基于深度学习的健美操力量训练器故障监测方法  

Fault detection method of Aerobics strength trainer based on deep learning

作　　者：周煜[1] 冯建强[1] ZHOU Yu;FENG Jianqiang(XI’AN UNIVERSITY OF TECHNOLOGY,Xi’an 710048,China)

机构地区：[1]西安理工大学,西安710048

出　　处：《自动化与仪器仪表》2022年第8期24-28,共5页Automation & Instrumentation

基　　金：陕西省体育局常规课题《新时期陕西省高校体育文明建设的理论与实践》(2021021)。

摘　　要：针对传统故障监测方法对精密的健美操力量训练器故障监测精度低的问题,通过优化卷积神经网络随机梯度下降算法中的梯度下降方式和学习率,并将提取的训练器振动信号特征作为改进卷积神经网络输入,提出一种基于深度学习的健美操力量训练器故障监测方法。仿真结果表明,所提方法对卷积神经网络改进有效,利用改进的卷积神经网络可识别监测健美操力量训练器不同类型、不同位置和不同程度的故障,平均监测准确率达到97.75%,对训练器正常状态和4根转子断裂的轴承故障监测准确率达到100%。相较于改进前卷积神经网络,所提监测方法的监测准确率提高了7.50%,相较于常用故障监测算法,所提方法的监测准确率均有不同程度的提升,具有一定的有效性和优越性,并采用改进卷积神经网络进行分类识别,可有效监测。Aiming at the problem of low fault detection accuracy of traditional fault detection methods for precision aerobics strength trainer,by optimizing the gradient descent mode and learning rate in the random gradient descent algorithm of convolution neural network,and taking the extracted vibration signal characteristics of trainer as the input of improved convolution neural network,A fault detection method of Aerobics strength trainer based on deep learning is proposed.The simulation results show that the proposed method is effective in improving the convolution neural network.The improved convolution neural network can identify and detect the faults of different types,positions and degrees of Aerobics strength trainer.The average detection accuracy is 97.75%,and the detection accuracy of bearing faults in the normal state of the trainer and four broken rotors is 100%.Compared with the improved convolution neural network,the detection accuracy of the proposed method is improved by 7.50%.Compared with the common fault detection algorithms,the detection accuracy of the proposed method is improved to varying degrees,which has certain effectiveness and superiority.,The improved convolution neural network is used for classification and recognition,which can detect effectively.

关 键 词：深度学习 健美操力量训练器 故障监测 卷积神经网络 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]