| dc.description.abstract |
Performing root cause analysis in technical systems is usually challenging owing to the complex failure associations which often exist between inter-connected system components. The recent adoption of maintenance management systems in industry has enhanced the collection of maintenance data which could assist practitioners derive meaningful failure associations embedded in the data. However, root cause analysis in the maintenance domain is dominated by the use of qualitative and semi-quantitative approaches. Such approaches, however, rely on expert elicitations whereof this elicitation process often introduces bias in the root cause analysis process. On the other hand, quantitative techniques for root cause analysis, for instance, fault trees and Bayesian networks are often limited to analyzing root causes in fairly simple systems. Moreover, the quantitative techniques seldom model the failure dependencies linked to the empirical failure events. Hence, to address these challenges, a novel data exploration methodology for root cause analysis is proposed which consists of four steps: 1) data collection and standardization step; 2) data exploration framework incorporating multivariate and cluster analysis; 3) causal mapping; and 4) maintenance strategy selection. The methodology is demonstrated in the application case of thermal power maintenance data. Moreover, the methodology is compared with two conventional qualitative root cause analysis techniques – Ishikawa cause-and-effect diagram, and the ‘5-whys’ analysis. A detailed discussion is presented whereof the added value of the methodology for maintenance decision support is demonstrated. |
en_US |