摘要
随着大数据时代的到来与人工智能的发展,故障诊断也朝着智能化方向发展。针对滚动轴承的故障诊断,提出基于短时倒频谱变换与卷积神经网络的故障诊断方法。首先,对实验采集到的滚动轴承原始信号进行短时倒频谱变换,得到二维的倒频谱,按照故障尺寸划分为不同标签的训练集和测试集,且每组数据集包含3种转速;然后,采用正交实验选取卷积神经网络最优训练参数,建立卷积神经网络模型对训练集进行训练;最后,利用训练好的卷积神经网络对测试集进行测试。结果表明,短时倒频谱变换作为卷积神经网络的输入,能够保留原始信号特征信息并预提取故障特征;基于短时倒频谱变化的卷积神经网络对轴承故障的测试准确率在一种转速和多种转速混合情况下,均可达到100%。
With the advent of the era of big data and the development of artificial intelligence,fault diagnosis is also developing towards intelligence.A fault diagnosis method based on short-time cepstrum transform(STCT)and convolutional neural network(CNN)is proposed for the fault diagnosis of rolling bearings.Firstly,the short-time cepstral transform of the original signal of the experimental rolling bearing was carried out to obtain the 2-D cepstrum spectrum,which was divided into training sets and test sets of different labels according to the fault size,and each data set contained three speeds.Then the optimal training parameters of the convolutional neural network were selected by orthogonal experiment,and the model of the convolutional neural network was established to train the training set.Finally,the trained convolutional neural network was used to test the test set.The results show that the short-time cepstrum transform as the input of the convolutional neural network can retain the original signal characteristic information and highlight the fault characteristics.The test accuracy of the convolutional neural network based on short-time cepstrum variation for bearing faults can reach 100%at one speed and a mixture of different speeds.
作者
王丹
金光灿
邱志
邢彦锋
WANG Dan;JIN Guang-can;QIU Zhi;XING Yan-feng(School of Mechanical and Automotive Engineering,Shanghai University of Engineering Science,Shanghai 201620,China;Schaeffler Trading(Shanghai)Co.,Ltd.,Shanghai 201804,China)
出处
《软件导刊》
2021年第8期44-48,共5页
Software Guide
基金
上海市自然科学基金项目(20ZR1422600)
上海市地方能力建设项目(19030501100)。
关键词
滚动轴承
故障诊断
短时倒频谱变换
卷积神经网络
rolling bearing
fault diagnosis
short-time cepstrum transform
convolutional neural network
作者简介
王丹(1997-),男,上海工程技术大学机械与汽车工程学院硕士研究生,研究方向为信号处理与智能故障诊断;通讯作者:金光灿(1981-),女,博士,上海工程技术大学机械与汽车工程学院副教授、硕士生导师,研究方向为振动与噪声、智能故障诊断。
