Optimization of system reliability over multiple periods under constrained maintenance resources

Along with the development of information technologies and control systems, large and sophisticated mechanical systems are able to perform unattended on production and service lines previously managed by humans. Consequently, upkeep of system condition has been a big concern for many organizations. Since the components deteriorate, system is often maintained periodically and its reliability value is evaluated in each period. The thesis models various systems over multiple periods to optimize their reliabilities under constraints and designs experiments to analyze the models and the systems. For this purpose, the system reliability and optimization techniques related literature is reviewed firstly. After different maintenance models are identified and classified, typical system structures and common solution approaches are summarized. Furthermore a selection of publications about system reliability optimization problem is presented. The main part of the thesis starts the discussion with the evaluation of system reliability from component reliabilities in identified system configuration. After developing the calculation formulas, a reliability calculation program is developed based on a defined nomenclature for system configurations using Visual Basic for Applications (VBA). In the next part the thesis models different systems as single machine maintenance problem (SMMP) or multiple machine maintenance problem (MMMP) under limited budgets. The modeled problems are solved using genetic algorithm (GA) and mixed integer nonlinear program (MINLP). After the demonstration of GA procedures, a complex system is used to analyze the models and the parameters of the system by design of experiments (DOE) to unveil the characteristics of the systems. For example, the effect of budget on the optimized value is examined; the relationship between the budget and number of periods when the objective values are the same. In the end the conclusions are summarized and possible further research is recommended