Feature extraction for periodic impacts based on singular gini index

Abstract Impulsivity and periodicity are key characteristics of bearing impact faults, which can be more advisable than energy-oriented methods when applied to develop effective singular component selection strategies. Therefore, unlike standard singular value decomposition methods, a fault feature extraction technique based on singular value sequence distribution and the Gini index is proposed. This approach employs the Gini index as a metric to assess both the impulsiveness and periodicity characteristics of individual components. It can describe the richness of fault information in singular components. Based on the proposed evaluation criteria, an effective singular component retention strategy is proposed, which can identify components with weak energy but rich fault information in vibration signals. Through simulation data analysis and application examples, the validity of the approach is confirmed.