Fault Diagnosis in a belt-drive system under non-stationary conditions. An industrial case study

Fault Diagnosis (FD) is a fundamental step to estimate the Remaining Useful Life (RUL) of a component and then achieve a reliable Predictive Maintenance (PdM) system. The signal-based diagnosis methods received wide research attention since their implementation relies basically on the processing of signals. This is an interesting feature but at the same time it has some limitations. Indeed, the signal-based methodology are usually developed "ad hoc" for the particular application and do not show the same validity under different assumptions. In particular most of the applications deal with systems working in steady-state. Differently, in this paper, is proposed a solution to overcome limitations related to non-stationary conditions. The proposed approach is theoretically evaluated and then tested in an industrial packaging machinery. The results evidence the validity of the strategy to deal with signals acquired in systems which run continuously in non-stationary conditions.