Failure mode classification for control valves for supporting data-driven fault detection

Significant losses of production due to unplanned downtimes are a major problem caused by technical failures of equipment. Existing approaches like failure mode and effect analysis try to identify possible equipment breakdowns, their causes and effects in order to quantify the reliability of the system. Yet, they are not used for the detection of faults. On the other hand, Industrie 4.0 and data mining aim to improve the total operating time of automated production systems. However, due to the complexity of automated production systems and the underlying physical phenomena, it is essential to formalize expert knowledge for usage during data analysis. In this contribution a classification table is proposed, in which the expert knowledge on failure modes, underlying parameters and detection features are summarized and presented. This knowledge is used to formulate appropriate detection models. The evaluation for detection of failure modes for control valves showed the usefulness of combination of expert knowledge and data-driven data analysis.