A bearing fault diagnosis and evaluation method based on knowledge feature fusion

Abstract

The present invention relates to the technical field of bearing fault diagnosis, and in particular to a bearing fault diagnosis and evaluation method based on knowledge feature fusion. The method comprises: extracting historical fault feature data; training a fault diagnosis model; using the output of the training data in the feature layer of the fault diagnosis model to introduce a central loss function to train an embedding layer network model; mapping the feature data with the embedding layer network model; weightedly fusing the mapped data with the knowledge features based on the data; calculating the local density distribution of adjacent data; and calculating the confidence of the diagnosis result. The present invention estimates the confidence of the model diagnosis result based on the local density distribution of the sample in the embedding space by fusing the features extracted during the diagnosis process of the fault diagnosis model and the features of the fault data itself, thereby solving the problems of difficulty in confidence evaluation of the fault diagnosis model and poor interpretability of the general evaluation process, and can obtain confidence evaluation results under complex models and complex data.

Description

Technical Field

[0001] This invention relates to the field of bearing fault diagnosis technology, and in particular to a bearing fault diagnosis and assessment method based on knowledge feature fusion. Background Technology

[0002] With the rapid advancement of technology and the rise of modern industry, industrial equipment needs to maintain reliable performance in diverse production environments, sometimes even requiring operation under extreme conditions. The working conditions of many components constituting mechanical devices are often complex and demanding, making accurate fault diagnosis of each component a key task to ensure smooth industrial production. Traditional fault diagnosis techniques primarily rely on time-domain, frequency-domain, and time-frequency-domain signal processing. Time-domain analysis assesses the health status of working components by analyzing the time series of signals. Frequency-domain analysis focuses on vibration signals, identifying characteristic frequencies of bearing faults by analyzing the frequency distribution of the signal, thereby achieving fault detection and diagnosis. While these methods have achieved some success in fault diagnosis, they have limitations in processing complex data. Traditional methods rely heavily on expert experience and prior knowledge, which not only increases the complexity of the analysis but also limits the universality and accuracy of the diagnostic methods.

[0003] To overcome these limitations, deep learning technology has been widely applied in fault diagnosis in recent years. Deep learning can automatically learn and extract features from large amounts of data without manual intervention or complex feature selection processes. This method utilizes deep neural networks, such as convolutional neural networks and recurrent neural networks, to identify bearing fault modes through deep features learned during training, significantly improving the accuracy and efficiency of fault detection. However, assessing the confidence level of deep learning diagnostic results remains a challenge. Due to the large number of parameters, complex structure, and poor interpretability of deep learning models, it is difficult to assess the confidence level of diagnostic results from a mechanistic perspective. Summary of the Invention

[0004] This invention discloses a bearing fault diagnosis and assessment method based on knowledge feature fusion, the specific method of which is as follows:

[0005] Acquire historical bearing fault data and extract historical fault feature data;

[0006] A fault diagnosis model is trained using historical fault feature data to obtain the fault diagnosis model;

[0007] Obtain the output of the previous layer of the fault diagnosis model as the feature data of the fault diagnosis model.

[0008] Based on the fault diagnosis model, an embedding layer network model is added, and the center loss is introduced to train the embedding layer network model. The embedding layer network model is then used to map the feature data of the fault diagnosis model.

[0009] Weighted fusion of mapped data and data-based knowledge features;

[0010] Search for the K nearest neighbor samples for each fused feature;

[0011] Based on the search results of the nearest neighbor samples, calculate the local density scores of different categories of its nearest neighbor samples;

[0012] We calculate the overall confidence level of the fault diagnosis model by weighting local density.

[0013] Furthermore, the historical fault characteristic data includes: mean, absolute mean, effective value, average power, root square amplitude, peak value, peak-to-peak value, variance, skewness, kurtosis, waveform, peak value, impulse, margin, skewness, and kurtosis index.

[0014] Furthermore, when introducing center loss to train the embedding layer network model, the formula for calculating center loss is as follows:

[0015]

[0016] In the formula, M represents the number of samples in each training mini-batch, and L... SO L represents the cross-entropy loss. Cen Indicates central loss; h ci For sample c i Center, c i For x i The label of the sample class, λ is a hyperparameter, x i ∈R D D is the dimension of the feature vector;

[0017] The formula for calculating cross-entropy loss is as follows:

[0018]

[0019] In the formula, y i It's a real label. is the model's predicted output probability, and log is the natural logarithm.

[0020] Furthermore, the K nearest neighbor samples for each fused feature are searched, specifically as follows:

[0021] We choose Euclidean distance as the distance metric between samples for each pair of sample points (x... i ,x j The corresponding distance metric calculation formula is as follows:

[0022]

[0023] In the formula, x ik and x jk It is the value of the sample on the k-th feature;

[0024] For each test sample fused feature, calculate its distance metric with all training samples;

[0025] Based on the calculated distances, the training samples are sorted in ascending order of distance, and then the K samples with the smallest distances are selected. These samples are the K nearest neighbors found.

[0026] Furthermore, the local density scores of different categories of its nearest neighbor samples are calculated using the following formula:

[0027]

[0028]

[0029] In the formula: where y represents the test sample y i The predicted value is given by c, where c represents the final classification result of the fault diagnosis by the model, j = 1, 2, 3, ..., k represents the k nearest neighbor training set samples, and f(·) represents the model after training, calculated according to the above formula. This will yield a local density fraction in the range of 0 to 1.

[0030] Furthermore, the overall confidence level of the fault diagnosis model is calculated using the following formula:

[0031]

[0032] In the formula: max(·) represents the score of the test set sample with the highest confidence obtained after mapping through the embedding network based on the fused features of the test samples, and C model This represents the confidence score of the fault diagnosis model.

[0033] Due to the adoption of the above technical solutions, the present invention has the following beneficial effects:

[0034] When evaluating the confidence of a fault diagnosis model with a certain degree of complexity, this invention takes a data-centric approach, fully considering the model's mapping and transformation process of the data as well as the characteristics of the dataset itself. It uses the calculation of local density scores to link the test set data and the model's training data, thereby accurately and quantitatively evaluating the model's diagnostic results and overall confidence.

[0035] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0036] The accompanying drawings of this invention are described below.

[0037] Figure 1 This is an overall flowchart of the method of the present invention.

[0038] Figure 2 This is a schematic diagram of the fault diagnosis model structure used in the implementation of this invention.

[0039] Figure 3 This is a graph showing the change in the loss function during the training process of the fault diagnosis model in this invention.

[0040] Figure 4 This is a graph showing the change in accuracy of the fault diagnosis model during the training process in this invention.

[0041] Figure 5 This is a graph showing the change in the loss function during the embedded network training process in this invention. Detailed Implementation

[0042] The present invention will be further described below with reference to the accompanying drawings and examples.

[0043] This embodiment uses the open-source Western Reserve University bearing fault diagnosis dataset. The basic information of the dataset is shown in Table 1, and the calculated fault data features are shown in Table 2.

[0044] Table 1. Basic Information of the Dataset

[0045]

[0046] Table 2 Fault Data Characteristics Calculation Table

[0047]

[0048] This embodiment focuses on the most commonly used fault diagnosis model in the field of fault diagnosis, which is based on convolutional neural networks. The model structure diagram is as follows: Figure 2 As shown, the model includes convolutional layers, pooling layers, flattening layers, fully connected layers, dropout layers, and an output layer. Specific model information is shown in Table 3. The model was trained using the training set, and the training performance was evaluated using the test set. The loss function value transformation curve during training is shown in the figure. Figure 3 As shown, the accuracy change curve is as follows: Figure 4 As shown;

[0049] Table 3 Fault Diagnosis Model Parameter Table

[0050]

[0051] A bearing fault diagnosis and assessment method based on knowledge feature fusion, such as Figure 1 As shown, the specific steps are as follows:

[0052] S1. Obtain historical bearing fault data and extract historical fault feature data, as shown in Table 2.

[0053] S2. Train the fault diagnosis model using historical fault feature data to obtain the fault diagnosis model. The model structure is as follows: Figure 2 As shown in Table 3, the specific parameters are as follows.

[0054] S3. Output the fully connected layer before the output layer of the fault diagnosis model to obtain the training and test set features extracted by the model.

[0055] S4. Input the features extracted from the training and test sets into the embedding layer for training to obtain the mapping representation of the training and test sets in the embedding space. During training, the aforementioned center loss is introduced to obtain a better data mapping representation. The curve of the loss function value change during training is shown in the figure. Figure 5 As shown.

[0056] S5. After obtaining the feature mapping expressions of the training set and test set extracted by the model, they are fused with the data features obtained based on knowledge computing. In order to fully reflect the decision-making process of the model in fault diagnosis, the feature weight extracted by the model is set to 0.8 and the data feature weight is set to 0.2 when setting the fusion weight, so as to obtain the fused features.

[0057] S6. Use the KNN algorithm to search for the K nearest neighbor training samples of each test set sample fusion feature, set the parameter K=15, and record the category of the nearest neighbor sample; taking the test set sample numbered 1 as an example, the information of the nearest neighbor sample after the search is shown in Table 4.

[0058] Table 4 Nearest neighbor search results for test samples

[0059]

[0060] S7. For each mapped test set feature data, calculate the local density scores of different categories of its nearest training samples based on the nearest neighbor sample search results. The calculation results are shown in Table 5.

[0061] Table 5 Calculation results of local density fraction

[0062]

[0063] S8. The overall confidence level C of the fault diagnosis model is obtained by weighting the local density scores of all test samples estimated through the test set. model The result is shown in the following formula:

[0064] C model =0.95632

[0065] Experimental results show that, when evaluating the confidence of fault diagnosis models with a certain degree of complexity, this invention, from a data perspective, fully considers the model's mapping and transformation process to the data, as well as the characteristics of the dataset itself. It uses the calculation of local density scores to correlate the test set data with the model's training data, thereby accurately and quantitatively evaluating the model's diagnostic results and overall confidence. In the example, the model achieved an overall confidence level of 95.632% after evaluation, consistent with its high accuracy on the validation set, demonstrating the rationality of this invention.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A bearing fault diagnosis and assessment method based on knowledge feature fusion, characterized in that, The specific method is as follows: Acquire historical bearing fault data and extract historical fault feature data; A fault diagnosis model is trained using historical fault feature data to obtain the fault diagnosis model; Obtain the output of the previous layer of the fault diagnosis model as the feature data of the fault diagnosis model. Based on the fault diagnosis model, an embedding layer network model is added, and the center loss is introduced to train the embedding layer network model. The embedding layer network model is then used to map the feature data of the fault diagnosis model. Weighted fusion of mapped data and data-based knowledge features; Search for the K nearest neighbor samples for each fused feature; Based on the search results of the nearest neighbor samples, calculate the local density scores of different categories of its nearest neighbor samples; Weighted local density is used to calculate the overall confidence level of the fault diagnosis model; The local density scores of its nearest neighbor samples for different categories are calculated using the following formula: In the formula: where y represents the test sample y i The predicted value is given by c, where c represents the final classification result of the fault diagnosis by the model, j = 1, 2, 3, ..., k represents the k nearest neighbor training set samples, and f(·) represents the model after training, calculated according to the above formula. This will yield a local density fraction in the range of 0 to 1.

2. The bearing fault diagnosis and assessment method based on knowledge feature fusion as described in claim 1, characterized in that, The historical fault characteristic data includes: mean, absolute mean, RMS value, average power, root square amplitude, peak value, peak-to-peak value, variance, skewness, kurtosis, waveform, impulse, margin, skewness and kurtosis index.

3. The bearing fault diagnosis and assessment method based on knowledge feature fusion as described in claim 1, characterized in that, When introducing center loss to train an embedding layer network model, the formula for calculating center loss is as follows: In the formula, M represents the number of samples in each training mini-batch, and L... SO L represents the cross-entropy loss. Cen Indicates central loss; h ci For sample c i Center, c i For x i The label of the sample class, λ is a hyperparameter, x i ∈R D D is the dimension of the feature vector; The formula for calculating cross-entropy loss is as follows: In the formula, y i It's a real label. is the model's predicted output probability, and log is the natural logarithm.

4. The bearing fault diagnosis and assessment method based on knowledge feature fusion as described in claim 3, characterized in that, The method for searching the K nearest neighbors of each fused feature is as follows: We choose Euclidean distance as the distance metric between samples for each pair of sample points (x... i ,x j The corresponding distance metric calculation formula is as follows: In the formula, x ik and x jk It is the value of the sample on the k-th feature; For each test sample fused feature, calculate its distance metric with all training samples; Based on the calculated distances, the training samples are sorted in ascending order of distance, and then the K samples with the smallest distances are selected. These samples are the K nearest neighbors found.

5. The bearing fault diagnosis and assessment method based on knowledge feature fusion as described in claim 1, characterized in that, The overall confidence score of the fault diagnosis model is calculated using the following formula: In the formula: max(·) represents the score of the test set sample with the highest confidence obtained after mapping through the embedding network based on the fused features of the test samples, and C model This represents the confidence score of the fault diagnosis model.

Images