针对基于机器学习模型的故障诊断存在依赖人工特征提取质量、维数灾难问题和卷积神经网络(CNN)模型构建缺乏自适应性等问题,提出了一种基于粒子群优化(PSO)算法的自适应CNN故障诊断方法,并将其应用于旋转机械故障诊断。将一维时域信号变成二维时频图像;使用PSO算法对CNN模型中的7个关键参数进行优化选取,以构建深度学习模型;将二维时频图像输入优化后的深度学习模型,对旋转机械故障进行诊断。结果表明,所提方法具有较高的准确率、稳定性和自适应性。
Aiming at solving the problems of fault diagnosis based on machine learning model, such as relying on manual feature extraction quality, dimension disaster, lack of selfadaptation of convolutional neural network (CNN) model construction,a fault diagnosis method of adaptive CNN based on particle swarm optimization (PSO) was proposed and applied to fault diagnosis of rotating machinery. Firstly, onedimensional timedomain signal was transformed into twodimensional timefrequency image; Then, the PSO algorithm was employed to optimize the seven key parameters in the CNN model to construct a deep learning model. Finally, the twodimensional timefrequency image was input into the optimized model to diagnose rotating machinery faults. The results show that the proposed method has high accuracy, stability and adaptability.