount of errors. No observation had a perfect score of zero errors, thus each instruction type is represented ten times in the chart. When the assembly errors exceeded 3 errors, the product was seen as unusable since the lenses position or the cardboard was so out of place that the VR-product’s functions was significantly affected. 29 4. Results The Video instruction has the smallest range, between 1 and 3 errors per assembly. 60% of all Video assemblies have only one error and no VR-product is unusable. T&P generated 1, 2, 3 or 5 errors per assembly, 20% of these products is regarded as unusable. The AR instruction resulted in everything between 1 and 5 errors, where 40% is considered unusable. The T&P and AR instructions have, to some extent, a similar error distribution since both stretches over 1 to 5 errors. However, the T&P median error is 2 while the AR median error is 3. Figure 4.9: A bar-chart of the achieved product quality. The number of quality errors per assembly is shown in relation to its frequency for all considered instruction types 4.2.4.2 Survey responses This section presents the results from the survey the participants filled in after the experiments. The first question was if they had any experience with assembly in work or school assignments. This was a yes or no question and 23 people (77%) responded yes. The rest of the questions were to be responded with a Likert scale ranging from 1 to 7. The questions was constructed so a low number in the Likert scale indicated a negative opinion and a high number indicated a positive opinion to the stated question. The questions are presented in Figure 4.10 - 4.14 as bar- charts with the Likert scale and type of instruction on the x-axis and the number of responses on the y-axis. All questions have also been compared in Table 4.2 in regards to their median and mode responses for each instruction type. The table will be explained continuously when the results from each question is presented. In order to rule out technical misconceptions of the instruction types the participants was asked how well they understood the instruction after they had completed the LEGO tutorial. A bar-chart of the responses can be seen in Figure 4.10. Generally, the participants seem to have understood the technologies well, since most responses is a five or higher. The exception is the AR instructions which 20% scored as a 3. 30 4. Results If the instruction types were to be ranked according to the Likert scale’s median or mode, the ranking would be; Rank Median or Mode (Q1): 1. Video 2. T&P 3. AR Table 4.2: Comparison of the instruction types in regards to mode and median of the Likert scales for each survey question. The Likert scales are from 1 to 7, where 1 indicates a negative response and 7 a positive response. Question Median Mode AR T&P Video AR T&P Video How well did you understand the instruction after the LEGO- assembly? (Q1) 6 6.5 7 6 7 7 How easy was the instruction to use during the VR-product as- sembly? (Q2) 5 4 6 5 4 6 How much would you like to as- semble products with the instruc- tions in the future? (Q3) 4 5 5.5 2 4 7 How amused were you by the as- sembly task? (Q4) 6 5 5.5 6 5 6 How stressful was the assembly task? (Q5) 5 5.5 5 5 6 6 Figure 4.10: Results to the question regarding how well the participants under- stood the instruction type after the LEGO tutorial (Q1). 31 4. Results The participants were also asked how easy they thought the instruction was to use during the VR-product assembly, see Figure 4.11 for results. This was to see differences in understanding a technology versus its perceived usability. This time the responses varied more greatly, ratings were from 2 to 7. The Video instruction was similar as to how well they understood the instruction during the LEGO tutorial, still ranging from 5 to 7. The AR instruction ranged from 2 to 7 and was therefore perceived as harder to use. The same development goes for the T&P instructions, but the range was from 3 to 7. The median or mode of the responses would result in the following rank for the instructions; Rank Median or Mode (Q2): 1. Video 2. AR 3. T&P Figure 4.11: Results to the question regarding how easy the participants thought the instruction was to use during the VR-product assembly (Q2). Figure 4.12: The results on how much the participants would like to use the instruction types for assembly tasks in the future (Q3). 32 4. Results The next question was related to how much they would like to use the instructions for assembly tasks in the future, Figure 4.12 presents the responses. The responses ranged from 2 to 7, and have a wide spread for all technologies. AR ranged from 2 to 7 and had 40% responding a 2, which indicate that many people did not like to use the technology. On the other hand, 30% responded a six indicating that several participants liked to use the technology. The T&P instruction is ranging from 4 to 7 where 40% scored it as a 4, while the rest of the responses is trending towards 7. The Video instruction is ranging from 3 to 7 and 30% scored it as a 7. The rest of the response distribution was almost uniform towards the score of 3. The ranking according to either median or mode would be; Rank Median or Mode (Q3): 1. Video 2. T&P 3. AR The participants was then asked how much they liked the assembly task, results can be found in Figure 4.13. The reason was to see if any technology could be more appreciated to use in industry for assembly tasks. The responses ranged from 4 to 7 and all instruction types had similar distributions. The T&P instructions ranged from 4 to 7 and peaked at 5 with 40% of the responses. AR peaked at 6 with 50% of the responses and ranged from 5 to 7. Video ranged from 4 to 7 and peaked at 6 with 40% of the responses. The ranking using median or mode resulted in two different rankings, where using mode resulted in two first places. The rankings are; Median (Q4): 1. AR 2. Video 3. T&P Mode (Q4): 1. Video 3. T&P1. AR Figure 4.13: The results when the participants were asked how amused they were during the assembly task (Q4). 33 4. Results Figure 4.14: The results from the participants when they were asked how stressful the assembly task was (Q5). The final question the participants was asked was how stressful they perceived the assembly task. The results can be found in Figure 4.14. This was to identify if there existed any differences on perceived stress between the instruction types. All responses ranged from 2 to 7. The AR and T&P instructions ranged from 3 to 7 while the Video instruction ranged from 2 to 7. The distributions of the responses for the instruction types does not seem to indicate any significant findings. If ranked to mode or median the rankings would be different and some would have several first and last places. The rankings according to mode and median would be; Median (Q5): 1. T&P 3.Video 3. AR Mode (Q5): 1. Video 3. AR1. T&P 34 5 Discussion This discussion chapter will comment on the methods used in the research, as well as on the results connected to theory, and future research. 5.1 Methods To consider both planning and presentation of instructions to produce effective work instructions is an interesting method [6]. It implies to look into and optimize several areas (elements in pre-work) that at first glance can seem to not affect the assembly instructions too much. Doing so changes the content the instructions need to present and should, in theory, result in effective and advantageous work instructions. The methods used have been iterative because the methods affect each other, changing one might affect previous steps. The development phase has been an on-going pro- cess where several previous steps has been reevaluated or adjusted. It has however worked well and the final results should be extensively optimized given the certain circumstances. The VR-product does differ quite a lot from a traditional product that are to be assembled in industry. There are often several parts that need to be assembled and nuts, screws and tool that need to be used in a certain order. The VR-product is simple, it is a single part that needs to be folded in a specific way to be assembled where no tools or similar is necessary. The assembly complexity of the VR-product could be considered much lower than a traditional product in assembly. The results might therefore be very product specific in that sense. The research methodology was designed with research quality in mind. The purpose was to compare the instructions in a fair way by using theoretical guidelines and vali- date with an improvement workshop to fix any misinterpretations. The experiments have not been triangulated since the methodology itself should ensure experiments with a high amount of trustworthiness. 5.2 Pre-work implications The pre-work results sets the foundation of the assembly instructions. It also makes sure that the instructions will be more effective than to just develop instructions of a specific type or to simply change instruction type to another. Properly ex- ecuted pre-work might reduce the number of components, screws, assembly steps 35 5. Discussion or similar needed which in turn will make the instructions simpler. It implies less work to execute for the same added value which raises not only productivity but also sustainability. Digitalization offers a lot of opportunities to efficiently inte- grate departments with systems that can connect the steps that affect instructions. The construction department could be fully integrated with digital instructions and therefore make changes in the instructions instantaneous when constructions are changed. The product design has been optimized according to the product requirements. One large factor during the product development was that every product would automat- ically indicate a lot of waste. It is though better in one way than the initial design from a sustainability accept since more cardboard of the sheet are waste and can be guaranteed to be recycled, instead of trusting the visitors to recycle the product after its life cycle. The poka-yoke solution on the other hand does require an addi- tional manufacturing step (to cut away the corner) but result in a lighter product to deliver and therefore less environmental impact. The cut corner can also be imme- diately recycled in the production plant and does not need to be transported back for the same purpose. On top of all of those things, it will of course make the waste removal operations easier and more intuitive. The improved product design together with the assembly sequence have indicated 20.6% lower assembly time based on MTM-SAM compared to the initial product design with the most appropriate assembly sequence. The reduced MTM-SAM time should make it easier and faster for operators to assemble the product, which would raise the economical and environmental sustainability since the same resources can be used for higher productivity. 5.3 Assembly instructions The assembly instructions have been designed according to guidelines from available assembly instruction theory [13, 21, 22, 23]. The intention was that each assembly instruction would be designed according to every guideline found in the theory, but some of the guidelines could not be incorporated into the design. This apply mostly to the AR instruction, where resources were lacking to employ the latest hardware and also knowledge in related software technology. The designed AR instruction could therefore e.g. not use graphics incorporated with the physical assembly object (object recognition) [40], instead all graphics were placed away from the assembly object, and the time lag generated from the mobile device could also not be reduced [23]. The instructions themselves does contain an animation of the VR-product showing the assembly sequence with the active folded parts highlighted in green. The instructions could however perhaps become more clear if more critical steps were highlighted with arrows, circles or similar during the animation. This was not done because of time issues. Time lag is, and will always be, an issue when de- vices with an ordinary camera and corresponding display are used. That is because the camera’s detection of the physical world needs to be processed and the virtual objects need to be laid over onto the detection before everything can be displayed. 36 5. Discussion This could be solved or at least be reduced by using devices that uses optical see- through video, where only the virtual objects are displayed on top of the physical world which demands less computations and therefore also generates less time lag. Microsoft’s Hololens1 is an example that uses optical see-through video in AR ex- periences. The research group also observed from the experiments that some of the partici- pants felt uncomfortable using the AR-instruction. They did not appreciate being in-closed within a HMD, largely limiting their field of view, losing depth-vision and sometimes causing dizziness afterwords. Using technology similar to Microsoft Hololens would enable better integration of available guidelines from theory and could also offer solutions to the HMD related problems because the see-through video device retain the user’s depth sense, which probably is a large factor causing nausea and similar. The AR instruction is sensitive to lights, sometimes the trackers is lost and the virtual objects disappears from the operator’s view. This is solved by getting closer with the HMD to the trackers until it is recognized again by the camera. This was also observed to be an issue during the experiments, sometimes the trackers were lost and the operator needed to come closer to let the app find the tracker before the assembly could continue, which did affect the quantitative mea- surements. It was also observed that some participants had a hard time to properly understand the functions and how to use them during the assembly. It would be interesting to extend the tutorial to several minutes to see if it would affect the AR instructions performance. Another factor that might have affected the results that became evident during the experiments was that several participants had ordinary glasses (due to refractive errors) and the HMD did not support that. The HMD did have a functionality to change the focus of the lenses, but several participants mentioned they could not change the focus to become clear enough. Having glasses might affect this further and therefore a future recommendation would be to use an HMD that supports wearing glasses or have a large focus range that removes the effect of refractive errors. Regarding the T&P and Video-instructions, almost all of the guidelines from theory were used, which means that they are both nearly complete solutions in relation to theory and therefore very interesting to further study and compare. The lay- out of the T&P instruction was designed to be easy for operators to understand and thus let operators experience low amount of extraneous cognitive load. The research group however observed an issue regarding the way in which the eight sub- sequences were presented on the A3 page. Some participants did not find the layout appropriate and it thereby caused confusion, which resulted in assembly errors. It is therefore suggested to design the T&P instruction with a book layout, if using landscape A3 format. It would also be beneficial if the instruction was presented on a e.g. computer screen instead of using paper-printed instruction. This would fully utilize the positive effects of digitalization. The Video instruction were designed according to the guidelines with the high focus 1https://www.microsoft.com/en-us/hololens 37 https://www.microsoft.com/en-us/hololens 5. Discussion to enhance operator control, since this has been the main reason bringing perfor- mance and operator acceptance levels down according to previous studies [24, 25], and the observations from the experiments were very positive. Participants men- tioned that they felt comfortable and in control during the whole assembly. Some participants mentioned that the Video instruction really facilitated the assembly process by showing how to place and adjust hands during the assembly. Seeing the hands moving in the video should not be underestimated since it is the participants first time assembling the VR-product and they otherwise have to fully or partly guess were to best place them, regarding the AR and T&P instructions respectively. 5.4 Experiments The experiments consisted of 30 participants, which were divided into three equal groups managing an instruction technology each. It is very important to notice that the experiment group were in total as a sample group to low to draw statistical conclusion, though the results from the measurements with observations and surveys show a good indication of the real underlying quantitative values and perceptions. The age difference, ranging from 15-55, between the participants may also have distorted the data since the sample group is low and the division into the three smaller groups were made using randomization, resulting in low chance of equal age distribution among the groups. The age may affect results with technical complex instructions, such as the AR instruction. It would thereby be interesting to for future studies to see if age have a impact or looking into the performance of specific age groups. 5.4.1 Assembly time The measurements of assembly time from the experiments showed that the Video instruction had the lowest average assembly time values in comparison to the other instruction types. The range (distribution) between measurements is also the lowest for the Video instruction in comparison to the others. One possible explanation to the lower values of the Video instruction could be that it showed the operator how to place its hands during assembly, which made the inexperienced operator feel more comfortable and sure that the assembly was made correctly. The technology used for the Video instruction is also more familiar then e.g. AR technology. The AR instruction had the largest time values and largest statistical distribution between measurements. An explanation may be that it may take some time of practice be- fore reaching proper usage of the AR technology, especially if you are not familiar with using AR before. The research group believes that age may have a big impact on these results and that this impact will be minimized in the future along with the technology development currently happening in the AR field. AR technology is probably more accepted in lower age groups, which will drive related technology development considering the demographically challenges industries are currently fac- ing. 38 5. Discussion The T&P instruction also had a large distribution regarding the assembly time measurements. Surprisingly it had the lowest measured assembly time, but also the longest (if disregarding the AR outlier). These measurements could be a result of the, for some, confusing layout of the pictures in the instruction. If the participants understood the layout directly, it generated a relatively low assembly time, otherwise quite poor performance in our experiments. These related observations show that there is a good underlying potential of the T&P instruction medium to perform well, if accurately designed. Good news for companies that utilize the paper instruction format and do not want to switch instruction technology into more digital solutions. 5.4.2 Quality errors Considering the measurements of the quality errors, similar pattern as for the as- sembly times appear. The Video instruction had fewest assembly error in total and non of the completed assemblies generated any severe damage to the VR-product’s functionality. The AR instruction had the largest amount of assembly errors and 40% of the completed assemblies had large quality damage, related to functionality. T&P instruction generated 20% serve quality assemblies. The results from the assembly time and quality errors are quite surprising con- sidering previous studies, which often concludes superiority of AR technology over text-based instructions regarding both assembly completion time and amount of quality error [43, 44, 45, 46]. Previous studies show that Video instructions are often slow or causes irritation among operators. However, no previous study have taken into consideration designing effective instructions that follow proven guidelines from available theory and applying this to all studied instructions. It seems that many studies often design one instruction appropriately and then test it against an improperly designed instruction [45, 46], often only in text form, which do not sup- port human cognitive processes very well. These studies often miss the underlying inherent potential of the individual instruction medium/technology. Since the AR instruction in this thesis could not utilize all available guidelines from theory in its design, it is possible that the AR instruction is improperly designed, like some in- structions from previous studies. Thus, industrial engineers must bear this in mind when considering AR technology in general for assembly work. 5.4.3 Survey results The results from the survey follow the same pattern, that the Video instruction overall perform most positively regarding the asked questions and that the AR in- struction least positive. Not surprisingly it seems that the Video instruction and the T&P instruction are easy to understand, this because of the fact that they utilize familiar technology and need therefore not much practice before using it. During the VR-product assembly, Video instruction is most preferred to use regarding usability, followed by AR. The T&P had problems with the layout and it showed when the participants stress level raised. Regarding stress and amusement levels during the assembly, participants experienced approximately the same level amount of stress 39 5. Discussion and amusement. The result that showed the most distinct differences between the instructions was the question about the future of the instruction technology, wher