摘要
针对现有基于深度学习的轴承故障诊断方法不适于小样本数据且超参数过多、计算开销大的问题,提出一种基于深度梯度下降森林模型(DSGDF)的轴承故障诊断方法。在凯斯西储大学轴承数据集上对该模型的性能进行验证,结果表明,DSGDF模型平均诊断正确率达99%以上。DSGDF模型较经典深度森林模型的收敛速度更快,较其他基于神经网络的深度学习模型计算开销更小。
In order to solve the problem that the existing bearing fault diagnosis methods based on deep learning are not suitable for small sample data,too many super parameters and high computational cost,a bearing fault diagnosis model based on Deep SGD-Forest(DSGDF)is proposed. The method is based on the deep forest model,and the gradient descent algorithm is integrated into the multi-grained scanning and the cascade layer to improve the convergence speed of the model. Experimental results on the bearing dataset of Case Western Reserve University show that the average diagnostic accuracy of the DSGDF model is more than 99%. Compared with the deep forest model,the DSGDF model has faster convergence speed and less computational overhead compared with other deep learning models based on neural networks.
作者
彭启明
邵星
王翠香
皋军
PENG Qi-ming;SHAO Xing;WANG Cui-xiang;GAO Jun(School of Mechanical Engineering,Yancheng Institute of Technology;School of Information Engineering,Yancheng Institute of Technology,Yancheng 224051,China)
出处
《软件导刊》
2022年第2期120-126,共7页
Software Guide
基金
国家自然科学基金项目(61502411,62076215)
江苏省自然科学基金项目(20150432)。
