Degree Thesis

HALMSTAD

UNIVERSITY

Master's Progamme in Mechanical Engineering, 60credits

Appropriate instructions for manualassembly workers in industrial manufacturingsettings: factors to consider

Mechanical Engineering, 15 credits

Nathanaël Kuipers

i

Preface This thesis is the capstone course comprising of 15 credits of the Master Program in Mechanical Engineering at Halmstad University to validate if the candidate Nathanaël Kuipers has mastered the field of studies and with this achievement is granted the title of master. The work has been carried out during the spring term 2020 and focuses on the use of manual assembly instructions in manufacturing settings as part of a larger research project. Although this work was not possible without the support of many parties, special thanks should be given to: Dr. Aron Chibba, supervisor of the thesis work from Halmstad University - For his feedback and help when I got stuck. Dr. Ari Kolbeinsson, supervisor from the research project from University of Skövde – For providing me with a project proposal and giving me the opportunity to work on this thesis. Dr. Peter Thorvald – For his valuable insights and confirming some of my own. My hope is that this work will be of use and provide a foundation for further research. Nathanaël Kuipers, May 2020

ii

Abstract Manual assembly workers have nowadays a much higher workload than before. Not only do they have to deal with many more product variants, but as a consequence they also receive many more information signals that they have to act upon. This study focuses on the information assembly workers receive through visual instructions. By conducting a literature review in the domains of product development with focus on design for assembly, cognition related to information in instructions and different instruction formats like paper, tablet, and augmented reality (AR), commonalities and differences could be identified. Assembly operations are generally divided in handling a part and joining a part, and instructions should inform the assembly worker about when what should be assembled where and how. Each of these aspects has an impact on the overall complexity of the assembly process. To realise which of these aspects is most critical for an assembly worker can be of help in creating and delivering effective, tailor made instructions. The main finding is that there is not a one size fits all solution when it comes to the effectiveness of instructions, but that the type of instructions and the way they are delivered should in the first place be adjusted to the complexity of the assembly operations and secondly - if possible - also adjusted to the experience of the worker. The outcomes mentioned in this document should help laying the foundation for rules and guidelines when it comes to manual assembly instructions and its factors to consider. Keywords: instructions, manual assembly, visual aid, cognition, augmented reality (AR), design for manufacturing and assembly (DFMA)

iii

Contents 1 Introduction ...................................................................................................................... 1

1.1 Background ............................................................................................................... 1

1.2 Problem definition .................................................................................................... 3

1.3 Limitations ................................................................................................................. 4

1.4 Study environment .................................................................................................... 5

2. Method ............................................................................................................................ 5

2.1 Possible methods ...................................................................................................... 5

2.2. Chosen methodology for this project ...................................................................... 6

3. Theoretical framework .................................................................................................... 8

3.1 Assembly task and DFA guidelines ............................................................................ 9

3.2 Information in manual assembly instructions ........................................................ 12

3.3 Instruction modes ................................................................................................... 15

4. Results of the literature review .................................................................................... 18

4.1 Commonalities ........................................................................................................ 18

5 Analysis and conclusion ................................................................................................. 21

6 Critical review and future work ...................................................................................... 24

6.1 Limitations in the work ........................................................................................... 24

6.2 Used material .......................................................................................................... 24

6.3 Future work ............................................................................................................. 26

6.4 Critical review .......................................................................................................... 27

References......................................................................................................................... 28

Appendix A. DFA Chart: manual handling times ............................................................... 32

Appendix B. DFA Chart: manual insertion times ............................................................... 33

iv

Appropriate instructions for manual assembly workers in industrial manufacturing settings: factors to consider

Glossary: Manual assembly process: the process to assemble a whole product Assembly cycle: the process from receiving one instruction to the next Operation cycle: handling, positioning and joining a part Liaison: a bond between two components Instruction: a piece of information that explains how something should be done Instruction format or instruction mode: the medium or vessel in which the information is delivered Effectiveness: how well something is done (quality) Efficiency: how fast something is done (time) Performance: The sum of effectiveness and efficiency Effort: the sum of physical and mental strength needed Cognition: any form of information processing, mental operation, or intellectual activity such as thinking, reasoning, remembering, imagining, or learning (Encyclopedia of neuroscience, 2009)

1

1 Introduction During the twentieth century manufacturers were looking for ways to increase their performance. Much focus was put on making production lines more effective and efficient leading to mass production factories that spit out products in large quantities. However at the end of that century there was an increasing pressure from consumers who wanted custom, tailor made products. This meant that production lines had to be adjusted allowing more flexibility and that the number of components increased and possible product combinations went up exponentially. To cope with customized products in a diversified product portfolio production systems have become more advanced with flexible production lines. As a consequence the amount of information manual assembly workers have to process and act upon daily has increased significantly compared to a decade ago. With more variables to take into consideration errors are prone to occur (Zhu et al. 2008). With the concept of smart factories also referred to as Industry 4.0 - because it is considered the fourth industrial revolution - where a computer system is registering and monitoring all activities in the production, a digital copy of the factory should be able to pre-emptively identify and avoid errors with the help of simulations. This Industry 4.0 digital transformation should also aid workers in filtering and processing information, reduce their cognitive load and lead to higher performance. The expectation is that with Artificial Intelligent (AI) systems, where all devices are connected and communicate with one another, the information workers receive could be targeted for specific times and moments. One such domain where performance can be gained is within assembly and assembly instructions. Reducing the time between processing the instructions and executing the task is favourable but at the same time instructions have to be adaptable due to shorter production cycles with more customization. Meanwhile workers come and go, and training them effectively and on time is key. So the challenge companies are facing today is how to optimize the performance during the assembly process and what type of instructions to use. In this area much is expected from digital technologies like Virtual Reality (VR) and Augmented Reality (AR) which should be able to project and communicate visual notes adapted to the circumstances.

1.1 Background Instructions can be seen as a form of information which someone has to act upon. Instructions can be given in different formats; auditory, tactile and visually, where the receiver uses senses like hearing, touch and sight. (Mijksenaar & Westendorp, 1999). Independent of which format, it is important that the receiver decodes correctly what the sender is asking for and acts accordingly. The situation on a manual assembly line can be considered similar; the worker receives instructions on what to do, has to decode and interpret what is asked, and then execute the task. The process then repeats itself where the worker either receives similar or a new set of instructions. The process between receiving one instruction to the next

2

Instructions decoding interpreting acting

Figure 1. The process of an assembly cycle.

can be referred to as an assembly cycle. A schematic overview of a cycle can be seen in figure 1. With the aim to get the highest overall performance during the whole assembly process it has to be both effective and efficient. The Lean philosophy defines effectiveness as doing the right thing, and efficiency as doing things the right way. Effectiveness is connected to quality and reducing errors, efficiency to reducing time and effort (George, 2004). Instructions play an important role in both of them during the assembly process. Instructions have to be well documented, clear and concise, easy to use but foremost are interpreted in the exact same way by different receivers to minimize the risk of confusion and errors and guarantee quality (Menn et al, 2015; Söderberg et al., 2014). However the quality of the instructions is not the only factor that can lead to confusion and errors. The instructions can be crystal clear, but if the assembly operation itself is causing problems, the worker might still question if either the instructions are incorrect or if they are misinterpreted, even though there is a chance that neither is true. In such situations the worker would like to know as soon as possible and with minimum effort what is causing this discrepancy. The focus of this thesis is to investigate the synergy between assembly instructions, assembly worker and assembly task in further detail and how different parameters like assembly complexity, the skills of the worker, and the way the instructions are displayed affect this process. 1.1.1 Client This study is part of a larger research project at Chalmers University called InsTruction innovAtion for Cognitive Optimisation (TACO) in collaboration with the University of Skövde, Volvo Car Corporation, Volvo GTO, Combitech, SAAB aeronautics and PTC. The whole project is planned to take two years and receives financial aid by Vinnova, a Swedish government agency with the mission to promote development of efficient and innovative Swedish systems within the areas of technology, transportation, communication and labour. During this phase and for this thesis the main partner is Volvo, a car manufacturer located in Sweden, Europe. To keep a large part of the production in Sweden the car manufacturer has to work smarter and more effective.

3

1.2 Problem definition Manufacturing companies have to cope with an increasing demand for customized products, but they do not want to lose any performance in their productivity to stay competitive. This means that despite having more product variations to deal with both the effectiveness – delivering the right products – and the efficiency – delivering the number of products – should not be compromised. The whole philosophy of lean is built around the concept of streamlining the value stream, where any type of waste during the product manufacturing process should be eliminated without compromising quality (George, 2004). Although the manual assembly process has been studied from various perspectives, the results of these studies seem not conclusive on what can be considered best practice to keep optimal performance with an increasing product variety. Assembly instructions should provide the worker with essential, easy to understand information on how to assemble the product at hand, but there is currently no consensus what information is essential and how to convey that information effectively, so that assembly workers can understand with minimal effort what it is they need to do. Some studies conclude that instructions in the form of AR is the future, whereas other studies conclude that many workers do not like AR and that paper instructions lead to better results. Also, when it comes to efficiency in regards to time and effectiveness in regards to errors the results seem not conclusive. Creating effective instructions for manual assembly requires collaboration between several domains. Tversky et al. (2006) mentions that - design engineering, graphic design, production, ergonomics and cognitive science all play their role, and as a consequence assembly work is therefore discussed in several research domains. The difficulty lies in identifying all the factors of influence on assembly performance, organizing and categorizing these factors and find the correlations on how they influence one another. The main research question (RQ) that has to be answered is: Main RQ: In what way should instructions be provided to the manual assembly worker to maximize assembly performance? To be able to answer this question, certain factors need to be taken in consideration each leading to a RQ on its own. These are the following: RQ1: What is the task the assembly worker needs to perform? RQ2: Which information needs to be conveyed to the assembly worker to perform the task? RQ3: Which information is essential to maximize assembly performance? RQ4: What format or mode is most effective in conveying essential information?

4

1.2.1 Aim of the study The aim of the study is to analyse and assess the factors that play a role in the effectiveness of instructions in manual assembly operations and to predict how to deliver instructions in the most effective way for certain manual assembly operations. The identified factors of influence are:

• The difficulty of the respective assembly operation • The information that needs to be conveyed through instructions • The format or mode that the information is delivered in • The experience and skills of the manual assembly worker

A quick analysis of the available literature suggests that in the research only one or two factors are considered simultaneously and that they often correspond with a specific domain of study. By conducting a thorough literature review from different domains the expectation is that commonalties and differences can be identified for each influencing factor.

1.3 Limitations Certain aspects will not be considered in this study. The main limitation in this study is the focus on no language specific, visual aid communication within manual assembly processes. Using communication in words or tactile matters is outside the scope of this study. The disadvantage of wording either verbal or written is that they are language dependent, meaning that translations need to be made, which is not desirable if the goal is to make the instructions universal. Ganier (2004) also points out that from a cognitive perspective building mental models from pictures requires less effort than text. Polanyi (1966) argues that tactile communication certainly does happen during assembly and plays a big role, but it occurs often exclusively in a tacit manner, meaning that experienced assembly workers usually feel if something is done correctly but cannot explain why. If experienced workers cannot make this explicit and codify it, it is considered difficult to instruct others using tactile instruction modes. Another aspect that is excluded is the design and the development of an artefact itself for which the instructions have to be made. If already from the start designers and engineers would take in consideration the assembly process and the instructions, the belief is that much of the assembly operations can be streamlined. In this study the main focus is on already developed artefacts that are currently manufactured and assembled. Although all aspects of sustainability are important, this study considers mainly the social sustainability aspect where the working conditions for the assembly worker are in the center. Economic and environmental sustainability of the instructions are in this case much more difficult to evaluate and will require its own study.

5

Similarly, the experience and the mentality of the assembly worker will only be considered to a certain extent. Ideally each individual worker will receive tailor made instructions for optimal performance, but that seems at this point unsustainable and difficult to implement.

1.4 Study environment The majority of this part of the project is performed at the University of Skövde and at the student's home. If possible, field studies at companies will be performed.

2. Method This part of the TACO project focuses on making suitable instructions for the assembly worker, with the objective to reduce the cognitive load as much as possible. The information presented shall - if possible - be standardized in all formats, from paper to AR, easy to use and understand, and be adaptable to both the task and the needs of the worker. Before recommendations can be made it is therefore of importance to analyse the correlation between the following parameters:

• assembly task, including the difficulty of operation • information in instructions • instruction mode or format in which the information is delivered • competence level of the worker

To increase the chance of a successful outcome for this project an appropriate research method need to be chosen. In this section several methods are discussed and an explanation is provided for the chosen method.

2.1 Possible methods When research is conducted it should be backed up by theory or empirical studies. Within assembly instructions the norm is to conduct tests with subjects, interview them regarding their experience, and evaluate the results. It does however require much time and resources to do these hands-on tests. Considering the diversity of the various parameters, another problem appears to be the unique setting and condition in which each test is done, where results are more on a case to case basis, making it difficult to state conclusions that can be generalized and lead to standardization. An alternative approach is to look at the diverse research that has been conducted so far in different areas through a literature review, compare the methods and results from the diverse studies, and deduce how the different parameters influence one another. Meanwhile it gives an opportunity to identify commonalities and differences between the different research areas, and could also give an indication of the gaps which still exists. This approach however cannot take in consideration the unique needs of each worker.

6

2.2. Chosen methodology for this project With the objective from the TACO project to distil general considerations and guidelines for increasing manual assembly performance the chosen methodology is to start combining the information from several fields of study through a literature review. The process exists of two phases, the gathering information phase, and the developing phase. The orientation phase is about understanding the boundaries and common concepts that are known within manual assembly instructions. The design phase is about synthesizing the information and develop some hypotheses and guidelines for how to prepare instructions. If time allows for this project, the hypotheses and guidelines are tested by field studies to further consider the unique needs of the workers. Figure 2 shows an overview of the steps involved in the process, in which the top row depicts the gathering information phase and the bottom row depicts the developing phase.

2.2.1 Gathering information Within the assembly process the following aspects are distinguished:

- Providing information for the assembly task - Interpreting the information for the assembly task - Performing the assembly task

Because the central theme is considered manual assembly instruction, these words were the starting point of the search for relevant articles on Google Scholar. The search was done in an exploratory way using the snowballing approach (Wohlin, 2014). The objective was to discover different research fields and at the same time find terms and researchers related to these fields. This initial search led to almost half a million hits. After a quick scan through the first couple of pages of the titles that seemed most relevant the article was downloaded and the abstract read. Afterwards the snowballing approach continued between articles, keywords and

Figure 2. Chosen method which consists of an information gathering phase (top row) and a developing phase (bottom row).

Synthesize information

•Map findings•Commonalities &

diffferences•Gaps

Discuss results

•Conclusions•Hypothesis•Guidelines

Conduct field studies

•Testing•Observations•Interviews

Scan literature

•Experts•References•Keywords

Define concepts

•Assembly task•Information in

instructions•Instruction mode

Pick articles

•Relevance•Diversity•Contemporary

7

references. If articles used similar keywords and references, they were considered in a similar category and likely of relevance. Keywords were noted and the list of references was consulted and cross referenced with the other articles. Also, the suggested related articles by Google Scholar were analysed. Names of researchers that popped up several times were considered experts in their respective field and other articles by them were investigated on their relevance. After the initial search new strings were used with other found terms similar to the word instruction, like: task, guideline, guidance, or (visual) aid. This led to a few more articles of interest but not to significant new discoveries. Similarly, other search tools like Worldcat and Sciencedirect didn’t provide new leads either. After analysing the results there were three main themes which could be identified. The search string manual assembly task quickly led to the field of design for manufacturing and assembly (DFMA) for which detailed literature is already available. The search string of manual assembly instruction or guidance split up in:

a. the field of information cognition with focus on how to reduce the cognitive load for assembly workers when presenting instructions;

b. the mode in which the information was presented, where the performance of contemporary instruction modes like AR and VR is compared with more traditional (paper) instructions.

Conclusively one more search was then conducted with the original search string manual assembly instruction but now with either word variants of cognition or word variations of augmented reality added to it. This resulted in five more articles of interest. Now the concepts were defined in more detail - assembly task operation, information in instructions and instruction mode -, the initial 53 sources were analysed more in depth to review how well they covered one or more of these concepts as well as their quality. The initial 53 sources were then reduced to a total 31 sources to work with: 10 on manual assembly task, 14 on manual assembly instruction information and 7 on instruction mode. The limited number of quotations that showed up in Google Scholar to the articles was seen as an indicator that these specific research fields are rather narrow and that searching further would not lead to significant new discoveries. Li et al. (2013) came to similar conclusion that there is in comparison little contemporary research or clear guidelines on (digital) assembly instructions. 2.2.2. Developing framework Information in itself is neutral, but it starts to mean something when it is organized and put in a context. Information can be structured in many different ways. Because the goal of this project is to gain a deeper understanding of the factors that influence assembly performance, the chosen context was the manual

8

assembly process in chronological order with focus on one assembly cycle in the overall assembly process. The assembly cycle runs from receiving the instruction, decoding the information, and acting upon it until the next piece of instruction is provided. Within this cycle the overall performance can according to the lean philosophy be broken down in effectiveness and efficiency. Effectiveness is measuring if the assembly task is executed correctly and efficiency is measured in the time and energy it takes to perform an assembly cycle. If an assembly task is done correctly depends largely on how the assembly worker has understood and interpreted the instructions and converted them into action. It is therefore argued that instructions in the first place have to be effective, presenting information in such a way that it is understood correctly. As a consequence it is important to comprehend what type of information is required during an assembly cycle, and, depending on the assembly task, valuing and ranking which information is either more or less important to avoid confusion and potential assembly mistakes. Once it is known which information is more or less important for a specific assembly task, instructions can be made more efficient, with the effectiveness ranking as input. Lastly the different instruction modes or formats are considered, analysing and identifying in which assembly situations they might perform best. The hypothesis is that there is a correlation between the assembly task, the information that needs to be conveyed, and the mode in which the information is presented. The different sources are analysed with this in mind to see which results and conclusions can be of relevance for further consideration. All the findings are then summarized and an overview is presented where each finding is coming from. This study should then conclude with insights in the correlation between these concepts and lead to some hypothetical guidelines to follow when considering instructions. Validating these hypothetical guidelines in the form of tests, observations and interviews will be the next step in this research project.

3. Theoretical framework In the manual assembly process the performance is mainly depending on the assembly worker. The worker has to be effective and efficient, meaning delivering a satisfactory product without errors in the shortest time possible. However it requires that the worker does not have a mental overload, which can quickly result in stress, loss of focus, inefficiency and errors (Bäckstrand et al., 2008). Due to the increase of processing information, much contemporary cognitive research focuses specifically on this domain. However cognitive overload can also occur if certain actions require a high physiological effort. As Galy et al. (2012) discusses performance is depending on both mental load and mental effort, in which case mental load is compared with the information that needs to be processed and the mental effort is related to the physiological activity. Furthermore, Galy et al. (2012) conclude that there are additive effects on the total workload from both task difficulty and time pressure. The National Aeronautics and Space Administration (NASA) developed a tool referred to as the task

9

load index (NASA-TLX) to assess the mental workload which rates six categories: mental demand, physical demand, temporal demand, effort, performance, and frustration level (Hart & Staveland, 1988). In section 3.1 the actual assembly task is discussed, what it demands physically from the assembly worker and which implications it has on the performance. The topic is initiated with DFA guidelines. In section 3.2 the focus then shifts to information in assembly instruction and how they mentally challenge the assembly worker, demanding cognitive effort. Subsequently section 3.3 considers the different modes in which instructions can be delivered and how they influence the demands on and the performance of the assembly worker. The assembly task, the information in instructions and the instruction mode are all important factors to consider when it comes to the overall assembly performance, which includes both a mental load aspect, interpreting the assembly instructions and mental effort aspect, performing the actual task, with tasks varying in complexity and difficulty as discussed by Falck et al. (2017).

3.1 Assembly task and DFA guidelines Already in the 1960’s it was noticed that simplifying the assembly process would have a big effect on performance. First and foremost the number of parts needs to be reduced as much as possible during the product development phase which would benefit overall quality and reduce assembly time and thus costs (Boothroyd et al., 2011). However other factors play an important role as well on the assembly difficulty level and efficiency. Here, although they are correlated, a distinction can be made between part attributes and assembly operations. Among part attributes are weight, dimensions, shape, and rigidity/ flexibility (Otto, 2001). Most of the design guidelines for parts to simplify the assembly process are shape related like the use of part symmetry, chamfers, snap fits, pegs, protrusions and holes. (Boothroyd et al., 2011) When it comes to the assembly operations themselves several phases can be distinguished. Otto (2001) identify the following operations:

1. Handling a part 2. Inserting a part 3. Joining a part;

Ullman (2016) on the other hand distinguishes the following operations:

1. Retrieving 2. Handling 3. Mating, also referred to as insertion

Some consider insertion and joining as one operation, whereas others treat them as two distinguishable operations. Otto (2001) argue that with insertion parts are put in the correct position and location, but they are not necessarily fixed and a special tool might be required.

10

Taking this distinction in consideration the combined list results in: 1. Retrieving; this considers where the part is located and coming from. In

certain cases there is an automatic feeder, in other cases the worker has to fetch the part from somewhere else.

2. Handling; this includes picking up a part and orientating it in the correct position.

3. Inserting; here the part is put in the correct location and position of the work piece.

4. Joining; this is the action required to fix the part on that location and position. The mating of two components or parts in an assembly task is also often referred to as a liaison (Lim & Hoffmann, 2014). A special tool or fastener might be needed here. In the case of a fastener they are also considered a unique part that has to be retrieved, handled and inserted.

The phases from retrieval of a part to the joining of it can be considered an operation cycle. To evaluate how effective such a cycle is several methods have been developed over the years. Some of these methods are:

• Boothroyd and Dewhurst DFA • LASeR • Lucas Engineering and Systems DFA • SEER DFM • Xerox Producibility Index (XPI)

However the focus of most of these methods is evaluating how difficult the actual liaison is, which includes the insertion and joining action. Here all the tools show a correlation between increasing difficulty and the extra time and effort it takes which reduces efficiency (Otto, 2001). But as discussed, an operation cycle consists of more than just the mounting. Also, the retrieval and handling play an important role and need to be considered. Of the aforementioned methods only Boothroyd and Dewhurst and Lucas Engineering consider the handling of a part (Otto, 2001; Zaeh et al., 2009). When it comes to difficulty of retrieval and handling the following four categories can be identified as having an influence:

1. Part attributes such as dimensions, weight and shape. 2. Handling distance that the assembler has to reach to get to the part. 3. Handling method, or what is required in terms of operators, tools, and

aids to handle the part. 4. Handling difficulties or conditions that make it difficult to get the part to

the assembly point. (Otto, 2001, p. 712, 713) Both the BoothRoyd and Dewhurst and the Lucas Engineering method seem much more thorough in their overall analysis when it comes to assembly time and effort. This is confirmed by tests by Otto (2001) in which they compared theoretical assembly operation times with actual assembly operation time. Both the Boothroyd and Dewhurst and the Lucas method predicted the actual assembly

11

operation time highly accurately. Due to their thoroughness these methods provide therefore more valuable information when it comes to assembly performance despite the fact that they were developed several decades ago. Considering that production lines have to be more agile and flexible to cope with the needs of modern society, manual assembly work will not disappear anytime soon and despite the fact that manufacturing has evolved over time, the basics of manual assembly work are still the same. Particular Boothroyd and Dewhurst is still much referred to in contemporary literature and articles confirm its usefulness even today. The reason why the Boothroyd and Dewhurst method is so accurate in estimating the time it takes to perform certain actions during the assembly process is because the values are based on empirical data and measurements. This data was converted into charts, where (the lack of) certain assembly features require extra time. The charts divide the assembly process in two main categories:

1. handling a part (appendix A) 2. insertion of a part (appendix B)

With handling is meant to pick up, orientate and put the part at the right location. With insertion is meant the activity to fixate the part. The handling and insertion time may vary greatly depending on different parameters that are considered to take extra effort. Because this method has mainly been developed to measure assembly efficiency in time, and is based on empirical data, it is not made to predict or assess assembly complexity. To assess assembly complexity Falck et al. (2012) uses a term to measure complexity for the car industry called basic assembly complexity (CXB). In here assembly tasks are ranked by personnel in low assembly complexity (LC) and high assembly complexity (HC) tasks. The results show a very strong correlation between the DFA guidelines and the perceived complexity among which:

• the size and features of the part. Smaller parts or larger parts may increase handling difficulties. Same with flexible or slippery parts;

• the weight of a part. If it is heavier it takes more effort and ergonomics might be jeopardized.

• how easy it is to align the parts. Some parts have self-guiding and aligning features which lead and keep the part in the correct position while the next step in the process is performed;

• if one of the parts has self-locking features. If this is the case then the assembly step can be completed in a much shorter time;

• whether the assembly process and operation steps are ambiguous or not. Preferably there is only one way of assembling the product.

• if special equipment or skill is needed to join parts. According to Sinha (2014) complexity is divided in complexity of product components (handling), complexity of the liaison (positioning and joining) and

12

complexity of the topology of the product related to the product architecture (intricateness or how interrelated components are on one another). This last aspect could have an influence on the possible number of sequences in which the product can be assembled. Based on this work Alkan et al. (2018) developed a model to assess assembly complexity already in product development using a formula based on Huckel’s molecular orbital theory. In summary Boothroyd et al. (2011) differentiate assembly operations and class them in different categories. The main categories in which operations fall are handling a component and joining it. The DFA timesheets show a strong correlation between time/ effort and the complexity of the part (handling) and the liaison (joining) as argued by Alkan et al. (2018). Furthermore, Alkan et al. (2018) adds that the product architecture that has an influence on the assembly sequence plays an important role on complexity too.

3.2 Information in manual assembly instructions With an increase of information that needs to be processed on a daily basis, the information presented in manual instructions has gained more attention over the years in particular in the field of cognitive science with focus on mental, cognitive load. Which information needs to be displayed and how can it be displayed in the most effective way for the worker to perform optimally? Mijksenaar and Westendorp (1999) bring forward aspects that each type of instruction needs to communicate. The essential elements are:

- identification of the object - composition, which shows how different parts or objects are combined or

related to one another in a bigger whole - location and orientation showing the exact position of an object

Other aspects which often have a certain relevance as well are: - movements which highlight the actions that are needed - sequence; certain actions need to be performed in a specific order to reach

a certain goal - measurements if certain units like weight, size or time play an important

role - warnings or highlights in case there are hazards or risks involved

Even though Mijksenaar and Westendorp (1999) their main focus is on instructions for consumer goods, they can be applied on instructions for assembly workers too, as Kaipa et al. (2012) state very similar parameters in their paper and they overlap to a large extent the assembly process as described by Boothroyd and Dewhurst. The main criteria that Kaipa et al. (2012) added to their list are certain feedback loops in the form of a check or control function if the liaison is performed correctly. They also refer to each aspect on the list as a sub-task. 3.2.1 Objective of the task Ganier (2004) outlines how information is processed when it comes to instructions and execution. The first aspect an instruction needs to communicate is a goal or target of what has to be achieved. The nearest set goal is then stored in

13

the working memory until the target is reached. This part of the memory can however only cope with a limited number of stimuli at the same time (Söderberg et al., 2014). To be able to reach that goal a plan is developed to divide the goal in smaller actions or sub-tasks. Novick et al. (2000) concluded that instructions presented in a step-by-step fashion with sub goals performed better than presenting the end goal. That step-by-step instructions are generally preferred is also supported by a test by Agrawala et al. (2003). In an article by Daniel and Tversky (2012) instructions are broken down in extrinsic actions, the desired end-result of what to do and intrinsic actions, smaller objectives on how to perform the task. 3.2.2 Hierarchies in assembly tasks When workers assemble parts both the parts and operations are mentally put in different hierarchies. A part is considered a part if it is disconnected from other parts. Parts are classified based on their size, features and function (Agrawala et al., 2003). Alkan et al. (2017) classifies three type of components: essential components, quasi-components and virtual components, in which the essential parts are considered the most important and main parts or assemblies, quasi-components are fasteners and virtual components are welded, soldered or glued joints. Parts that are the same or have a similar function are grouped. (Agrawala et al., 2003). Grouped parts should preferably be mounted parallel or sequential, without steps in between (Tversky & Hemenway, 1984). Similarly, assembly operations are also put in hierarchies. The assembly is divided in a string of actions. There is the main assembly, which can consist of many subassemblies. This goes then down to component level, where one part is added at the time. In this case parts are classified into different categories, depending on their size and function (Agrawala et al., 2003). 3.2.3 Planning of assembly sequence In many cases an artefact can be assembled in several different sequences, but it seems difficult to determine which sequence is best from a performance perspective. In a test by Lim and Hoffmann (2014) inexperienced subjects had to assemble a hacksaw, but could develop their own strategy in which order they preferred to assemble the parts. After about a dozen of trials the subjects settled for a sequence that they considered most efficient. Although there was not a clear preference in sequence, a pattern was identified that subjects started with the larger, heavier parts. Another important note to consider is that some liaisons are conditional, meaning that certain parts cannot be assembled before a certain condition is fulfilled. Identifying which steps or parts in the assembly process play a key role is important in planning the sequence. Baldwin (1991) rooted his test on research by Bourjault (1987) in which first has to be determined if a part physically can be attached without obstructions in the mounting path. With the possible sequences that are then generated using an algorithm containing certain basic rules to eliminate other impossible combinations Baldwin (1991) then explains how the possible combinations that are left can be evaluated by ease, reliability and

14

fixturing. The designer or user can then decide which combinations have most potential. Agrawala et al. (2003) also used an algorithm to define possible assembly sequences, some of which have a similar purpose as the rules from Baldwin (1991), like if parts can physically be assembled. However Agrawala et al. (2003) put more emphasis on how to communicate the information for the assembly process effectively whereas Baldwin (1991) considers mainly the actual liaison. Furthermore, Agrawala et al. (2003) differentiates significant and less significant parts but the research does not consider subassemblies. Baldwin (1991) does not consider hierarchy of parts, but the paper does take in consideration the hierarchy of assembly operations including several subassembly levels. 3.2.4 Decoding instructions According to Tversky et al. (2006) instructions are often cluttered and depict too many details, making it harder to decode and increase mental load. When people need to decode instructions they are looking for clear landmarks or references (Tversky et al., 2006; Zaeh et al., 2009). In such cases it’s less about accuracy but more about building a mental model of how the elements are located in relation to one another. Parts with distinguishing features seem easier to recognize than small details. This might also explain the findings of Lim and Hoffmann (2014) where subjects started with assembling the bigger more recognizable parts of a hacksaw providing a good reference base. That unnecessary details should be avoided in instructions is also stated by Söderberg et al. (2014), but they also suggest that photo quality pictures are preferred. Another important aspect to consider is when certain parts are used during the assembly process which look similar, but are different for example screws. In such situations a clear label is required to distinguish the parts but also to inform the worker to pay extra attention. 3.2.5 Visualizing assembly information As mentioned earlier there seems to be a clear preference for step-by-step instructions. The instructions can then either be presented in structural or action diagrams. Structural diagrams present the parts in their final position whereas action diagrams show the parts separated from the assembly piece with guideline indicators on where to mount them, similarly to exploded views (Agrawala et al., 2003). Agrawala et al. (2003) finds that action diagrams are superior because they add information on the joining operation which is excluded from structural diagrams. In a test by Heiser et al. (2004) this notion was supported that action diagrams are preferable. Another aspect that Agrawala et al. (2003) addresses is that both the assembly part and location should be clearly visible in the instructions. This is again in agreement with DFA principles that recommend that the assembly should be performed without any visible obstruction. However a difference can be identified as well. Whereas Agrawala et al. (2003) advocate to change the perspective and rotate the view if this leads to an easier decoding of the liaison, Boothroyd et al. (2011) strongly object to any type of rotation unless absolutely necessary, because

15

any rotation step is considered a waste of time which negatively impacts overall assembly efficiency. 3.2.6 Skills and experience When information is presented a distinction should be made between novel and trained people. Ganier (2004) identified a trend that novel persons presented with a task want the information to be instruction-based whereas trained more experienced persons want the information to be task-based. The difference here is that instruction-based information is presented in a linear fashion to get to a predefined goal, whereas task-based information is only consulted by the more experienced user when a specific issue occurs and the person wants to solve it. Both knowledge and skill acquisition play hereby an important role in how workers value and interact with instructions (Menn et al., 2015). In summary there are some commonalities on how assembly information can best be displayed in visual instructions to reduce the mental load. Many of the principles are in line with the DFA guidelines suggested by Boothroyd and Dewhurst. This includes information on which part to assemble, where and how to position it and how to join or assemble the part. This field of research also adds assembly planning or in other words when or in which order to assemble. Some liaisons are conditional, where certain steps must precede. Furthermore, the field explains the concept of assembly hierarchies, with tasks and sub-tasks stored in the working memory, as well as identifying that people can have different acquisition skills from beginners to advanced, and that these people have different working methods and requirements. Another aspect that is touched upon is feedback during the assembly process.

3.3 Instruction modes How effective instructions are not only depends on what information is presented but also in which mode or format it is presented, because they influence how information can be displayed. The traditional printed paper instruction as visual aid had been the preferred format for many years. The printed instructions then shifted to digital formats being displayed on tablets and smart phones. Smarter products with higher processing power also propelled AR and VR technologies forwards and much is expected form these formats within the assembly process. But although quite some tests with AR and VR have been conducted over the years, there is still no consensus on what benefits these technologies bring exactly compared to the more traditional picture based step-by-step (paper) instructions. In this section an analysis is conducted on different ways instructions can be presented. Several perspectives - sometimes unique to only one instruction mode - are discussed with benefits and drawbacks for each of them. 3.3.1 Static or dynamic One of the main disadvantages of paper instructions is that they poorly communicate how to assemble an artefact. That’s why they often need to be

16

complemented with written text to explain this process (Daniel & Tversky, 2012). With the digital portable tools that are readily available nowadays, visual instructions don’t longer have to be static, but can be dynamic, showing animations of the assembly process. However this does not automatically lead to better comprehension, but much depend on the way the instructions are presented (Tversky et al., 2006). Another disadvantage of animations is often that they cannot be reinspected at the viewers’ own pace (Tversky et al., 2006). Though to be able to see an animation of the way a part should be assembled in the right place and in the correct orientation can be valuable for the worker (Sääski et al., 2008). Using digital tools, objects and artefacts can even be manipulated in a virtual, three-dimensional space, where they can be analysed from different perspectives. 3.3.2 Integrated or remote Instructions are in many cases presented remotely, through a different medium that the worker than has to translate to the real world. This switching back and forth from the medium to the real world negatively impacts the mental load and the efficiency (Lampen et al., 2019). They also mention that external devices often require manual input, which means that workers cannot have both hands free at all times. An advantage with AR is that it can also be used to project instructions in situ, meaning directly in the workspace where the assembly takes place (Funk et al., 2017), whereby diffusing the boundaries between virtual and reality. Being able to project instructions on the actual location comes with the disadvantage that a large space has to be prepared for the equipment and as a consequence this mode is therefore not particularly flexible or mobile. Remote on the other hand is not so dependent on the location and can easily be transported. 3.3.3 Passive or adaptive Another aspect that can be achieved due to artificial intelligence and connectivity nowadays is the possibility to monitor the assembly process and anticipate on the situation at hand. When the system is able to assess the environment and interact with it, it is called context-aware. When the system is passive, it is referred to as context-free (Boud et al., 1999). For the system to be context-aware it needs to be able to scan the environment with the help of sensors, often cameras and software to process that information. Paper instructions are therefore per default classified as context-free due to their analogue nature. Funk et al. (2017) conducted a multi-day test with in-situ AR, collecting data from each subject for at least three full workdays using an intelligent system. Workers generally did not seem to appreciate that they are monitored over a longer period of time. The in-situ system also had a considerable negative impact on the task completion time where subjects felt like the system was slowing them down after a while and became a nuisance. However the system was first much appreciated for learning the assembly steps, but once that became a routine the system became a distraction instead of an aid.

17

3.3.4 Efficiency and effectiveness The main reason why the empirical studies on different formats are conducted is to measure the influence on efficiency and effectiveness while maintaining a sustainable workload. The results are not conclusive because they seem to be case dependent each using unique parameters. Some reoccurring trends however can be identified. Head mounted displays (HMD’s) do not seem to provide clear advantages in regards to time and number of errors (Blattgerste et al., 2017; Funk et al, 2016; Syberfeldt et al., 2015; Zheng et al., 2015) Tests by Blattgerste et al. (2017) and Zheng et al. (2015) conclude that HMD increased completion time either marginally or significantly in comparison to paper instructions. What is of particular interest in their findings is that locating the correct position for the assembly task is not faster with HMD compared to paper instructions no matter if the information was depicted remote or in situ. Syberfeldt et al. (2015) also mentions significant longer assembly times with HMD compared to paper instructions, although their goal was more to evaluate acceptance rather than efficiency. Also, regarding number of errors HMD seems to perform worse than other modes (Blattgerste et al, 2017; Funk et al., 2016). However there is one aspect that sticks out positively in both of their findings; in situ projected instructions reduces errors in picking the correct part significantly compared to other modes. In summary there are several modes in which instructions can be presented. The traditional (paper) instructions are the most common ones where the instructions are presented remotely and the assembly worker uses it as reference on what the result should look like. This type of instructions either on paper or tablet is mainly passive, meaning that the worker decides the pace and checks if things are done correctly. On the other side of the spectrum are the integrated modes which include VR and AR. Because these modes are using digital models of the real world and should be able to register things in real time they can adapt to the circumstances and correct if needed. AR can project objects in the real world through a HMD or in situ. HMD is not performing consistently well compared to in situ, and the latter is therefore preferred. VR has the disadvantage that it is depending completely on HMD technology, making it only interesting for assembly training simulation. The table below provides an overview of the different modes and their identified properties (table 1). Table 1. Instruction modes and identified properties. Paper Tablet AR VR Static or Dynamic

Static Can handle both

Dynamic Dynamic

Integrated or remote

Remote Remote Integrated (HMD and in situ)

Integrated (HMD)

Passive or adaptive

Passive Both, but mostly passive

Adaptive Adaptive

18

4. Results of the literature review The literature study suggests that there is at the moment no consensus on what can be considered best practice when it comes to manual assembly instructions. An explanation could be that the conducted tests all had different characteristics with different test parameters. The data from Boothroyd et al. (2011) on assembly times is possibly the most accurate considering all the practical data it is based on. Also, the way paper instructions can be presented most effectively as discussed by Agrawala et al. (2003) seems to have a thorough foundation. Still it states more about the complexity of the instructions than the difficulty of the assembly task. Another issue that could be of influence is that DFA clearly differentiates between the different operations, retrieving, handling, positioning and joining, or in Alkan’s paper (2018) components, liaisons, and topology of the product, where each aspect gets a simplicity/ complexity factor. Few articles of the other research domains seem to take these different aspects in consideration. Either the study does take in consideration the different aspects, but does not seem to link the factors of simplicity or complexity to it (Funk et al., 2016; Blattgerste et al., 2017) or the factors of complexity are taken in consideration, but then the papers seem not to segregate the different operations (Agrawala et al, 2003; Tversky et al., 2008). Even though there are some gaps after analysing, comparing and mapping the input from the different research fields, a certain commonality and overlap can be identified.

4.1 Commonalities Although the study went in three different directions, namely:

1. DFA and the assembly task, 2. Interpreting and decoding instructions, and 3. Different instruction modes,

they have several things in common. The analysed literature seems to be inline regarding the following when it comes to manual assembly and instructions:

1. Performance is measured in efficiency (time) and effectiveness (number of errors).

2. The whole assembly process can be divided in the operations retrieving a part, handling a part, positioning a part and joining a part.

3. Operations can be of different complexity, depending on the complexity of the part, the liaison and the intricateness of the product. This complexity requires different levels of mental effort.

4. A strong correlation exists between increasing complexity of the task and decreasing performance in other words an increase in time and errors.

5. The assembly process has to be channelled and explained through instructions: what has to go where and how.

6. A set of instructions have to be decoded and interpreted by the worker, and then acted upon, which require mental load.

7. Mental load and mental effort are both needed in an assembly cycle and are additive.

19

8. Tasks are divided in subtasks towards a goal, resulting in a string of actions.

9. Beginners perform better with step-by-step instruction based information, while trained co-workers are prone to err more using the same information.

10. Conventional (paper) step-by-step instructions still perform well in comparison with modern technologies.

11. Current HMD products and technology generally do not lead to higher performance and the results indicate that they increase the mental load experienced by the user, even though many users find it an interesting medium to use.

An overview of the findings and in which sources they are most prevalent is shown in table 2. In this table each number corresponds to the numbers of the findings above, whereas the letters A, B, and C respectively correspond to the different domains researched in the literature:

A. Assembly task B. Information in instructions C. Instruction mode

In some cases the literature does touch upon other aspects from the list as well, but they did not seem prevalent. However all findings have at least a couple of references. An interesting observation here is that there seems not that much overlap between the domains, which could explain the earlier identified gap between the different operations and the difficulty. It can also be seen that finding number 6 regarding decoding of instructions is well represented, which it should be considering the central role it plays in this study.

20

Table 2. Overview of which finding is correlated to which source. Sources within field A: assembly task, B: information in instruction and cognition, C: instruction mode.

field 1 2 3 4 5 6 7 8 9 10 11 Agrawala et al. (2003)

B x x x x x Alkan et al. (2017) A x x Baldwin et al. (1991)

A x Blattgerste et al. (2017)

C x x x x Boothroyd et al. (2011)

A x x Bäckstrand et al. (2008)

B x x x Daniel & Tversky (2012)

B x x Falck et al. (2012) A x Falck et al. (2017) A x Funk et al. (2016) C x x Funk et al. (2017) C x x x x Galy et al. (2012) B x x Ganier (2004) B x Heiser et al. (2004) B x x x x Kaipa et al. (2012) B x Lampen et al. (2019)

C x x Lim & Hoffmann (2015)

B x x x x Menn et al. (2015) B x x x Mijksenaar & Westendorp (1999)

B x x Novick et al. (2000)

B x x Otto (2003) A x x Sääski et al. (2008) C x x Sinha (2014) A x Söderberg et al. (2012)

B x x x Syberfeldt et al. (2012)

C x x Tversky & Hemenway (1984)

B x Tversky et al. (2008)

B x Ullman (2016) A x x Zaeh et al. (2009) A x x x x x Zheng et al. (2015) C x x Zhu et al. (2008) A x

21

5 Analysis and conclusion At the start of this thesis the main RQ was: In what way should instructions be provided to the manual assembly worker to maximize assembly performance? To be able to answer this question, the topic had to be dissected in several subcategories each leading to a RQ on its own. After analysing the literature some of the RQ’s can be answered quick and concise whereas others are more complicated and conditional. Below follows each RQ with an answer: RQ1: What is the task the assembly worker needs to perform? The task an assembly worker needs to perform can be divided in read instructions, retrieve the part, handle the part, position the part and join or create the liaison. RQ2: Which information needs to be conveyed to the assembly worker to perform the task? Considering the task that needs to be performed the information that needs to be conveyed is when what part has to go where and how it should be joined. RQ3: Which information is essential to maximize assembly performance? Answer: Whereas RQ1 and RQ2 are straight forward to answer based on the literature, RQ3 becomes more conditional depending on the circumstances. All the information mentioned above is important, but it depends on the complexity of the handling of the part, the complexity of the liaison and the complexity of the product architecture which information is essential or critical to display. Figure 3 displays the different aspects of complexity within product assembly, divided in part complexity, which relates to retrieval and handling, and the complexity of the liaison, which includes positioning and joining. Both the part and the liaison are an intricate aspect of the product architecture which in itself also has a complexity level. With increasing complexity of the product features much more consideration should be given on which information to display and how much. At the same time when the assembly process requires little effort and is mostly self-explanatory information could be reduced. Information could also be reduced when workers are familiar with the task and work mostly on routine. RQ4: What format or mode is most effective in conveying essential information? Answer: Considering that this RQ builds upon the conditions from RQ3, deciding which format or mode to use should be based on which information is essential for the assembly task. HMD’s have been used in several studies, but results have been varying greatly, and therefore no recommendations can be made at this point. In situ projection based instructions on the other hand have been generally generated promising results showing its potential. It is particularly performing strongly in the part picking process in comparison to other modes. Traditional step-by-step (paper) instructions have been compared in many tests and they seem

22

still the overall best performing mode, despite not always being the preferred mode by the test subjects. It is however unclear if the performance is so strong due to the mode itself or because it is just so well known.

When it comes to the main RQ: In what way should instructions be provided to the manual assembly worker to maximize assembly performance, research has shown that there is not a simple answer to this questions because there are many factors of influence. As discussed complexity plays a key role in performance, where a higher complexity is related to longer operation times and an increase of errors. Complexity during the assembly process can be related to the part (what, as in retrieval and handling), the liaison (where and how, as in positioning and joining) and the product architecture (how many interconnections there are which can influence the when, as in sequence of actions). To be able to answer the question firstly it should be determined which of these aspects is considered most complex. DFA worksheets, the Lucas Method or the formula that Alkan (2017) proposes can be of support in estimating the complexity of the different aspects. Once the complexity is evaluated for the product including each operation cycle, instructions have to be adjusted accordingly, putting most emphasis on the most complex part of an operation cycle. If it is considered very difficult to assess for the worker where the part is located for retrieval or how to handle it due to geometry or other characteristics this should be considered in the way the instructions are displayed. The same can be said for the liaison of a part. If it is tricky to put the part in the right location or join it, this should receive extra attention. In such cases an action diagram or animation is preferred rather than structural diagrams. When

Architecture (when) Complexity

Part (what)

retreival

handling

Liaison

positioning(where)

joining(how)

Figure 3. The different dimensions of complexity within the assembly process.

23

animations are preferred, digital modes have a strong advantage over (analogue) paper instructions. If it becomes known that the handling and the liaison are both of a high complexity during the same assembly cycle, it is recommended to split the operation cycle in smaller subtasks and present each action on its own so that the cognitive load does not peak and lead to stress. Similarly, if the analysis shows that the product architecture is of high complexity and intricateness, the possibility of splitting it up in different submodules with subassemblies should be investigated. On the opposite side of the spectrum, if certain tasks are in comparison simple to perform then, if the product architecture allows it, they could be grouped and perhaps done and shown as one step. This is especially preferred when the tasks are identical or highly similar. As an example it could be numerous screws in a product or several shelves in a cabinet, where they should often be mounted in the exact same way. As discussed there are several different aspects to consider when it comes to assembly performance, and due to the complex nature it is difficult to develop a set of common rules. However a first draft has been made to present an overview with some considerations on complexity and guidelines when it comes to information and instruction mode which are of interest when designing instructions (table 3). Table 3. Guidelines on information and instruction mode with different types of complexity.

Complexity Type of information

Recommended Instruction mode

comment

Retrieval Which part In situ AR Mental load could be reduced by indicating correct bin, so the worker does not have to search or wonder if the part is correct.

Handling Manipulation orientation

Remote, to be able to compare as reference

Mental effort could be reduced by automated vibration feeder for quasi components.

Positioning Structural diagram showing relational references

Remote or in situ, showing location on product as well as provide point of reference.

If possible, a jig should be considered which can reduce both mental load and effort.

Joining Action diagram

Digital animation If the joining requires an unusual action the worker should be clearly instructed about it.

24

6 Critical review and future work

6.1 Limitations in the work At the start of the search process it seemed that only a few of the articles were related to manual assembly instructions, but when the work progressed, more details surfaced. In retrospect the search for relevant articles could have been more thorough and the chosen articles could have been picked with more consideration. Having said this, the articles mentioned in this document seem to cover a wide span on the topic of manual assembly instructions, and it is therefore safe to conclude that no big new discoveries or surprises will turn up. At the same time the chosen articles have proven that when it comes to manual assembly instructions there is no clear recipe of what can be considered best practice. The performed tests in the studies are done on a case-to-case basis which present data and insights, but it seems difficult to extrapolate them to more generic rules and guidelines. It seems possible to identify some common trends, although it is difficult to say whether or not this is due to specific test conditions. There are still so many unknowns and this also means that the subject has much potential to grow and mature. Assembly performance is to a large extent depending on product complexity. Product complexity is already established during earlier phases of product development, in particularly during embodiment design and detailing. If product designers, who know the product best, would already consider manual assembly instructions from the start, there is a good chance that many of the assembly complexities could be simplified. This requires however that product designers are more aware of and take in consideration the manufacturing conditions. However, simplifying the assembly process might not be enough. Also, reducing the chance for mistakes should be actively pursued. Here the use of Poke-Yoke principles already during the design phase could be considered. Although the objective is to create common rules and guidelines for manual assembly instructions, they can only to a certain extent consider the different skill levels of assembly workers. As several papers state, the difference between novice and experience assembly workers is quite large and they have different needs when it comes to information and instructions. But even when the different needs can be satisfied, then still there is the challenge to cope with the change of performance during shorter periods of time (Galy et al., 2012).

6.2 Used material During the analysis several questions came up which are open for debate in particular when it comes to measuring complexity and performance. 6.2.1 Target group and repeatability Much of the articles and literature regarding testing the usefulness of instructions are considering and testing with persons with little to no experience with the task,

25

or in this case assembling a product, where test subjects had to assemble a unique configuration one time. This provides great data insights and results for this target group, but trained assembly operators work day in day out with similar tasks and gain knowledge and skills over time. Therefore it can be argued how well the results of the different studies will fit them, because trained and experienced people can handle a different level of mental load and effort. The long term study by Funk et al. (2017) indicates that beginners require different aid than trained and experienced assembly workers. 6.2.2 Test medium Quite some assembly tests in the articles were simulating the assembly process using LEGO bricks. It can be challenged if this medium is representative for real life assembly tasks.

• LEGO bricks have some common physical properties from the DFA principles which components of other products often lack: 1. Classic LEGO bricks have tubes on the bottom and studs on the top

side, making them both easier to decode and as a consequence also more intuitive to orientate them correctly.

2. LEGO bricks are mostly simple shaped and many of them are symmetrical in at least one plane, helping to decode the orientation further.

3. LEGO bricks are part of a system that is built on a grid. This makes it more difficult to locate the correct position for a brick, because they can be placed almost anywhere in the grid.

4. LEGO bricks have been produced with great precision where dimension tolerances are extremely tight. As a result they fit/ join perfectly, which is less likely with other type of product assemblies.

5. LEGO bricks are self-locking because of clutch power and do not need special tools to assemble. Compare this with an assembly process in real-life and there is a good chance that some parts need to be fixed with fasteners and tools.

The first two points have a positive effect on the handling, the third point negatively influences insertion (positioning) which is supported by the findings of Funk et al. (2016) and Blattgerste et al. (2017), and the last two points benefit the joining process. • A few tests used more than one LEGO model, and compared those,

claiming they have a similar level of complexity, but it is unclear how this was assessed. It would be interesting to use a DFA worksheet or Alkan’s (2018) formula to verify if the models have a similar level of complexity.

• Although not directly influencing instructions, it is also possible that a lot of effort is needed for an assembly action due to weight, size, or loose tolerances of a part for which both hands or an aiding device are required, whereas LEGO parts can usually be handled with one hand.

• From some tests it seemed that certain shaped bricks were used in distinguished colours, which could be an extra indicator for the assembly worker that the correct part is picked.

26

The above arguments might lead to skewed results in particular when the bias is not identified and the data not adjusted accordingly. 6.2.3 Feedback Some articles state that the system they are testing is monitoring, meaning that it can detect when something is wrong and does not continue until it is ratified (Syberfeldt et al., 2015; Funk et al., 2017). Also, Kaipa (20) mentions check and control moments during the assembly process, but in these articles it is unclear how the situation is remedied if something is done incorrectly. If the assembly worker only got to know that something is incorrect, but does not receive feedback on when or where the error occurred, this can lead to mental stress, in particular if the subject cannot figure out what is wrong. The importance of feedback during the assembly process might be underestimated when it comes to mental load. In particular positive feedback in the form of confirmation should not be taken for granted as this will help the worker to promptly assess when things go wrong. Information from a tool like Failure Mode and Effect Analysis (FMEA) which helps to assess the severity, occurrence and detection of an error during the process could be useful here for an instruction developer so the assembly worker gets immediate feedback when it is required most.

6.3 Future work The analysis of the literature has provided insights on manual assembly work and instructions. Based on the conclusions a draft for some guidelines has been proposed. A field that has not been explored is tutorials or instructions in game design. Considering the trend to use more digital tools for instructions, it could be interesting to look into game design and development to analyse how this industry implements instructions, in particular in VR games, where subjects have to perform physical motions. This could provide a different perspective on how to develop instructions. Observations and interviews are necessary to further investigate assembly complexity and performance, as well as to evaluate the proposed guidelines. The proposed guidelines should furthermore be tested and verified. Much thought needs to be put in how to design and conduct the tests to reduce bias and singularity of the results. Assembly sequence and how it influences overall complexity and performance is an area that has not received much research. The literature has shown that in some cases components have to be assembled in a fixed sequence and that this can be calculated, but if a product allows for more sequences, little is known on what can be considered best practice from an assembly workers perspective.

27

6.4 Critical review The main objective of this project has been on optimizing manual assembly performance in which the worker plays a key role. It can be argued that social sustainability will benefit from this, providing more sustainable working conditions for the assembly worker by reducing their cognitive load. Reducing the cognitive load should also reduce the number of errors made, and consequently increase product quality which should benefit the company from an economic perspective. Reducing the cognitive load should furthermore increase mental health, and consequently benefit the society as a whole when people feel better and less psychological aid is required. Although this sounds all commendable in theory there are still various aspects that need to be debated. Because the goal is to maximize performance as a mean to increase profits this also means that companies expect an increased output from their workers. So it could be argued if the total workload will actually go down or if one type of load is just replaced by some other one. With the rise of the smart factories where everything is monitored, including the performance of the workers, more pressure could be experienced. Furthermore, it could be questioned if it is ethical to monitor workers during their shifts. In one of the tests workers already expressed that they felt more like machines than humans. The risk when humans start to work more like machines is also that they are less likely to think for themselves. This can actually hinder innovation, because nobody questions the status quo anymore. As discussed, instructions should show when, what should go where and how, but does not explain the why. If workers not only follow the instructions but also want to understand why they are doing things the way they do, they are more prone to come up with viable suggestions for improvement. Last but not least the impact on environment should be discussed. Instructions require some format or mode in which they are presented. All these different modes accumulate energy during their total lifetime from raw material, manufacture, transport, use, and end of life. Paper instructions require a lot of paper and ink, but don’t use energy when consulted. Every change however impacts the environment significantly with new prints, and the old instructions are considered waste. A digital format like tablets on the other hand take much more energy to produce and also use energy when consulted. The benefit is that updates can be done digitally and be implemented almost instantaneously compared to paper instructions. So even though certain instruction modes can be more helpful for assembly workers than other, they might require much sources to either develop, drive, maintain or update. How much value is added depends on the costs versus the benefits and in particular the impact on the environment should be further investigated.

28

References Agrawala, M., Phan, D., Heiser, J., Haymaker, J., Klingner, J., Hanrahan, P., & Tversky, B. (2003). Designing effective step-by-step assembly instructions. ACM Transactions on Graphics (TOG), 22(3), 828-837.

Alkan, B., Vera, D., Ahmad, B., & Harrison, R. (2017). A method to assess assembly complexity of industrial products in early design phase. IEEE Access, 6, 989-999.

Bäckstrand, G., P. Thorwald, L. J. De Vin, and D. Högberg. 2008. The Impact of Information Presentation on Work Environment and Product Quality: A Case Study. The 40th Annual Nordic Ergonomics Society Conference, Reykjavik, August 11–13. Baldwin, D. F., Abell, T. E., Lui, M. C., De Fazio, T. L., & Whitney, D. E. (1991). An integrated computer aid for generating and evaluating assembly sequences for mechanical products. IEEE transactions on robotics and automation, 7(1), 78-94. Blattgerste, J., Strenge, B., Renner, P., Pfeiffer, T., & Essig, K. (2017, June). Comparing conventional and augmented reality instructions for manual assembly tasks. In Proceedings of the 10th international conference on pervasive technologies related to assistive environments (pp. 75-82). Boothroyd, G., Knight, W. A., & Dewhurst, P. (2011). Product design for manufacture and assembly (3. ed.). CRC Press. Bourjault, A. (1988). Methodology of assembly automation: A new approach. In Robotics and Factories of the Future’87 (pp. 37-45). Springer, Berlin, Heidelberg. Daniel, M. P., & Tversky, B. (2012). How to put things together. Cognitive processing, 13(4), 303-319.

Falck, A. C., Örtengren, R., & Rosenqvist, M. (2012). Relationship between complexity in manual assembly work, ergonomics and assembly quality. In Ergonomics for Sustainability and Growth, NES 2012 (Nordiska Ergonomisällskapet) konferens, Saltsjöbaden, Stockholm, 19-22 augusti, 2012 (No. Book of abstracts, Abstract B4: 2).

Falck, A. C., Örtengren, R., Rosenqvist, M., & Söderberg, R. (2017). Proactive assessment of basic complexity in manual assembly: development of a tool to predict and control operator-induced quality errors. International Journal of Production Research, 55(15), 4248-4260.

29

Funk, M., Kosch, T., & Schmidt, A. (2016, September). Interactive worker assistance: comparing the effects of in-situ projection, head-mounted displays, tablet, and paper instructions. In Proceedings of the 2016 ACM International Joint Conference on Pervasive and Ubiquitous Computing (pp. 934-939).

Funk, M., Bächler, A., Bächler, L., Kosch, T., Heidenreich, T., & Schmidt, A. (2017, June). Working with augmented reality? A long-term analysis of in-situ instructions at the assembly workplace. In Proceedings of the 10th International Conference on PErvasive Technologies Related to Assistive Environments (pp. 222-229). Galy, E., Cariou, M., & Mélan, C. (2012). What is the relationship between mental workload factors and cognitive load types?. International Journal of Psychophysiology, 83(3), 269-275.

Ganier, F. (2004). Factors affecting the processing of procedural instructions: implications for document design. IEEE Transactions on Professional Communication, 47(1), 15-26. George, M. L., Maxey, J., Rowlands, D. T., & Upton, M. (2004). Lean six sigma pocket toolbook. McGraw-Hill Professional Publishing. Hart, S. G., & Staveland, L. E. (1988). Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. In Advances in psychology (Vol. 52, pp. 139-183). North-Holland. Heiser, J., Phan, D., Agrawala, M., Tversky, B., & Hanrahan, P. (2004, May). Identification and validation of cognitive design principles for automated generation of assembly instructions. In Proceedings of the working conference on Advanced visual interfaces (pp. 311-319). Kaipa, K., Morato, C., Zhao, B., & Gupta, S. K. (2012, August). Instruction generation for assembly operations performed by humans. In ASME 2012 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference (pp. 1121-1130). American Society of Mechanical Engineers Digital Collection.

Lampen, E., Teuber, J., Gaisbauer, F., Bär, T., Pfeiffe, T., & Wachsmuth, S. (2019). Combining Simulation and Augmented Reality Methods for Enhanced Worker Assistance in Manual Assembly. Procedia CIRP, 81, 588. Li, D., Cassidy,T., Bromilow, D. (2013) The Design of Product Instructions, licensee InTech, 101-114, Leeds

30

Lim, J. T., & Hoffmann, E. R. (2015). Strategies in performing a manual assembly task. International Journal of Industrial Ergonomics, 50, 121-129.

Menn, J. P., Müller, B. C., Seliger, G., & Ulbrich, C. (2015). Generic model for international assembly instructions for special machinery assembly. In Proceedings of 43rd Annual SEFI Conference, Orléans (p. 176). Mijksenaar, P., Westendorp, P. (1999). Open here: the art of instructional design. Thames and Hudson, London Otto, K. N. (2003). Product design: techniques in reverse engineering and new product development. 清华大学出版社有限公司. Polanyi, M., The Tacit Dimension, Routledge and KeganPaul, London, UK, 1966

Sääski, J., Salonen, T., Liinasuo, M., Pakkanen, J., Vanhatalo, M., & Riitahuhta, A. (2008). Augmented reality efficiency in manufacturing industry: a case study. In DS 50: Proceedings of NordDesign 2008 Conference, Tallinn, Estonia, 21.-23.08. 2008 (pp. 99-109).

Sinha, K. (2014). Structural complexity and its implications for design of cyber-physical systems (Doctoral dissertation, Massachusetts Institute of Technology). Söderberg, C., Johansson, A., & Mattsson, S. (2014). Design of simple guidelines to improve assembly instructions and operator performance. In The sixth Swedish Production Symposium. Syberfeldt, A., Danielsson, O., Holm, M., & Wang, L. (2015). Visual assembling guidance using augmented reality. Procedia Manufacturing, 1, 98-109. Tversky, B., & Hemenway, K. (1984). Objects, parts, and categories. Journal of experimental psychology: General, 113(2), 169. Tversky, B., Agrawala, M., Heiser, J., Lee, P., Hanrahan, P., Phan, D., ... & Daniel, M. P. (2006). Cognitive design principles for automated generation of visualizations. Applied spatial cognition: from research to cognitive technology, 53-75. Ullman, D. G. (2016). The mechanical design process (5. ed.). McGraw-Hill Education. Wohlin, C. (2014, May). Guidelines for snowballing in systematic literature studies and a replication in software engineering. In Proceedings of the 18th international conference on evaluation and assessment in software engineering (pp. 1-10).

31

Zaeh, M. F., M. Wiesbeck, S. Stork, and A. Schubö. 2009. A Multi-Dimensional Measure for Determining the Complexity of Manual Assembly Operations. Production Engineering 3: 489–496. Zheng, X. S., Foucault, C., Matos da Silva, P., Dasari, S., Yang, T., & Goose, S. (2015, April). Eye-wearable technology for machine maintenance: Effects of display position and hands-free operation. In Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems (pp. 2125-2134). Zhu, X., J. Hu, Y. Koren, and S. Martin. 2008. Modeling of Manufacturing Complexity in Mixed-Model Assembly Lines. Journal of Manufacturing Science and Engineering 130: 051013-1–051013-10.

32

Appendix A. DFA Chart: manual handling times

33

Appendix B. DFA Chart: manual insertion times

PO Box 823, SE-301 18 HalmstadPhone: +35 46 16 71 00E-mail: registrator@hh.sewww.hh.se

Name: Nathanaël KuipersMaster in Mechanical EngineeringEmail: kuipers_n@hotmail.comPhone no.: +46 (0)727016750