An Evidential Reasoning Rule-Based Quality State Assessment Method of Complex Systems Considering Feature Selection

Quality state assessment (QSA) is a top priority to ensure the quality and reliability of complex system repairs. In complex systems, a large number of features can reflect complex nonlinear associations with multiple information sources. However, the current research on the QSA of complex systems based on the evidential reasoning (ER) rule generally ignores the complex relationship between features, increasing maintenance difficulty. Therefore, a novel ER rule-based QSA approach considering feature selection (FS) for a complex system is proposed in this article. The correlation and redundancy factors are defined by introducing the Spearman correlation coefficient and mutual information. A hybrid weight factor is employed to balance the significance of redundancy and relevance, which can measure the correlation and redundancy between features. Second, a hybrid weight coefficient is introduced into the ER rule, and a QSA model considering FS is proposed. Third, a parameter optimization model based on the ER rule is proposed, and the optimized ER rule model can adaptively select informative and balanced features. Finally, an ER rule-based QSA fusion mechanism is put forward to fuse the selected features to obtain the quality state of the complex system. A numerical study is taken out to demonstrate the validity and performance of the approach. Furthermore, an applicable experiment of an inertial navigation system (INS) is introduced to prove the application of the ER rule-based QSA approach.