Statistical characteristics clustering–based adaptive debris signature identification method for inductive sensors with dual phase

Abstract Wear monitoring plays an important role in the early warning of mechanical equipment failures and in predicting the operational life. Inductive sensors provide data support for wear analysis by monitoring and extracting key features of oil debris online. However, the low identification accuracy of tiny metal particles under complex interference remains a critical factor that limits the detection sensitivity. The sensors with dual probes (DPS) utilize a double-induction structure to enhance noise reduction and debris perception using correlation analysis. Nevertheless, the performance of debris signature identification still faces challenges related to dependence on previous knowledge, insufficient sensitivity, destruction of features, and weak generalizability. In this study, we propose a novel debris signature identification method, named GIM-SCC. By constructing a global independence metric (GIM), time series samples are transformed into characterization vectors. Next, debris identification is achieved by statistical characteristic clustering (SCC). Using numerical simulations and experiments, we demonstrate the advantages of this method in terms of signature identification accuracy, robustness, feature protection ability, and generalization capability via algorithm comparison. This contribution is expected to provide reliable technical support for the accurate extraction of debris signatures via inductive sensors with DPS.