Visualizing Logical Architecture of Electrical and Electronic (E/E) Systems in Automotive Industry

Visualizing Logical Architecture of

Electrical and Electronic (E/E) Systems

in Automotive Industry

By

KAZIMGÖKBERKNUR

gokberknur@gmail.com

Paper Within Informatics, 30 credits

JÖNKÖPING UNIVERSITY

Supervisor: Domina Robert Kiunsi Examiner: He Tan

January 2020

JÖNKÖPING UNIVERSITY Tekniska Högskolan

A

UTHOR

’

S DECLARATION

T

his exam work has been carried out at the School of Engineering in Jönköping in the subject area User Experience Design & IT Architecture. The work is a part of the two-year university diploma programme, of the Master of Science programme. The authors take full responsibility for opinions, conclusions and findings presented.

A

BSTRACT

M

odern vehicles equipped with many hardware and software systems. To develop new functionalities and maintain existing features of the vehicles, engineers have to check relationship between software and hardware systems. Due to complexity of these rela-tionships, visual representation of how systems are working together is required to make R&D process easier. At Scania, engineers using node-link diagrams to represent logical function archi-tecture of the Scania vehicles. Logical function archiarchi-tecture is a part of electrical and electronic systems in the automotive industry, due to size of these systems they are considered as a large network. Visualizing large networks by diagrams is not a new problem in literature. At past, researches published about diagram drawing and algorithms have been developed to generate good looking diagram. However, sometimes due to complexity of the data, having complex and unreadable diagrams are unavoidable and they are hard to understand.

Previous studies investigated how diagrams should be drawn, however focus was not how users should interact with the diagrams. In node-link diagrams users follows edges to understand relationships between components. Having edges in the diagram heavily affects the diagram drawing time and also required space for the diagram. In this paper I developed an artefact which is not using edges to visualize LFA at Scania. Artefact usability has been tested with Scania engineers by giving some tasks to them. In the tests, artefact without edges achieved better results than node-link diagram and 426% improvement achieved by comparing task completion times in seconds. The artefact proved that it can be powerful alternative to classic node-link diagram visualization.

Keywords:Information visualization, network visualization, graph drawing, diagram us-ability, complex information, electrical and electronic (E/E) systems architecture

D

EDICATION AND ACKNOWLEDGEMENTS

T

his degree project was conducted with cooperation with Scania in Södertälje, Sweden. I would like to express my gratitude to Sofia Cassel for giving me the opportunity to conduct this thesis work at Scania and her support along the way. Also I would like to thank Marcus Dahlberg and EPID team during my time at Scania, it was pleasure to work with you. Since I was interested in automotive industry it was a very good experience to be at Scania and with curiosity I learned many things about vehicle manufacturing process.

A

BBREVIATIONS

AE: Allocation Element

CAN: Controller Area Network E/E: Electrical and Electronic ECU: Electronic Control Unit FV: Functional Variable

LFA: Logical Function Architecture

SESAMM: Scania Electrical System Architecture made for Modularization and Maintenance UF: User Function

T

ABLE OF

C

ONTENTS

Page

List of Tables xi

List of Figures xiii

1 Introduction 1

1.1 Background . . . 1

1.2 Problem Statement . . . 3

1.3 Purpose and Research Questions . . . 4

1.4 Delimitations . . . 4 1.5 Outline . . . 4 2 Theoretical Background 5 2.1 Information Visualization . . . 5 2.2 Diagram Layout . . . 7 2.3 Layering Method . . . 9 2.4 Usability . . . 9

2.5 Logical Function Architecture At Scania and SESAMM Tool . . . 10

2.6 Summarizing Chapter . . . 14

3 Method and Implementation 15 3.1 Design Science . . . 15

3.2 Data Collection . . . 18

3.3 Data Analysis . . . 18

3.3.1 Interview Analysis . . . 18

3.3.2 Literature Review Analysis . . . 19

4 Findings and Analysis 21 4.1 Interviews . . . 21

4.2 Interview Results Overview . . . 21

4.2.1 Categorisation . . . 22

TABLE OF CONTENTS

4.3.1 Categorisation . . . 25

5 Artefact Design and Development 27 5.1 Influences . . . 28

5.2 Artefact Development . . . 29

5.3 Artefact Evaluation . . . 32

6 Discussion and Conclusions 35 6.1 Discussion of Method . . . 35

6.2 Discussion of Findings . . . 36

6.2.1 RQ1: What are the problems when users interact with complex logical function architecture diagrams in the automotive industry? . . . 36

6.2.2 RQ2: How can the usability of logical function architecture diagrams in E/E systems can be improved? . . . 37

6.3 Conclusions . . . 38

6.3.1 Limitations of The Study . . . 38

6.3.2 Future Research . . . 39 Bibliography 41 A Artefact Screens 45 B Interviews 49 B.1 Interview Questions . . . 49 B.2 Interview Answers . . . 49

L

IST OF

T

ABLES

TABLE Page

4.1 Categorisation After Interviews . . . 22 4.2 Categorisation After Interviews . . . 25

L

IST OF

F

IGURES

FIGURE Page

1.1 Diagram Elements . . . 2

1.2 A Logical Architecture Diagram at Scania . . . 3

2.1 Example Node Link Diagram . . . 6

2.2 Sample Data set Visualization . . . 7

2.3 Different Diagram Layouts. Layout Names From Left to Right: . . . 7

2.4 Layered UML Diagram Example . . . 9

2.5 LFA Diagram Example . . . 11

2.6 Manually Drawn Diagram in SESAMM Tool 1 . . . 13

2.7 Auto-generated Diagram in SESAMM Tool 2 . . . 13

3.1 Design Science Activities . . . 16

4.1 Tree View in SESAMM Tool . . . 23

4.2 User Function Diagram in SESAMM Tool 2 . . . 24

5.1 Search Functionality in SESAMM Tool 2 . . . 27

5.2 Simple User Function Diagram in SESAMM Tool 2 . . . 28

5.3 Possible knight movements on the chess board . . . 28

5.4 Artefact Screen with 4 different layered view . . . 29

5.5 Artefact Screen with hexagonal representation . . . 30

5.6 Artefact screen highlighted selection in hexagonal view . . . 31

5.7 Artefact after 2nd iteration . . . 32

5.8 User Scores . . . 34

A.1 Real Auto Generated Diagram in SESAMM Tool 2 (My artefact used same data as shown in this diagram) . . . 45

A.2 Artefact main screen . . . 46

A.3 Artefact ECU filtering . . . 47

A.4 Artefact AE filtering . . . 47

LIST OFFIGURES

C

H A P T E R

1

I

NTRODUCTION

1.1

Background

I

n the automotive industry, the usage of new electronic units is increasing. As a result, software complexity and lines of code of the systems is also increases. Today an average vehicle has around 100 million lines of source code [1]. The softwares that runs inside these vehicles are small computers called ECUs (Electronic Control Unit) [2]. A vehicle can have more than 100 ECUs. Each ECU controls various vehicle functions and behaviours, for example, an Anti-lock Braking System (ABS) handles the braking of a vehicle and is controlled by an ECU. Distinct ECUs are responsible for different functionalities and vehicle behaviours. In some cases, several ECUs are required to work together for changing vehicle behaviour.

ECUs are connected via a network called CAN Bus (Controller Area Network) [3], and all of these connections are represented via diagrams. These diagrams are called Logical Function Architecture (LFA) in E/E systems in the automotive industry [4]. LFA diagrams are basically data-flow diagrams that represent input and output signals of the vehicle components. The diagrams show how a vehicle function works and what kind of components are required in a vehicle. Due to a high number of components and their connections, the diagrams become large and are hard to handle. The increasing complexity of E/E architecture diagrams affects the diagram understandability, traceability and their usability. As a result, it has become time consuming for the users to interact with these diagrams.

At Scania, engineers use LFA diagrams that are automatically generated by a software. Due to the large size of LFA diagrams, it is time consuming for the engineers to understand their use and functionality. For example, if they need to analyze the diagrams for vehicle functional improvement or diagnostics it consumes most of their time. Another problem with auto-generated

CHAPTER 1. INTRODUCTION

diagrams is that when a new element is introduced, the diagram design changes, this is because the software algorithms re-analyzes components and re-positions them. This redesigning of diagrams leads to increase in learning curve as a user has to learn a new diagram each time a new component is introduced.

The current autogenerated diagrams consists of three main elements: Nodes, Links(Edge) and Property of Nodes as shown in Figure 1.1. The Edges can be either directed or undirected but at Scania, all LFA diagrams have directed edges. With the integration of new technology such as Internet of Things (IoT) in cars, wireless updates and autonomous driving, LFA diagram complexity will increase in the future in terms of the nodes available and their connections [5] .

node 1

_{link (edge)}

node 2



directed

undirected

property property

Figure 1.1: Diagram Elements

Surprisingly, for LFA diagrams there are no alternative visualization methods exists besides node-link diagrams. Despite their complexity and size, they are widely used and accepted in the industry. During my research I could not find any alternative method that mentioned against node-link diagrams. In this paper I am challenging this gap and trying to re-invent the wheel for visualization of the LFA diagrams.

In order to deal with said complexity, this paper proposes the generation of automatic diagrams without using edges. The aim is to increase the usability of these complex diagrams making it more understandable for the users without any hike in the learning curve. This would provide users with information in a much faster way irrespective of data size. So far, different drawing algorithms have been created for handling complex diagrams to provide users with the best possible diagram design. These algorithms have been loyal to the diagram elements for creating their rulesets. However, removing one of the fundamental elements of a diagram has not been considered yet.

1.2. PROBLEM STATEMENT

1.2

Problem Statement

EPID team at Scania decided to implement a module in their software, that draws E/E architec-ture diagrams automatically. Goal of this module is to provide readable and easy to use diagrams. This goal was missed due to usability issues. When auto-generated diagrams are very large, interaction by mouse panning and zooming is required due to limited screen size. If a diagram contains many components, texts are not visible without zooming and users generally do not know where the information they need is. This creates a usability issue among the users and decreases user satisfaction within the tool [6].

Figure 1.2: A Logical Architecture Diagram at Scania

E/E systems are evolving, and complexity is increasing. Due to complexity and increasing number of components, drawing the logical architecture diagrams requires more time.

The auto-generated diagram in this paper means, a set of components drawn by a computer and shown associations between components. There is of course an algorithm behind a computer created diagram, but algorithms will not be discussed in this paper in detail.

Logical function architecture diagrams of the E/E systems have their specific rules, e.g. some elements cannot exist by themselves alone without any associations, or some elements cannot

CHAPTER 1. INTRODUCTION

exist in the same diagram according to their specification and documentation. Scania uses specific algorithms and libraries to produce auto-generated diagrams. Diagram drawing algorithms are generic and focused on producing diagrams by some rules, not by visualizing information in a user friendly way or to increase user satisfaction.

1.3

Purpose and Research Questions

This thesis wants to find a way to visualize LFA within E/E systems and increase user satisfaction. Due to collaboration with Scania in this thesis, problem domain will be the automotive industry. Currently, interaction between users and diagrammatic visualization at current tool (SESAMM Tool) is not user friendly and user satisfaction is low [6] .

RQ1: What are the problems when users interact with complex logical function architecture diagrams in the automotive industry?

RQ2: How can the usability of logical function architecture diagrams in E/E systems can be improved?

1.4

Delimitations

This thesis scope will be mainly user experience side of the complex diagrams and their usability. Literature search about different algorithms and the different visualization methods can be referred to but investigating algorithms of the visualization methods will be out of scope.

1.5

Outline

The rest of the report will give brief information about the problem domain, and some history about the theoretical background, related work and possible influences from other domains. Method of how the thesis aims to investigate the problem domain and tries to offer a solution, students’ findings and analysis, conclusions and future work. In chapter 2 related information visualization methods will be introduced briefly and how the diagrammatic information repre-sentation is used. Chapter 3 gives information about how author of this thesis will approach the problem and used methods will be explained.

C

H A P T E R

2

T

HEORETICAL

B

ACKGROUND

I

n this chapter, the author explains, how auto-generated diagrams are created by computers. and how it has been used in the E/E systems of Scania. Since auto-generated diagrams represents information about a system, information visualization, graph drawing, data visualization, network visualization are the keywords that have been used in this thesis during research.

2.1

Information Visualization

Information visualization is a persistent topic in many problem domains. It has been applied in different fields of daily human lives, from simple primary school textbooks to complex engineering problems. When there is too much information to visualize, the usability of this visualization is open to discussion. Due to the limited amount of screen size, representing a big amount of data becomes a challenge [6] . Many types of research have been done when it comes to drawing complex diagrams efficiently with computers and different layout algorithms have been discussed.

In literature visualizing information or a network by connecting nodes and edges called graph drawing [7] . Graph visualization is an area of both mathematics and computer science. However, graph theory within mathematics is not related to what this thesis is looking for.

For visualizing data, properties of the data should be well understood and then visualization method choice should be made.

According to data models, visualization of large systems is made to reveal structure and help users to understand the system. Size of the data affects the understandability of the visualization models. When there is a lot of data and their relationships to show, computers have been used for their visualization. These visualizations are often called as diagrams. Figure 2.1 shows elements

CHAPTER 2. THEORETICAL BACKGROUND

of a node-link diagram. So far, node-link diagrams are one of the most common diagram types [7] .

node 1

_{link (edge)}

node 2



directed

undirected

property property

Figure 2.1: Example Node Link Diagram

• Nodes: An entry that users care the most.

• Link (Edge): Shows relationships between nodes. Links are also called as edges in graph drawing.

• Property: Descriptive characteristics of the nodes or relationships.

In node-link diagrams, links can be either directed or undirected. Directed link means, link has a direction from one node to another, while there is no direction in undirected links.

To simplify large networks, diagrammatic representation is a common method to help users understand data. Simple data set given as an example below and then visualized as by a diagram. This data set is a social network and edges shows friendships among the people [7] .

Andre ↔ Beverly,Andre ↔ Diane,Andre ↔ Fernando,Beverly ↔ Garth,Beverly ↔ Ed,Carol↔Andre,Carol ↔Diane,Carol ↔Fernando,Diane ↔Beverly,Diane ↔Garth,Diane ↔Ed,Farid↔Aadil,Farid ↔Latif,Farid ↔Izdihar,Fernando ↔Diane,Fernando ↔Garth,Fernando ↔ Heather,Garth ↔ Ed,Garth ↔ Heather,Heather ↔ Jane,Izdihar ↔ Mawsil,Jane ↔ Farid,Jane ↔ Aadil,Latif ↔ Aadil,Mawsil ↔ Latif

Reading the text is relatively harder than diagrammatic representation to understand data. Figure 2.2 shows visualization of the same data set [7] .

In the dataset, we see “Aadil,Mawsil ↔ Latif” which means Aadil and Mawsil is friend of Latif, and Latif is friend of Aadil and Mawsil. In the diagram we see the association between Latif and Aadil and Mawsil via edges, that indicates Latif has friends called Aadil and Mawsil. This visualization is another example for Node-Link Diagrams, where the edges are undirected. This sample data set consist of 14 nodes, and edges between nodes are still visible by eye within the computer screen. If data set became very large e.g. 100000 nodes, nodes and their edges would be too complex. The visualization of such a big data set, may not be visible by eye and nodes would look very small on the computer screen.

2.2. DIAGRAM LAYOUT

Figure 2.2: Sample Data set Visualization [7]

2.2

Diagram Layout

The layout is basically the design of the diagrams, that means how nodes and edges will be placed in the given area. With the increasing data size and complexity, layout aesthetics started to being considered and discussions have been made to achieving the best looking diagram [8], [9] . While best looking diagram may sound a subjective statement, researchers put an effort to define some rule sets to define best looking diagram.

Figure 2.3 shows effect of the different layout choices in diagram design.

Figure 2.3: Different Diagram Layouts. Layout Names From Left to Right:

(a) Straight-edge, (b) Polyline, (c) Orthogonal, (d) Hierarchical

[10]

Network diagrams described set of edges connecting the nodes [10] . The way how nodes and edges will be placed on the area depends on the layout choice, selected aesthetic criteria’s, edge lengths and size of the area where the diagram will be drawn.

If we compare straight-edge and polyline representations, straight-edge uses smaller space to represent nodes and edges with some edge crossings while polyline minimizes edge-crossings but sacrifices the usage of required space to drawing the same diagram.

CHAPTER 2. THEORETICAL BACKGROUND

Using design space efficiently is also a consideration, especially when there are many nodes and edges to visualize. In the worst-case scenario for a large diagram, the diagram may be too big to fit the screen and exploring relationships between nodes and following edges may consume too much time. If some of the aesthetic rules are followed when drawing the diagram, the quality and understandability of the diagram is increasing [9] .

In [11], a graph drawing experiment have been made with 17 participants. Researches gave sample graph to participants and asked them to re-draw it according to their preferences. In this experiment, participants draw graphs by using a modelling tool. After the experiment, researchers analyzed the drawings and they found that minimizing edge crossings was the most common aesthetic criteria that participants followed.

The experiment mentioned in [11], proves that minimizing the amount edge crossings is the most crucial aesthetic property to generate a good looking diagram for humans.

With minimizing edge crossing, edge bending is also frequently mentioned in literature as an important aesthetic property for the quality of the layout [7], [8], [9] .

Those two aesthetic properties are not the only aesthetic properties that researchers focused in graph drawing. Different researches have been made to investigate other aesthetic properties and their effects on the diagram, but those two were the most common aesthetic properties mentioned almost every paper that I found in the literature. Also, due to time limitations of this thesis it is not possible to dive deep into every aesthetic criteria in graph drawing.

• Edge Bends: Increasing number of edge bends decreasing understandability.

• Edge Crosses: Increasing number of edge crossings decreasing the understandability

It has been mentioned that size of the graphs is exponentially increasing year by year [7], as a simple example Facebook has more than 2 billion active users per month in 2017 If we think about visualization of the Facebook network via a graph, size of this graph will be very huge. Another limitation is computational power while visualizing this network. Even with the fastest algorithms, it may take hours to find a layout for visualization [7]. However, that is not the only problem. When computer produces such a large diagram, the computer screen has millions of pixels to show the diagram to the user, in that case with the billions of nodes and edges there will not be enough pixels to render the diagrams.

The size of the LFA diagrams at Scania, may not be big as Facebook example. However, Scania engineers facing the same problem when graphs are very large. Users can not read the diagram element names without zooming in and they can not follow the edges without panning. In addition to that, size of the LFA diagrams will increase by using more components in the vehicles.

At this point, my research findings became repetitive and there are no major changes in graph drawing methods in order to improve usability of large diagrams.

2.3. LAYERING METHOD

2.3

Layering Method

Unified Modelling Language (UML) diagrams are widely used in software engineering to visualize software components, they are also used for visualize embedded hardware systems [12] .

In order to deal with complexity and increase usability for large UML class diagrams layering method have been suggested in [13] . According to results of layering approach, applying layering method increased understandability and usability of complex diagrams.

This layering approach is identical to Adobe’s image processing software Photoshop [14]. In the layering method, each layer encapsulates similar type of items in the diagram and enables to create different abstraction levels of the diagram. With checkboxes, users can show or hide different layers according to abstraction level they desire. 2.4 shows how layering method applied on UML diagram.

Figure 2.4: Layered UML Diagram Example [13]

Since layering method is a very new approach and published only a year ago (2018) while this thesis being written, it is applied on only an example data not tested with real large complex systems.

2.4

Usability

To be able to answer my second research question properly, and measure the usability I believe definition of usability is required at this point. Usability may have different meanings in different

CHAPTER 2. THEORETICAL BACKGROUND

domains. According to ISO 9241-210 [15] definition, usability in Human-centered design for interactive systems, within a product or tool users should achieve their goal in an efficient and satisfactory way. Nielsen Norman Group [16], sees usability as a quality attribute. Then they define 5 quality attributes within the usability topic: Learnability, Efficiency, Memorability, Errors and Satisfaction.

• Learnability: How easy is it for users to accomplish basic tasks the first time they en-counter the design?

• Efficiency: Once users have learned the design, how quickly can they perform tasks?

• Memorability: When users return to the design after a period of not using it, how easily can they re-establish proficiency?

• Errors: How many errors do users make, how severe are these errors, and how easily can they recover from the errors?

• Satisfaction: How pleasant is it to use the design?

2.5

Logical Function Architecture At Scania and SESAMM Tool

In Scania vehicles all ECUs are connected to each other via a network called CAN bus. CAN bus was invented by Robert Bosch GmbH in 1980s, and became dominant network protocol in the automotive industry [3] . Currently CAN bus have 4 different segments in Scania vehicles, red, orange, yellow and green. ECU connections are made via these segments and ECUs are communicating with each other by messages. That means an ECU sends an object to another ECU, this object carrying some data to be processed. Figure 2.5 shows an example communication between two ECUs. Each message object can belong to only one segment, either red, orange, yellow or green. They represent prioritization levels respectively. Color of the edges (links) in the diagrams represents CAN segments.

2.5. LOGICAL FUNCTION ARCHITECTURE AT SCANIA AND SESAMM TOOL

ECU 1

_{ECU 2}

 

Segment

FV

signal1

signal2

signal3

signal4

AE 1

AE 1

AE 2

AE 2

FV

FV

     

FV:

Transferred object that contains data.

AE:

Software module within ecu

Figure 2.5: LFA Diagram Example

ECUs are physical units and their connection made by cables. However, each ECU have different software modules inside to process data. These modules called Allocation Elements (AE) at Scania. That means LFA diagrams consist of both physical (Hardware) and logical (Software) units. These diagrams used for developing new functionalities (or change them) or analyzing diagnostics for Scania vehicles.

In diagrams each transferred object called Functional Variable (FV). FVs can be CAN mes-sages, Sensors, Direct wires or internal messages inside the ECUs. That means they can be both physical and logical elements. Sensors, direct wires and internal messages do not belong to any CAN segment. Therefore they do not have any color specification, and their edges are colored as black in the diagrams.

To sum up LFA diagrams at Scania made of three main elements:

• Electronic Control Units (ECU)

• Allocation Elements (AE)

CHAPTER 2. THEORETICAL BACKGROUND

AEs can be described as a software pieces inside the ECUs and they communicate with other ECUs by sending or receiving FVs. FVs can mean the carried signal, carried message, sensors. There is no exact definition for FVs as they vary on the different diagrams.

To draw and analyze LFA diagrams, Scania engineers are using SESAMM Tool developed by EPID team at Scania [6] .

SESAMM Tool used by three user groups.

• System Architects: Engineers who is creating diagrams and creating electrical architec-ture.

• System Owners: Engineers who is responsible for adding new functions to Scania vehicles and maintaining them.

• System Owners: Engineers who is responsible for each specific ECU.

Those three user groups are working together, if a new function needs to be added, function owners are designing a function and communicating with system owners and system architects. System owners are responsible for adapting their ECUs to a new function, System architects are responsible for creating electrical architecture of the new function and deciding which CAN segments, signals, messages shall be used and which elements are required to realize a function.

Until now, system architects at Scania were using SESAMM Tool to draw diagrams by manually and not following any of the aesthetic rules in diagram drawing. System owners and function owners were reading information inside the these manually drawn diagrams. With increasing complexity of LFA diagrams, EPID team at Scania decided to develop new version of SESAMM Tool that draw diagrams automatically to save time for diagram drawing and make diagrams more usable by system owners and function owners.

In the new version of SESAMM Tool, LFA diagrams drawn automatically by using existing data. This automatically generated diagrams are prioritizing the minimizing edge crossing and edge bending aesthetic properties. However, due to number of the components in the diagrams, diagrams became very large and users complaining about usability of the diagrams.

The LFA diagrams are called User Function (UF) diagrams in SESAMM Tool. According to Scania‘s own description a User Function described as:

"All functionality in the electrical system shall, at its highest level, be defined by a User Function (UF). A User Function is defined as functionality that is useful to the driver, entails the electrical system in some way and that either can be influenced by direct input, or is noticeable to the driver (by indications or vehicle behaviour)" -Scania Lexicon

Below, two figures shows how user function diagrams drawn manually in SESAMM Tool 1, and in SESAMM Tool 2 automatically. These two figures represents two different variants of

2.5. LOGICAL FUNCTION ARCHITECTURE AT SCANIA AND SESAMM TOOL

the same user function. That means its the same user function but belongs to different truck model/variant so items in the diagram may vary, according to truck design. This user function is showing how fuel level is being displayed to the driver of the truck.

Figure 2.6: Manually Drawn Diagram in SESAMM Tool 1

CHAPTER 2. THEORETICAL BACKGROUND

Even though, showing fuel level is may sound like a very simple function it requires more than one ECUs and AEs. Some UF diagrams in SESAMM Tool may consists more than 1000 nodes.

2.6

Summarizing Chapter

In this chapter, diagram drawing rules have been explained by mentioning different layouts and aesthetic properties. SESAMM Tool diagrams at Scania are also explained to make readers to understand what kind of data this thesis going to work with. In addition, definition of usability is explained to make clear what I mean with the term called usability in my second research question.

C

H A P T E R

3

M

ETHOD AND

I

MPLEMENTATION

T

his chapter gives information about which research method is applied and gives expla-nation about it. Then how data will be collected, analyzed and how the artefact will be created during this research.

3.1

Design Science

Design science studies suitable for designing an artefact and investigating them in their con-text [17] . In this paper the current issues with Scania‘s logical architecture diagrams will be investigated. Due to size of the diagrams and many connections between software and hardware components causes some usability issues and reading information from diagrams gets harder according to users.

This thesis work, will design a prototype for Scania and then evaluate this artefact with the users, if the artefact provides higher usability or not. SESAMM Tool 2 diagrams, will be the example data set for designing the prototype in this paper. Therefore stakeholders will be the Scania employees.

In the context of this problem, human interaction is involved. Therefore, while designing and evaluating the artefact usability guidelines will be used from literature. According to [17], research goal of the design studies described as evaluation of the implementation. This imple-mentation means in this thesis work an example solution for the problem domain. Design science studies focuses on improving problematic situations. To apply a design science methodology a real-world situation is required as literature mentions.

In design science methodology five activities are presented [18], explicate problem, define requirements, design and develop artefact, demonstrate artefact and evaluate artefact. This

CHAPTER 3. METHOD AND IMPLEMENTATION

thesis will follow these steps to apply design science methodology for creating an artefact at Scania. Figure 3.1 illustrates how these activities take input from each other and then proceeding the next activity.

In this thesis, I have a practical problem that Scania engineers are experiencing, and produc-ing artefacts is the main goal for the design science studies. Design science methodology is very similar to software development process in the IT industry.

Figure 3.1: Design Science Activities

Producing artefact will be the most crucial thing in this thesis due to answering my research questions are directly linked to artefact development and evaluation. First three steps as shown in the figure 3.1, is related to my first research question. Without proceeding those steps artefact creation is not possible, that means answering my second research question would not be possible. After the evaluation of the artefact, my second research question will be answered. As I am looking for usability improvement in my artefact, the goal of the artefact is to produce the knowledge about how can diagrams represented in a more usable form. This is an expected output for answering my second research question due to how I formulated this research question. However output of the artefact can be negative as well. In that case, how I created the artefact can be a valuable knowledge of how not to represent diagrams in a more usable form.

As a first step, problems will be identified via interviews with Scania employees within current diagrammatic representation. In parallel, literature review will be done related to this domain. However reading information from very large diagrams are not specific problem to Scania or automotive industry. Therefore, it enables me to explore more information in different domains as well, but of course Scania have its own specific requirements which will affect the

3.1. DESIGN SCIENCE

artefact creation.

After understanding the problems of users, requirement analysis will be performed to under-stand what my artefact is supposed to do. These requirements will be mostly Scania or automotive industry specific needs due to data type that my prototype will work with.

Then artefact will be developed, in this paper the artefact will be a prototype that going to use real-world data from Scania‘s SESAMM tool.

After creation of the prototype, it will be tested with users. Users will be system owners and function owners at Scania who are mostly reading information from diagrams. Artefact will be discussed with SESAMM tool developers as well to validate its applicability and if it satisfies the functional requirements. According to feedback from users and developers, previous steps will be performed iteratively.

As a final step after testing, performance of the prototype will be evaluated according to users test score and their feedbacks. Users test score will be measured by time in seconds for the given task.

In Scania problem domain is auto-generated diagrams within the software SESAM Tool 2. SESAMM Tool 2 produces auto-generated diagrams to represent information. These diagrams represents logical architecture of the E/E systems within the Scania trucks and buses. SESAMM Tool 2 aims to increase user satisfaction and diagrams usability by auto-generated diagrams. So far, SESAMM Tool 2 is still under development and only alpha release is available to limited amount of users. Usability of complex auto-generated diagrams and their usability in SESAMM Tool 2 is the main problem in this thesis.

As an example template of design problem given below [17].

• Improve <a problem context>

• by <(re)designing an artifact>

• that satisfies <some requirements>

• in order to <help stakeholders achieve some goals

With following this template this thesis design problem shaped like below according to research questions:

• Improve usability of auto-generated diagrams

• by (re)designing users interaction with diagrams

• that satisfies users requirements

CHAPTER 3. METHOD AND IMPLEMENTATION

3.2

Data Collection

To create an artefact data collection is required in order to identify problems and evaluating the solution. For design studies different 5 data collection methods are presented (Questionnaires, Interviews, Focus Groups, Observations, Documents) [18]. However, in this paper only interviews and documents will be used due to lack of number of users that eliminates other methods.

In this thesis target users will be function owners and system owners at Scania. These two user groups are the ones who are mainly reading information in the diagrams.

Semi structured interviews will be a conducted with target users to understand their needs and revealing problem with the auto-generated diagrams in SESAMM Tool 2. Interviews are helpful to go deep inside the subject and also have high response rate [18]. Even though they might be time consuming sometimes and there is always risk with hearing biased opinions about facts. For the purpose of gaining more idea this trade off can be made. It is important to get deeper understanding of the problem area from the real-world users.

Literature review (documents) will be done in parallel with interviews. There is high possibil-ity to gain different ideas for solution of the problem. Within this problem context academical resources will be the mainly trusted sources. Literature search will be done in graph drawing, information visualization and usability keywords. Different solution applications to similar prob-lem areas may influence this work as an outcome of this data collection method. This method with interviews will be helpful to answer my first research question, as well as will effect to artefact development.

3.3

Data Analysis

With interviews some qualitative data will be gathered from users. Literature reviews from similar studies could influence the creation of an artefact, hence such will be applied in this thesis.

According to [19], qualitative researches focuses on understanding behaviors and subjective opinions of users for a computer systems, artefacts and inventions. Therefore, it is important to pay attention to Scania engineers personal opinions about current diagrammatic representation and how they feel when they are using the SESAMM Tool.

3.3.1 Interview Analysis

After conducting interviews with Scania engineers, interviews will be transcribed to text. Content analysis will be performed over the transcribed text. Aim of content analysis is picking the important keywords or sentences from the text and classify them into categories [18]. Most frequent keywords will be labeled in usability and requirements categories in order to answer

3.3. DATA ANALYSIS

the research questions. This will help to focus on prioritizing levels of the problems according to Scania engineers’ answers.

Denscombe [20], suggests following six steps to perform content analysis. These are:

1. Choose an appropriate sample of texts or images:

From interview responses, I will have a list of answers according to questions. Since answers will not be long paragraphs sampling is not needed.

2. Break the text down into smaller component unit:

According to answers, units might be just a keyword or expressions of users.

3. Develop relevant categories for analysing the data:

Category names will be chosen according to my research questions. Since my first research question is about identifying problems within its context, there will be a category name called problems.

My second research question is about improving usability. To be able to answer this question I need to understand what the system should do, after that I need to figure out how the system should behave. Therefore, there will be functionality and usability categories.

4. Code the units in line with the categories:

The units identified in step two, will be put into categories that I defined in step three. A unit may belong to more than one category.

5. Count the Frequency of the Units for Each Category:

In this step for each category number of unit’s frequency will be calculated.

6. Analyse the text in terms of the frequency of the units and their relationship with other units that occur in the text:

After coding units into categories, it is possible perform sophisticated analysis over the units and their reason why they occur frequently.

3.3.2 Literature Review Analysis

Like interview analysis in this part content analysis will be made according to findings in the literature. Identifying the most frequent keywords or texts will shape the focus of this thesis. According to the findings from literature review, successful applications will be discussed with engineers at Scania.

Based on the collected data and analysis, a new diagram prototype (artefact) will be produced. This prototype will include a real data from one of the user function diagrams in SESAMM Tool and will be tested with the users. Interaction with the new diagram prototype will be offered to

CHAPTER 3. METHOD AND IMPLEMENTATION

users. A prototyping software will be used in this process: Adobe Xd and Balsamiq. Balsamiq is useful tool to create low fidelity mockups in a very short time, however for testing and high fidelity mockups, Adobe Xd will be used. Adobe Xd, is personal choice of the author.

C

H A P T E R

4

F

INDINGS AND

A

NALYSIS

4.1

Interviews

To identify problems within SESAMM tool, semi-structured interviews were held with Scania engineers and EPID team. There are three main user groups who are using SESAMM Tool: System Architects, Function Owners and System Owners.

Several interviews have been made with these three user groups and target user groups have been chosen according to these interviews. System owners and the Function owners are the two target groups using SESAMM Tool for information exploration, while System Architects are responsible for making changes and editing information inside the SESAMM Tool. Therefore, system architects are not considered a target user group.

Each interview was conducted one-by-one with system owners and function owners, at each interview same questions have been asked and interviews were recorded by audio. After answering questions, users have been asked how they use SESAMM Tool and observations have been made while users interacted with the tool.

According to users’ answers and their behaviour with the SESAMM Tool, one use-case has been determined for creating the artefact. Interviews have been held continuously during the artefact development. After a testable artefact was created, it got tested with users. For measuring usability, a task was given to users to execute in both SESAMM Tool 2 and in my artefact.

4.2

Interview Results Overview

Interviews have been made with 10 users. Those users were included in this thesis’ target user group, and these 10 users consisted of 5 system owners and 5 function owners. According to interview results most frequent units (keywords) are identified. In the interviews users used

CHAPTER 4. FINDINGS AND ANALYSIS

some units frequently when answering questions. Filtered the units from the full sentences represented below: • CAN Segments • CAN Messages • CAN Signals • ECU • I use Tree-view • I use diagrams • Search Function • Filtering • Large diagrams • Complexity

• Many things to show

4.2.1 Categorisation

Units placed in categories during content analysis. Category names have been chosen to represent the essence of each of the research questions.

Problems: the problems users complained about during the interviews will be labelled into this category. Functionality: task related answers labelled into this category, units in this category describes what the system should provide to its users.

Usability: difficulties of user’s desired task and how users wants to achieve their tasks will be labelled into this category. Units in this category should present how system represents its functionality currently.

Problems Functionality Usability

Large Diagrams Search Function Tree View

Too much Unrelated Information Filtering Search Function Components are not visible Show CAN Segments Large diagrams Diagrams are hard to follow Show CAN Messages Complexity

Show CAN Signals Many things visualized at once Show ECU Components are not visible Show Visualization Usage of diagrams should be easier

Filtering Table 4.1: Categorisation After Interviews

4.2. INTERVIEW RESULTS OVERVIEW

After analyzing response sentences of these keywords where frequently used by users, it was revealed that users used these keywords when they described their tasks. They explained what they do and how they do it. The interviews were specifically designed to get an answer for what and how questions to get an understanding of the users’ behaviour.

According to answers of the interview participants, most users have similar tasks when they interact with diagrams, these tasks can be summarized as below: Users wants to:

• See how ECUs communicate with each other

• See which CAN Segments are used within a user function

• See what CAN messages and signals are involved in user function

• See CAN messages’ and signals’ details

• See if diagram elements have changed in time

This list reveals basic requirements of what diagrams should provide to users. These are the main requirements users expect from SESAMM tool; however, users complain about ease of use of SESAMM tool and they think it is hard to learn functionalities of the tool.

Majority of the users claimed they are familiar with using UML diagrammatic representation. However, some users are not using the diagrams to read information, especially when they get too large. Instead they choose text-based tree-view representation in SESAMM Tool within user functions. As the diagrams get bigger users tends to use tree-view.

CHAPTER 4. FINDINGS AND ANALYSIS

Figure 4.2: User Function Diagram in SESAMM Tool 2

Users desire to have search and filtering functions within the diagrams instead of zooming in/out and panning over the diagrams.

All participants complained about the steep learning curve of using SESAMM Tool. They also think they spend more time than needed to accomplish their tasks. They want to spend less time for each task.

4.3

Literature Review Overview

As I discussed in Chapter 2, for auto-generated diagrams, minimizing edge bindings and edge crossings are known as the most common aesthetic attributes to have better quality diagram. Also, layering method as mentioned aims to reduce diagram size and increase understandability. Most frequent units we filtered out from literature reviews are:

• Diagram size

• Diagram readability

4.3. LITERATURE REVIEW OVERVIEW

• Edge crossing

• Quality of the diagram

• Diagram layout

Findings in literature review labelled into usability and problems category, since these findings seems more related to how to represent information and what problems are exist in diagram drawing.

4.3.1 Categorisation

Usability Problem Quality of the diagram Diagram Size

Diagram layout Edge bending Diagram readability Edge crossing Table 4.2: Categorisation After Interviews

For generating good diagrams, some aesthetic rules must be satisfied, however applying aesthetic rules affects required area for diagram drawing. So following the aesthetic rules to get good diagrams may not be beneficial for large diagrams, since humans interact with these on computer screens.

C

H A P T E R

5

A

RTEFACT

D

ESIGN AND

D

EVELOPMENT

A

ccording to gathered data from the interviews, basic requirements of my artefact is determined. At this stage I know what the artefact is supposed to do. However, to decide how my artefact should offer its functionalities some brainstorming preceded the development. Categorised units, represented in chapter 4 became the foundation of this artefact. During artefact development, interviews were continuosly conducted and weekly meetings were held with the SESAMM Tool developers. According to interviews with users, searchable diagram elements appeared as a one of the most desired features. While this thesis was written EPID team was developing new version of SESAMM Tool and they were also aware of these requirements from users. SESAMM Tool 2, as today has already implemented search function for diagrams, it gave me the possibility to test search function within the large diagrams.

Figure 5.1: Search Functionality in SESAMM Tool 2

CHAPTER 5. ARTEFACT DESIGN AND DEVELOPMENT

This enables me to test my artefact and SESAMM Tool 2 with users who have no experience of SESAMM Tool 2 and my artefact. Therefore, I have a chance to test usability of both interfaces. Usability defined by 5 quality attributes according to NN Group [16].

At brainstorming phase author focused on edge bending and edge crossings. In node-link diagrams associations between components are made via edges, as figure 5.2 illustrates.

Figure 5.2: Simple User Function Diagram in SESAMM Tool 2

5.1

Influences

I thought if it is possible to remove edges in diagrams, then there will be no edge crossing or edge bending problem at all. To do that one influence I found from chess game. In computers or mobile phones chess games, illustrates where each piece can move around the board. Figure 5.3 illustrates where a knight (a piece from a chess game) can move according to from its position on the chess board.

Figure 5.3: Possible knight movements on the chess board

As figure 5.3 shows, if knight piece is clicked, computer illustrates the squares that where knight can move. I decided to consider this sort of illustration on diagrams to represent connec-tions between components.

5.2. ARTEFACT DEVELOPMENT

Another influence is taken from the study of Harald Störrle [13]. As mentioned in literature review overview, layered diagrams approach had a positive impact on diagram understanding. To decide what possible layers could exist in SESAMM Tool and LFA diagrams, I considered ECUs, AEs and FVs as 3 different layers.

5.2

Artefact Development

I considered large diagrams are not fitting to screen and users must zoom in zoom out or pan by mouse. Even though if I apply the influences that I mentioned in previous subchapter to my artefact, size of the diagrams may not get smaller and may still look complex to users.

To use screen size efficiently, I decided to create different views for each layer, and I divided screen by 4 different views according to layers. This approach came to my mind from CAD and 3D modelling softwares where users can see different dimensions and details in divided screens [21] .

Tx

Rx

ECU

AE

FV

FV

1

2

3

4

Figure 5.4: Artefact Screen with 4 different layered view

In figure 5.4, top left view, which is marked with number 1, represents ECU view, that will show only ECUs in a selected User Function.

In screen number 2, only AEs will be shown.

In Screen number 3 and 4 FVs are represented. FVs are divided by 2 categories that are Rx (Receiving) and Tx (Transmitting).

CHAPTER 5. ARTEFACT DESIGN AND DEVELOPMENT

In Scania‘s electrical architecture all these 3 components can be either a transmitting a signal or receiving a signal or doing both. Since users’ main goal with diagrams is to see which FVs (CAN Message, sensor or direct wire) are used between components, I allocated more space to FVs than ECU and AE layers.

These layers and their location is decided based on use case scenarios from the users. A system owner starts looking at the diagram by locating the ECU first and then which AE elements involved and then checks which FVs are used within those ECUs and AEs.

Briefly, ECUs encapsulates AEs and AEs encapsulates FVs. AE cannot exist without an ECU, FV cannot exist without an AE. They have similar relationships compared to class-subclass relationship in software development.

After putting the idea of a 4 layered view in a mock-up, added the required filtering elements and search function, but without data, skeleton of the artefact was created.

Figure 5.5: Artefact Screen with hexagonal representation

Hexagons were in the mock-up used to visualize ECUs and AEs instead of rectangle drawings. This decision was based on efficiency of the hexagon shape within the given area. Honeycomb style hexagonal representation can be seen in nature as well, and efficiency of the hexagons proven in 2001, by T.C Hales [22]. Idea of this skeleton was combining both diagrammatic representation and text-based view in single screen. Text-based view is identical to tree-view in SESAMM tool but only limited to FV elements (See figure 4.1 for tree-view of SESAMM tool).

ECUs and AEs are connected to each other, but FVs are not connected to each other. FVs are just representing how ECUs and AEs are connected. Therefore, FVs are removed from the

5.2. ARTEFACT DEVELOPMENT

diagrammatic representation.

Another reason why FVs represented as a list but not visualized, number of the FV com-ponents is generally greater than ECUs and AEs. I did not want to include them into the diagrammatic representation.

Removing FVs, reduces the required space of the diagram drawing.

Figure 5.6 shows the artefact that is filled with real data from SESAMM Tool. As chessboard example, when user clicks on an ECU on ECU layer, the selected item is highlighted with blue color, connected ECUs highlighted with red, and white color indicates that there is no connection related to selection.

Figure 5.6: Artefact screen highlighted selection in hexagonal view

According to user’s selection of ECU, in the AE view, related AEs are highlighted with blue color. On FV views, list shows which FVs are related to selection. If any item in the list belongs to the specific CAN segment, they are coloured according to their segment color.

Figure 5.6 shows FV elements that belongs to Rx layer. List items coloured with orange indicates that they belong to orange CAN segment. Those are not colored indicates they are not on the CAN segment, which means they are either sensors, direct wires or internal elements.

After putting real data into the artefact, it has been tested with some target users and presentation has been made to EPID team and manager.

CHAPTER 5. ARTEFACT DESIGN AND DEVELOPMENT

In the interviews, since some users claimed that they are familiar with using UML style rectangle diagrammatic representations, hexagonal shapes were criticized by at least half of the users and developers. Users claimed hexagonal style may take time to learn and they feel more comfortable with UML style rectangle box representations. After that feedback, top two layers (ECU and AE layers) of the artefact changed to the rectangle box representations in the second iteration of the artefact development.

In addition, below search bars of ECU and AE layer views, Rx and Tx lists were added to indicate if that component is a Rx unit or Tx unit. For increasing realistic view and giving users the feeling that they can interact with the tool, a side bar and top bar have been added.

Figure 5.7 shows high fidelity mock-up where those changes have been made and interactable artefact has been produced. In this artefact, items are clickable, where users can accomplish the given task. 18 ECU 18 ECU 18 ECU Search  ECU 1 ECU 2 ECU 3 ECU 4 ECU 5 ECU 6 ECU 7 EC U8 ECU 9 ECU 10 EC U 11 ECU1 2 ECU 13 ECU1 4 ECU1 5 ECU1 6 ECU 17 ECU 18 Rx Tx UF485                 Label  Label  Label  Label  Label  Label  ECU1 ECU6 ECU11 ECU2 ECU7 ECU12 ECU3 ECU8 ECU13 ECU4 ECU9 ECU14 ECU5 ECU10 ECU15 ECU16 ECU17 ECU18

Search RED ORANGE YELLOW GREEN SENSOR ALL

All MESSAGES INTERNAL DIRECT WIRES SIGNALS  

Rx RED ORANGE YELLOW GREEN SENSOR ALL Search

All MESSAGES INTERNAL DIRECT WIRES SIGNALS   MESSAGE1  MESSAGE2  MESSAGE3  MESSAGE4  MESSAGE5  MESSAGE6  MESSAGE7  MESSAGE8  MESSAGE9  MESSAGE10  MESSAGE11  MESSAGE12  Search a e1 a e 2 a e3 a e 4 a e 5 a e 6 ae 7 a E8 a e 9 a e 10 ae 11 ae 12 a e 13 ae 14 ae 15 a e1 6 a e 17 a e1 8 a e1 9 a e2 0 ae 21 ae 22 a e2 3 ae 24 ae 25 a e 26 a e2 7 a e 28 a e 29 a e 30 A E 31 Rx Tx AE    

AE1 AE2 AE3 AE4 AE5 AE6 AE7 AE8 AE9 AE10 AE11 AE12 AE13 AE14 AE15 AE16 AE17 AE18 AE19 AE20 AE21 AE22 AE23 AE24 AE25 AE26 AE28 AE29 AE30 AE31

MESSAGE1  MESSAGE2  MESSAGE3  MESSAGE4  MESSAGE5  MESSAGE6  MESSAGE7  MESSAGE8  MESSAGE9  MESSAGE10  MESSAGE11  MESSAGE12  Tx    

Figure 5.7: Artefact after 2nd iteration

5.3

Artefact Evaluation

To measure usability NN Group [23] suggests to measure these attributes:

• success rate (whether users can perform the task at all),

• the time a task requires,

5.3. ARTEFACT EVALUATION

• users’ subjective satisfaction.

I followed this guideline to measure usability of my artefact. After creation of high-fidelity mock-up, artefact was tested with users. For testing I created a task list to apply in both my artefact and auto-generated diagrams in SESAMM Tool 2.

Task list was created according to answers of users in interviews. Findings show that revealing CAN messages (including CAN signals) in the diagram is the most crucial information for the users.

The task was based on real data from SESAMM Tool 2, I exported the same data to my artefact. In the task I picked one of the largest diagrams in SESAMM Tool 2, that is UF485 (User Function 485 -Advanced Emergency Braking, see original diagram in appendix section).

UF485 diagram consist of 18 ECUs, 129 AEs, 5 CAN Segments, 286 Signals, 72 Messages. Before starting the test, I introduced SESAMM Tool 2 features to users. E.g. how to apply filter, how to search within diagram. I did same introduction to my artefact as well.

After the introduction, I gave tasks to the user in written format and explained verbally what users were supposed to do.

Task list derived from a possible use case scenario from system owners and function owners interview answers.

Task list:

1. Find ECU8 component (ECU8 is an ECU name)

2. Find which FVs ECU8 receives and transmits and show which CAN segment is being used

3. Find which components (ECU) ECU8 is interacts with

4. Find AE17 and show which FVs is used

5. Find MESSAGE9

6. Find detailed information about MESSAGE9.SUBMESSAGE4.PROPERTIES

Users interacted with real data in Scania‘s E/E architecture, due to confidential agreement with Scania, in this paper task list is replaced with representative data names.

Task list indicates a single use case scenario from daily work of users. Therefore, whole steps count as a single task.

I run tests with 10 different users and these are the same users that I interviewed: 5 function owners, 5 system owners. Each test was performed by individually in quiet meeting rooms where users did not have any external interruptions. I ran stopwatch when users started doing task and I stopped the timer at the end of the task list. Each user was informed about their right to stop doing the task if they wished.

CHAPTER 5. ARTEFACT DESIGN AND DEVELOPMENT

To measure success NN Group [23] suggests to compare task completion times, as in "how long does it take users to complete the task" for both designs. In the tests, my artefact performed better than auto-generated diagrams in SESAMM Tool 2. For the same task, users spent less time to finish in the artefact. Success rate was calculated by comparing average task completion times. Figure 5.8 shows the task completion times for both interfaces, in minutes.

Figure 5.8: User Scores

Average task completion times:

• SESAMM Tool 2: 341 seconds

• Artefact: 80 seconds

• Success rate 426.25%

In the artefact fastest user was able to finish the task in 56 seconds and slowest user was able to finish in 1 minute 52 seconds respectively. All users finished the task.

In auto-generated diagrams fastest user was able to finish the task in 3 minutes and 18 seconds while slowest user was able to finish in 9 minutes and 36 seconds. One user gave up doing the tasks after certain amount of time, that is not included in the average time.

After the test users got asked for both systems about their subjective opinion on user satisfac-tion. According to users’ opinions artefact has higher user satisfaction (See Appendix C).

C

H A P T E R

6

D

ISCUSSION AND

C

ONCLUSIONS

6.1

Discussion of Method

Design science was chosen as method in this thesis due to implementation requirements from Scania. Design science allowed me to brainstorm in this thesis. I had chance to test my idea with users and improved my artefact iteratively. Due to limited amount of time the artefact had only two iterations. Before and after artefact development I contacted with users, and I wanted to make my artefact as innovative as possible.

As mentioned in [18], design science aims study artefacts and create them. By saying that, Design science was a good methodology to use in this thesis.

According to Gregor and Hevner [24], validity refers to the artefacts working or doing what they are supposed to do.

[25], explains the threats for validity of the design science studies in information systems as artefact instantiation space, artefact medium and distance, artefact complexity, artefact cost and technological progress. For the present study, the complexity of artefacts was the greatest threat. I handled this threat by putting extra effort into understanding the current system before the production of an artefact. [25], also argues that the data selection by the researcher in complex information systems is one of the biggest threat to validate the artefact and that using equivalent data is one way to deal with this threat. Following the recommendations of these authors, I used real data to eliminate that risk. Instead of showing only important parts or some parts of the system in my artefact, during the test phase, the users in the current study faced with the same data in both existing system and the artefact.

In the content analysis, since I am the only person who is writing this paper. I did coding by myself to create categories from the interview results. In [26], authors says that there is a risk

CHAPTER 6. DISCUSSION AND CONCLUSIONS

of subjective opinion of a person who is doing coding during the content analysis. They suggest, more than one person to do coding to eliminate this. To reduce this risk, I checked previous thesis works at Scania which are conducted about same tool and in the same team. I decided which units to code in my categories during the content analysis by using those thesis datas and also talking with developers of the tool. Frequency of the words that I highlighted for categorization were similar in the previous studies as well. That way I reduced my subjective opinion during the content analysis. To increase reliability, conducting this thesis work with one another person or more people would be beneficial.

Since the complexity of visualization of the LFA is an industry problem, my interviews and artefact tests were limited to Scania engineers. I have tried to contact with other manufacturers and try to book an interview, unfortunately response rate was very low and I was able to do only one interview from another manufacturer. I have interviewed with Markus Kühl who is a E/E architect from Porche AG, and he was able to confirm there is no existing solution for them to make E/E diagrams more usable, (they are also using node-link diagrams for their LFA) and this is valid for Volkswagen Group [27] (Volkswagen group have ownership in many automotive brands including: Scania, MAN, Porche, Audi, Skoda, Lamborghini, Bentley, Seat etc. [28] ). After my interview with Markus Kühl, it is safe to say rules of the LFA are not different to another manufacturer and they have to design their system by following the same rules. This input was important for the artefact design and validity of the artefact.

At the end, creating an artefact helped me to answer my research questions in this paper. I had a chance to test my concept idea and compare with existing E/E architecture visualization of the Scania buses and trucks.

Content analysis has helped me to understand main points that I should be aware of when developing the artefact. This helped to answer both my how and what research questions, as well as what my artefact is supposed to do.

6.2

Discussion of Findings

6.2.1 RQ1: What are the problems when users interact with complex logical function architecture diagrams in the automotive industry?

I tried to answer my first research question during the first two stages of design science method-ology and also with literature review. I found similarities between the literature review and my interviews with users about complex diagrams. Literature review, expanded the scale of the problem beyond Scania and helped me to see main issues with node-link diagrams. Output from this question lead me to answer my second research question.

Size of the diagram has impact on diagram readability and traceability. With the case at Scania, I found large diagrams have negative impact on user satisfaction as well. By saying that I discovered users does not want to interact with diagrams just with panning and zooming for

6.2. DISCUSSION OF FINDINGS

information exploration. In most cases they need to read some parts of the diagram instead of the whole diagram. This means they want to see reduced size of the diagrams according to their needs. However, even with filtering feature the reduced diagram might still be too complex to users. In these cases, required amount of time for a specific task is considered very high by users. They believe they should be able to achieve their goals within few minutes without contacting another person.

Some users must interact with diagrams every day while other users do the same once a week or month. I considered these two user groups as: experienced users and not experienced users. During my tests there was a noticeable time difference between these two user groups for solving the given task.

6.2.2 RQ2: How can the usability of logical function architecture diagrams in E/E systems can be improved?

Within all stages of chosen method, I put an effort to answer my second research question. In literature I found that diagram size is a common problem for auto-generated diagrams, in addition to that there are many factors affecting diagram size. Minimizing edge bending and edge crossing increases the readability and traceability of the diagrams. However, for large diagrams users still have problems with reading and tracing the information. Another issue with auto-generated diagrams is adding new elements to diagram, changing the diagram elements position due to new edge bendings and crossings inside the diagram. Therefore, satisfying aesthetic rules is alone not enough for increasing user satisfaction and usability of auto-generated diagrams. Searching and filtering feature within the diagram has more impact on user satisfaction than having aesthetic rule-based diagrams.

I tried to improve usability of large diagrams by creating a compact, layered view without edges in my artefact. Evaluation of my artefact proves that a compact view is much more useful than a larger diagram.

In my artefact I showed that a layered view makes it possible to reduce diagram size signifi-cantly and it increases the usability of the diagrams.