可解释性
计算机科学
机器学习
人工智能
过程(计算)
国家(计算机科学)
数据挖掘
算法
操作系统
作者
Xianyong Yin,Wei He,You Cao,Ning Ma,Guohui Zhou,Hongyu Li
标识
DOI:10.1016/j.ress.2023.109744
摘要
Health state assessment is an important way to maintain system safety. The belief rule base (BRB) is a rule-based model that embeds expert knowledge and has the advantage of interpretability. However, BRB-based models face two problems in interpretability: Lack of reasonable methods to improve the interpretability of the optimization phase and BRB that only considers accuracy tends to favor black-box models. To address the above two problems for developing a highly interpretable model, this paper proposes a new health state assessment method based on interpretable BRB with bimetric balance (IBRB-b). First, interpretability criteria for health state assessment are proposed, and based on the fact that some interpretability criteria may be broken in the optimization process, two interpretable maintenance strategies were developed. Second, the reliability calculation method of expert knowledge is given and considered in the interpretable optimization methods. In addition, the bimetric-oriented population selection method is proposed to balance the accuracy and interpretability of the BRB model. In the case study, the effectiveness and superiority of the IBRB-b method were analyzed and verified through the health state assessment experiments of NASA lithium-ion batteries and milling.
科研通智能强力驱动
Strongly Powered by AbleSci AI