齿轮箱故障诊断的小波包-ICA分析方法.docx

该【齿轮箱故障诊断的小波包-ICA分析方法 】是由【wz_198613】上传分享，文档一共【4】页，该文档可以免费在线阅读，需要了解更多关于【齿轮箱故障诊断的小波包-ICA分析方法 】的内容，可以使用淘豆网的站内搜索功能，选择自己适合的文档，以下文字是截取该文章内的部分文字，如需要获得完整电子版，请下载此文档到您的设备，方便您编辑和打印。齿轮箱故障诊断的小波包-ICA分析方法
Abstract
In recent years, the diagnosis of gearboxes has been a growing field of research as it plays an important role in ensuring the smooth operation of industrial machines. The traditional methods of gear fault diagnosis are generally based on signal processing techniques, but the complexity of gear signals still poses a number of challenges to traditional methods. In this study, we propose a wavelet packet independent component analysis (ICA) method to diagnose gearbox faults. Firstly, the wavelet packet transform is performed to decompose the gearbox vibration signal into multiple wavelet packet components, and then ICA is used to extract features of the fault signal. Finally, the extracted features are classified using a back propagation neural network (BPNN). The proposed method is tested on a gearbox fault dataset, and the results show that the proposed method has high accuracy in gearbox fault diagnosis.
Keywords: Gearbox fault diagnosis; Wavelet packet transform; Independent component analysis; Back propagation neural network.
Introduction
Gearboxes are widely used in a variety of industrial applications, and their faults can significantly affect the operation of machines. Therefore, it is essential to diagnose gearbox faults in order to avoid equipment downtime and maintain equipment safety. Various methods have been developed for gearbox fault diagnosis, such as time-domain analysis, frequency-domain analysis, and time-frequency analysis. However, traditional methods still face some limitations because of the complexity of gearbox signals. Therefore, it is necessary to develop a new method to overcome these limitations.
Wavelet packet transform (WPT) is a signal processing technique that has been widely used in gearbox fault diagnosis. Wavelet packet decomposition can provide a more comprehensive representation of the signal, which can facilitate the identification of fault features. Independent component analysis (ICA) is a nonlinear signal processing method that can extract latent variables that are independent of each other and can provide a more compact representation of the signal than wavelet packet analysis. In this paper, we propose a new method that combines WPT and ICA to diagnose gearbox faults.
The proposed method consists of three main steps: signal pre-processing, feature extraction, and fault classification. In the pre-processing step, the raw vibration signal of the gearbox is collected, and then the WPT is performed to decompose the signal into multiple wavelet packet components. In the feature extraction step, ICA is performed on each wavelet packet component to extract independent features of the fault signal. Finally, the extracted features are used for fault classification by the back propagation neural network (BPNN).
Methodology
Wavelet Packet Transform (WPT)
Wavelet packet decomposition is a time-frequency analysis method that decomposes the signal into different frequency band components with different time resolutions. The decomposition process can be carried out using various wavelet bases, such as Daubechies, Symlets, and Coiflets. In this study, we use the Daubechies wavelet function because it has better performance in signal decomposition.
For an N-point signal xn, the wavelet packet decomposition can be described as follows:
1. Set the level j = 0 and k = 0.
2. Calculate the wavelet packet co-efficients for each level j and node k. The wavelet packet coefficients can be defined as follows:
W_j,k = abs(convolve(xn, wavelet_j,k))^2
where wavelet_j,k is the wavelet basis function at level j and node k, and convolvep(x,y) is the convolution operation.
3. At the end of each level, choose the node with the largest wavelet packet coefficient and split it into two child nodes.
4. Increase the level by one and go back to step 2 until reaching a certain level.
The wavelet packet decomposition generates a set of frequency band components at different time resolutions. Each component contains different frequency information and is able to reveal different characteristics of the signal.
Independent Component Analysis (ICA)
ICA is a signal processing technique that extracts independent features of the signal by assuming that the signals are a linear combination of independent sources. The objective of the ICA algorithm is to find a set of linearly independent components that can represent the original signals. The most common ICA algorithm is FastICA, which is an optimization algorithm based on a maximum likelihood estimation.
The ICA algorithm can be described as follows:
1. Collect the signals and standardize them to have zero mean and unit variance.
2. Initialize the weight matrix W with random values.
3. For each signal, estimate its independent component by calculating the covariance of the signal and the weight matrix W.
4. Calculate the negentropy of the estimated independent component.
5. Update the weight matrix W using the learning rule in order to maximize the negentropy.
6. Check the convergence of the algorithm, and if the convergence criteria are met, terminate the algorithm; otherwise, go back to step 3.
The back propagation neural network (BPNN)
The back propagation neural network (BPNN) is a supervised learning algorithm that can be used for classification and regression problems. The BPNN consists of input layer, hidden layer, and output layer. The input layer receives the features extracted in the previous step, the hidden layer performs nonlinear transformations on the input, and the output layer provides the classification result.
The BPNN training process can be described as follows:
1. Initialize the weights of the network randomly.
2. Present the training data to the network.
3. Calculate the output of the network and the error between the predicted output and the actual output.
4. Update the weights based on the error using a learning rate and a momentum term.
5. Check the convergence of the training process, and if the convergence criteria are met, terminate the training process; otherwise, go back to step 2.
Experimental Results
The proposed method is tested on a gearbox fault dataset. The dataset contains five types of faults: gear crack, gear tooth wear, gear corrosion, bearing damage, and shaft misalignment. The dataset consists of 500 samples for each fault type, and the sampling rate is 12,000 Hz. The proposed method is compared with traditional methods, including envelope analysis and WPT-based methods.
The diagnostic performance is evaluated by calculating the accuracy, sensitivity, and specificity of the methods. The accuracy is defined as the ratio of the correct diagnoses to the total number of diagnoses. The sensitivity is defined as the ratio of correct positive diagnoses to the total number of actual positive cases. The specificity is defined as the ratio of correct negative diagnoses to the total number of actual negative cases.
The results show that the proposed method achieves the best diagnostic performance among all the methods tested. The accuracy, sensitivity, and specificity of the proposed method are %, %, and %, respectively. The traditional methods have an accuracy of approximately 90%, and their sensitivity and specificity are lower than the proposed method. The comparison results are shown in Table 1.
Table 1. Comparison of diagnostic performance of different methods.
Method Accuracy Sensitivity Specificity
Envelope analysis % % %
WPT-based method % % %
Proposed method % % %
Conclusion
In this study, we propose a wavelet packet independent component analysis (ICA) method for gearbox fault diagnosis. The proposed method decomposes the vibration signal using the WPT and extracts independent features using the ICA. The extracted features are then classified using a BPNN. The proposed method is tested on a gearbox fault dataset, and the results show that the proposed method has better diagnostic performance than traditional methods. The accuracy, sensitivity, and specificity of the proposed method are higher than those of traditional methods. The proposed method can be used as a reliable and effective method for gearbox fault diagnosis.