Multi-state System Reliability Modeling And Optimization With Considering Dynamic Load Distribution Strategy

With the development of advanced industrial techniques, systems and devices are designed with diverse functionalities and complexities. Reliability as a critical measure of products’ quality becomes a basic attribute of products nowadays, and it is oftentimes strictly required in engineered system design and operation. Nevertheless, the traditional binary-state reliability methods fail to provide an accurate characterization of complex degradation/failure patterns manifested by engineered systems with high accuracy, large physical scale, and complexity. The multi-state system reliability methods enrich the reliability theory since it outperforms the traditional binary-state reliability methods in terms of revealing the multi-state nature of complex systems.Most reported works on multi-state system reliability modeling and assessment are developed on the premise the failure processes between any pair of components are statistically independent. However, such hypothesis may not be always held. Components in a system may influence one another during their failure processes, making the deterioration process of the entire system even more complicated. This thesis aims at studying the failure dependency between components under the dynamic load distribution strategy. The major contributions of the thesis include:(1) Constructing a new reliability model for multi-state systems with dynamic load distribution strategy. A general Markov model is built to characterize the mechanism of dynamic load distribution for multi-state systems. The relation between the load imposed on units, reliability of the entire system, and system structure is studied.(2) Jointly optimizing the dynamic load distribution strategy and system configuration to achieve a higher system reliability/performance. A two-level genetic algorithm is put forth to jointly optimize the dynamic load distribution strategy and system configuration with the aim of maximizing the mean time to failure or the expected cumulative performance rate of the entire multi-state system. As the mean time to failure and the expected cumulative performance rate are two conflicting objectives, the non-dominated sorting genetic algorithm is employed to seek a set of Pareto optimal solutions for the resulting multi-objective optimization problem.(3) Taking account of the imperfect fault coverage in multi-state system reliability models. The reliability model for multi-state systems with jointly considering the dynamic load distribution strategy and the imperfect fault coverage is developed. The optimal load distribution strategy and system configuration in this case are also investigated.