Decision models can help companies determine the value of maintenance. There are many decision models designed to determine optimium maintenance policies in the literature (Horenbeek, Liliane Pintelon, and Muchiri 2010). Maintenance optimization models are mathematical models which quanitfies both the drawbacks (costs) and benefits of maintenance, and allows an optimum balance between the two to be found. Indeed, Dekker (1996) defines maintenance optimization models as follows:
Maintenance optimization models are mathemetical models whose aim it is to find the optimum balance beween costs and benefits of maintenance, while taking all sorts of constraints into account.
However, the gap between academic models and application in a business specific context is currently the biggest problem encountered in the field of maintenance optimization (Horenbeek, Liliane Pintelon, and Muchiri 2010). One reason for this gap is that many papers have been written for maths purposes only. Mathematical analysis and techniques, rather than solutions to real problems, have been the focus of many papers on maintenance optimization models (Dekker 1996). It can be argued that maintenance optimization should not start with developing a maintenance model and trying to fit an application to it, but it should start with an application and try to fit a maintenance optimization model to it.
Maintenance optimization consists of more than a maintenance optimization model. The main ingredients of a maintenance optimization method are the chosen performance indicators, their predictive models, the selected types of maintenance, the optimization problem formulation, and the solution technique (Okasha and Frangopol 2009). In the following text, performance indicators/optimization criteria are discussed in Section 1.3.1, multi-component maintenance optimization models in Section 1.3.2, and optimization problem formulation is part of the discussion on decision support systems in Section 1.3.3. Simulation-based solution techniques is the topic of Section 1.5.
Maintenance objectives can be divided into five categories (Dekker 1996; Horenbeek, Liliane Pintelon, and Muchiri 2010); ensuring system function (availability, efficiency and product quality), ensuring system life (asset management), ensuring safety, ensuring human well-being and ensuring optimal capital replacement decisions.
The maintenance objectives are quantified by corresponding maintenance performance indicators, or a combination of maintenance performance indicators. Maintenance performance indicators should ideally be selected to represent only the performance of the maintenance organization, and be independent of non-related effects such as differences in production volumes or changes to the operating environment. This is a tricky task, as the benefits of maintenance improvement tend to show up in other areas like production, quality or inventory. The maintenance organization itself may show, for example, higher cost. Another problem is knowing which maintenance criteria are the most relevant for a particular maintenance organization. In an attempt to alleviate these problems Horenbeek, Liliane Pintelon, and Muchiri (2010) has created a generic list of maintenance optimization criteria to serve as a starting point when selecting objectives for a particular problem:
A large number of maintenance optimization models have been published over the years (Cho and Parlar 1991; Dekker and Wildeman 1997; Wang 2002; Horenbeek, Liliane Pintelon, and Muchiri 2010), but the vast majority of them focus on a single optimization objective. Multi-objective optimization models, in particular those considering multiple-components, is an underexplored area of maintenance optimization.
While there are plenty of maintence optimization models focusing on single components, typically to determine optimal replacement intervals of that component, any real world maintenance management situation requires a focus on multiple components and/or multiple systems. With increasing complexity of the maintained equipment, maintenance optimization models need to be able to describe the interaction between multiple components and units. Cho and Parlar (1991) defined multi-component maintenance optimization models as follows:
Multi-component maintenance models are concerned with optimal maintenance policies for a system consisting of several units of machines or many pieces of equipment, which may or may not depend on each other (economically/stochastically/structurally).
The fact that up to 10 years ago, the vast majority of the maintenance models were concerned with one single piece of equipment operating in a fixed environment, was considered as an intrinsic barrier for applications (Wang 2002). Consequently, many articles in the area of multi-component maintenance optimization models have been published as a result of recent research. However, currently, there seem to be very few models that consider multiple types of dependencies within the same model (Nicolai and Dekker 2008). There are some recent works, but it is a relatively unexplored area of maintenance modeling.
Economic dependence implies that costs can be saved when several components are jointly maintained instead of separately. Alternatively, the opposite can be true so that simultaneous downtime of components is undesirable and hence maintenance must be spread out over time as much as possible (Dekker 1996). In the first case, economic dependence means that benefits can be achieved by grouping maintenance actions together. In the second, grouping maintenance likely leads to higher costs.
Stochastic dependence occurs if the condition of some components influences the lifetime distribution of other components. Stochastic dependence is sometimes referred to as failure interaction or probabilistic dependence (Nicolai and Dekker 2008). The degradation or failure of one component will in turn cause failure, degradation or an increased wear rate of stochastically dependent components.
Structural dependence applies if components structurally form a part, so that maintenance of a failed component implies maintenance of working components (Nicolai and Dekker 2008). Structural dependence thus interferes with a maintenance managers options when it comes to grouping of maintenance actions, that may be desirable from an economic dependence point of view. One example of structural dependence is replacement units, where multiple parts are treated as one at the time of replacement.
Decision support systems can help an individual make better decisions through enhanced situational awareness; problem recognition, problem structure, information management, statistical tools, and by suggesting solutions through application of knowledge and optimization techniques. However, the nature of building a decision support system for a specific decision process, and tailoring it to specific managers, makes it hard to identify a generalized approach to building decision support systems (Santana 1995). To help structure the area, decision support systems can be classified into passive, active or cooperative decision support systems. A passive decision support system is a system that aids the process of decision making, but cannot bring out explicit suggestions or solutions. An active decision support system has this ability, while a cooperative decision support system allows the decision maker to modify and refine suggestions provided by the system (Falk 2008).
Passive decision support systems assists the decision maker, for example by collating and filtering relevant information, but does not explicitly provide suggested improvements or solutions. An abundance of examples of passive decision support systems can be found among the large number of computerized maintenance management systems, devoted to help managing maintenance activities, listed on the Web site www.plant-maintenance.com (over 360 software packages). However, maintenance optimization is typically not included as a feature in these software packages, which makes them excellent databases to track repair orders and maintaining appropriate book-keeping (Nguyen and Bagajewicz 2010).
Active decision support systems explicitly suggests improved solutions. If a maintenance optimization using a single-objective model is performed, a single best solution would be presented as the outcome. However, computers can not think about reality. They work with a model and human operators are required to relate the model to reality (Falk 2008).
A major shortcoming of most maintenance decision support systems is that they act like a black box (Dekker 1996). Since each maintenance problem is likely to be different, it is only the user (and not the decision support system) who can validate the calculations and convince his/her management of their value. Similar concerns have been raised among the Swedish armed forces, where several decision support systems have have already been developed and marketed but not always with success. Interviews performed with Swedish officers suggested one reason for these failures; human contributions can not be excluded from the analysis and are often vital for the result. An officer expressed his experience succinctly: “A decision support system should give me time to think” (Falk 2008).
A cooperative decision support system allows the decision maker to modify and refine suggestions provided by the system. This helps alleviating the problems users experience in applying the results of active decision support systems to the real world situation, by facilitating integration of the human contributions into the decision making process.
By utilizing a multi-criteria maintenance optimization model, and a corresponding multi-criteria optimization algorithm, a Pareto-optimal set of suggested solutions can be generated. The benefits of simultaneous optimization of different criteria has been noted in the literature. For example, Wang (2002) notes that to achieve the best operating performance, an optimal maintenance policy needs to consider both maintenance cost and reliability measures simultaneously, and D. Murthy, Atrens, and Eccleston (2002) argue that operating load and maintenance strategies need to be optimized jointly since the load degrades the equipment and maintenance actions control the degradation. By creating and presenting a Pareto-optimal set of solutions, the user will be able to perform a trade-off analysis among the modeled criteria, and selecting candidate solutions for implementation by integrating real world knowledge not included in the model.
Decision models can help companies determine the value of maintenance. Maintenance optimization models are mathemetical models whose aim it is to find the optimum balance beween costs and benefits of maintenance, while taking all sorts of constraints into account. The main ingredients of a maintenance optimization method are the chosen performance indicators, their predictive models, the selected types of maintenance, the optimization problem formulation, and the solution technique (Okasha and Frangopol 2009). This section discusses performance indicators/optimization criteria, multi-component maintenance optimization models and optimization problem formulation as part of creating decision support systems. Simulation-based solution techniques is the topic of Section 1.5.
Multi-objective optimization models, in particular those considering multiple components, is an underexplored area of maintenance optimization. Multi-component maintenance models are concerned with optimal maintenance policies for a system consisting of several units of machines or many pieces of equipment, which may or may not depend on each other (economically / stochastically / structurally) (Cho and Parlar 1991). Decision support systems can be classified into passive, active or cooperative decision support systems. A passive decision support system is a system that aids the process of decision making, but cannot bring out explicit suggestions or solutions. An active decision support system has this ability, while a cooperative decision support system allows the decision maker to modify and refine suggestions provided by the system (Falk 2008). Creation of a cooperative decision support system will be facilitated by the application of multi-objective optimization to create a Pareto-optimal set of solutions.