A simulation model of the optimal maintenance strategy: conceptual agent-based approach

Kokoteksti

(1)Lappeenranta University of Technology School of Industrial Management and Engineering Department of Innovation Management. Master’s Thesis. A SIMULATION MODEL OF THE OPTIMAL MAINTENANCE STRATEGY: CONCEPTUAL AGENT-BASED APPROACH. Reviewers: Professor Tuomo Kässi, D.Sc. (Tech.) Professor Timo Kärri, D.Sc. (Tech.). Supervisor: Post-doctoral researcher Samuli Kortelainen, D.Sc. (Tech.).

(2) ABSTRACT Author: Pontus Huotari Name: A simulation model of the optimal maintenance strategy: conceptual agentbased approach Department: Innovation Management Year: 2013. Place: Lappeenranta. Master’s thesis Lappeenranta University of Technology 86 pages, 19 figures, 16 tables, and 4 appendices Keywords: maintenance strategy, maintenance management, maintenance optimization, asset management, life cycle management, agent-based modeling, simulation. In this Master’s thesis agent-based modeling has been used to analyze maintenance strategy related phenomena. The main research question that has been answered was: what does the agent-based model made for this study tell us about how different maintenance strategy decisions affect profitability of equipment owners and maintenance service providers? Thus, the main outcome of this study is an analysis of how profitability can be increased in industrial maintenance context. To answer that question, first, a literature review of maintenance strategy, agentbased modeling and maintenance modeling and optimization was conducted. This review provided the basis for making the agent-based model. Making the model followed a standard simulation modeling procedure. With the simulation results from the agent-based model the research question was answered. Specifically, the results of the modeling and this study are: (1) optimizing the point in which a machine is maintained increases profitability for the owner of the machine and also the maintainer with certain conditions; (2) time-based pricing of maintenance services leads to a zero-sum game between the parties; (3) value-based pricing of maintenance services leads to a win-win game between the parties, if the owners of the machines share a substantial amount of their value to the maintainers; and (4) error in machine condition measurement is a critical parameter to optimizing maintenance strategy, and there is real systemic value in having more accurate machine condition measurement systems..

(3) TIIVISTELMÄ Tekijä: Pontus Huotari Nimi: Optimaalisen kunnossapitostrategian simulaatiomalli: konseptuaalinen agenttipohjainen lähestymistapa Osasto: Innovaatiojohtaminen Vuosi: 2013. Paikka: Lappeenranta. Diplomityö Lappeenrannan teknillinen yliopisto 86 sivua, 19 kuvaa, 16 taulukkoa, and 4 liitettä Avainsanat: kunnossapidon johtaminen, kunnossapitostrategia, kunnossapitostrategian optimointi, omaisuuden hallinta, elinkaarijohtaminen, agenttipohjainen mallinnus, simulointi. Tässä diplomityössä tutkittiin kunnossapitostrategiaa agenttipohjaisen mallinnusmenetelmän avulla. Päätutkimuskysymys oli: mitä mallinnuksen tulokset kertovat siitä, miten eri kunnossapitostrategiset päätökset vaikuttavat omaisuuden haltijoiden ja huoltopalveluita tarjoavien yritysten kannattavuuteen? Täten tutkimuksen päätulos on analyysi siitä, miten kannattavuutta voidaan parantaa kunnossapitoalalla. Työn teoreettisessa osassa käsiteltiin kunnossapitostrategiaa ja agenttipohjaista mallinnusta yleisellä tasolla, ja lopuksi paneuduttiin kunnossapitostrategian mallintamiseen ja optimointiin. Kirjallisuuskatsaus toimi pohjana agenttipohjaisen mallin rakentamiselle. Mallin tekeminen noudatti erästä standardimallinnusprosessia. Mallin ja sen avulla simuloitujen tulosten avulla vastattiin päätutkimuskysymykseen. Tarkemmin sanottuna, mallinnuksen ja samalla tutkimuksen tulokset olivat: (1) koneen huoltopisteen optimoinnilla on positiivinen vaikutus koneen omistajan ja huoltajan kannattavuuteen, kunhan tietyt ehdot täyttyvät; (2) kunnossapitopalveluiden hinnoitteleminen aikaperusteisesti johtaa nollasummapeliin osapuolten välillä; (3) kunnossapitopalveluiden hinnoitteleminen arvoperusteisesti johtaa yhteiseen etuun, jos omaisuuden haltijat jakavat merkittävän osan arvostaan koneidensa huoltajille; ja (4) virheen suuruus koneen kunnon mittauksessa on kriittinen parametri optimoitaessa kunnossapitostrategiaa, ja tarkemmat koneen kunnon mittaamisjärjestelmät tuottavat systeemistä lisäarvoa..

(4) PREFACE The following phrase is a straight quote from one of the very first course assignments I produced in studying Industrial Management at LUT: “a Master of Industrial Management is usually a dynamic multi-talent…” Now, after four years of writing that phrase, one must be critical: as “dynamic” means “change over time” explicitly, it is as if a Master of Industrial Management changes himself from being a talent of something to being the talent of everything, or nothing. Honestly speaking, the phrase is ridiculous and means nothing in reality, but it just sounded great at that time…. But there was to be a hidden meaning (just kidding) in that phrase: I actually ended up being most interested in studying the dynamics in industrial management. The moment of enlightenment came on the “system dynamics in industrial management” course: no one else seemed to care at all, but I realized immediately in the beginning that these kinds of simulation tools enable us to really understand the real world. The beauty in using simulation methods is that we can build a dynamic system that takes all the relevant aspects into account regarding the system behavior. Ideally, we could build a model that takes almost everything of reality into account. Well, in practice it is impossible. But still, I love the idea and possibility that not that many stupid assumptions and simplifications must be made when simulating.. In retrospect, system dynamics was nothing compared to the power of agent-based modeling. I have to say that agent-based modeling is one of the most valuable skills I learned at LUT. First, my thanks go to the reviewers, Tuomo Kässi and Timo Kärri, who allowed me the possibility to learn this skill while making my Master’s thesis. Secondly, special thanks go to the supervisor, Samuli Kortelainen, who pushed me to learn more and more. Finally, and most importantly, I thank myself.. Pontus Huotari 1.10.2013, Lappeenranta.

(5) TABLE OF CONTENTS 1. 2. INTRODUCTION ................................................................................................ 1 1.1. Background and motives for the thesis .......................................................... 1. 1.2. Research questions and outcomes .................................................................. 3. 1.3. Research method ............................................................................................ 4. 1.4. Restrictions .................................................................................................... 7. 1.5. The structure of the thesis .............................................................................. 7. MAINTENANCE STRATEGIES AND AGENT-BASED MODELING ........... 8 2.1. 2.1.1. Strategic dimensions of maintenance management ................................ 9. 2.1.2. Strategic options for conducting maintenance ...................................... 13. 2.1.3. Choosing of a maintenance strategy ..................................................... 16. 2.2. The basics of agent-based modeling ............................................................ 18. 2.2.1. What agent-based modeling is about .................................................... 18. 2.2.2. Agent-based modeling principles ......................................................... 21. 2.2.3. Main benefits of using agent-based modeling in economic studies ..... 25. 2.3. 3. An overview of maintenance strategy............................................................ 9. Modeling and optimization of maintenance strategies ................................ 27. 2.3.1. Motives for modeling and optimization ............................................... 28. 2.3.2. Principles of modeling and optimization .............................................. 28. 2.3.3. Other than agent-based models ............................................................. 32. 2.3.4. Previous agent-based models ................................................................ 34. THE AGENT-BASED MODEL DESCRIPTION ............................................. 38 3.1. The purpose of the model ............................................................................ 38. 3.2. Defining the boundaries of the model .......................................................... 39. 3.3. The dynamic hypothesis .............................................................................. 41. 3.4. The structure of the model ........................................................................... 43. 3.5. Simulating and testing the model ................................................................. 49.

(6) 4. 5. 6. THE RESULTS OF THE MODELING ............................................................. 50 4.1. Profitability when pricing is time-based ...................................................... 54. 4.2. Profitability when pricing is value-based..................................................... 55. 4.3. Condition measurement error and profitability in time-based pricing ......... 57. 4.4. Condition measurement error and profitability in value-based pricing ....... 59. 4.5. Implications of the results ............................................................................ 61. CONCLUSIONS AND DISCUSSION .............................................................. 64 5.1. The main outcomes ...................................................................................... 64. 5.2. Limitations ................................................................................................... 68. 5.3. Possibilities for future research .................................................................... 70. SUMMARY........................................................................................................ 71. REFERENCES ........................................................................................................... 76 APPENDICES ............................................................................................................ 87.

(7) LIST OF FIGURES Figure 1. A standard procedure for simulation modeling. ............................................ 4 Figure 2. Investment life cycle and driving forces of asset management. .................. 10 Figure 3. Key elements for SMM. .............................................................................. 11 Figure 4. A framework for maintenance concept development. ................................ 17 Figure 5. An overall abstraction of an AB system...................................................... 20 Figure 6. A conceptual view of maintenance activities over time. ............................. 30 Figure 7. The conceptual view of the overall model structure. .................................. 41 Figure 8. The process view of the tasks that are executed on each time step and the order of execution. ...................................................................................................... 46 Figure 9. Principal relationships between machine condition and related variables. . 47 Figure 10. Expected values regarding machine conditions. ....................................... 48 Figure 11. The overall degradation of a machine through time. ................................ 48 Figure 12. The effect of the price for maintenance (time-based) and the minimum accepted condition for machines on the average cumulative profits for the customers .................................................................................................................................... 54 Figure 13. The effect of the price for maintenance (time-based) and the minimum accepted condition for machines on the average cumulative profits for the maintenance company. ............................................................................................... 55 Figure 14. The effect of the price for maintenance (value-based) and the minimum accepted condition for machines on the average cumulative profits for the customers .................................................................................................................................... 56 Figure 15. The effect of the price for maintenance (value-based) and the minimum accepted condition for machines on the average cumulative profits for the maintenance company. ............................................................................................... 57 Figure 16. The effect of the error range in condition measurement and the minimum accepted condition for machines on the average cumulative profits for the customers when pricing is time-based. ........................................................................................ 58.

(8) Figure 17. The effect of the error range in condition measurement and the minimum accepted condition for machines on the average cumulative profits for the maintenance company when pricing is time-based. ................................................... 59 Figure 18. The effect of the error range in condition measurement and the minimum accepted condition for machines on the average cumulative profits for the customers when pricing is value-based........................................................................................ 60 Figure 19. The effect of the error range in condition measurement and the minimum accepted condition for machines on the average cumulative profits for the maintenance company when pricing is value-based. .................................................. 61.

(9) LIST OF TABLES Table 1. The questions to be answered with the simulation model. ............................. 4 Table 2. The structure of the study. .............................................................................. 8 Table 3. Maintenance strategy decision elements ...................................................... 13 Table 4. Different maintenance strategies, methodologies and techniques combined and briefly explained. ................................................................................................. 14 Table 5. The key aspects to be specified in an AB model. ......................................... 22 Table 6. A standard protocol for describing an AB model. ........................................ 23 Table 7. Maintenance optimization variables and related aspects. ............................. 31 Table 8. Different approaches to maintenance modeling and their simplified definitions and examples of utilization. ...................................................................... 33 Table 9. Comparison of two AB models on maintenance management from ABM’s point of view. .............................................................................................................. 36 Table 10. Comparison of two AB models on maintenance management, considering AB design concepts and the extent to which maintenance strategy elements have been taken into account. ...................................................................................................... 37 Table 11. How maintenance strategy decision elements are or are not taken into account in the model. .................................................................................................. 40 Table 12. How the AB design concepts have been taken into account in the model. 43 Table 13. The agents’ beliefs, interactions, goals, actions, state variables and scales 45 Table 14. The AB model details. ................................................................................ 52 Table 15. The actual simulations that were run with the model with the given input parameters (the ones that were varied, while others being constant). ........................ 53 Table 16. The results of the modeling and their practical implications...................... 66.

(10) LIST OF APPENDICES Appendix 1. The AB model parameters and variables explained. ............................. 87 Appendix 2. The most crucial algorithms in the model being visualized. ................. 92 Appendix 3. Additional graphs of the system response in different settings. ............ 95 Appendix 4. Some screenshots of the model in AnyLogic. ..................................... 103.

(11) ABBREVIATIONS AB. Agent-based. ABM. Agent-based modeling. CM. Corrective maintenance. CMMS. Computerized maintenance management systems. CBM. Condition-based maintenance. EB. Equation-based. EBM. Equation-based modeling. ECM. Effectiveness-centered maintenance. MO. Maintenance outsourcing. PCM. Profit-centered maintenance. PdM. Predictive maintenance. PM. Preventive maintenance. RBM. Risk-based maintenance. RCM. Reliability-centered maintenance. RTF. Run-to-failure. SMM. Strategic maintenance management. TBM. Time-based maintenance. TPM. Total productive maintenance.

(12) 1. 1. INTRODUCTION. This study adds to the research on industrial maintenance strategies by applying the agent-based (AB) methodology in simulating maintenance strategy. The main outcome of this thesis is an analysis of how profitability of equipment owners and maintenance service providers can be increased in industrial maintenance context. The analysis is based on the simulation results from the conceptual AB model that has its foundations in literature findings. The results are theoretical, but insightful for the practitioners also, as they reveal fundamental logics that affect profitability in maintenance context. Profitability-centered focus was chosen, as this has not been the main focus in most of the previous studies. It is argued that maximizing profitability should be one of the main goals when optimizing strategy. In addition, ABM has not been applied many times in determining optimal maintenance strategies, although ABM is a powerful simulation method that can reveal new insights in this area. 1.1. Background and motives for the thesis. This study serves and provides new knowledge for industrial maintenance research in using AB methodology for simulating maintenance strategies. It was found while reviewing the literature that we are lacking business-oriented views on maintenance management, and this is why that view is taken in this study. In addition, ABM has not been used many times in modeling maintenance management related phenomena. However, it is argued and shown in this study that using ABM yields interesting and new insights for maintenance management. Therefore, developing an AB model for simulating maintenance strategies should be of interest. The objective of maintenance is “to preserve the condition of products so as to fulfill their required functions throughout their life cycle”. When life cycle management of products is considered, maintenance is one of the primary functions in the value chain. In the past, maintenance was only seen as a “necessary evil”, which had to be done when a product got broken. Nowadays, however, managing the whole product.

(13) 2. life cycle and preserving functionality of a product has become more and more important. (Takata et al. 2004) Thus, maintenance should not be seen just as a cost, but a service that can increase the value of a product throughout its life cycle, and increase the overall profitability of a company (Alsyouf 2007, p. 70; Aurich et al. 2006, p. 480).. While reading on the literature, it was found that in most cases maintenance strategy optimization models deal with technical aspects of maintenance. A common and generalized example would be a study, in which the interest was to model a technical system in accurate technical detail, and analyze what the optimal maintenance schedule is for preserving technical system performance. However, as shown later, maintenance has a very distinctive effect on profitability. To put it simply, maximizing profitability is the core goal of a company. This is why it is argued that maintenance management models should incorporate profitability goals in the optimization formulas. Therefore, in this study, finding ways for maximizing profitability for the owners of technical equipment and for the maintainers is of main interest.. Why is ABM rather than any other simulation technique used in this study? It is found that lots of optimization models in industrial maintenance have been made. However, few of those models are AB models, although ABM is a powerful simulation method. Therefore, it is of interest to explore the possibilities of ABM in maintenance management and optimization. It is argued and shown in this study that using ABM can yield some new insights in maintenance management.. Compared to other simulation approaches, ABM offers several benefits in simulating systems consisting of many active objects, such as people, business units, projects, stocks and products. In general, it offers the tools to analyze the most complex systems and dynamics. In addition, it makes possible constructing models in the absence of the knowledge about the global interdependencies. It allows modeling of microscale phenomena, and constructing models bottom-up. Bottom-up explanations are.

(14) 3. ultimately the only way to explain and predict both the micro and macro phenomena accurately. Thus, AB models are structurally realistic. Modeling is flexible and responsive. ABM is tailored for simulating thousands of repeated experiments, which allows revealing all the dynamics of a system. Finally, AB models are easy to maintain, as model refinements typically result in very local, not global changes. (Bonabeau 2002b, p. 6; Borshchev et al. 2004, p. 1-7; Luck et al. 2003, p. 29-31; Macal et al. 2007, p. 96) 1.2. Research questions and outcomes. The main research question being answered in this study is: what does the AB model made for this study tell us about how different maintenance strategy decisions affect profitability of equipment owners and maintenance service providers? The main outcomes of this research are explanations on how optimizing maintenance strategy in terms of maximizing profitability is possible. The explanations are based on the simulation results from the AB model. In practice, answering the main research question is divided into two parts. In the first part (the theoretical part of this thesis), the following questions are being answered based on literature findings: . What are the overall aspects of maintenance strategy that should be taken into account in this study?. . What is ABM, how is the method used, and what are the benefits of using the method in this study?. . How have maintenance strategies been modeled in the previous studies?. . How have AB models been used before in simulating maintenance strategies?. In the second part (the empirical part) of answering the main research question, an AB model has been made. The literature review provides the foundation for the model. With the model it is possible to analyze how different maintenance strategy decisions affect profitability of equipment owners and maintenance service providers. Specifically, the simulations answer to the following questions in the table 1..

(15) 4. Table 1. The questions to be answered with the simulation model. Question: how do the aspects below affect profitability of equipment owners and maintainers?  Price for maintenance  Pricing logic (time- or valuebased)  Minimum accepted conditions for machines.   . 1.3. Error in machine condition measurement Pricing logic (same as above) Minimum accepted condition for machines. Brief explanation. The maintainer (i.e. the maintenance company) sets different prices for its maintenance services, while the equipment owners (i.e. customers) set different minimum accepted conditions (i.e. points in which the machines have to be maintained) for their machines. It is analyzed, how these parameters affect profitability of both parties. The prices for maintenance are either time-based (a X % margin on top of the cost of work time) or value-based (a X % share of the profits of the customers is offered to the maintainers). Otherwise the same situation as above, but the prices are kept constant (pricing logic is varied) and the error in machine condition measurement is varied instead. The interest is to analyze, how different values for errors and minimum accepted conditions affect profitability of both operating parties.. Research method. In the figure 1, a standard procedure for simulation modeling is presented. This is followed in order to make the AB model “the right way”. The more exact procedures are elaborated on later. The literature review (chapter 2) serves as the basis for the problem articulation and whole model design. Chapter 3 covers all the steps from 1 to 4, by revealing the model in full detail. Finally, in the chapter 4, the results from the simulations are analyzed and the implications of these results are presented.. 1. Problem articulation (boundary selection) 5. Policy formulation & evaluation. 4. Testing. 2. Dynamic hypothesis. 3. Formulation. Figure 1. A standard procedure for simulation modeling. (Sterman 2000, p. 87).

(16) 5. Sterman (2000, p. 37-38) points out the need for simulation in a quite compact but effective way: simulation is essential, because “the complexity of our mental models vastly exceeds our capacity to understand their implications”. That is, humans are very poor in explaining how dynamic systems behave without the aid of proper tools. Although it is possible that simulation cannot lead to fundamentally new conceptions (if our mental models were correct in the beginning), it is still likely that both the mental and formal models change when simulating. Formalizing qualitative models and testing them via simulation often leads to radical changes in our perception of reality. Without simulation, it becomes all too easy for our mental models to be driven by ideology or unconscious bias. Especially when experimentation in the real world is not possible, simulation becomes the main and perhaps the only way to understand how real systems behave. However, there is of course a danger that formalizing our mental models can lead to omitting certain aspects of reality (due to, say, enabling theorems to be proved). By choosing the right simulation method for the task at hand guarantees that the simulation model will be realistic.. It is important to emphasize that modeling is iterative, not a linear process of sequential steps. Models go through constant iteration, continual questioning, testing and refinement. However, the first and most important part of modeling is the problem articulation or boundary selection. Models are always simplifications from the real world: without boundary selection, models cannot be simplified from the real world, which is necessary for understanding a problem at hand. The art of modeling is to know, which features to cut-out and which features are the key to explaining the behavior of a system. In modeling dynamic systems, determining reference modes (descriptive graphs and data to refer to throughout the modeling process) and time horizon (what is the total timeframe for a problem to emerge and in which should it be examined?) are essential. (Sterman 2000, p. 87-94). Secondly, the dynamic hypothesis must be formulated. It is an assumed explanation for the behavior of a system over time. It is a hypothesis, as it is subject to revision.

(17) 6. and abandonment throughout the modeling process. Formulating a dynamic hypothesis is about underlining the assumptions behind a model, proposing the initial system structure, proposing initial causal loops between variables, and proposing decision making structures in the model. (Sterman 2000, p. 94-102). After formulating the dynamic hypothesis, simulation and testing can begin. They go hand in hand, as a system is put to test through simulating different alternatives of behavior. The system should be tested with various kinds of alternative runs, and especially with extreme ones, because they can reveal hidden dynamics not observable in the real world. If not already in defining the dynamic hypothesis, at last at simulation and testing stages initial explanations of the system behavior most probably change. This is because humans are very poor in explaining dynamic phenomena, and thus, models tend not to be good on the first try. Simulation forces a modeler to analyze system behavior thoroughly, and thus, better understanding can be gained. Finally, after confidence has been built through modeling and simulation, policy design and evaluation can be done. Usually, modeling and simulation will reveal unpredicted and significant behavior, which necessitates structural changes in the real world system design. (Sterman 2000, p. 102-104). For constructing the AB model in this study, AnyLogic software has been used. And for making appropriate presentations (graphs) from simulation data, MATLAB has been used. AnyLogic is a multifunctional simulation modeling tool, which supports many simulation approaches, such as ABM, discrete event simulation and system dynamics (AnyLogic 2013). MATLAB is an environment for numerical computation, visualization, and programming (MathWorks 2013). These programs were chosen for the task as these were technically eligible and available for use. There are other programs that would have qualified also, but these were not considered this time..

(18) 7. 1.4. Restrictions. Firstly, not all the aspects related to maintenance management can be taken into account in the model. It is defined later how different maintenance strategy elements (table 3, chapter 2.1.1) and other aspects have been taken into account in the model. Simply put, the AB model is an optimization model, and thus this study falls into the category of maintenance optimization studies.. Secondly, only the strategies related to industrial maintenance are of interest. In general, strategy is a diverse topic in industrial management, and it is not covered as a whole. This means that all the other strategies except maintenance strategy are left out of consideration. In turn, most of the operational aspects of industrial maintenance are excluded. This means that guidelines cannot be given on, say, designing the equipment to be maintained, or how the equipment should be maintained in technical respect. Thus, this research deals with maintenance on the upper, business-oriented level.. Thirdly, but maybe most importantly, the AB model is a conceptual model. It was not possible to make a fully featured and precise model based on accurate quantitative data, partly because the gathering of this data was not practically possible in this study. Thus, the structures and parameters used in the model are based on the literature findings. In order to increase structural realism and confidence for the model, qualitative feedback from the supervisor and reviewers of this thesis was used for refining the model. 1.5. The structure of the thesis. The structure of the thesis is presented in the table 2. The main inputs for each major chapter are presented on the left column, the chapter name on the center column, and the main outputs of each chapter on the right column..

(19) 8. Table 2. The structure of the study. Input  . Background Research motives. Theory of:  Maintenance strategies  ABM  Maintenance optimization modeling  Theoretical framework of maintenance strategies and ABM  Structures and assumptions for the model based on literature review  Feedback from the supervisor and reviewers of this thesis  The AB model  The simulated results  Research questions  Theoretical findings  Results from the simulations  Experiences from the project  Theory  Research results  Conclusions. 2. Chapter 1 Introduction. Output      . Research questions Research outcomes Research method Restrictions Definition of concepts Background and assumptions for the AB model. 3 The AB model description. . The AB model itself. 4 The results of the modeling 5 Conclusions and discussion.     . 6 Summary. . Analysis of the results Implications of the results Conclusions of the study Limitations of the results Recommendations for the future research Summary of the research. 2 Maintenance strategies and ABM. MAINTENANCE STRATEGIES AND AGENT-BASED MODELING. In this chapter, the literature review on maintenance strategies and ABM is provided. Firstly, strategic dimensions of maintenance management are investigated. Then, ABM is introduced, procedures for AB modeling, and benefits of using ABM are shown. Finally, previous studies on modeling maintenance strategies with and without AB methodology are looked through. This literature review provides the foundation for the AB model made for this study..

(20) 9. 2.1. An overview of maintenance strategy. First of all, we need to investigate and define what maintenance strategies are in order to model them. Strategic dimensions of maintenance management are introduced first, to get an overview on the subject. Secondly, different options for maintenance strategies are shown. It is found that the term “maintenance strategy” is not that clearly defined in the literature, as different authors declare different things as strategies or something else. Finally, it is shown how to choose a right maintenance strategy depending on the situation. 2.1.1. Strategic dimensions of maintenance management. If we put maintenance into a larger context, it is part of asset management. Asset management is basically about managing physical assets optimally over their life cycle in order meet the stated business objectives. The figure 2 clarifies the life cycle perspective and the driving forces of asset management, in which maintenance is part of operations over the life cycle. (Komonen et al. 2012, p. 49-50) In this study, we are only interested in “operations and maintenance” part of the figure. That is, no preceding or following stages of the life cycle, such as installation or disposal of equipment, are of interest..

(21) 10. Investment. Use of investment    . Asset management of plants in operation: focus on upkeeping of plants’ productivity and profit making capability, and improving plants in changing business environment. Concept design. Design of process and equipment. Manufacturing. Asset management of new installations: focus on the optimization of the life cycle profits of plants. Installation and commencing. Changing demand Changing competitive environment Modified product Economic obsolescence. Operations and maintenance.    . Disposal. Changing operational requirements Wear and aging Technical obsolescence Environmental obsolescence. Effects of changes in the environment at the corporate, plant, process and equipment level. Figure 2. Investment life cycle and driving forces of asset management. (Komonen et al. 2012, p. 50). And to put maintenance management into a bit more focused context, it is directly linked to manufacturing management. The main objective of maintenance is to preserve the condition of manufacturing equipment through operational life cycle. Thus, maintenance strategy and manufacturing strategy have to be directly interrelated. This is not the case in many situations, as the importance of maintenance is not recognized. The strategic level of maintenance management is often ignored, and the focus is set only on tactical and operational aspects, such as how to maintain equipment in technical respect. This can actually lead to reduction in manufacturing performance. And in turn – if maintenance and manufacturing are managed strategically, in parallel, and strategies and operations establish a coherent link between each other – this can lead to increased performance in manufacturing. (Robson 2010, p. 206-208). Murthy et al. (2002, p. 291) have proposed the key elements for strategic maintenance management (SMM). These can be expressed in visual format (figure 3), also.

(22) 11. including some additions based on the findings made by Robson (2010, p. 75). On the upper level, it is recognized that the business objectives are dependent not only on the internal objectives but also on the external environment. And the business objectives are the input to determining maintenance and manufacturing strategies. Maintenance strategies are for ensuring that the load requirements of equipment are met and that the equipment state is appropriate. Maintenance strategy determines how maintenance is conducted on the operational level, and implementation of strategy results in some performance (i.e. equipment state is restored or improved). The linkage between maintenance and manufacturing is direct as manufacturing is dependent on equipment to be maintained. Manufacturing outputs serve the market, and the reactions from the market feedback to decision making.. Inputs to decision making from the external environment Business objectives. Maintenance strategies. Operating load of equipment. Manufacturing strategies. Implementation of maintenance strategy. Equipment state. Implementation of manufacturing strategy. Maintenance output performance. Manufacturing output performance. Internal environment. Corporate Customers New entrants Competitors Low cost Shareholders Market share Marketing / sales Suppliers Competitive advantage Etc.. Manufacturing outputs to the market. External environment. Figure 3. Key elements for SMM. (Robson 2010, p. 75; Murthy et al. 2002, p. 291). When taking a strategic approach to maintenance management, two key aspects need to be emphasized: (1) maintenance management is a core business function for overall business survival and success, and therefore it should be managed strategically;.

(23) 12. and (2) maintenance management needs to be based on quantitative models, which integrate maintenance decisions with other strategic decisions such as production planning. In SMM approach, maintenance is a multi-disciplinary activity, which involves: . Understanding of degradation mechanisms of equipment and linking it with data collection and analysis, in order to assess the state of the equipment.. . Building quantitative models to predict the effects of different maintenance actions on degradation.. . Managing maintenance from strategic perspective. (Murthy et al. 2002, p. 290-291). Pinjala et al. (2006, p. 216-220) have summarized maintenance strategy decision elements, which are presented in the table 3. A very similar classification has been made by Tsang (2002, p. 10-23, 26-35). He has defined the four strategic dimensions of maintenance management as (1) service-delivery options, (2) organizational design and structuring of maintenance work, (3) support systems, (4) and maintenance methodologies. Both of these classifications are essentially same.. The first four decisions are called structural decisions, as these decisions cannot be undone fast. For example, if maintenance has been outsourced, it is practically impossible to bring it back in-house immediately. The last six decisions are generally linked to specific operating aspects of a company, such as production process. Both the structural and infrastructural decisions are closely interrelated. They can have a major impact on the maintenance function’s ability to implement and support the overall business strategy. (Pinjala et al. 2006, p. 220).

(24) 13. Table 3. Maintenance strategy decision elements. (Adapted from Pinjala et al. 2006,. INFRASTRUCTURE DECISION ELEMENTS. STRUCTURAL DECISION ELEMENTS. p. 219; Tsang 2002, p. 10-23, 26-35). 2.1.2. Maintenance capacity Maintenance facilities Maintenance technology Vertical integration Maintenance organization Maintenance policy and concepts Maintenance planning and control systems Human resources Maintenance modifications Maintenance performance measurement and reward systems. Capacity in terms of work force, supervisory and management staff. Tools, equipment, spares, workforce specialization, location of workforce. Condition monitoring technology, expert systems, e-maintenance systems In-house maintenance versus outsourcing, and relationships with suppliers. Organizational structure, responsibilities. Policies and concepts (see chapter 2.1.2, table 4). Activity planning, scheduling, spare parts inventory control, costs. Computerized maintenance management systems (CMMS). Recruitment policies, training and development of workforce and staff. Culture and management style. Modifications in equipment, design improvements, new installations and new machine design support. Performance recognition, reporting and reward systems.. Strategic options for conducting maintenance. While reading the literature on maintenance management, it becomes clear that many papers refer to the principally same subjects with varying terminology. Especially the distinction between maintenance strategy, methodology, technique or policy is not clear. If different papers are compared, for example, Tsang (2002, p. 23-26) refers to the terms below (in the table 4) as methodologies, Wang et al. (2007, p. 153-154) define them maintenance strategies, and Garg et al. (2006, p. 214) declare them techniques. Anyway, table 4 summarizes the relevant strategies, methodologies and techniques found in the field of maintenance and a short description is given on what each term means in principal terms. This classification in the table 4 is mainly based on the classification made by Garg et al. (2006, p. 214-219), and the definitions for each term are adapted from a few different sources..

(25) 14. Table 4. Different maintenance strategies, methodologies and techniques combined and briefly explained. Options CM and RTF. CMMS. CBM and PdM. ECM. MO. PCM. PM. RBM. RCM. SMM. TPM. Explanations Corrective maintenance means maintaining equipment after a failure has happened, aiming to restore the equipment to a specific condition. (Wang 2002, p. 470) Another term called run-to-failure (RTF) is linked to CM, as it means that only routine maintenance is being conducted until a system fails. (Tsang 2002, p. 23) Computerized maintenance management systems provide information technology tools for storing, retrieving and analyzing information. They also facilitate communication and coordination of activities. (Garg et al. 2002, p. 217; Swanson 1997, p. 11) Condition-based maintenance means making maintenance decisions based on the state of the system, which is being monitored or inspected constantly. Maintenance tasks are aimed to be done prior to predicted failures. (Chen et al. 2002, p. 43; Garg et al. 2002, p. 216; Grall et al. 2002, p. 167; Jardine et al. 2006, p. 1483; Marseguerra et al. 2002, p. 151-152) In addition to CBM, predictive maintenance (PdM) is also about predicting the need for maintenance based on data collected through monitoring systems. (Chu et al. 1998, p. 285; Garg et al. 2002, p. 217-218; McKone et al. 2002, p. 109-110) That means these two terms are somewhat interchangeable. Effectiveness-centered maintenance puts emphasis on doing the right maintenance tasks in terms of systems functions and customer service. The basic goal is to prioritize maintenance tasks based on the importance of system functions. ECM is composed of people participation, quality improvement, maintenance strategy development, and performance measurement. (Garg et al. 2006, p. 218; Pun et al. 2002; p. 346-347) Maintenance outsourcing refers to outsourcing those maintenance activities, which are not reasonable to conduct in-house. (Garg et al. 2002, p. 218; Martin 1997, p. 83-84; Murthy et al. 1999, p. 259261) Profit-centered maintenance puts emphasis on reducing the need for maintenance and re-engineering of maintenance activities, in order to eliminate non-value adding activities and maximizing profits and minimizing maintenance costs. (Bond et al. 1997) Preventive maintenance refers to maintaining equipment before a breakdown, thus preventing a system failure. The frequency of maintenance is dictated by the passage of time (time-based maintenance, TBM), the amount of production, or machine condition (CBM). (Garg et al. 2002, p. 214-215; McCall 1965, p. 496; Takata et al. 2004; Wang 2002, p. 470) Risk-based maintenance means taking into account the related risks of equipment failures and risks are being minimized or optimized in parallel with other objectives for maintenance, primarily cost effectiveness. (Garg et al. 2002, p. 218-219; Khan et al. 2003, p. 561-566) Reliability-centered maintenance means determining a maintenance plan for an asset during its lifecycle, while balancing the reliability criteria for the asset and cost effectiveness of maintenance tasks. The main focus in RCM is minimizing costs through focus on the most important functions of the equipment and avoiding or removing maintenance actions that are not strictly necessary. (Frangopol et al. 2001, p. 27; Garg et al. 2002, p. 217; Rausand 1998, p. 121-122; Tsang 2002, p. 24-25) Strategic maintenance management views maintenance as a multidisciplinary and a core business activity. Decisions on maintenance activities should be made in parallel with other strategic business decision. SMM takes a quantitative approach to modeling maintenance activities. (Garg et al. 2002, p. 218; Murty et al. 2002, p. 290) Total-productive maintenance takes a comprehensive view on maintenance: a system, covering the whole life-cycle of equipment and all aspects related to maintenance tasks, is established in a company. All employees from top-management to line staff and all company departments are encouraged to participate in promoting TPM. Maintenance is conducted through autonomous cross-functional team work. TPM aims not only to minimize costs of maintenance, but to maximize equipment effectiveness and profitability from customer’s perspective. (Chan et al. 2005, p. 72; Garg et al. 2002, p. 216-217; McKone et al. 1999, p. 124; Tsang 2002, p. 25-26). As the above terms are referred to with differing terminology, maintenance policies are a bit more clearly defined and referred concept. However, we can notice some.

(26) 15. overlap between policies and the above terms as well. Examples of maintenance policies are age replacement, block replacement, repair limit, failure limit, sequential and repairs counting policy. Age replacement policy means a unit is repaired or replaced in relation to its age. Block replacement policy means individual parts of a system are replaced at prescribed times. Repair limit policy means a unit is repaired if the cost of repair is acceptable. Failure limit policy means repair is conducted if failure rate of a system rises above an unacceptable level. Sequential PM policy means a unit is preventively maintained at unequal and usually shortening time intervals as time passes. Repairs counting policy means a unit is replaced after some number of failures have occurred. (Wang 2002, p. 469-483). Now, if we look at the presented strategies, methodologies, techniques and policies in parallel, the conclusion is that they all deal with the same subject from quite similar viewpoints. All of them can be regarded as strategic options for conducting maintenance. And all of these take into account several maintenance strategy elements presented in the table 3 (previous chapter), but not all if any can be regarded as a complete maintenance strategy. In this study, it is not thoroughly argued, whether any of the above strategies is really a maintenance strategy or rather a methodology, technique or policy. However, they all are just regarded as strategic options. And later in the AB model description, the strategic dimensions and options being modeled are clearly defined, based on the concepts presented in the previous tables.. In general, proactive maintenance strategies such as CBM, PM, PdM, and TPM are better in terms of performance compared to reactive maintenance strategies such as CM. By performance it is meant improved quality of products, equipment availability and reduction in production costs. (McKone et al. 2001, p. 39-58; Swanson 2001, p. 239-243).

(27) 16. 2.1.3. Choosing of a maintenance strategy. The overall aim in determining maintenance strategy is to find the right methods for conducting maintenance in terms of objectives for maintenance. Ideally, the goal is to find the optimal maintenance strategy in terms of maintenance objectives. (Wang 2002, p. 482) An optimal maintenance strategy mix can increase availability of equipment, improve reliability levels and reduce unnecessary investments in equipment. The choosing of an optimal maintenance strategy is a typical multi-criteria decision making problem. (Wang et al. 2007, p. 160). Waeyenbergh et al. (2002, p. 311-313) have developed a framework (figure 4) for maintenance concept development. The step-by-step process constitutes of: identification of objectives and resources for maintenance; identification of the most important support systems; decision making on the maintenance strategy and its optimization; monitoring the performance of the chosen strategy (referred as policy by Waeyenbergh et al.); and applying continuous improvement throughout the whole process. It should also be noted that there are feedback and feedforward loops between identification of objectives and maintenance strategy decisions, which is obvious as these steps are clearly interrelated. Waeyenbergh et al. (2004, p. 404) studied the applicability of the framework in a real-life situation, and they state that the framework helped a company to consider and develop maintenance concepts appropriate for its situation. Overall, the framework provides a structural view to choosing maintenance strategy..

(28) 17. Identification of objectives and resources. Identification of the most important systems. Continuous improvement. Performance measurement. Maintenance strategy decision and optimization. Figure 4. A framework for maintenance concept development. (Adapted from Waeyenbergh et al. 2002, p. 312). Information systems are of key importance to maintenance. They are the key to CBM: with monitoring systems the condition of a system can be known and thus the need for maintenance can be optimized (Nilsson 2007, p. 223). Pure predictions based on failure data without knowing current condition of a system do not allow optimal maintenance strategies to be found. (Heng et al. 2009, p. 726-731; Lee et al. 2006, p. 476) Nowadays, we can even talk about the concept of e-maintenance. In addition to being the key for monitoring purposes, information systems make other critical maintenance processes possible, such as remote maintenance decision making, collaborative maintenance, immediate online maintenance and predictive maintenance. (Muller et al. 2008, p. 1165-1172).

(29) 18. As the framework in the figure 4 points out, based on the overall feedback structure, maintenance strategy should not be static but dynamic. However, in modeling respect, it is hard to model and include all the elements of the feedback structure. In the AB model made for this study, each of the elements is simplified, and maintenance strategy is kept static in a simulation. Objectives and resources are imposed in the beginning of a simulation. The most important system, the one for condition measurement, is also given so no identification or change happens. Maintenance strategy decision and optimization is not done automatically by the model, because the intention is to simulate all the possible strategies (i.e. the ones the model can simulate) and determine which one produced the best results after all simulations. Performance measurement is not directing decision making within the model, but allows observing how the agents succeeded. Overall, of these four elements, the condition measurement system is of most interest as it provides the agents critical information that affects their decisions. How it affects agents’ decision is elaborated on later, but at this point it is already mentioned that the modeled maintenance strategies represent CBM the most. 2.2. The basics of agent-based modeling. In this chapter ABM is introduced, the principles of ABM are outlined, and the benefits of using ABM are emphasized. This chapter serves as a short conclusion of the main concepts in ABM and provides some core frameworks for designing and explaining AB models. 2.2.1. What agent-based modeling is about. In ABM the key component to be understood is the “agent”. Agents are the building blocks in ABM and functionality of an AB model is directly related to the functionality of the agents. Agents have certain characteristics that clarify what they are: . identifiable problem solving units with defined boundaries, with a set of characteristics and rules governing their behavior and decision-making capabilities;.

(30) 19. . situated in a defined environment, receiving inputs related to the state of the environment and other agents, and responding to the environment;. . designed to meet certain goal(s) or objective(s);. . autonomous, as they have control over their internal state and own behavior at least to some defined extent;. . flexible problem solvers that act both reactively on the changes in the environment and behavior of the other agents, and proactively based on their objectives. (Bonabeau 2002a, p. 7280; Jennings 2000, p. 280; Macal et al. 2006, p. 74; Wooldridge et al. 1995, p. 116-118). ABM is basically about modeling the behavior of agents with defined objectives in an environment and over time. One key aspect worth emphasizing is that in an AB model there are often multiple agents and they interact with each other. Agents communicate and decide on what goals are being achieved, at what time and by whom agent. Interactions can also evolve over time: that is, the nature of interactions can change, new interactions can be formed and existing connections can disappear. In addition, agents have only partial control and visibility in their environment. That means that agents’ decisions and interactions are context-dependent and emerging over time. Finally, agents can act either as individuals or they can share common goals. If agents have common goals, there is usually some organizational context to agents’ interactions. Thus, the key abstractions that define AB models are agents, interactions and organizations. (Jennings 2001, p. 36-37) An overall abstraction of an AB system is presented in the figure 5..

(31) 20. Organizational relationships. Agents. An interaction. Environment Sphere of visibility and influence in the environment. Figure 5. An overall abstraction of an AB system. (Adapted from Jennings 2000, p. 281; Jennings 2001, p. 37). ABM is most suitable to studying processes that lack central coordination, which means top-down order of behavior. Rather, AB models focus on bottom-up problems: how simple and predictable local patterns can lead to familiar but highly complex and enigmatic global patterns, such as diffusion of information and participation in collective action. ABM provides theoretical leverage where the global patterns are more than the aggregation of individual attributes, and at the same time, the foundation for explaining the emergent patterns with bottom-up dynamical model. (Macy et al. 2002, p. 148). Of course there are situations in which ABM is not the best approach. However, according to Macal et al. (2007, p. 104-105) the following situations support or even necessitate using ABM: . When it is natural to represent a system as agents..

(32) 21. . When decisions and behaviors in a system can be defined discretely (with boundaries).. . When it is important that agents adapt and change their behavior.. . When it is important that agents learn and engage in dynamic strategic behaviors.. . When it is important that agents have dynamic relationships with each other, and these relationships form and dissolve.. . When it is important that agents form organizations, and adaptation and learning occur on the organizational level.. . When it is important that agents’ behaviors and interactions have a spatial component.. . When the past cannot predict the future.. . When scaling-up to arbitrary levels is important.. . When the result of the modeling has to be a structural change.. 2.2.2. Agent-based modeling principles. When specifying an AB system, at least the following aspects presented in the table 5 must be captured (Wooldridge 1997, p. 26-37). Next, in the table 6 a standard protocol for describing an AB model is presented (Grimm et al. 2006, p. 117-119)..

(33) 22. Table 5. The key aspects to be specified in an AB model. (Adapted from Wooldridge 1997, p. 26-37) Aspects to be specified when specifying an AB system The beliefs agents have. Description. The ongoing interaction between agents and their environment. Reactive systems do not terminate, so it is required that the reactive nature of agents is specified. This requires using logic, which is capable of reasoning ongoing behavior. The basic idea is that an agent’s behavior is described in ways that “if something happens, do something, and if something else happens thereafter, then do something different”.. The goals that agents try to achieve. The behavior of agents is goal-directed. That does not mean that agents must explicitly generate and represent goals. Goals may be regarded as commitments or obligations as well. The goal-directed behavior does not necessarily mean that agents meet their goals, but their actions are directed by goals.. The actions that agents perform and the effects of these actions. Agents do not typically have full control over their environment. But they are able to influence their environment through actions, which means they may have reliable control over some portions of environment.. Agents have some information about their environment. Firstly, information may be imperfect. Agents’ sensors may be faulty, information may be out of date, or agents’ can be deliberately given false information. Secondly, the information an agent has is not directly available to others, because agents do not share and do not have private access to data structures of other agents. Thirdly, environment contains other agents, each having its own information about the environment..

(34) 23. Table 6. A standard protocol for describing an AB model. (Adapted from Grimm et al. 2006, p. 117-119) Purpose. DETAILS. DESIGN CONCEPTS. OVERVIEW. State variables and scales. The purpose of a model has to be stated first. The purpose is the reason why a model is made, what the core functionality is (what the model can do), and why some aspects of reality are included while others are ignored. First, the full set of state variables should be described. State variables characterize low-level entities in a model, i.e. individuals or habitat units. Second, higher-level entities should be described, for example a population consisting of individuals. An important distinction between low-level state variables and auxiliary or aggregated variables has to be done: auxiliary variables are deduced from low-level entities and their low-level state variables. Low-level state variables describe elementary properties of the model’s entities. Auxiliary variables aggregate information from the model entities.. Finally, in addition to state variables, the numerical scales used should be stated. These include length of time steps and time horizon, size of habitat cells (if the model is grid-based), and extent of the model world (if the model is spatially explicit). Choosing a scale is a fundamental decision determining the overall design of a model. The dimensions must be clearly defined for all parameters and variables. Process A conceptual overview of individual and environmental processes should be given. overview This is done by giving a verbal description of each process and its effects. Also, the and sched- scheduling of the model processes should be described. This deals with the order of uling processes and the order in which the state variables are updated. Design concepts provide a common framework for designing and communicating AB models. At least these design concepts should be described (if they are relevant in a model): (1) emergence: which system-level phenomena emerge from individual traits, and which are imposed; (2) adaptation: how do agents adapt their behavior in relation to other agents and environment; (3) fitness or objectives: how is goal-directed behavior being modeled; (4) prediction: how do agents predict future conditions and how this affects their behavior; (5) sensing: what internal and environmental state variables are agents assumed to sense and consider in their adaptive decisions; (6) interaction: what kind of interactions among agents are assumed; (7) stochasticity: if stochasticity is part of the model, what are the reasons for this and how does it emerge; (8) collectives: are agents grouped into some sort of hierarchical groups; and (9) observation: how are data collected from the model for testing, understanding and analyzing? InitializaThis deals with questions as: how are environment and agents created at the start of tion a simulation run, and what are the initial values of state variables? Is initialization always the same or varied between runs? How were the initial values chosen? Input The dynamics of many AB models are driven by some environmental conditions that change over time and space. Environmental conditions are referred as input to the model outputs, in other words, output gives the response of the model to the input. Thus, it is vital to specify, what inputs are used, how they are generated, and how they can be generated or obtained. Submodels All submodels representing the processes described in the overview-stages are explained in detail, including the parameterization of the model. As this usually requires a lot of space (dependent on the model), and mathematical expressions are harder to understand than verbal ones, it is meaningful to divide the explanations of submodels into two: the first is a “mathematical skeleton”, consisting only of equations, rules and parameters, and their very short and simplified explanations. And the second is a full model description, presented in the very same structure as the mathematical skeleton, but including fully detailed verbal explanations..

(35) 24. There are a few possibilities for designing agents technically. “The production system” is the simplest: it consists of a set of rules, a working memory and a rule interpreter. Rules consist of two parts: a condition, which specifies when the rule is executed; and an action part, which determines what happens when the rule is executed. The working memory keeps track of an agent’s state, which may affect its behavior. The rule interpreter considers each rule in turn, fires those whose conditions are met, performs related actions, and repeats this cycle indefinitely. Though the production system is a simple concept, by using it, it is relatively easy to build reactive agents that respond to stimuli from the environment with some action. In the production system agents have potential to learn about their environment and from the other agents by adding knowledge to their working memories. However, the rules always remain unchanged. By using adaptive algorithms, it is possible to change agents’ rules, which simulates learning in a more fundamental level. In addition, agents are almost always modeled as operating in some environment that consists of a network of interactions with other agents. Sometimes, it is also useful to model a physical environment that imposes constraints on the agents’ location. (Gilbert et al. 2000, p. 62-63). One key question regarding modeling is: how can we say that the modeled results are believable? And this question yields a few key sub-questions to be answered: what is the correct level of resolution in a model; does a certain modeled behavior explain an observed pattern in the real world; and how do we cope with uncertainty regarding model parameters? These questions can be answered by applying the so called pattern-oriented modeling methodology: 1. Finding the correct level of resolution in a model is done by observing multiple patterns in the real world, and designing the model based on these patterns. This makes the model structurally more realistic, i.e. its ability to produce independent predictions that match observed patterns is increased. 2. To determine whether a certain modeled behavior explains an observed behavior in the real world, it is needed to formulate alternative theories, which all are being modeled. If a theory being modeled is incapable of producing.

(36) 25. realistic results, it is rejected. If a theory can produce realistic results in modeling, it is much more likely to be the right theory, as alternative explanations are rejected. 3. Uncertainty regarding model parameters is reduced by making the model structurally more realistic (the first point), and this helps parameters to interact in ways similar to real mechanisms. It is therefore possible to test whether different parameter sets can produce all the patterns observed in the real world. If a parameter set cannot produce all the observed patterns, it is ruled out. All the parameter sets, which are able to produce realistic results, are accepted and thus, uncertainty regarding parameters is reduced. (Grimm et al. 2005, p. 987-991) 2.2.3. Main benefits of using agent-based modeling in economic studies. We will see later in the chapter 2.3.3 that mathematical or equation-based (EB) models are the most popular type among scientists in modeling maintenance strategies. There are, however, some shortcomings in using equation-based modeling (EBM) techniques in studying economic phenomena, and a few key benefits in using ABM instead. This is why ABM and EBM are compared here.. If compared to traditional EBM techniques such as system dynamics, ABM is fundamentally different: in ABM, internal behavior of an individual agent is represented, and (usually) not by equations but by step-by-step processes and condition decisions, for example. This behavior may be affected by other individuals, but it is not (usually) directly modified by others. In EBM, however, a system is represented by equations that (usually) relate individuals to each other and their behavior is changed according to other individuals. The bracketed ‘usually’ in the previous phrases points out that AB and EB models could be made differently, but these are the core principles presented above. And these differences between ABM and EBM favor ABM in two key aspects in business-related applications:.

(37) 26. 1. Each firm / operational unit can have its own agents and their behaviors do not have to visible to the rest of the system, so the units can maintain proprietary information about their internal operations. In EBM, making this happen would require disclosure of relationships between units, which does not naturally represent commercially important boundaries between units. 2. In many cases, simulation of a system is part of a larger project whose desired outcome is a control scheme. That scheme regulates the behavior of the entire system. For example in supply networks, different firms try to meet their own targets for, say, inventory levels in order to optimize overall inventory levels in the whole supply-chain. In ABM this can be naturally modeled, as individual agents try to meet objectives related to control scheme, and by interacting with each other, they may be able to reduce overall inventory levels of all individuals. If this scheme is considered with EBM, the control of all inventory levels is handled by an external control equation, which has to handle all specific relationships between firms and factors related to inventory levels. (Parunak et al. 1998, p. 11-14). Thus, the main weakness of EBM is that equations do not represent decision making in the real world. As modeling a problem by equations requires abstraction and simplification (modeling in general requires this, but there are more and less powerful methods to cope with this), the results from modeling are limited. And if abstraction and simplification are tried to be kept at minimum, the resulting equations are usually so hard to calculate that there is no way to acquire new knowledge. Examples of the shortcomings of EBM are seen in the classical micro-economic theories: they oversimplify many things, such as the assumption of perfect decision making by homogenous consumers, what obviously contrasts with the reality. These assumptions are basically made because only then it is possible to analyze these systems with EB methodologies. (Gilbert et al. 2000, p. 57-58).

(38) 27. In ABM, the real world structures are represented in a very similar way in a model, which makes AB models structurally more realistic compared to EB models. ABM offers a methodology to connect macroeconomic and microeconomic phenomena to each other, without imposing top-down constraints as in EBM. AB models offer endogenous explanations to phenomena, rising from individual agent’s behavior and interaction between agents, rather than from exogenous equations. In addition, use of ABM can facilitate the development and experimental testing of integrated theories built on studies from many different fields of social science. Moreover, ABM can be used to test economic theories developed by using more standard methodologies. Finally, ABM can be used to test for the possibility that multiple distinct microstructures can support a given macro regularity. (Tesfatsion 2002, p. 55-82; Tesfatsion 2003, p. 263-264). On the fundamental level, ABM can answer for the requirements in modeling social and collective phenomena. First of all, the real foundation of all sociality, e.g. cooperation, competition, groups and organizations, is the individual social action and mind. It is impossible to connect action at the collective level to action at the individual level without considering individual characteristics. Secondly, important levels of coordination and cooperation necessarily arise from individual behavior (beliefs, desires, intentions etc.). Thirdly, cooperation does not only arise from deterministic behavior of individuals, but also from emergent pre-cognitive structures and constraints of individuals. (Castelfranchi 1998, p. 157-158) 2.3. Modeling and optimization of maintenance strategies. In this chapter, modeling and optimization of maintenance strategies is investigated. First, the motives for modeling and optimization, and the basics of maintenance models are summarized. Then, a few other than AB models and previous AB models on maintenance are studied. Worth mentioning is that lots of maintenance optimization models have been made through the years. It is not possible and not of interest to ana-.

(39) 28. lyze many models in this study. The study on some previous models helps in determining the perspective and foundations of the AB model made for this study. 2.3.1. Motives for modeling and optimization. Maintenance optimization models allow strategies to be evaluated and compared to each other. In addition, models can reveal optimal strategies in a given situation. (Dekker 1996, p. 231-232) However, in modeling there is always a risk of going “too deep”: that means, often the real-world data and applicability of models are not taken into account, but the focus is rather on pure theoretic trickery. However, as the modeling of maintenance management is certainly applied science, it should be directed by real-world applications, if it is wanted to be valuable for practitioners. Although, especially mathematical modeling in maintenance faces a serious problem of absence of sufficient data related to the problem to be solved. The solution to this is to collaborate more with managers and engineers in order to gather not only quantitative data but also qualitative information of maintenance processes. Often the qualitative information is even more valuable than quantitative: for example, determining a deterioration function for a system based on failure rates, and without knowing how the system works, cannot capture the real deterioration process. (Scarf 1997, p. 493-494). It is a fact, however, that optimization models are much more precise and can offer much more to maintenance decision making than “purely qualitative” approaches such as RCM and TPM, yet at a cost of increased complexity. Problem structuring and sharing with others, well-organized data collection and analysis, development of models with the problem owners and application of information technology are necessities for taking advantage of optimization models. Also, integration of optimization and qualitative techniques should be made. (Dekker et al. 1998, p. 118) 2.3.2. Principles of modeling and optimization. In general, maintenance optimization models cover four aspects: (1) the description of a technical system, how it functions and what the importance of it is; (2) how the.

(40) 29. system deteriorates over time and what the consequences of this deterioration are to other system components; (3) a description of the available information regarding the system and open possibilities of actions for management; and (4) an objective functions of the operators in the model and an optimization technique which helps in finding the best balance. (Dekker 1996, p. 231-232). Every maintenance model incorporates prediction or extrapolation of future performance of a system, whether it is deterministic or probabilistic. Uncertainty of the future performance is always present in the real life situations, at least to some degree, which means probabilistic approach to modeling is necessary. (Frangopol et al. 2004, p. 197-198) Mathematical and probability-based maintenance models are complex, but they usually provide the most accurate results in forecasting and optimizing maintenance strategies. (IEEE / PES Taskforce 2001, p. 643-644). In the figure 6 it is shown a conceptual view of maintenance activities over time. The figure illustrates how a system may operate over time and how maintenance acts on or responds to functional condition of the system. Worth noticing is that degradation time of a system is dependent on the technical aspects of the system, and also on the age of the system. Usually, degradation time decreases (i.e. degradation is faster) as the age of a system increases. That is due to imperfect maintenance: usually, a system cannot be restored to initial condition but the system is younger after each maintenance task. This is mainly due to repairing a wrong part, only partial repair, unintended damaging of a system during maintenance, incorrect assessment of maintenance requirements, wrong timing of maintenance, hidden faults, human errors and / or replacements with faulty parts. (Pham et al. 1996, p. 425) There are actually many optimization models, which study optimal maintenance strategies on infinite time span. This is the case basically because it is theoretically easier to study optimal strategies on infinite rather than finite time span. In reality, however, infinite operating time is impossible and thus, optimization models should take into account the age of equipment. (Nakagawa et al. 2009, p. 89-90).

(41) 30. Functional level. Upgrade Degradation time 2. Degradation time 1. Improvement. CM. PM. Required functional level. Change of the required functional level. Functional degradation Initial malfunction. Failure Delays in maintenance service. Time. Figure 6. A conceptual view of maintenance activities over time. (Adapted from Takata et al. 2004). In the table 7, the most relevant maintenance optimization variables are summarized and related aspects are outlined. Not all of these variables are being optimized in AB model made for this study, and actually in most maintenance optimization models only one variable is being optimized. In this study, the optimal maintenance strategy is defined to be the one that maximizes profitability. In addition, there are quality and risk related variables in the model that affect the profitability optimization..