“…Basically, existing methodologies can be categorized into the simulation and the analytical methods. The simulation methods are outstanding in their wide applicability to a variety of scenarios [23,28]; whereas the analytical methods, including binary decision diagram (BDD)-based method [17,24,25], multivalued decision diagram (MDD)-based method [13,16], Markov chains-based method [18], Markov reward model-universal generating function (UGF) technique [7], Bayesian networks approach [4], recursive algorithm [3], have advantages in obtaining accurate results with high efficiency, but may not be suitable in large-scale PMS with complex dynamic behaviors.…”