Adaptive Multiscale Convolution Manifold Embedding Networks for Intelligent Fault Diagnosis of Servo Motor-Cylindrical Rolling Bearing Under Variable Working Conditions

The rolling bearing of the servo motor is widely used in precision-controlled mechanical systems. It usually works at variable speed and load, possibly resulting in partial bearing failure. Meanwhile, the varying conditions may cause the smearing of classable features, increasing the diagnostic difficulty. To this end, an intelligent fault diagnosis method of servo motor-cylindrical roller bearings based on adaptive multiscale convolution manifold embedding networks (AMCMENet) under variable working conditions is proposed. The core of the proposed algorithm is to apply the designed intraclass and interclass constraints to reprocess the feature extracted by designed multiscale convolutional neural networks (MSCNN). In this way, the distribution differences of samples could be improved. The training sample under variable conditions is first input to the designed MSCNN for initial feature extraction. Afterward, the constructed locality sensitive discriminant analysis algorithm module is used, which is adjusted to optimal parameters by the particle swarm optimization algorithm, to enlarge the heterogeneous distance and narrow the homogeneous distance of the extracted feature. Finally, the testing subset is provided to the trained AMCMENet algorithm for fault diagnosis. The experimental results of two datasets demonstrate that the proposed intelligent fault diagnosis method performs better under cross working conditions.