基于WOA优化卷积神经网络的齿轮箱故障诊断研究

Research on Gearbox Fault Diagnosis Based on WOA Optimized Convolutional Neural Network

针对传统齿轮箱故障诊断准确率不高、诊断算法耗时长、模型容易陷入局部最优等问题,提出一种基于鲸鱼算法优化卷积神经网络的WOA-CNN模型,结合WOA算法结构简单、参数少、搜索能力强且易于实现等特点,对CNN卷积核的大小、个数、学习率等参数进行寻优,利用所得全局最优参数重构WOA-CNN模型并进行齿轮箱故障诊断。实验结果表明,相比一般神经网络优化方法,WOA-CNN算法能够从原始信号中提取更多有效故障特征,取得更好的识别准确率,有一定的工程应用价值。

Low accuracy,time-consuming diagnosis algorithms,and models that are prone to fall into local optimality are the challenges of the traditional gearbox fault diagnosis.To cope with these problems,a convolutional neural network(CNN)based on Whale Optimization Algorithm(WOA)is proposed in this work.The WOA algorithm with advantages in terms of simple structure,few parameters,strong search ability,and easy implementation is adopted in the developed WOA-CNN model.The size,number,learning rate and other parameters of the CNN convolution kernel are inserted in the program to find the global optimal parameter,to reconstruct the WOA-CNN model and realize the gearbox fault diagnosis.The experimental results show that,the WOA-CNN algorithm can extract more effective fault features from the original signal,obtain better recognition accuracy,and give out more accurate value for engineering application,than the general neural network optimization method.