【摘要】 针对滚动轴承故障诊断中的时频特征自适应提取与智能诊断问题，提出一种基于优化变模态分解和卷积神经网络的旋转机械故障诊断方法。通过粒子群算法优化VMD算法得到核心参数的最优组合，经过PSO优化的VMD算法将原始轴承信号分解为一系列不同的本征模态函数，提取其中有效的IMF分量，再将有效的IMF分量输入到CNN网络里训练，利用CNN网络进行自主特征学习和模式识别。提出的方法克服了传统VMD分解而引起的过度分解或分解不彻底问题，有效提高了分类精度，获得的故障诊断准确率最高。更多还原

【Abstract】 In view of the problems of time-frequency feature adaptive extraction and intelligent diagnosis in bearing fault diagnosis, a rotating machinery fault diagnosis method based on optimized variable mode decomposition(VMD) algorithm and convolution neural network(CNN) is proposed, in which VMD algorithm is optimized through particle swarm optimization(PSO) to obtain the optimal combination of core parameters, then the original bearing signal is decomposed into a series of different intrinsic mode functions(IMF) through VMD algorithm optimized by PSO, the effective IMF component is extracted, the effective IMF component is input into CNN network for training, and CNN network is used for autonomous feature learning and pattern recognition. The proposed method overcomes the over-decomposition or incomplete decomposition caused by traditional VMD decomposition, and it effectively improves the classification accuracy, achieving the highest fault diagnosis accuracy.更多还原