基于改进Laplace小波和改进卷积神经网络的压裂车动力端轴承故障识别

Fault identification of fracturing vehicle power end bearing based on enhanced ELW and enhanced ECNN

在强背景噪声工况下,压裂车动力端轴承振动信号故障特征较微弱,导致轴承故障诊断的准确率较低。针对这一问题,提出了一种基于改进Laplace小波(ELW)和改进卷积神经网络(ECNN)的压裂车动力端轴承故障识别方法。首先,采用了一种Laplace小波振荡频率参数选取策略,使Laplace小波搜寻到了最佳频率参数;然后,采用改进Laplace小波,对采集到的压裂车动力端轴承故障振动信号进行了降噪处理,并在卷积神经网络(CNN)的基础上引入了自注意力机制和编码器、解码器结构,设计出了改进卷积神经网络(ECNN)模型;最后,将压裂车动力端轴承降噪后的信号输入改进卷积神经网络,进行了自动特征提取和故障识别;为了验证该方法的有效性和先进性,将其与其他方法(模型)进行了对比分析。研究结果表明:采用基于改进Laplace小波与和改进卷积神经网络的方法(模型),对压裂车动力端轴承故障进行识别的准确率可高达99.67%,单个样本的测试时间仅为0.14 s;在识别准确率、召回率、F1得分和统计检验等方面,与其他方法(模型)相比,基于改进Laplace小波与改进卷积神经网络的组合模型具有更为优秀的故障识别性能。

Under the condition of strong background noise,the fault characteristics of the power end bearing of the fracturing vehicle were weak,resulting in a low accuracy of bearing fault diagnosis.To solve this problem,a fault identification method for the power end bearing of the fracturing vehicle based on enhanced Laplace wavelet(ELW)and enhanced convolutional neural network(ECNN)was proposed.Firstly,a selection strategy of Laplace wavelet oscillation frequency parameter was proposed,so that the best frequency parameters were searched by the Laplace wavelet.Secondly,the collected vibration signals of the power end of the fracturing truck were de-noised by enhanced Laplace wavelet,and the self-attention mechanism and encoder structure and decoder structure were introduced on the basis of the one-dimensional convolutional neural network.Then,the enhanced convolutional neural network model was constructed.Finally,the de-noised signals of the power end of the fracturing vehicle were fed into enhanced convolutional neural network for automatic feature extraction and fault identification.In order to verify the effectiveness and advancement of the method,it was compared with other methods(models).The research results show that,the combined model of the enhanced Laplace wavelet(ELW)and enhanced convolutional neural network(ECNN)has an accuracy rate of 99.67%in fault identification of the power end of the fracturing vehicle,and the testing time of individual sample is 0.14 s.The combined model of the enhanced Laplace wavelet and enhanced convolutional neural network has superior fault recognition performance in terms of recognition accuracy,recall,F1 score and statistical test.