Cost Efficient Automobile System Architecture

Transcription

1 Cost Efficient Automobile System Architecture Erik Fröstad Jakob Tivell Master of Science Thesis INDEK 2008:93 KTH Industrial Engineering and Management Industrial Management SE STOCKHOLM

2 Examensarbete INDEK 2008:93 Kostnadseffektiv bilelektronik Godkänt Examinator Thomas Sandberg Erik Fröstad Jakob Tivell Handledare Thomas Sandberg Uppdragsgivare CEDES Kontaktperson Håkan Edler Sammanfattning CEDES är ett forskningsprojekt inom IVSS (Intelligent Vehicle Safety Systems) där forskning bedrivs inom kostnadseffektiv och tillförlitlig bilelektronik. Mängden elektronik och programvara ökar kraftigt inom bilindustrin. Detta examensarbete syftar till att belysa de kostnadsaspekter som kan skönjas av ny elarkitektur. Vägledande frågor för studien har varit, Vad är kostnadseffektivitet? Vilka kostnadsdrivande faktorer finns? Hur kan en modell för kvalitativ bedömning av kostnadseffektivitet konstrueras? För att besvara frågorna har en litteraturstudie genomförts inom områdena systemarkitektur och kostnader. Baserat på litteraturstudien utvecklades en modell för utvärdering av kostnadseffektivitet. Modellen bygger på Cost Benefit Analys och kan användas vid utvärdering av kostnadseffektivitet av elarkitektur inom bilindustrin. Modellen belyser kostnadsaspekter inom alla steg av en bils livscykel, från utveckling och inköp, till produktion och försäljning, samt eftermarknad och samhällseffekter. För att testa och verifiera modellen har ett antal intervjuer genomförts i den svenska bilindustrin. Modellen har även applicerats på det specifika fallet brake-by-wire. Resultatet visar att elarkitektur baserad på programvara och elektronik är en nödvändighet för att klara av de krav på funktionalitet som bilköpare ställer. Nyckelord för kostnadseffektivitet är: standardisering, modularisering, flexibla plattformar, och långsiktighet.

3 Master of Science Thesis INDEK 2008:93 Cost Efficient Automobile System Architecture Approved Examiner Erik Fröstad Jakob Tivell Supervisor Thomas Sandberg Commissioner CEDES Thomas Sandberg Contact person Håkan Edler Abstract CEDES is a research project within IVSS (Intelligent Vehicle Safety System) focusing on cost efficient and dependable electronics system. The use of electronics and software is increasing rapidly in the automotive industry. This thesis aims to highlight the cost aspects in relation to automobile electrical architecture. Guiding questions throughout the study have been: What is cost efficiency? What cost drivers exist? How can a model for qualitative assessment of cost efficiency be constructed? To answer these questions a literature study was conducted in the fields of system architecture and costs. As a result from the literature study a model for evaluating cost efficiency was developed. The model is based on Cost Benefit Analysis and can be used for evaluating cost efficiency of automobile electrical architecture in general. The model includes cost aspects from all steps of the life cycle of an automobile, from development and sourcing, through production and sales, to aftermarket and externalities. To test and verify the model a number of interviews were conducted in the Swedish automotive industry. The model has also been applied on the specific case of brake-by-wire. The results show that electrical architecture based on software and electronics is necessary to meet the customer demand for functionality. Key words for cost efficiency are: standardization, modularisation, flexible electrical platforms, and long-term perspective.

4 i Outline of the report 1 Introduction This opening chapter introduces the problem with background, gives some information about the initiator of the study, and describes the scope of the study with delimitations. 2 Methodology In this chapter the methods used throughout the study are introduced. The chapter gives answer to questions such as, how respondents were selected and why, and how validity and reliability are ensured. It also presents the relationship between the theoretical framework, the model, and the empirical observations. 3 Model A key part of this thesis is the development of a model for qualitative evaluation of cost efficiency for automobile system architecture. This chapter presents the model, how it is used, and gives rationale to why it is reasonable to evaluate cost efficiency this way. 4 Theoretical framework This chapter introduces theories and terms that are used in the thesis. The presentation is to be regarded as a theoretical framework that supports the study. Theory is presented on system architecture, cost and cost analysis models, and gives an introduction to system architecture and cost in automobile applications. Readers who are already familiar with the areas may wish to skip this chapter. 5 Empirical observations This chapter presents findings from the empirical observations. The material was collected, and is presented according to the structure of the model presented in chapter 3. Some further discussion on issues raised in this chapter can be found in chapter 7. 6 Applying the model on a brake-by-wire system This chapter applies the model on a concrete example, brake-by-wire. Following the structure of the model, answers to each area are presented. The presentation is in line with how the model is to be used in a real world example. 7 Discussion on findings When evaluating cost efficiency it is healthy to reflect on some of the basic assumptions that the evaluation relies on. In this chapter some topics encountered during the empirical study are discussed. The discussion is to be seen as a complement to the empirical observation and the conclusion. 8 Conclusion This final chapter presents the conclusions drawn from applying the model on the Swedish automobile industry. It also presents some proposals for future research.

5 ii Vocabulary ABC Activity Based Costing ABM Activity Based Management Application A specific functionality. Not necessarily a separate system CAN Controller-Area Network CBA Cost Benefit Analysis CEDES Cost Efficient Dependable Electronic Systems Component The smallest parts of a system, could be either computer software, hardware or mechanical parts CVP Cost-Volume-Profit ECU Electronic Control Unit E/E system Electrical/Electronic system ESP Electronic Stability Program HMI Human Machine Interface IVSS Intelligent Vehicle Safety Systems LIN Local Interconnect Network Module Hardware and computer software that aims to fulfill some functionality MOST Media Oriented Systems Transport OEM Original Equipment Manufacturer, for example Volvo Car Corporation

6 iii Table of contents Outline of the report... i Vocabulary... iii 1 Introduction Problem Background CEDES Aim of the thesis Scope Research questions Delimitations Methodology What is scientific method? Relation between theory and empirical observations Collecting information Validity and reliability Designing a model Research design Model The model Rationale Applicability Theoretical framework System architecture Basic concept of automobile system architecture Dysfunction Modularization Cost Introduction to cost Efficiency Cost analysis models Cost and automotive system architecture... 21

7 iv 4.3 Automobile applications Automotive challenges Safety functions Brake-by-Wire Brake-by-Wire and dysfunction Repairs and recalls Infotainment Externalities Road safety Empirical observations Development What drives technology development? Effects of computer software Trends Standardization Connection to the model Sourcing Separating hardware from computer software effects of standardization Price development Use of materials Connection to the model Production Software download Other production aspects Connection to the model Pricing Connection to the model Aftermarket Repair shops Components Connection to the model Externalities... 49

8 v Connection to the model Applying the model on a brake-by-wire system Discussion Discussion of observations Discussion of model aspects Conclusion and future research Conclusion Future research References Published sources Interviews Appendix A, Questions used in interviews... A

9 1 1 Introduction This opening chapter introduces the problem with background, gives some information about the initiator of the study, and describes the scope of the study with delimitations. 1.1 Problem In the aviation industry, safety-critical applications are often based on software, but with a hardware-based backup. Due to heavy requirements on cost efficiency, such setups are not possible in the automobile industry. That may seem straightforward, but what are the factors that affect cost related to automobile electrical architecture? 1.2 Background Until recently the research project that initiated this thesis, CEDES, had assumed software to be more cost efficient than hardware, and focused on technology development instead of studying cost drivers related to electrical architecture. A previous thesis (Hoff & Sandberg, 2007) at Chalmers University of Technology started to investigate cost drivers in development and production of automobile electronics. That study left questions yet to be answered, which is why CEDES now seeks the answer to how automobile electrical architecture affects cost efficiency over the life cycle of the product. Some previous research has been done in this field (Fröberg, 2004; Larses, 2005). These doctoral theses discuss aspects of electrical architecture and as a part of the general discussion also studied costs related to electrical architecture. They give general guidelines on how costs are affected by different aspects of an electrical architecture, but the topic is not examined in detail. As automobiles are equipped with more safety-critical E/E systems, the demand on dependability increase, but costs are not allowed to increase. To supply these needs, dependable, cost efficient system architecture is required. The result of this thesis is expected to be used as input for future research in automobile system architecture. There are several reasons as to why it is interesting to study the cost of electrical architecture throughout the entire life cycle of the product. As volumes increase, the development cost of software measured by per unit diminishes. For a global OEM with many different models using the same software this statement may be true, but for a small OEM it may very well not be. A second issue stems from automobile quality and experienced quality. Previous attempts to move the frontier of automobile safety applications has had its drawbacks. One such example is Mercedes which launched by-wire braking systems and two years later had to recall a large number of vehicles, imposing the company large costs and indirect damage. Time spent on development may well become cost saved in a later phase. The point is that all phases of the life cycle of the product affect each other, and that similar factors may have influence throughout the life of an automobile. 1.3 CEDES The Vision Zero program is the basis for all road safety work in Sweden. It was initiated by the Swedish Road Administration and accepted by the Swedish Parliament in Vision Zero

10 2 states that fatalities and severe injuries in traffic accidents are unacceptable. The program has three areas of attention, safe roads, safe automobiles, and safe drivers. One way to improve the overall road safety is to equip automobiles with functionality that assists the driver in emergency situations. This is the foundation of the CEDES research program. CEDES is short for Cost Efficient Dependable Electronic Systems. As the name implies, CEDES aims to develop techniques and methods to design and build cost efficient, yet dependable, automobile electronic systems that enable safety-critical functionality. The project is a cooperation between industry and academia. The partners are Volvo Car Corporation, Volvo AB, Autoliv Electronics AB, Chalmers University of Technology, and SP Technical Research Institute of Sweden. The aim of the program is to study cost efficient technology that enables safetycritical functionality in vehicles, but also to educate and graduate a number of technical PhD students. 1.4 Aim of the thesis To investigate how safety-critical automobile electrical architecture affects cost efficiency throughout the whole life cycle of a product. 1.5 Scope System architecture and electrical architecture are sometimes used as interchangeable terms. In the context of this thesis the term system architecture refers to automobile electrical architecture. Since CEDES is the initiator of the study, the focus of the thesis is to examine how safety-critical electrical architecture affects costs. With a different questioner, the study could have included other fields of interest such as further evaluating how fuel consumption and user economics are affected by electrical architecture. The prevailing trend in the automotive industry is functionality based on software rather than mechanics. There are two contributors of this trend. One is replacing existing mechanical functions with software and electronics. The other is increasing the functionality of the vehicle by using software and electronics. Evaluating how the shift towards the use of software and electronics affects costs is the foundation of this thesis. As an attempt to concretize, brake-by-wire is evaluated as an enacting example. Brake-by-wire is of special interest since it represents the shift from mechanics and hydraulics to electronics and computer software, but also because the technology has been released in consumer vehicles, caused a large amount of recalls, and then been replaced by traditional technology. Despite previous failures on the automobile market, brake-by-wire will be introduced on the 2009 Honda CBR600R super sport motorcycle. The dream scenario is a report that gives guidance on how to build electronic systems that lower costs. Results from the study will be used as input to future research within the field of automobile electrical architecture.

11 3 1.6 Research questions The aim and scope of the thesis have resulted in three research questions. - What is cost efficient automobile electrical architecture? - How can a model for qualitative evaluation of cost efficiency be formulated? - What cost drivers can be seen throughout the life cycle of the electrical architecture of an automobile? 1.7 Delimitations The report aims to give a broad perspective on electrical architecture and cost efficiency using the following delimitations. In the automobile industry, production costs are prognosticated over large time spans why alternative hardware based systems need not be studied in detail. However, the model presented could potentially be used for evaluating cost efficiency related to mechanical architecture. Choosing level of abstraction of what is studied represents a clear delimitation. Electrical architecture can be studied in detail as copper wire, or it can be studied in more general terms as effects of standardization and modularization. System engineers at OEMs optimize systems in amount of wiring and use of material, but for the purpose of this thesis a high level of abstraction is chosen. This since the aim scope is to evaluate cost efficiency in a broad perspective rather than in detail, but also since the results are to be used as input for future research on systems that do not yet exist.

12 4 2 Methodology In this chapter the methodology of the thesis is presented. The chapter gives answers to questions such as, how are respondents selected and why, how is validity and reliability ensured, and what methods were used during the study. 2.1 What is scientific method? The empirical part of the research relies on a foundation of systematic qualitative interviews with different stakeholders of the automotive industry, together with studies of technical documentation. Systematic empiricism is the foundation of scientific studies (Eriksson & Wiedersheim-Paul, 2006). ). The aim of the study is to create a model for evaluation of cost efficiency that is valid not only for one automobile electrical architecture but for automobile electrical architectures in general. This is what separates investigation from scientific investigation (Eriksson & Wiedersheim-Paul, 2006). On the other hand the usability of the model is verified through interviews with the affected stakeholders, which is more typical for an investigation. 2.2 Relation between theory and empirical observations Observing an event and assuming that event to be true for all similar situations is called inductive conclusion (Gustavsson, 2002). This usually leads to attempts to verify what is believed in, rather than to question it. Scientific knowledge is dependent on continuous questioning of what is known (Gustavsson, 2002). The opposite of inductive conclusion is deductive conclusion. That is, testing a hypothesis on specific data, and concluding from what is observed (Gustavsson, 2002). A deductive study starts with a general hypothesis and tests it on the specific case (Gustavsson, 2002). In this thesis an initial literature review resulted in a model that represents the general hypothesis that is tested on an example. The approach has been to refine and extend previous theory, create a model, and to test that model through interviews. Testing a hypothesis on data is typical for a deductive approach. The theoretical framework presented in chapter 4 is to be considered as an introduction to the fields of system architecture, cost, cost analysis, and system architecture in the context of the automotive industry. Conducting the literature review was an enabler to the first step of the thesis, to develop a model for evaluation of cost efficiency. The model is an extension of the findings of Larses (Presented in section (Larses, 2005)). After the model was developed the second step was to test the model on real world data. As a result of this procedure the empirical observations are structured using the approach of the model. The theoretical framework acts to support the model and the empirical study, but is not structured in the same way. Thus, the model and the empirical observations are connected through structure, and the theoretical framework supports the understanding of the two. 2.3 Collecting information Traditionally, social sciences conduct qualitative interviews in order to understand human behavior. The general approach in scientific studies is to do quantitative studies (Gustavsson,

13 5 2002). Evaluating cost efficiency related to system architecture can be seen as somewhat of a mix of the two areas. The selection of methodology falls on qualitative interviews to give input to the analysis that will validate or falsify the specified model. A quantitative approach could have been selected, but since the topic and answers were considered complex, a qualitative approach was chosen. Another reason for choosing qualitative onsite interviews was that the answers were expected to result in new questions (Eriksson & Wiedersheim-Paul, 2006). Thus a semi-structured approach was taken with questions that were adapted to the progress of the interview. Before the interviews, the questions were carefully evaluated to prevent them from being leading. The interviews were conducted in groups, and with individuals. There were always two interviewers present, who both took notes. After the interviews the notes were compared, and the respondents were asked to verify that the notes were a correct illustration of their answers, so that the effect of misinterpretations would be reduced. Despite the fact that the study used qualitative research methods, the aim of the study has been rather technical. The main concern for individual integrity has been to ensure that the source of sensitive information is not revealed. In general, the automobile industry keeps most of its business secret. Due to this way of working, many of the respondents clearly stated that they wished not be referred to in specific. In order to accommodate for the integrity of the respondents, some facts will not be referred to in specific. Instead of referring in detail some findings will be described as general observations from discussions with automotive industry parties. For the interviews, respondents were chosen to represent the entire life cycle of system architecture. The respondents work in the fields of, system development, software development, sourcing, production, product planning, aftermarket sourcing, aftermarket service, brake system development, OEM system supply, and economic aspects of road safety. As the aim was to test the model, respondents from different competing companies were chosen. Organizations that participated were Denso, Mecel, Mercedes-Benz Sweden, SAAB Automobile, Scania, Volvo Car Corporation, and VTI. Before the empirical study took place, a literature review was conducted. The result is to be found in the chapter theoretical framework. Literature was selected to cover the fields of system architecture, cost and cost models, and to introduce system architecture in the field of automobiles. The aim of the literature study was to give a proper understanding of the mentioned areas. In cases where the literature did not give a comprehensive understanding it was complemented with findings from interviews. 2.4 Validity and reliability How well a measurement is able to measure what it is supposed to measure is called validity. If efficiency is studied, the result should indicate the level of efficiency (Gustavsson, 2002). In this thesis, validity corresponds to giving an indication of what cost efficient system architecture in automobiles is. During the interviews the respondents were asked to give their concrete opinion on what the definition of cost efficiency is. The results were used to interpret and calibrate the

14 6 answers. The fact that many different respondents gave the same answers is an indication of validity. The findings in this thesis are also clearly in line with earlier findings (Larses, 2005). Can the results be reconstructed? Would other researchers have come to the same conclusion if they had used the same approach? Reliability is related to how well measurements give reliable and stable results. Reliability is an issue especially in interpreting research. To ensure the reliability of this thesis, questions and answers from interviews are well documented. The aim has been to meet people in different positions, in different organizations. As discussed in section 2.3 the respondents were chosen with reliability in mind. 2.5 Designing a model A model is a map of the actual landscape. When developing a model there are different aspects to consider. For example, for a model to have practical use it needs to be robust. A robust model is applicable in different environments and not only the one it was developed for. For this particular reason, the model presented in this thesis is based on Cost Benefit Analysis that allows unequal weighting of factors. A second aspect of a model is precision versus coverage. The level of abstraction chosen as the scope of this thesis is rather high. As a result, the model presented tends towards coverage rather than precision, but as the model is structured, the user who is not helped by the coverage can reduce the model to his or her purpose. As described in chapter 3, the model is based on findings of earlier research (Larses, 2005). It does not contradict existing theory, but rather broadens the perspective e from trucks to automobiles. Thus it is consistent with and complementary to previous findings. 2.6 Research design The research was conducted as seen in the illustration below. Figure 1 - Illustration of the research phases The first step was to start the literature review that is the foundation of the theoretical framework of the thesis. Shortly after the literature review had commenced, interviews were held to familiarize with industry and test some ideas that developed during the literature study. Based on the findings from the literature and the early interviews, the model was specified. The

15 7 model was then tested throughout the remaining interviews. Continuously with the interviews the analysis was conducted. The reader who is interested in learning more of how the interviews were conducted is referred to the appendix where the interview questions are presented. The interview transcripts were used as basis when the chapter on empirical observation was formulated.

16 8 3 Model A key part of this thesis is the development of a model for qualitative evaluation of cost efficiency for automobile system architecture. This chapter describes the model, how it is used, and gives rationale to why it is reasonable to evaluate cost efficiency this way. Please see chapter 4, Theoretical framework, for further understanding of the terms and theories used in this chapter. 3.1 The model The factors that are believed to affect cost efficiency in the context of automobile system architecture are: development cost, sourcing cost, requirements on production, pricing opportunities, aftermarket effects, and externalities (see section 4.2.4). Figure 2 - The model for cost efficiency evaluation When performing the analysis, a central question (or questions) needs to be answered for each factor. The answer to all the central questions will give a basis for evaluating if the architecture in question is cost efficient. Since cost efficiency is a relative term that depends on the user and the purpose, the model can only give a basis for evaluation of cost efficiency, not an answer (see section 4.2.2). It is only the user who can give the answer. The central question for each area is given below. Corresponding to each of those are several other questions that need to be answered in order to find the answer to the central question (see chapter 5). The questions that were used in the empirical part of this thesis are to be found in the appendix. - What is the driving force for implementing system architecture? - How is the use of components and material affected by the system architecture? - How are prices on ingoing components expected to change over time? - How must existing production facilities be adapted to handle the system architecture? - How does the system architecture affect pricing opportunities? - How does the system architecture affect cost in aftermarket service? - What can be said of the effect on the volume of warranty repairs in long and short run?

17 9 - What external effects can be seen through the introduction of the system architecture in question? When all central questions have been answered, the evaluator has the basis for evaluating if a solution is cost efficient. Cost Benefit Analysis as used conventionally, means summing up cost and benefits in real numbers, but since the answers are unlikely to include actual numbers, the summing of factors is a somewhat abstract moment (see section 4.2.3). To perform the analysis, advantages and disadvantages are to be presented jointly as illustrated in the fictive example below. The model allows unequal weighting of factors. If the evaluator is not interested in one or several factors, those factors can be given lower weights. In a Cost Benefit Analysis with real numbers, these weights can be assigned any real number. When performing a qualitative analysis on the other hand, the weights assigned are zero or one. Figure 3 - Example of model application Comparing the result with results from evaluations of alternative architectures will give the answer to which solution is more beneficial, or more cost efficient. Different solutions may be cost efficient, but only one of the alternatives is the most cost efficient. 3.2 Rationale As described in the methodology chapter, the presented model is based on the findings of earlier research (Larses, 2005). The addition of this thesis is to extend and refine the previous findings. Larses states that the factors affecting cost related to truck system architecture are: development, production, maintenance, and availability. The differences between the findings of Larses and the model presented in this thesis are that availability is toned down, price development and sourcing are separated from production, maintenance is called aftermarket, and benefits from introducing the architecture are included. The reason for toning down the effects of availability in the model is that this is an issue primarily for commercial vehicles, not consumer vehicles. Price development is separated from production as this has influence not only on production cost but on other factors as well. The

18 10 same goes for sourcing. Changes required to the production facilities are separated from development as it is judged to be an important factor in itself. As will be seen in later chapters, development costs are large in real numbers but compared to production costs and tool investments they are rather small. The market that is created once an automobile is sold is usually called aftermarket or service market by OEMs. The term aftermarket is chosen instead of maintenance as it more intuitive for OEMs and automobile industry parties. Larses describes factors that drive cost. The model presented in this thesis aims at evaluating cost efficiency. Thus benefits of different options need to be included. Cost Benefit Analysis was chosen as foundation to the model since it allows the model to be adjusted according to the preferences of the evaluator. This is believed to broaden the applicability of the model beyond OEM decision making. The reason for creating a qualitative model instead of a quantitative is that absolute numbers are not always available. The assigned weights are to be seen as qualitative measures and the different factors are to be evaluated in the same way. The central aim of the model is to give an understanding of areas that need to be examined. This is also part of the reason to why many of the alternative cost analysis models were disqualified for the task. Activity Based Costing would have been an appropriate selection if the level of abstraction were lower (see section for more background on cost analysis models). That is, if the purpose of the thesis were to minimize the resources of an electrical architecture the model would have been more appropriate. Activity Based Management has an interesting approach towards cost efficiency, but requires known cash flows. As Cost-Volume- Profit analysis is based on a scenario with both cost and income, it is disqualified as there is no income involved when selecting electrical architecture, or that the revenue created is difficult to derive to the architecture in specific. An alternative approach to a model for evaluating cost efficiency is to study factors like quality, and modularization rather than product life cycle phases. Doing this would however cause a practical implication. In most OEM organizations it is difficult to find a specific person to ask questions about such factors. Many different people would be fit to, but who would be most appropriate? Throughout the stages of the product life cycle there are different organizations with responsibility for each stage. Thus, finding the appropriate people to ask questions is a much easier task. Factors such as modularization and quality are still part of the analysis despite the fact that they are not factors of their own. An important finding of the empirical studies is that a system perspective of the life cycle of an electrical architecture is a factor that has great influence on costs. With a person who is allocated to seeing the life cycle effects of factors such as modularization can save cost in many phases of the life cycle. Modularization for example, does not only cut costs in production, but also on the aftermarket. As the user of the model is expected to evaluate the different steps of the life cycle, this person will have the system perspective. Including externalities as a factor in the model reflects the point of view from which the question is asked. From an OEM perspective, external effects are of interest if they affect sales volume in real terms. For reasons like this, the model is selected so that the weights can be changed for the purpose of the user.

19 11 At the same time, the statistical value of life can be seen as a reflection of people s will to pay for reduced risk and thus be an indicator of pricing opportunities of safety-critical functionality. 3.3 Applicability When applying the model, the evaluator is supposed to answer the questions based on information from an OEM. The questions are designed so that the answers give a general basis for decision. As the answers are collected, it is up to the evaluator to compare the findings from different options and to make a decision based on the comparison. A risk of the model is that ingoing parameters become speculative. To avoid this, the parameters must be collected and selected carefully so that they represent true measures. Another aspect of the model is that it studies avoidable cost related to an option. Unavoidable cost is to be disregarded in the analysis since it will affect all options equally.

20 12 4 Theoretical framework This chapter introduces theories and terms that are used in the thesis. The presentation is to be regarded as a theoretical framework that supports the study. Theory is presented on system architecture, cost and cost analysis models, and also gives an introduction system architecture and cost in automobile applications. Readers who are already familiar with the areas may wish to skip this chapter. Section 4.1 gives an introduction to system architecture in general and to some applications of automobile system architecture. Section 4.2 introduces cost and cost analysis models. Pay special attention to the section on cost analysis as the model presented in his thesis is based on the theory presented there. Section 4.3 discusses some future challenges and practical experience from the automotive industry. In section 4.4 externalities are introduced. This section is somewhat different from other sections as it contains both theory and some analysis. 4.1 System architecture As mentioned in section 1.5, the term system architecture can have different meanings. System architecture can for example refer to mechanical, organizational and electrical structures. For the purposes of this thesis, system architecture refers to the electrical architecture of an automobile; but first comes an introduction to system architecture in general. This section aims to give an introduction to the field of system architecture and how it relates to automobiles. What is a system? The word system is usually used to describe a collection of related components that interact as a whole. A large system is often made up of many smaller systems, which in turn can each be made up of many smaller systems (Denton, 2004 ). Figure 4 - System and subsystems with boundaries System engineering can be described as engineering of an entire system, rather than of its consisting parts. This means designing all at once rather than designing components separately (Fröberg, 2004). As previously mentioned, the term system can take different meanings, for example, some sources include organizational systems such as enterprises to what objects can

21 13 be subject to system engineering (Fröberg, 2004). In automobiles, system architecture usually refers to architecting of the electrical/electronic architecture (Pernstål, 2008). Good system architecture is where design decisions are balanced to meet different stakeholder requirements. For automobiles, the large amount of stakeholders with different demands acts as a driving force towards more complex systems. Customer satisfaction, safety requirements, legislation and large number of suppliers are examples of factors that impose requirements on the electrical architecture (Fröberg, 2004). The latter is something of special interest as the percentage of an automobile contributed by external suppliers could be as much as 50 percent (Pernstål, 2008). Architecture design is the process of negotiating these sometimes conflicting requirements in order to make correct trade-offs, and still fulfilling the business case (Fröberg, 2004). The electrical/electronic system (E/E system) in an automobile is an application of a mechatronic system. Key technologies of mechatronic systems are software, electronics, sensors and actuators. The most common setup is software together with microcontrollers. This setup has given rise to the term embedded systems. Since production in general, and automobile production in specific, is highly pressured by cost efficiency, embedded control systems are highly resource constrained (Larses, 2005) Basic concept of automobile system architecture The system architecture of an automobile can be described as a map of the layout of the different components and how they are interconnected (Pernstål, 2008). As to illustrate how complex the map can seem, a sketch of an E/E system is presented in figure 5.

22 14 Figure 5 - E/E system in the VW Phaeton (Leohold, 2004) By using the system approach it is possible to split complex technical entities into manageable parts. It is important to note that the links between the parts and their boundaries are not to be forgotten as they overlap (as illustrated in figure 4) in many cases (Denton, 2004). Choosing how to split the vehicle system architecture is not an easy task as this can be done in many different ways. One possible split is between mechanical and E/E systems. This division however, can cause just as many problems as it solves. For example, in which half should anti-lock brakes be put, mechanical or electrical? The answer is both (Denton, 2004 ). Once a complex system has been systemized, the function or performance of each part can be examined in more detail. Functional analysis determines what each part of the system should do and, in turn, can determine how each part actually works (Denton, 2004 ). To examine subsystems separately can however be misleading as the system, when working as one entity, may act different from what its subsystems do when separated. To analyze a system, further consideration needs to be given to the inputs and outputs of the system. Many of the complex electronic systems in a vehicle lend themselves to this form of analysis. The recommended approach is to consider the Electronic Control Unit (ECU) of the system as the control element, and look at its inputs and outputs (Denton, 2004 ). The following sections present different classes of E/E system in automobiles. Power train and chassis systems. These include systems that are highly critical for the vehicle s functionality such as control functions for engine, brakes, and steering. The systems are characterized by high demands on safety, reliability, and real-time constraints (Åkerholm, 2005).

23 15 Cabin system. Cabin systems are less critical systems, but at the same time these systems play an important part of the driver experience of the vehicle. Examples of such systems are dashboard instruments, electrically powered windows, and air-conditioning (Åkerholm, 2005). Infotainment systems. These systems are dedicated to information processing, for example entertainment, but also communication with systems outside the vehicle, such as radio, video, and satellite navigation systems. Normally the systems are not closely integrated with the core functionality of the vehicle and are easier to replace, supplement and remove (Åkerholm, 2005). The physical architecture of the E/E system in vehicles is a complex distributed computer system. The computer nodes are designated ECUs, and are often developed by different vendors and use different hardware (Åkerholm, 2005). The ECUs are interconnected by one or several networks often using different network technologies within the same vehicle. As an example, the Volvo Truck Corporation uses two different network technologies and has six to eight different suppliers of ECUs. The Volvo Car Corporation uses four different network technologies, depending on model, and has more than ten different suppliers of ECUs (Åkerholm, 2005). The location of each ECU is determined by where the controlled object is allocated in the vehicle, aiming to minimize the length of wiring to sensors and actuators. Interconnection networks allow collaboration between several ECUs. One such example is the Electronic Stability Program (ESP) that uses ECUs to control the engine and brakes to assist the driver in emergency situations. Figure 6 - Schematic picture of E/E system in a Volvo XC90 (Axelsson, 2008) Figure 6 gives a schematic overview of the E/E system in a Volvo XC90. About 40 ECUs are interconnected through various networks. The networks in this automobile are high speed CAN (Controller-Area Network), low speed CAN, MOST (Media Oriented Systems Transport), and LIN

24 16 (Local Interconnect Network). The safety-critical modules such as the engine control module and the brake control unit are allocated on the high speed CAN-network (red line). On the low speed CAN-network (green line) functions such as the climate control module are found. From the low speed CAN a special network for infotainment is connected (MOST). This network provides enough bandwidth to support streaming of large data content, video for example. There are also some smaller LIN networks that interconnect different modules in an easy way Dysfunction As will be seen in later chapters, identifying the source of error in a system has practical implications, and can be associated with large costs for automotive industry parties. In this section an introduction to different aspects of dysfunction is given, but let us start by stating what is meant by a safety-critical system. The occurrence of dysfunction in a system becomes more severe when the system is safety-critical. Systems that have safety requirements are usually referred to as safety-critical (Larses, 2005). Brake-by-wire is a good example of a safetycritical system. When designing safety-critical systems, early security analysis indicates what parts of the construction classify as safety-critical. Every component that is connected to a safety-critical component is in itself considered critical (Rehnström & Näsström, 2002). A system can be in three states of dysfunction: failure, error, and fault (Storey, 1996). System failure is the highest level and means that the system fails to perform the required action. Before the system shows loss of functionality there are two states of dysfunction. A fault is a defect within the system. As long as the faulty part of the system is not activated, the fault remains a fault and does not cause any real harm. When the faulty part of the system is activated, the system responds by producing an error. This means that a deviation from the required operation has occurred. If the deviation prevents the system from performing an action system failure has occurred (Storey, 1996). Figure 7 - States of dysfunction (Larses, 2005) If the system failure is created in a subsystem, the failure represents a fault in the main system, until it is activated (Edler, 2008). All faults are the effect of failures in subsystems that are, or have delivered service to the observed system level (Larses, 2005). To put this is some context it can be mentioned that brake-by-wire is an example of a subsystem, where the automobile E/E system represents the main system (Pernstål, 2008). Depending on the situation it can be useful to differentiate between faults by studying their background. For example, faults can be described by origin, cause or nature.

25 17 Random faults stem from hardware component failure and external influence. Systematic faults are designed into the system through human error. These faults are sometimes called design faults (Larses, 2005). Another possible differentiation is between transient, intermittent and permanent faults. Transient faults are temporary faults caused by the physical environment. Intermittent faults are temporary faults that result from a set of unique conditions that are hard to reproduce. Permanent faults are ever present and easily reproduced (Larses, 2005). Systematic faults are permanent faults. For responsibility issues another distinction may be of interest. Primary faults and failures occur within the specification of the system and are usually the result of incorrect design, manufacture, or construction (Larses, 2005). Secondary faults and failures result from extraordinary environmental stress that exceeds the system specifications. Command faults are improper operations of the system that lead to failure. They are created by human operation or by a defect control unit (Larses, 2005). A system free from faults is nearly impossible because of two main reasons. Firstly, components are subject to random failure due to wear, ageing and external circumstances, and secondly, perfection in design is unachievable. Thus safety-critical systems need to be architected with fault management. Fault management is performed in four stages: fault avoidance, fault removal, fault detection and fault tolerance. Fault avoidance is a part of the design phase which aims at designing fault free computer software. Fault removal attempts to find faults before they are activated, and includes both hardware and software testing. Fault detection techniques are used during service where they try to minimize the effects of detected faults (Storey, 1996). Fault tolerance techniques are designed to allow the system to operate correctly even in the presence of faults (Storey, 1996). A system with full fault tolerance is unaffected by any fault on the system level (Larses, 2005) Modularization To modularize electronics means to divide the system in modules which are preassembled separately (Danielsson, 2000). According to different sources, modularization of electronic systems gives shorter time to market, lower development costs, better quality and lower production costs (Danielsson, 2000). In a fully modularized system, each module can be tested separately in testing stations (Danielsson, 2000). Production costs can be cut as the modules can be assembled concurrently in different production lines, after which the system is assembled (Larses, 2005). Different modules could for example be assembled at different suppliers, and the OEM would only have to lay the finishing touch. If hardware and software is bought from one supplier, it is better to bundle that module at that supplier rather than to bundle in-house specified hardware with bought software (Larses, 2005). Another benefit of modularization is that development cost can be cut as new models of the system need not be developed from scratch as some modules can

26 18 be used from previous versions (Danielsson, 2000). Modularization also improves the possibility of variety management as modules can be altered separately. Other benefits of modularization are seen within logistics and procurement (Larses, 2005). 4.2 Cost In this section, terms and theories on cost and cost analysis models are presented. As the aim of the thesis has been to evaluate cost efficiency in the context of system architecture, it is considered appropriate to give a basic introduction to terms and theories related to cost. The reader who is already familiar with economic terms such as variable cost and opportunity cost may skip to section where cost analysis models are introduced. Pay special attention to Cost Benefit Analysis as this method is used as basis for the model presented in the previous chapter. The rationale for selecting CBA is to be found in chapter Introduction to cost In order to evaluate cost efficiency it is appropriate to start by discussing the terms cost and efficiency separately. The most general definition of cost is the value of used resources during a specific period in time (Andersson, 1992). Cost is usually divided in variable cost and fixed cost. Variable cost varies with produced volume, fixed cost does not. Measured by unit, variable cost remains fixed, while fixed cost per unit falls with increasing volume (Andersson, 1992). How cost is affected by volume varies from situation to situation and usually depends on the production facilities. To describe this phenomenon, variable cost can be divided in three groups. Proportionally variable cost increases proportionally to a change in volume, progressively variable cost increases/decreases more than the volume change, and digressively variable cost changes slower than volume (Andersson, 1992). Fixed cost can also be divided in different categories. Entirely fixed cost has the same value regardless of the situation, operation related fixed cost is fixed when production is running (for example heating or electricity), and partially fixed cost is fixed at a given volume but increases stepwise (Andersson, 1992). Salary is often assumed to be variable cost. In practice this is however not always the case. In short-term decision making it is more correct to treat salary as a fixed cost. In the long-term salary varies with production volume. If the workforce is flexible and can be transferred between production elements, salary is variable also in the short-term (Andersson, 1992). Depending on the situation, cost can be labeled differently. For decision making purposes one way of labeling may be appropriate, whereas for other purposes that specific way of labeling may be inappropriate. For costing purposes, cost can be divided in direct and indirect cost. Direct cost is explicitly associated with a cost unit such as a product. Indirect cost is not directly traceable to a product or cost unit. Indirect cost is accounted to different cost centers before being added to the cost