Data driven approach for fault detection and identification using competitive learning

Abstract

Condition Based Maintenance (CBM) is the process of executing repairs or taking corrective action when the objective evidence indicates the need for such actions or in other words when anomalies or faults are detected in a control system. The objective of Fault Detection and Identification (FDI) is to detect, isolate and identify these faults so that the system performance can be improved. When condition based maintenance needs to be performed based on just the data available from a control system then Data Driven approach is utilized. The thesis is focused on the data driven approach for fault detection and would use: (i) Unsupervised Competitive Learning, (ii) Frequency Sensitive Competitive Learning, (iii) Conscience Learning and (iv) Self Organizing Maps for FDI purpose. This approach would provide an effective Data reduction technique for FDI so that instead of using the complete data set available from a control system, pre-processing of the available data would be done using vector quantization and clustering approach. The effectiveness of the developed algorithms is tested using the data available from a Vertical Take off and Landing (VTOL) aircraft model.