A New Belief Rule Base Considering Explainability Optimization for Complex System Health State Assessment

Health status assessment of complex systems is a critical method to ensure their safety. The belief rule base (BRB) is essentially a white-box expert system. However, experts cannot fully understand the mechanism of complex systems, the parameters of the BRB need to be adjusted through optimization algorithms. Due to the optimization algorithm has a random exploration mechanism, the optimized expert knowledge cannot be effectively retained, which reduces the explainability of BRB. Therefore, a new belief rule base considering explainability optimization (BRB-EO) for complex system health status assessment is proposed. In BRB-EO model, a modified whale optimization algorithm (WOA) for explainable BRB is proposed. Firstly, a new scattering method centered on expert knowledge is proposed to effectively utilize expert knowledge. Then, an explainability constraint is proposed to preserve the expert knowledge after optimization. Finally, the effectiveness of the BRB-EO is demonstrated through a NASA lithium battery case study.