An evidential network approach to reliability assessment by aggregating system‐level imprecise knowledge

Abstract

AbstractModern complex engineering systems oftentimes possess hierarchical structures which can be physically divided into several levels. Traditional reliability assessment methods were conducted using field data or time‐to‐failure data. For a hierarchical system, different levels’ reliability‐related data from monitoring or observations can provide additional information to update the system's reliability. However, with the limitation of sensing and monitoring technologies and vague judgments from experts, the reliability data inevitably contain epistemic uncertainty. Properly fusing the imprecise data can still improve the accuracy of system reliability assessment. In this article, we propose a reliability assessment approach based on the evidential network (EN) model to update the hierarchical systems’ reliability by aggregating the system‐level imprecise knowledge. First, the evidence theory is applied to quantify the epistemic uncertainty associated with the imprecise system‐level reliability data. Then, an EN model is created to realize the uncertainty propagation and fusion of imprecise data. Finally, at any particular time instant, the system reliability can be updated with the aid of imprecise data based on the proposed EN model. The effectiveness of the proposed method is validated by a real engineering case of a wind turbine system. Overall, the results shown from the case study support that the proposed method can update the reliability of hierarchical systems by using system‐level imprecise data.