Designing an Improved VMD-CNNs-DS Algorithm for the Port Crane Bearing Fault Diagnosis

Abstract

Accurate diagnosis of bearing faults is crucial for detecting faults in rotating machinery. However, the variability and complexity of recorded vibration signals present a significant challenge in distinguishing unique characteristic fault features. To address this issue, we proposed VMD-CNNs-DS, a novel bearing fault diagnosis algorithm based on data-level and decision-level bipolar fusion network. The adaptive variational mode decomposition (VMD) method based on parameter optimization is employed to transform multiple sensor raw vibration signals into time-frequency analysis. The convolutional neural network (CNN) takes as input a graph of the features of the vibration signals from the two sensors. Improve DS theory for decision blending. An extensive evaluation of the fault diagnosis performance of VMD-CNNs-DS on various data sets shows that VMD-CNNs-DS achieves better results than a single CNN network model by fusing complementary or conflicting evidences from different models and sensors and adapting to different load conditions.