Failure Prognosis using Dynamic Bayesian Networks and Decision Trees

Production managers commonly need to assess the reliability of production equipment throughout its life cycle. Since equipment is typically composed of multiple components, failures may stem from dependencies due to interactions. We have found that the literature does not adequately address these dependencies in the dynamic reliability models of the studied systems. The purpose of this study is to utilize decision trees (DT) and Dynamic Bayesian Networks (DBN) to approximate the failure probability of multi-component systems. We present a failure prediction method and utilize a parametric estimation approach grounded in a priori laws. The DBN learning process is assessed via a Monte-Carlo simulation method and a parametric compliance examination. Finally, this study illustrates the significant impact of using Dynamic Bayesian Networks (DBNs) in conjunction with Decision Trees (DTs) to evaluate the state of a water production system.