Machine tool feed shaft health monitoring method based on optimized condition auto-encoder

Abstract

The invention belongs to the field of machine tool health state monitoring, and discloses a machine tool feed shaft health monitoring method based on an optimization condition auto-encoder. The sensor collects vibration signals and power signals of the machine tool feed shaft in a healthy state and a partial fault state; preprocessing and denoising the vibration signal and the power signal through wavelet packet decomposition; after denoising, calculating time domain and frequency domain eigenvalues to form a machine tool eigenvector, generating a time-frequency matrix, converting the time-frequency matrix into an RGB image, and constructing a machine tool multi-source data set; a machine tool multi-source data set is divided, a training set, a verification set and a test set are obtained, and the training set only contains health data; the machine tool feature vector and the training set are input into a machine tool feed shaft health state monitoring model based on an optimization condition variational auto-encoder for training, and the trained model is used for machine tool feed shaft health state monitoring. According to the method, the health state of the feed shaft is monitored only by using the health data, so that the influence of imbalance of the fault data and the health data on the data-driven monitoring method is reduced.

Description

Technical Field

[0001] The present invention relates to the field of machine tool health status monitoring, and in particular to a machine tool feed axis health monitoring method based on an optimized condition autoencoder. Background Art

[0002] The feed axis of CNC machine tools is an important part of CNC machine tools. During long-term processing and operation, the performance of the feed axis gradually declines over time under the action of external loads, resulting in reduced processing accuracy, processing quality and qualified rate of parts. In severe cases, it may cause machine tool failures and affect normal production operations. Therefore, it is of great significance to realize real-time monitoring of the health status of machine tool feed axes.

[0003] In recent years, with the rise of neural networks, more and more neural network-based health status diagnosis methods for key machine tool components have been proposed. In the patent "A CNC machine tool health monitoring method based on digital twin" (CN118192423A), cutting force sensors, vibration sensors, single temperature sensors and temperature and humidity sensors are used to construct a multi-sensor data set for CNC machine tools, and the CNN-LSTM neural network model is used to infer and judge the current health status and milling status of the CNC machine tool, and output the health prediction results. In the patent "A tool health monitoring method for aviation CNC machine tools based on OPCUA" (CN116372665A), discrete wavelet transform is used to perform symmetric wavelet transform on the force signal, acceleration signal, and acoustic emission signal during the CNC machine tool processing process to construct a model data set, and the LSTM-CNN neural network model is used to predict tool wear. The patent "A rolling bearing RUL prediction method based on deep learning" (CN117634300A) performs wavelet transform on the acquired bearing vibration signal to obtain a wavelet power spectrum to characterize the bearing degradation state, and uses CNN-Transformer to predict the remaining service life of the bearing.

[0004] The above research shows that neural networks have achieved good diagnostic results in the field of health status monitoring. However, in actual engineering applications, the feed shaft needs to run for a long time from healthy to faulty, and the time cost of acquiring fault data is high, resulting in a serious imbalance between fault data and health data. The traditional data-driven neural network detection method is difficult to achieve the desired effect. The present invention only uses health data to realize feed shaft health status monitoring, which is beneficial for enterprises to maintain and manage equipment in a timely manner and has practical significance. Summary of the invention

[0005] The purpose of the present invention is to provide a machine tool feed axis health monitoring method based on an optimized conditional autoencoder, which is used to solve the problem that the machine tool feed axis fault data is difficult to obtain and the health status is difficult to diagnose. Only health data is used for training to realize the machine tool feed axis health status monitoring.

[0006] The technical solution of the present invention is as follows: A machine tool feed axis health monitoring method based on an optimized condition autoencoder comprises the following steps:

[0007] S1. Collecting the power signal and vibration signal of the machine tool feed axis in a healthy state and the vibration signal and power signal in a partial fault state through sensors;

[0008] S2. Perform signal preprocessing and denoising on the vibration signal and the power signal by wavelet packet decomposition to obtain a denoised vibration signal and a denoised power signal;

[0009] S3. Calculate the time domain eigenvalues ​​and frequency domain eigenvalues ​​of the denoised vibration signal and the denoised power signal to form a machine tool feature vector; perform wavelet transform on the denoised vibration signal and the denoised power signal to generate a vibration signal time-frequency matrix and a power signal time-frequency matrix, and obtain a vibration time-frequency grayscale map and a power time-frequency grayscale map; draw a denoised vibration signal grayscale map, input the vibration time-frequency grayscale map into the red channel, the power time-frequency grayscale map into the green channel, and the denoised vibration signal grayscale map into the blue channel, generate a machine tool status RGB image, and construct a machine tool multi-source data set;

[0010] S4. Divide the multi-source data set of the machine tool to obtain a training set, a validation set, and a test set; the training set and the validation set only contain healthy data, and the test set contains both healthy data and fault data; build a machine tool feed axis health status monitoring model based on an optimized conditional variational autoencoder, input the machine tool feature vector and the training set into the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder for training, iteratively update the network parameters by minimizing the loss function, and obtain the trained machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder;

[0011] Furthermore, the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder takes the conditional variational autoencoder as the main framework, and includes two parts: an encoder and a decoder;

[0012] The machine tool feature vector and the training set are input to the encoder; the machine tool feature vector is used as the conditional input of the conditional variational autoencoder, and the latent space is constrained by the machine tool feature vector, so that the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder generates a corresponding mapping space with the machine tool feature vector as the label during the training process, and each subsequent health diagnosis finds the corresponding mapping space through the machine tool feature vector to reconstruct the data; the machine tool status RGB image is used as the feature input of the conditional variational autoencoder;

[0013] The encoder is mainly composed of convolutional layers, residual blocks, and SENet attention modules. The input machine tool status RGB image is sequentially passed through the encoder. The convolutional layers are used for preliminary feature extraction, and residual blocks are added between the convolutional layers. The SENet attention module is used to increase the weight of important features and suppress irrelevant features. Finally, the flattening layer is used to compress the features into a one-dimensional vector. ,Will and machine tool feature vector After splicing, access the distributed computing layer;

[0014] The distribution calculation layer consists of two fully connected layers, which are used to calculate the mean of the data in the latent space. With log variance , the latent variable is analyzed by the mean and log variance Take samples, ;

[0015] Will and machine tool feature vector Merge to get the encoder output , and then the decoder reconstructs the encoded features into data of the same size as the original input, where the structure of the decoder is symmetrical with that of the encoder.

[0016] S5. The validation set is input into the trained machine tool feed axis health status monitoring model based on optimized conditional variational autoencoder to calculate the reconstruction error. According to the 3σ principle, the health status judgment threshold is set by the validation set reconstruction error. When the data reconstruction error exceeds the health status judgment threshold, the machine tool deviates from the healthy state and is judged to be faulty. When the data reconstruction error is less than or equal to the health status judgment threshold, the machine tool is judged to be still in a healthy state.

[0017] Furthermore, the sensor includes an acceleration sensor and a power sensor; the acceleration sensor is arranged at the proximal bearing of the machine tool feed shaft, and is used to collect vibration signals of the machine tool feed shaft under different health states; the power sensor is connected to the feed shaft servo system, and is used to collect power signals of the machine tool feed shaft under different health states; the vibration signal is expressed as , the power signal is expressed as , Represents the number of sampling points, and each signal acquisition time is min, the sampling frequency is .

[0018] Furthermore, the signal preprocessing and denoising is specifically to segment the vibration signal and the power signal respectively, and the length of each sub-signal after segmentation is , for each sub-signal Layer wavelet packet decomposition, the wavelet basis function is , and obtain a wavelet packet tree; Tier The energy at a node is expressed as ,in The vibration signal Tier The wavelet coefficient or power signal at the node Tier The wavelet coefficients under the node The energy of all nodes in the layer is summed up to get Total energy of layer nodes , calculate the wavelet denoising threshold , nodes with energy greater than or equal to the wavelet denoising threshold are defined as valid nodes, and nodes with energy less than the wavelet denoising threshold are defined as noise nodes, where is the threshold adjustment coefficient, which is used to adjust the sensitivity of noise signal judgment; according to the remaining valid nodes, the vibration signal or power signal is reconstructed to obtain the denoised vibration signal With the denoised power signal .

[0019] Furthermore, the machine tool feature vector is expressed as ,in , Representing the time domain eigenvalues ​​and frequency domain eigenvalues ​​of different signals; the time domain eigenvalues ​​include rectified mean, root mean square, standard deviation, and kurtosis; the frequency domain eigenvalues ​​include spectral amplitude mean, spectral amplitude root mean square, spectral amplitude standard deviation, spectral frequency centroid, spectral frequency standard deviation, spectral frequency root mean square, spectral frequency skewness, and spectral coefficient of variation.

[0020] Furthermore, the specific calculation method of the vibration signal time-frequency matrix and the power signal time-frequency matrix is ​​to perform continuous wavelet transform on the denoised vibration signal and the denoised power signal respectively:

[0021]

[0022] In the formula, is the vibration wavelet transform coefficient or power wavelet transform coefficients , To denoise the vibration signal or denoised power signal , is the scale parameter of the wavelet function, is the translation parameter of the wavelet function, is the conjugate function of the wavelet basis function;

[0023] right and Take the two norms respectively and get the vibration signal time-frequency matrix And the power signal time-frequency matrix .

[0024] Furthermore, the method for generating the machine tool status RGB image is to convert the denoised vibration signal into a two-dimensional matrix with n rows and n columns: ; Normalize the values ​​in the power signal time-frequency matrix, the vibration signal time-frequency matrix and the two-dimensional matrix to the range of [0, 255], and then map the matrices to grayscale images respectively to obtain the vibration time-frequency grayscale image, the power time-frequency grayscale image, and the denoised vibration signal grayscale image, and uniformly set the size of each grayscale image to W*H:

[0025]

[0026] In the formula, is the vibration time-frequency matrix , power time-frequency matrix or a two-dimensional matrix , is the converted grayscale image; the vibration time-frequency grayscale image, the power time-frequency grayscale image and the denoised vibration signal grayscale image are input into the red, green and blue color channels in sequence to generate the machine tool status RGB image.

[0027] Furthermore, the machine tool multi-source data set is divided into 1 %Healthy data is used as a training set for model training, P 2 %Healthy data is used as a validation set to calculate the health status judgment threshold, all fault data and the remaining P 3 %The healthy data is used as a test set to judge the ability of the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder to distinguish the healthy state and fault state of the machine tool feed axis.

[0028] Furthermore, the loss function is:

[0029]

[0030] in is the KL divergence of the encoder output mean and variance for the standard normal distribution, , , They are the red, green and blue channels of the original machine tool status RGB image; , , Reconstruct the red, green and blue channels of the RGB image respectively. and The width and height of the image.

[0031] Furthermore, the health status judgment threshold is the sum of the mean value of the validation set reconstruction error and 3 times the standard deviation of the validation set reconstruction error. The specific formula is:

[0032]

[0033] In the formula The mean error of the reconstruction for the validation set is The standard deviation of the reconstruction error of the validation set. If the data reconstruction error is greater than the health status judgment threshold, it means that the machine tool deviates from the healthy state and is judged to be faulty. If the data reconstruction error is less than or equal to the health status judgment threshold, it means that the machine tool is still in a healthy state.

[0034] Beneficial effects of the present invention:

[0035] (1) Use the time domain eigenvalues ​​and frequency domain eigenvalues ​​of the machine tool feed shaft vibration signal and power signal to generate the machine tool state feature vector as the conditional input of the conditional variational autoencoder, so that the model can generate the corresponding mapping space with the machine tool feature vector as the label during the training process. At the same time, when reconstructing the data, the model can find the corresponding mapping space based on the machine tool feature vector to reconstruct the data, thereby reducing the probability of model misjudgment;

[0036] (2) The monitoring method of the present invention, by arranging a variety of sensors for monitoring the abnormal state of the machine tool feed axis, converts the time domain signal and time-frequency domain signal of the data into a grayscale image and synthesizes the RGB image of the machine tool state, thereby realizing multi-source fusion of the machine tool signal, so that the model can simultaneously capture the time domain characteristics and frequency domain characteristics of the signal, thereby improving the model's ability to understand and identify the machine tool state and enhancing its diagnostic effect under complex working conditions;

[0037] (3) The existing technology not only requires manual labeling of training data, but also requires that the number of normal samples and the number of abnormal samples remain close. However, in actual engineering applications, the feed shaft needs to run for a long time from healthy to faulty, and the time cost of obtaining fault data is high, resulting in a serious imbalance between fault data and healthy data. This method uses an unsupervised conditional variational autoencoder, which only requires normal samples for training, without the need for abnormal data and manual labeling, eliminating the need for manual labeling, thereby effectively overcoming the training difficulties caused by the scarcity of fault samples;

[0038] (4) This method has a certain degree of universality, and the health status monitoring method provided by this method can be extended to other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 The figure is a flow chart of the machine tool feed axis health monitoring method based on the optimized condition autoencoder.

[0040] Figure 2Schematic diagram of the signal processing flow of the machine tool feed axis health monitoring method based on the optimized conditional autoencoder.

[0041] Figure 3 This is a structural diagram of the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings.

[0043] The solution provided, a machine tool feed axis health monitoring method based on an optimized condition autoencoder, has the following steps:

[0044] Taking the Y axis of a certain model of five-axis vertical machining center as an example, the implementation method of the present invention is described in detail. The travel range of the five-axis vertical machining center is 0~500mm (machine tool coordinate value), and the maximum feed speed is 40000mm / min. The process of the machine tool feed axis health monitoring method based on the optimized condition autoencoder is as follows Figure 1 shown.

[0045] S1. Collect the power signal and vibration signal of the machine tool feed axis in a healthy state through sensors, and obtain the vibration signal and power signal in a partial fault state to verify the effectiveness of the model;

[0046] The sensor includes an acceleration sensor and a power sensor; the acceleration sensor is arranged on the proximal bearing of the machine tool feed shaft, the power sensor is connected to the feed shaft servo system, the sampling frequency is 10000Hz, the machine tool movement stroke is 0~420mm, the feed speed is 10000mm / min, and the acquisition time is 180min.

[0047] S2: performing signal preprocessing and denoising on the vibration signal and the power signal by wavelet packet decomposition to obtain a denoised vibration signal and a denoised power signal;

[0048] The signal preprocessing denoising process diagram is as follows Figure 2 As shown in the figure, the specific method is to first segment the collected vibration signal and power signal respectively, and the length of each sub-signal after segmentation is 4096. Then, each sub-signal is decomposed into 4 layers of wavelet packets, and the wavelet basis function selects the "dmey" wavelet to obtain a wavelet packet tree. The energy at a node can be expressed as ,in For vibration signal layer 4 The wavelet coefficients or power signals under the nodes of the 4th layer The wavelet coefficients of the nodes are summed up for all nodes in the 4th layer to get the total energy of the nodes in the 4th layer. , calculate the wavelet denoising threshold , nodes with energy greater than or equal to the wavelet denoising threshold are defined as valid nodes, and nodes with energy less than the wavelet denoising threshold are defined as noise nodes; is the threshold adjustment coefficient, which is used to adjust the noise signal judgment sensitivity. is 0.5; the vibration signal or power signal is reconstructed according to the remaining valid nodes to obtain the denoised vibration signal With the denoised power signal .

[0049] S3. Calculate the time domain eigenvalues ​​and frequency domain eigenvalues ​​of the denoised vibration signal and the denoised power signal to form a machine tool eigenvector; perform wavelet transform on the denoised vibration signal and the denoised power signal to generate a vibration signal time-frequency matrix and a power signal time-frequency matrix, and obtain a vibration time-frequency grayscale map and a power time-frequency grayscale map; draw a denoised vibration signal grayscale map, input the vibration time-frequency grayscale map into the red channel, the power time-frequency grayscale map into the green channel, and the denoised vibration signal grayscale map into the blue channel, generate an RGB image of the machine tool state, and construct a multi-source data set for the machine tool;

[0050] The machine tool feature vector is expressed as ,in , They are the time domain eigenvalues ​​of the denoised vibration signal: rectified mean, root mean square, standard deviation, and kurtosis; They are the time domain characteristic values ​​of the denoised power signal: rectified mean, root mean square, standard deviation, and kurtosis; They are the frequency domain characteristic values ​​of the denoised vibration signal: spectrum amplitude mean, spectrum amplitude root mean square, spectrum amplitude standard deviation, spectrum frequency centroid, spectrum frequency standard deviation, spectrum frequency root mean square, spectrum frequency skewness, and spectrum variation coefficient. They are the frequency domain eigenvalues ​​of the denoised power signal: spectrum variation coefficient and spectrum frequency centroid. The mathematical expressions of each eigenvalue are as follows:

[0051] (1) Rectified average

[0052]

[0053] (2) Root mean square

[0054]

[0055] (3) Standard deviation

[0056]

[0057] (4) Kurtosis

[0058]

[0059] (5) Spectral amplitude mean

[0060]

[0061] (6) Spectral amplitude root mean square

[0062]

[0063] (7) Spectral amplitude standard deviation

[0064]

[0065] (8) Spectral frequency center of gravity

[0066]

[0067] (9) Spectral frequency standard deviation

[0068]

[0069] (10) Spectral frequency RMS

[0070]

[0071] (11) Spectral frequency skewness

[0072]

[0073] (12) Spectral variation coefficient

[0074]

[0075] in is a denoised vibration signal or a denoised power signal, is the frequency domain amplitude of the denoised vibration signal after fast Fourier transform or the frequency domain amplitude of the denoised power signal after fast Fourier transform, For The corresponding frequency.

[0076] The specific calculation method of the vibration signal time-frequency matrix and the power signal time-frequency matrix is ​​to perform continuous wavelet transform on the denoised vibration signal and the denoised power signal respectively:

[0077]

[0078] In the formula, is the vibration wavelet transform coefficient or power wavelet transform coefficients , is a denoised vibration signal or a denoised power signal, is the scale parameter of the wavelet function, is the translation parameter of the wavelet function, is the conjugate function of the wavelet basis function;

[0079] right and Take the two norms respectively and get the vibration signal time-frequency matrix And the power signal time-frequency matrix .

[0080] The method for generating the machine tool status RGB image is to first convert the denoised vibration signal into a two-dimensional matrix with 64 rows and 64 columns: ; Normalize the values ​​in the power signal time-frequency matrix, the vibration signal time-frequency matrix and the two-dimensional matrix to the range of [0,255], and then map the matrices to grayscale images respectively to obtain the vibration time-frequency grayscale image, the power time-frequency grayscale image, and the denoised vibration signal grayscale image, and at the same time, uniformly set the size of each grayscale image to 64*64:

[0081]

[0082] In the formula, is the vibration time-frequency matrix , power time-frequency matrix or a two-dimensional matrix , The vibration time-frequency grayscale image, power time-frequency grayscale image and denoised vibration signal grayscale image are sequentially input into the red, green and blue color channels to generate the machine tool status RGB image.

[0083] S4. Divide the multi-source data set of the machine tool to obtain a training set, a validation set, and a test set; the training set and the validation set only contain healthy data, and the test set contains both healthy data and fault data; build a machine tool feed axis health status monitoring model based on an optimized conditional variational autoencoder, input the machine tool feature vector and the training set into the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder for training, iteratively update the network parameters by minimizing the loss function, and obtain the trained machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder;

[0084] The machine tool multi-source data set is divided, where 60% of the healthy data is used as a training set for model training, 20% of the healthy data is used as a validation set for calculating the health status judgment threshold, and all fault data and the remaining 20% ​​of the healthy data are used as a test set to judge the ability of the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder to distinguish between the health status and fault status of the machine tool feed axis.

[0085] The schematic diagram of the machine tool feed axis health monitoring model based on the optimized conditional variational autoencoder is as follows: Figure 3 As shown, the machine tool feed axis health monitoring model based on optimized conditional variational autoencoder takes the conditional variational autoencoder as the main framework, including an encoder and a decoder;

[0086] The machine tool feature vector and the training set are input to the encoder; the machine tool feature vector is used as the conditional input of the conditional variational autoencoder, and the machine tool state RGB image is used as the feature input of the conditional variational autoencoder;

[0087] The encoder is mainly composed of a convolution layer, a residual block, and a SENet attention module. The input machine tool status RGB image passes through three convolution layers in the encoder in sequence. The convolution kernel size of each layer is 3*3, and the number of channels is 16, 32, and 64, respectively. Preliminary feature extraction is performed. A residual block is introduced after the second and third convolution layers to prevent the gradient of the deep network from exploding or disappearing. The SENet attention module is used to increase the weight of important features and suppress irrelevant features. Finally, the flattening layer is used to compress the features into a one-dimensional vector. ,Will and machine tool feature vector After splicing, access the distributed computing layer;

[0088] The distribution calculation layer consists of two fully connected layers, which are used to calculate the mean of the data in the latent space. With log variance , the latent variable is analyzed by the mean and log variance Take samples, ;

[0089] Will and machine tool feature vector Merge to get the encoder output The decoder reconstructs the encoded features into data of the same size as the original input. The decoder module reconstructs the data using a symmetrical structure with the encoder. The convolution kernel sizes of the transposed convolution layers Tconv1, Tconv2, and TConv3 are all 3*3, and the number of channels is 64, 32, and 16, respectively. Finally, the transposed convolution layer TConv4 with a size of 3*3 and a number of channels of 3 is used to output the reconstructed image. The maximum pooling window size in the network is 2, and the step size is 1.

[0090] The loss function is:

[0091]

[0092] in is the KL divergence of the encoder output mean and variance for the standard normal distribution, , , They are the red, green and blue channels of the original machine tool status RGB image; , , They are the red, green and blue channels of the reconstructed RGB image respectively. In this example, the image width and height are both 64.

[0093] S5. The validation set is input into the trained machine tool feed axis health status monitoring model based on optimized conditional variational autoencoder to calculate the reconstruction error; according to the 3σ principle, the health status judgment threshold is set by the validation set reconstruction error; the data to be tested is input into the trained machine tool feed axis health status monitoring model based on optimized conditional variational autoencoder. When the data reconstruction error exceeds the health status judgment threshold, the machine tool deviates from the healthy state and the machine tool is judged to be faulty. When the data reconstruction error is less than or equal to the health status judgment threshold, the machine tool is judged to be still in a healthy state.

[0094] The health status judgment threshold is the sum of the mean of the validation set reconstruction error and 3 times the standard deviation of the validation set reconstruction error. The specific formula is:

[0095]

[0096] In the formula The mean error of the reconstruction for the validation set is The standard deviation of the reconstruction error of the validation set. If the data reconstruction error is greater than the health status judgment threshold, it means that the machine tool deviates from the healthy state and is judged to be faulty. If it is less than or equal to the health status judgment threshold, it means that the machine tool is still in a healthy state.

[0097] It should be noted that the above specific embodiments of the present invention are only used to exemplify the principles and processes of the present invention and do not constitute a limitation of the present invention. Therefore, any modifications and equivalent substitutions made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention.

Claims

1. A machine tool feed axis health monitoring method based on an optimized condition autoencoder, characterized in that: The steps include: S1. Collecting the power signal and vibration signal of the machine tool feed axis in a healthy state and the vibration signal and power signal in a partial fault state through sensors; S2. Perform signal preprocessing and denoising on the vibration signal and the power signal by wavelet packet decomposition to obtain a denoised vibration signal and a denoised power signal; S3. Calculate the time domain eigenvalues ​​and frequency domain eigenvalues ​​of the denoised vibration signal and the denoised power signal to form a machine tool feature vector; perform wavelet transform on the denoised vibration signal and the denoised power signal to generate a vibration signal time-frequency matrix and a power signal time-frequency matrix, and obtain a vibration time-frequency grayscale map and a power time-frequency grayscale map; draw a denoised vibration signal grayscale map, input the vibration time-frequency grayscale map into the red channel, the power time-frequency grayscale map into the green channel, and the denoised vibration signal grayscale map into the blue channel, generate a machine tool status RGB image, and construct a machine tool multi-source data set; S4. Divide the multi-source data set of the machine tool into a training set, a validation set, and a test set; The training set and the validation set only contain healthy data, and the test set contains both healthy data and fault data; a machine tool feed axis health status monitoring model based on an optimized conditional variational autoencoder is constructed, and the machine tool feature vector and the training set are input into the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder for training, and the network parameters are iteratively updated by minimizing the loss function to obtain a trained machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder; The machine tool feed axis health status monitoring model based on optimized conditional variational autoencoder takes the conditional variational autoencoder as the main framework, including an encoder and a decoder; The machine tool feature vector and the training set are input to the encoder; the machine tool feature vector is used as the conditional input of the conditional variational autoencoder, and the latent space is constrained by the machine tool feature vector, so that the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder generates a corresponding mapping space with the machine tool feature vector as the label during the training process, and each subsequent health diagnosis finds the corresponding mapping space through the machine tool feature vector to reconstruct the data; the machine tool status RGB image is used as the feature input of the conditional variational autoencoder; The encoder is mainly composed of convolutional layers, residual blocks, and SENet attention modules. The input machine tool status RGB image is sequentially passed through the encoder. The convolutional layers are used for preliminary feature extraction, and residual blocks are added between the convolutional layers. The SENet attention module is used to increase the weight of important features and suppress irrelevant features. Finally, the flattening layer is used to compress the features into a one-dimensional vector. ,Will and machine tool feature vector After splicing, access the distributed computing layer; The distribution calculation layer consists of two fully connected layers, which are used to calculate the mean of the data in the latent space. With log variance , the latent variable is analyzed by the mean and log variance Take samples, ; Will and machine tool feature vector Merge to get the encoder output , then the decoder reconstructs the encoded features into data of the same size as the original input, where the structure of the decoder is symmetrical with that of the encoder; S5. The validation set is input into the trained machine tool feed axis health status monitoring model based on optimized conditional variational autoencoder to calculate the reconstruction error; according to the 3σ principle, the health status judgment threshold is set by the validation set reconstruction error; When the data reconstruction error exceeds the health status judgment threshold, the machine tool deviates from the healthy state and is judged to be faulty. When the data reconstruction error is less than or equal to the health status judgment threshold, the machine tool is judged to be still in a healthy state.

2. The machine tool feed axis health monitoring method based on the optimized condition autoencoder according to claim 1 is characterized in that: The sensor includes an acceleration sensor and a power sensor; the acceleration sensor is arranged at the proximal bearing of the machine tool feed shaft, and is used to collect vibration signals of the machine tool feed shaft under different health states; the power sensor is connected to the feed shaft servo system, and is used to collect power signals of the machine tool feed shaft under different health states; the vibration signal is expressed as , the power signal is expressed as , Represents the number of sampling points, and each signal acquisition time is min, the sampling frequency is .

3. The machine tool feed axis health monitoring method based on the optimized condition autoencoder according to claim 1 is characterized in that: The signal preprocessing and denoising is specifically to segment the vibration signal and the power signal respectively, and the length of each sub-signal after segmentation is , for each sub-signal Layer wavelet packet decomposition, the wavelet basis function is , and obtain a wavelet packet tree; Tier The energy at a node is expressed as ,in The vibration signal Tier The wavelet coefficient or power signal at the node Tier The wavelet coefficients under the node The energy of all nodes in the layer is summed up to get Total energy of layer nodes , calculate the wavelet denoising threshold , nodes with energy greater than or equal to the wavelet denoising threshold are defined as valid nodes, and nodes with energy less than the wavelet denoising threshold are defined as noise nodes, where is the threshold adjustment coefficient, which is used to adjust the sensitivity of noise signal judgment; the vibration signal or power signal is reconstructed according to the remaining valid nodes to obtain the denoised vibration signal With the denoised power signal .

4. The machine tool feed axis health monitoring method based on the optimized condition autoencoder according to claim 1 is characterized in that: The machine tool feature vector is expressed as ,in , Represent the time domain eigenvalues ​​and frequency domain eigenvalues ​​of different signals; The time domain eigenvalues ​​include rectified mean, root mean square, standard deviation, and kurtosis; the frequency domain eigenvalues ​​include spectrum amplitude mean, spectrum amplitude root mean square, spectrum amplitude standard deviation, spectrum frequency centroid, spectrum frequency standard deviation, spectrum frequency root mean square, spectrum frequency skewness, and spectrum variation coefficient.

5. The machine tool feed axis health monitoring method based on the optimized condition autoencoder according to claim 1 is characterized in that: The specific calculation method of the vibration signal time-frequency matrix and the power signal time-frequency matrix is ​​to perform continuous wavelet transform on the denoised vibration signal and the denoised power signal respectively: In the formula, is the vibration wavelet transform coefficient or power wavelet transform coefficients , To denoise the vibration signal or denoised power signal , is the scale parameter of the wavelet function, is the translation parameter of the wavelet function, is the conjugate function of the wavelet basis function; right and Take the two norms respectively and get the vibration signal time-frequency matrix And the power signal time-frequency matrix .

6. The machine tool feed axis health monitoring method based on the optimized condition autoencoder according to claim 1 is characterized in that: The method for generating the machine tool status RGB image is to convert the denoised vibration signal into a two-dimensional matrix with n rows and n columns: ; Normalize the values ​​in the power signal time-frequency matrix, the vibration signal time-frequency matrix and the two-dimensional matrix to the range of [0, 255], and then map the matrices to grayscale images respectively to obtain the vibration time-frequency grayscale image, the power time-frequency grayscale image, and the denoised vibration signal grayscale image, and uniformly set the size of each grayscale image to W*H: In the formula, is the vibration time-frequency matrix , power time-frequency matrix or a two-dimensional matrix , is the converted grayscale image; the vibration time-frequency grayscale image, the power time-frequency grayscale image and the denoised vibration signal grayscale image are input into the red, green and blue color channels in sequence to generate the machine tool status RGB image.

7. The machine tool feed axis health monitoring method based on the optimized condition autoencoder according to claim 1 is characterized in that: The machine tool multi-source data set is divided, wherein P1% healthy data is used as a training set for model training, P2% healthy data is used as a validation set for calculating the health status judgment threshold, and all fault data and the remaining P3% healthy data are used as a test set to judge the ability of the machine tool feed axis health status monitoring model based on the optimized conditional variational autoencoder to distinguish between the health status and fault status of the machine tool feed axis.

8. The machine tool feed axis health monitoring method based on the optimized condition autoencoder according to claim 1 is characterized in that: The loss function is: in is the KL divergence of the encoder output mean and variance for the standard normal distribution, , , They are the red, green and blue channels of the original machine tool status RGB image; , , Reconstruct the red, green and blue channels of the RGB image respectively. and The width and height of the image.

9. The machine tool feed axis health monitoring method based on the optimized condition autoencoder according to claim 1 is characterized in that: The health status judgment threshold is the sum of the mean of the validation set reconstruction error and 3 times the standard deviation of the validation set reconstruction error. The specific formula is: In the formula The mean error of the reconstruction for the validation set is The standard deviation of the reconstruction error of the validation set. If the data reconstruction error is greater than the health status judgment threshold, it means that the machine tool deviates from the healthy state and is judged to be faulty. If the data reconstruction error is less than or equal to the health status judgment threshold, it means that the machine tool is still in a healthy state.

Images