Rolling Bearing Compound Fault Diagnosis Based on Parameter Optimization MCKD and Convolutional Neural Network

• Published in: IEEE transactions on instrumentation and measurement Vol. 71; pp. 1 - 8
• Main Authors: Gao, Shuzhi, Shi, Shuo, Zhang, Yimin
• Format: Journal Article
• Language: English
• Published: New York IEEE 2022; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Artificial neural networks; Background noise; Compound fault; Compounds; Convolution; convolutional neural network (CNN); Convolutional neural networks; Diagnostic systems; Fault detection; Fault diagnosis; Feature extraction; Kurtosis; maximum correlated kurtosis deconvolution (MCKD); multi-strategy cuckoo search algorithm (MSACS); Neural networks; Noise reduction; Optimization; Parameters; Particle swarm optimization; Roller bearings; rolling bearing; Rolling bearings; Search algorithms; Vibrations
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2022.3158379

For the sake of solving the problem of the difficulty of extracting fault features under the background of noise and accurately identify the state of the bearing, a compound fault diagnosis method of rolling bearing based on parameter optimization maximum correlated kurtosis deconvolution (MCKD) and convolutional neural network (CNN) is proposed. First, the adaptive multi-strategy cuckoo search algorithm (MSACS) is used to iteratively optimize the important parameters of MCKD. Second, the optimized MCKD is used to filter and denoise the rolling bearing fault signal, and the denoised signal is obtained. Finally, the denoised signal is input to the CNN model for training and testing to obtain the classification result of fault diagnosis. Through the test and evaluation of the fault dataset, the proposed method is compared with particle swarm optimization (PSO) parameter optimization method (PSO-MCKD-CNN) and CNN method without noise reduction. At the same time, it is compared with other advanced methods. The experimental results shows that this method improves the diagnostic performance of the neural network, obtains higher diagnostic accuracy, and is more conducive to the detection of compound faults.