Fault detection in a network of similar machines using clustering approach

摘要: Fault detection, which involves the estimation of condition, health or degradation an equipment a process and decision logic to determine whether event that can be considered as fault has occurred, is integral component in prognostics management because it essential indicator when perform diagnosis isolation, also precedes any performance prediction methodology. The implementation data-driven detection generally been reliant on unit-specific models less effective with insufficient training data used applications non-stationary working conditions. aforementioned scenarios alleviated by leveraging from similar units experiencing comparable operating regimes. This dissertation investigates formulation, development cluster-based fleet machines. A two-step approach introduced: clustering local cluster detection. Fleet verifies, discovers identifies group structure network Afterwhich, each unit assessed using peer-to-peer comparison. developed this validated two case studies: industrial welding robots automotive manufacturing facility wind turbines several farms.