An efficient reliability evaluation method for large-scale multi-performance multi-state series-parallel systems considering multi-dimensional approximation

Multi-performance multi-state series-parallel systems (MPSPSs) considering performance conversion process have been currently observed for representing engineering systems, where different types of performances can be converted into one another. The multi-dimensional universal generating function (UGF) technique is one of the most-frequently used methods for exact reliabilities evaluation of MPSPSs. However, different performance variables and performance conversion characteristics in the large-scale MPSPS enlarge the computational complexity of exact reliability evaluation. Therefore, to efficiently evaluate the reliability of large-scale MPSPS considering performance conversion, this paper proposes an approximate reliability evaluation method, named as multi-dimensional approximation (MDA). Procedures to build the MPSPS model are introduced, whose components are classified into two types corresponding to different performance conversion characteristics. The multi-dimensional gaussian and clustering approximation theories are employed in the MDA, where the states and performance conversion process of different components can be approximated, respectively. Finally, the multi-dimension UGF method is modified to evaluate the system approximate reliability. Illustrative results show that MDA method has advantages in computational efficiency with satisfactory accuracy.