Diagnosis Approach for Incipient Faults of Rolling Bearings Under Sectional Jumping Speed

Since the characteristics of the incipient faults for the rolling bearing under the sectional jumping speed are difficult to be extracted, the conventional fault diagnostic approaches usually with poor monitoring performance and low diagnostic accuracy (i.e., the ratio of samples correctly classified to total samples). This paper proposes a novel cooperative diagnostic approach for the incipient faults of the rolling bearing based on the optimized local mean decomposition (LMD) and support vector machine (SVM). First, to resolve the problem of selecting the appropriate number of product function (PF) components, an optimally weighted fusion model of PF components is established by introducing a genetic algorithm, aiming to maximize the correlation with the original incipient fault signal. Besides, in this model, a novel rule is set to calculate the weight coefficients of fusion. Second, considering the sparsity of the incipient fault characteristics caused by the sectional jumping speed, from the perspective of energy distribution, a novel characteristic extraction model is constructed based on the equal interval energy projection. This characteristic extraction model can correctly extract the fault characteristics and effectively eliminate redundant information. Moreover, the SVM has collaborated with the above-mentioned characteristic extraction model to diagnosis incipient faults. Finally, the effectiveness and correctness of the proposed approach are verified by the experimental simulation results. The comparison and analysis show that the proposed algorithm cannot only correctly extract fault characteristics but also have a high accuracy of fault characteristics recognition with good operability and scalability.