Evidential reasoning rule with dynamic correlation for system reliability prediction

In engineering, reliability prediction holds crucial significance for ensuring the normal operation of complex systems. Since the reliability prediction involves both quantitative data and qualitative knowledge, the evidential reasoning (ER) rule emerges as a promising prediction approach. However, the ER rule-based prediction model assumes that there is independence or static correlation between different past instants, which is inconsistent with engineering practice. In light of this, a new prediction model based on the ER rule with dynamic correlation is proposed in this paper. In this model, the exponential distributions with temporal information are employed to describe the dynamic correlations between different time instants. Subsequently, the dynamic correlations are utilized to discount the initial evidence and the prediction results are obtained through the nonlinear fusion of multiple pieces of evidence. Besides, several interpretability criteria are set after analyzing the physical meanings of model parameters. Moreover, these criteria are transformed into corresponding parameter constraints, contributing to establishing an optimization objective with interpretability. This can ensure the prediction accuracy while preserving the model interpretability as much as possible. Two engineering examples are carried out to verify the validity of the proposed model.