A data- and ontology-driven text mining-based construction of reliability model to analyze and predict component failures

A real-life reliability system is proposed by fusing the field warranty failure data with the failure modes extracted from unstructured repair verbatim data by using the ontology-based natural language processing technique to facilitate accurate estimation of component reliability. Traditionally, the reliability estimation process uses the warranty data, but it provides limited support to handle the “failure confounding” problem, whereby different failure modes associated with a component failure are confounded into a single failure mode. The resulting reliability estimation lacks the required level of precision. Because our model takes into account textual failure modes associated with component failures, it enhances the overall reliability estimation. The performance of our system is evaluated with the baseline system for predicting absolute errors by using the real-life data from the automotive domain, e.g., headlamp failure, collected at different miles exposures. In the best case, the absolute errors predicted by our model showed an improvement of 97 % with respect to the baseline model (without considering the failure modes), while in worst case, it was 71 %.