Fine tuning machine fault diagnosis system towards mission critical applications

Abstract

Machine condition monitoring has become increasingly important in enhancing productivity, reducing maintenance cost, and casualties in mission critical applications (for example, aerospace). It may be noted that we cannot afford to miss an alarm (fault) condition in a mission critical application as the cost of a failure can be fatal and very expensive, and this means that we need to have a high alarm accuracy, i.e., sensitivity. A better alarm accuracy is desired in mission critical applications, even at the expense of a lower no-alarm detection accuracy, i.e., specificity. A lower specificity means more false alarms. In this work, we propose a novel approach to fine tune a decision tree based machine fault identification system towards a mission critical application like aerospace by improving the sensitivity. We experiment the proposed approach with a three phase 3kVA synchronous generator to diagnose the inter-turn short circuit faults. The proposed approach outperforms the baseline system by an absolute improvement in sensitivity of 1.08%, 0.82%, and 0.53% for the R, Y, and B phase faults respectively.