Special motor for turning and milling with hydraulic brake and control method thereof

Abstract

Disclosed is a special motor for turning and milling with hydraulic brake and a control method thereof, which uses artificial intelligence monitoring technology based on deep learning to extract multi-scale dynamic feature information of the power signal of the special motor for turning and milling in the time dimension, and uses the dynamic feature of the power to obtain a generated vibration signal, which is respectively input into a twin network with the same network structure as the actual collected vibration signal to mine the source domain image features of the two, and further extracts the difference feature information of the source domain image by the difference comparison of the obtained detection vibration feature and the generated vibration feature, and uses this to detect abnormal working conditions of the special motor for turning and milling. In this way, a hydraulic emergency braking command is generated when the working condition of the special motor for turning and milling is detected to ensure the safety of operation while ensuring the working performance of the special motor for turning and milling.

Description

Technical Field

[0001] The present application relates to the technical field of turning and milling, and more specifically to a turning and milling compound motor with hydraulic braking and a control method thereof. Background Art

[0002] Traditional machine tools can only process workpieces in a single way, such as milling or turning. With the improvement of social needs, the structure of workpieces is becoming more and more complex, and a single processing method can no longer meet the requirements of workpiece processing. When two or more processing methods are used on a workpiece, such as using a turning tool mechanism to turn the flatter part of the workpiece, and then using a milling cutter mechanism to mill the narrower part of the workpiece, two machine tools are required and they must be clamped twice to complete the processing. This method requires secondary clamping, which has low production efficiency, and secondary clamping is easy to reduce the processing accuracy due to changes in the positioning reference. Therefore, many CNC machining centers that can perform multiple processing are now designed.

[0003] The milling machine, also known as the milling center, is used to complete the cutting of various shaft and disc parts such as internal and external cylindrical surfaces, conical surfaces, grooving, chamfering, etc., and can also realize various processes such as carving, milling, drilling, etc. During the use of the milling machine, it needs to be equipped with a milling motor to provide power, but the power equipment usually needs to work stably for a long time. At present, the maintenance of the milling motor is only carried out manually on a regular basis. This not only fails to timely discover the safety hazards in the milling motor, but also reduces its working performance, thereby affecting the quality of the milling process.

[0004] Therefore, an optimized control scheme for a special turning-milling motor is desired, which can perform real-time intelligent monitoring of the working state of the special turning-milling motor, so as to control and generate a braking command when an abnormal working condition of the special turning-milling motor is found. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a special turning and milling compound motor with hydraulic brake and a control method thereof, which adopts artificial intelligence monitoring technology based on deep learning to extract the multi-scale dynamic feature information of the power signal of the special turning and milling compound motor in the time dimension, and uses the dynamic characteristics of the power to obtain the generated vibration signal, which is respectively input into a twin network with the same network structure as the actual collected vibration signal to mine the source domain image features of the two, and further extracts the difference feature information of the source domain image by the difference comparison of the obtained detection vibration feature and the generated vibration feature, and uses this to detect the abnormal working condition of the special turning and milling compound motor. In this way, a hydraulic emergency braking command is generated when the working condition of the special turning and milling compound motor is detected to ensure the safety of operation while ensuring the working performance of the special turning and milling compound motor.

[0006] According to one aspect of the present application, a special motor for turning and milling with hydraulic braking is provided, which comprises:

[0007] A working condition monitoring module, used for acquiring power values ​​of the turning-milling compound motor at a plurality of predetermined time points within a predetermined time period and a vibration signal of the turning-milling compound motor within the predetermined time period;

[0008] A power feature extraction module, used for arranging the power values ​​of the turning-milling compound special motor at the plurality of predetermined time points according to the time dimension as an input vector and then passing the vector through a multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector;

[0009] A power feature enhancement module, used for performing feature-level data enhancement on the multi-scale power feature vector based on a Gaussian density map to obtain a power feature matrix;

[0010] A vibration generation module, used for passing the power feature matrix through a vibration generator based on an adversarial generation network to generate a vibration signal;

[0011] A twin detection module, configured to pass the vibration signal and the generated vibration signal through a twin network model including a first image encoder and a second image encoder to obtain a detected vibration feature vector and a generated vibration feature vector, wherein the first image encoder and the second image encoder have the same network structure;

[0012] A difference module, used for calculating a difference feature vector between the detected vibration feature vector and the generated vibration feature vector, and performing a distribution transfer optimization oriented to a quasi-bounded closed domain on the difference feature vector to obtain an optimized difference feature vector;

[0013] A working condition monitoring result generating module, used for passing the optimized differential feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the working condition of the turning-milling compound special motor is normal; and

[0014] The hydraulic brake control module is used to determine whether to generate a hydraulic emergency brake instruction based on the classification result.

[0015] In the above-mentioned special motor for turning and milling with hydraulic brakes, the power feature extraction module includes: a first scale unit, used to input the input vector into the first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first-scale power feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel of a first length; a second scale unit, used to input the input vector into the second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second-scale power feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and a multi-scale cascade unit, used to cascade the first-scale power feature vector and the second-scale power feature vector to obtain the multi-scale power feature vector.

[0016] In the above-mentioned hydraulic brake milling compound special motor, the first scale unit is further used to: use the first convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the input vector according to the following formula to obtain the first scale power feature vector; wherein the formula is:

[0017]

[0018] Wherein, a is the width of the first convolution kernel in the x direction, F(a) is the first convolution kernel parameter vector, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents the input vector; the second scale unit is further used to: use the second convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the input vector according to the following formula to obtain the second scale power feature vector; wherein the formula is:

[0019]

[0020] Among them, b is the width of the second convolution kernel in the x direction, F(b) is the second convolution kernel parameter vector, G(xb) is the local vector matrix operated with the convolution kernel function, m is the size of the second convolution kernel, and X represents the input vector.

[0021] In the above-mentioned hydraulic brake milling compound special motor, the power feature enhancement module includes: a Gaussian density map construction unit, which is used to: construct a self-Gaussian density map of the multi-scale power feature vector according to the following formula; wherein the formula is:

[0022]

[0023] Wherein, μ represents the multi-scale power feature vector, and the value of each position of σ represents the variance between the eigenvalues ​​of the corresponding two positions in the multi-scale power feature vector; and a Gaussian discretization unit is used to perform Gaussian discretization on the Gaussian distribution of each position in the self-Gaussian density map to obtain a power feature matrix.

[0024] In the above-mentioned special turning-milling motor with hydraulic brake, the vibration generator based on the adversarial generation network includes a discriminator and a generator.

[0025] In the above-mentioned special turning-milling motor with hydraulic brake, the first image encoder and the second image encoder are convolutional neural network models as filters.

[0026] In the above-mentioned milling and turning compound motor with hydraulic brake, the differential module is further used to calculate the differential feature vector between the detected vibration feature vector and the generated vibration feature vector using the following formula; wherein the formula is:

[0027]

[0028] Among them, V d is the differential eigenvector, V 1 is the detected vibration feature vector, V 2 Generate a vibration eigenvector for the It represents the subtraction of the elements at corresponding positions of the detected vibration feature vector and the generated vibration feature vector.

[0029] In the above-mentioned special motor for turning and milling with hydraulic brakes, the working condition monitoring result generation module includes: a fully connected encoding unit, which is used to use the fully connected layer of the classifier to fully connect the optimized differential feature vector to obtain a coded classification feature vector; and a classification result generation unit, which is used to input the coded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0030] According to another aspect of the present application, a control method for a special turning-milling motor with hydraulic brake is also provided, which comprises:

[0031] Acquire power values ​​of the turning-milling compound motor at a plurality of predetermined time points within a predetermined time period and a vibration signal of the turning-milling compound motor in the predetermined time period;

[0032] The power values ​​of the milling-turning motor at the plurality of predetermined time points are arranged as input vectors according to the time dimension and then passed through a multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector;

[0033] Performing feature-level data enhancement on the multi-scale power feature vector based on a Gaussian density map to obtain a power feature matrix;

[0034] Passing the power feature matrix through a vibration generator based on a generative adversarial network to generate a vibration signal;

[0035] Passing the vibration signal and the generated vibration signal through a twin network model including a first image encoder and a second image encoder to obtain a detected vibration feature vector and a generated vibration feature vector, wherein the first image encoder and the second image encoder have the same network structure;

[0036] Calculating a differential feature vector between the detected vibration feature vector and the generated vibration feature vector, and performing a distribution transfer optimization oriented to a quasi-bounded closed domain on the differential feature vector to obtain an optimized differential feature vector;

[0037] Passing the optimized differential feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the working condition of the milling-turning motor is normal;

[0038] Based on the classification result, it is determined whether to generate a hydraulic emergency brake command.

[0039] In the above-mentioned control method of the special turning-milling compound motor with hydraulic brake, the power values ​​of the special turning-milling compound motor at the multiple predetermined time points are arranged as input vectors according to the time dimension and then passed through a multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector, including: inputting the input vector into the first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first-scale power feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel of a first length; inputting the input vector into the second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second-scale power feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and, cascading the first-scale power feature vector and the second-scale power feature vector to obtain the multi-scale power feature vector.

[0040] In the control method of the above-mentioned hydraulic brake milling composite special motor, the multi-scale power feature vector is enhanced at the feature level based on the Gaussian density map to obtain the power feature matrix, including: constructing the self-Gaussian density map of the multi-scale power feature vector according to the following formula; wherein the formula is:

[0041]

[0042] Wherein, μ represents the multi-scale power feature vector, and the value of each position of σ represents the variance between the eigenvalues ​​of the corresponding two positions in the multi-scale power feature vector; and the Gaussian distribution of each position in the self-Gaussian density map is Gaussian discretized to obtain a power feature matrix.

[0043] In the above-mentioned control method of the special turning-milling motor with hydraulic brake, the vibration generator based on the adversarial generation network includes a discriminator and a generator.

[0044] In the above-mentioned control method of the special turning-milling compound motor with hydraulic brake, the first image encoder and the second image encoder are convolutional neural network models as filters.

[0045] In the control method of the above-mentioned hydraulic brake milling composite special motor, the calculation of the differential feature vector between the detected vibration feature vector and the generated vibration feature vector, and the distribution transfer optimization of the differential feature vector oriented to the bounded closed domain to obtain the optimized differential feature vector, includes: calculating the differential feature vector between the detected vibration feature vector and the generated vibration feature vector by the following formula; wherein the formula is:

[0046]

[0047] Among them, V d is the differential eigenvector, V 1 is the detected vibration feature vector, V 2 Generate a vibration eigenvector for the It represents the subtraction of the elements at corresponding positions of the detected vibration feature vector and the generated vibration feature vector.

[0048] In the control method of the above-mentioned hydraulic brake milling composite special motor, the calculation of the differential feature vector between the detected vibration feature vector and the generated vibration feature vector, and the distribution transfer optimization of the differential feature vector oriented to a class bounded closed domain to obtain an optimized differential feature vector, also includes: the distribution transfer optimization of the differential feature vector oriented to a class bounded closed domain to obtain an optimized differential feature vector according to the following formula; wherein the formula is:

[0049]

[0050] Among them, v i is the i-th eigenvalue of the differential eigenvector, L is the length of the differential eigenvector, and log represents the logarithm with base 2.

[0051] In the above-mentioned control method of the special turning-milling compound motor with hydraulic brake, the optimized differential feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the working condition of the special turning-milling compound motor is normal, including: using the fully connected layer of the classifier to fully connect encode the optimized differential feature vector to obtain an encoded classification feature vector; and, inputting the encoded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0052] Compared with the prior art, the present application provides a special turning and milling motor with hydraulic brake and a control method thereof, which uses artificial intelligence monitoring technology based on deep learning to extract the multi-scale dynamic feature information of the power signal of the special turning and milling motor in the time dimension, and uses the dynamic features of the power to obtain the generated vibration signal, which is respectively input into a twin network with the same network structure as the actual collected vibration signal to mine the source domain image features of the two, and further extracts the difference feature information of the source domain image by the difference comparison of the obtained detection vibration feature and the generated vibration feature, and uses this to detect the abnormal working condition of the special turning and milling motor. In this way, a hydraulic emergency braking command is generated when the working condition of the special turning and milling motor is detected to ensure the working performance of the special turning and milling motor while ensuring the safety of operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0054] Figure 1 This is an application scenario diagram of a special turning-milling compound motor with hydraulic braking according to an embodiment of the present application.

[0055] Figure 2 It is a block diagram of a special motor for turning and milling with hydraulic braking according to an embodiment of the present application.

[0056] Figure 3 This is a structural diagram of a special motor for turning and milling with hydraulic braking according to an embodiment of the present application.

[0057] Figure 4It is a block diagram of a power feature extraction module in a special turning-milling motor with hydraulic brake according to an embodiment of the present application.

[0058] Figure 5 This is a flow chart of a control method for a special turning-milling motor with hydraulic braking according to an embodiment of the present application. DETAILED DESCRIPTION

[0059] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0060] Application Overview

[0061] As mentioned in the background technology above, the turning-milling compound machine is also called the turning-milling compound center, which is used to complete the cutting processing of internal and external cylindrical surfaces, conical surfaces, grooving, chamfering and other processes on various shaft and disk parts, and can realize various processing such as carving, milling, drilling, etc. During the use of the turning-milling compound machine, it is necessary to equip it with a turning-milling compound motor to provide power, but the power equipment usually needs to work stably for a long time. At present, the maintenance of the turning-milling compound motor only relies on manual regular maintenance, which not only cannot timely discover the safety hazards in the turning-milling compound motor, but also reduces its working performance, thereby affecting the quality of the turning-milling compound processing. Therefore, an optimized turning-milling compound motor control scheme is desired, which can perform real-time intelligent monitoring of the working status of the turning-milling compound motor, so as to control the generation of braking instructions when the working condition of the turning-milling compound motor is found to be abnormal.

[0062] At present, deep learning and neural networks have been widely used in computer vision, natural language processing, speech signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields.

[0063] In recent years, the development of deep learning and neural networks has provided new solutions and plans for the intelligent control of special motors for turning and milling.

[0064] Accordingly, considering that when the working state of the milling machine is monitored, when the milling machine is in normal working condition, the vibration signal generated by it will show periodic regular changes. However, when the dynamic characteristics of the vibration signal are actually used to detect the abnormal working state of the milling machine, the vibration signal will be affected by external environmental factors and the vibration of the milling machine itself. The feature extraction of the vibration signal of the milling machine will bring difficulties to the abnormal working state detection of the milling machine. In addition, considering that the power information of the milling machine during operation also reflects its working state during operation, if the dynamic characteristics of the working power of the milling machine can be further introduced to compare the differences in the characteristics, it will obviously improve the accuracy of abnormal detection.

[0065] Specifically, in the technical solution of the present application, an artificial intelligence monitoring technology based on deep learning is used to extract the multi-scale dynamic feature information of the power signal of the turning-milling compound motor in the time dimension, and the dynamic feature of the power is used to obtain the generated vibration signal, which is respectively input into a twin network with the same network structure as the actual collected vibration signal to mine the source domain image features of the two, and further the difference feature information of the source domain image is extracted by the difference comparison of the obtained detection vibration feature and the generated vibration feature, and the abnormal working condition of the turning-milling compound motor is detected by this. In this way, a hydraulic emergency braking command is generated when the working condition of the turning-milling compound motor is detected to ensure the safety of operation while ensuring the working performance of the turning-milling compound motor.

[0066] More specifically, in the technical solution of the present application, first, the power values ​​of the special turning and milling motor at multiple predetermined time points within a predetermined time period and the vibration signals of the special turning and milling motor in the predetermined time period are obtained. Then, since the power value of the special turning and milling motor has different pattern characteristics under different time period spans, the power values ​​of the special turning and milling motor at the multiple predetermined time points are arranged as input vectors according to the time dimension and then feature extraction is performed in a multi-scale neighborhood feature extraction module to extract the multi-scale neighborhood correlation features of the power of the special turning and milling motor under different time spans, thereby obtaining a multi-scale power feature vector.

[0067] Then, considering that the power value of the special turning and milling motor has volatility within the predetermined time period, in order to improve the accuracy of detecting abnormal working state of the special turning and milling motor, it is necessary to perform feature-level data enhancement on the power dynamic characteristics of the special turning and milling motor in the high-dimensional feature space. It should be understood that as a learning target of the neural network model, the Gaussian density map can represent the joint distribution of a single eigenvalue of the feature distribution due to its probability density when multiple eigenvalues ​​constitute an overall distribution, that is, the feature distribution is used as a prior distribution to obtain the probability density of each prior distribution position due to the correlation of other prior distribution positions as a posterior distribution, thereby more accurately describing the feature distribution in a higher dimension. Therefore, in the technical solution of the present application, the dynamic multi-scale neighborhood correlation characteristics of the power of the special turning and milling motor can be enhanced by the prior distribution of the power value, that is, the Gaussian distribution. Specifically, a Gaussian density map of the multi-scale power feature vector is constructed, and the Gaussian density map is Gaussian discretized so as not to generate information loss when the data feature is augmented, thereby obtaining a power feature matrix.

[0068] Furthermore, the power feature matrix is ​​passed through a vibration generator based on an adversarial generative network to obtain a generated vibration signal. It is worth mentioning that, here, the vibration generator based on the adversarial generative network includes a generator model and a discriminator model, the generator model of the vibration generator can utilize the dynamic characteristics of the power value of the turning-milling compound motor in the time series dimension to generate a generated vibration signal that characterizes the working characteristics of the turning-milling compound motor; and the discriminator model of the vibration generator can make the feature information extracted from the generated vibration signal and the reference vibration signal similar by "cheating" the discriminator, thereby enabling the generated vibration signal to characterize the vibration signal excited by the working characteristics of the turning-milling compound motor.

[0069] Next, the generated vibration signal and the collected vibration signal are further processed in a twin network model including a first image encoder and a second image encoder to obtain a detected vibration feature vector and a generated vibration feature vector. That is, the first image encoder of the twin network model is used to perform image encoding on the waveform of the vibration signal to extract the feature distribution information of the local implicit features of the waveform of the vibration signal in a high-dimensional space, thereby obtaining the detected vibration feature vector; and the second image encoder of the twin network model is used to encode the generated vibration signal to extract the local implicit feature distribution information of the generated vibration signal, thereby obtaining the generated vibration feature vector. In particular, here, the first image encoder and the second image encoder have the same network structure. It should be understood that here, the twin network model is used to perform convolution encoding of the same network structure on the vibration signal and the generated vibration signal, respectively, so as to extract feature information whose feature differences are not obvious in the source domain image, so as to focus more on the vibration hidden feature information of the turning-milling composite motor in the source domain images of the two.

[0070] Then, the differential feature vector between the detected vibration feature vector and the generated vibration feature vector is calculated to represent the difference feature information between the vibration hidden feature and the power hidden feature of the turning-milling compound motor, and the differential feature vector is used as the classification feature vector to perform classification processing in the classifier, so as to obtain a classification result indicating whether the working condition of the turning-milling compound motor is normal. Furthermore, based on the classification result, determine whether to generate a hydraulic emergency braking command. That is, in response to the classification result that the working condition of the turning-milling compound motor is abnormal, a hydraulic emergency braking command is generated. In this way, the working state of the turning-milling compound motor can be detected in real time and accurately, so as to ensure the working performance of the turning-milling compound motor while ensuring its operation safety.

[0071] In particular, in the technical solution of the present application, when the multi-scale power feature vector is subjected to feature-level data enhancement based on the Gaussian density map to obtain the power feature matrix, due to the randomness during Gaussian discretization, some abnormal distributions that deviate from the natural state will be introduced into the power feature matrix. These abnormal distributions will be amplified after passing through the vibration generator based on the adversarial generative network and the second image encoder, thereby causing the feature distribution of the generated vibration feature vector to be abnormal, so that the differential feature vector obtained by calculating the difference between the detected vibration feature vector and the generated vibration feature vector has inductive divergence when passing through the classifier, affecting the training speed of the classifier and the accuracy of the classification results.

[0072] Therefore, preferably, the differential feature vector is subjected to distribution transfer optimization oriented to a class-bounded closed domain:

[0073]

[0074] v i is the i-th eigenvalue of the differential eigenvector, L is the length of the differential eigenvector, and log represents the logarithm with base 2.

[0075] Here, the class-bounded closed domain-oriented distribution transfer optimization targets the possible inductive divergence of the high-dimensional feature distribution represented by the differential feature vector when it is transferred to the target domain of the classification problem. The feature distribution is converged toward the bounded closed domain of the feature set through structured information constraints based on conditional classification, so that the feature distribution of the differential feature vector is transferred to a range with stable structured boundaries in the target domain, thereby improving the stability of the inductive iteration of the classification solution, that is, improving the training speed of the classifier and the accuracy of the classification results.

[0076] Based on this, the present application proposes a special turning and milling compound motor with hydraulic brake, which includes: a working condition monitoring module, used to obtain the power value of the special turning and milling compound motor at multiple predetermined time points within a predetermined time period and the vibration signal of the special turning and milling compound motor in the predetermined time period; a power feature extraction module, used to arrange the power values ​​of the special turning and milling compound motor at the multiple predetermined time points as input vectors according to the time dimension and then pass them through a multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector; a power feature enhancement module, used to perform feature-level data enhancement on the multi-scale power feature vector based on a Gaussian density map to obtain a power feature matrix; a vibration generation module, used to pass the power feature matrix through a vibration generator based on an adversarial generative network to obtain a generated vibration signal; a twin detection module, used to pass the vibration signal and the generated vibration signal through a twin network model including a first image encoder and a second image encoder to obtain a detected vibration feature vector and a generated vibration feature vector, wherein the first image encoder and the second image encoder have the same network structure; a differential module, used to calculate the differential feature vector between the detected vibration feature vector and the generated vibration feature vector; a working condition monitoring result generation module, used to pass the differential feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the working condition of the milling compound motor is normal; and a hydraulic brake control module, used to determine whether to generate a hydraulic emergency brake command based on the classification result.

[0077] Figure 1 This is an application scenario diagram of a special motor for milling and turning with hydraulic brake according to an embodiment of the present application. Figure 1 As shown, in this application scenario, firstly, a power sensor (for example, Figure 1 Se1) as shown in the figure collects data of a milling machine special motor (for example, Figure 1 The power value M) shown in FIG. 1 and the power value M) provided by the vibration sensor (e.g. Figure 1 Se2 as shown in the figure) collects the vibration signal of the milling machine motor in the predetermined time period. Then, the power value of the milling machine motor at multiple predetermined time points in the predetermined time period and the vibration signal of the milling machine motor in the predetermined time period are input into a data processor of the milling machine motor with hydraulic brake (for example, Figure 1 P) as shown, wherein the data processor of the special turning-milling compound motor with hydraulic braking is capable of processing the power values ​​of the special turning-milling compound motor at multiple predetermined time points within the predetermined time period and the vibration signals of the special turning-milling compound motor in the predetermined time period to obtain a classification result indicating whether the working condition of the special turning-milling compound motor is normal, and based on the classification result, determine whether to generate a hydraulic emergency braking command.

[0078] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0079] Exemplary Systems

[0080] Figure 2 FIG. 1 is a block diagram of a milling and turning motor with hydraulic brake according to an embodiment of the present application. Figure 2As shown, according to the embodiment of the present application, the special turning and milling compound motor 100 with hydraulic brake includes: a working condition monitoring module 110, which is used to obtain the power value of the special turning and milling compound motor at multiple predetermined time points within a predetermined time period and the vibration signal of the special turning and milling compound motor in the predetermined time period; a power feature extraction module 120, which is used to arrange the power values ​​of the special turning and milling compound motor at the multiple predetermined time points as input vectors according to the time dimension and then pass them through a multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector; a power feature enhancement module 130, which is used to perform feature-level data enhancement on the multi-scale power feature vector based on a Gaussian density map to obtain a power feature matrix; a vibration generation module 140, which is used to pass the power feature matrix through a vibration generator based on an adversarial generative network to obtain a generated vibration signal; a twin detection module 150 , used for passing the vibration signal and the generated vibration signal through a twin network model including a first image encoder and a second image encoder to obtain a detected vibration feature vector and a generated vibration feature vector, wherein the first image encoder and the second image encoder have the same network structure; a differential module 160, used for calculating the differential feature vector between the detected vibration feature vector and the generated vibration feature vector, and performing a distribution transfer optimization for a class-bounded closed domain on the differential feature vector to obtain an optimized differential feature vector; a working condition monitoring result generation module 170, used for passing the optimized differential feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the working condition of the milling-turning composite motor is normal; and a hydraulic brake control module 180, used for determining whether to generate a hydraulic emergency brake command based on the classification result.

[0081] Figure 3 FIG. 1 is a schematic diagram of a milling and turning motor with hydraulic brake according to an embodiment of the present application. Figure 3As shown, in the architecture of the special turning and milling motor with hydraulic brake, first, the power value of the special turning and milling motor at multiple predetermined time points within a predetermined time period and the vibration signal of the special turning and milling motor in the predetermined time period are obtained. Then, the power values ​​of the special turning and milling motor at the multiple predetermined time points are arranged as input vectors according to the time dimension and then passed through the multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector. Then, the multi-scale power feature vector is subjected to feature-level data enhancement based on the Gaussian density map to obtain a power feature matrix. Furthermore, the power feature matrix is ​​passed through a vibration generator based on an adversarial generative network to obtain a generated vibration signal. Then, the vibration signal and the generated vibration signal are passed through a twin network model including a first image encoder and a second image encoder to obtain a detection vibration feature vector and a generated vibration feature vector, and the first image encoder and the second image encoder have the same network structure. Then, the differential feature vector between the detection vibration feature vector and the generated vibration feature vector is calculated, and the differential feature vector is subjected to distribution transfer optimization for a class-bounded closed domain to obtain an optimized differential feature vector. The optimized differential feature vector is then passed through a classifier to obtain a classification result, which is used to indicate whether the working condition of the milling machine motor is normal. Further, based on the classification result, it is determined whether to generate a hydraulic emergency brake instruction.

[0082] In the above-mentioned milling machine with hydraulic brake 100, the working condition monitoring module 110 is used to obtain the power value of the milling machine at multiple predetermined time points within a predetermined time period and the vibration signal of the milling machine in the predetermined time period. As mentioned in the above-mentioned background technology, the milling machine is also called the milling machine center, which is used to complete the cutting process of internal and external cylindrical surfaces, conical surfaces, grooving, chamfering and other processes on various shaft and disk parts, and can realize various processes such as carving, milling, drilling, etc. During the use of the milling machine, it is necessary to equip it with a milling machine to provide power, but the power equipment usually needs to work stably for a long time. At present, the maintenance of the milling machine is only carried out manually on a regular basis, which not only cannot timely discover the safety hazards in the milling machine, but also reduces its working performance, thereby affecting the quality of the milling machine processing. Therefore, an optimized milling machine control scheme is expected, which can perform real-time intelligent monitoring of the working state of the milling machine, so as to control the generation of braking instructions when the working condition of the milling machine is found to be abnormal.

[0083] At present, deep learning and neural networks have been widely used in computer vision, natural language processing, speech signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. In recent years, the development of deep learning and neural networks has provided new solutions and solutions for the intelligent control of milling motors.

[0084] Accordingly, considering that when the working state of the milling machine is monitored, when the milling machine is in normal working condition, the vibration signal generated by it will show periodic regular changes. However, when the dynamic characteristics of the vibration signal are actually used to detect the abnormal working state of the milling machine, the vibration signal will be affected by external environmental factors and the vibration of the milling machine itself. The feature extraction of the vibration signal of the milling machine will bring difficulties to the abnormal working state detection of the milling machine. In addition, considering that the power information of the milling machine during operation also reflects its working state during operation, if the dynamic characteristics of the working power of the milling machine can be further introduced to compare the differences in the characteristics, it will obviously improve the accuracy of abnormal detection.

[0085] Specifically, in the technical solution of the present application, an artificial intelligence monitoring technology based on deep learning is used to extract the multi-scale dynamic feature information of the power signal of the turning-milling compound motor in the time dimension, and the dynamic feature of the power is used to obtain the generated vibration signal, which is respectively input into a twin network with the same network structure as the actual collected vibration signal to mine the source domain image features of the two, and further the difference feature information of the source domain image is extracted by the difference comparison of the obtained detection vibration feature and the generated vibration feature, and the abnormal working condition of the turning-milling compound motor is detected by this. In this way, a hydraulic emergency braking command is generated when the working condition of the turning-milling compound motor is detected to ensure the safety of operation while ensuring the working performance of the turning-milling compound motor.

[0086] More specifically, in the technical solution of the present application, first, the power values ​​of the turning-milling compound motor at multiple predetermined time points within a predetermined time period and the vibration signal of the turning-milling compound motor in the predetermined time period are obtained. Here, the power values ​​of the turning-milling compound motor at multiple predetermined time points within the predetermined time period can be acquired by a power sensor, and the vibration signal of the turning-milling compound motor in the predetermined time period can be acquired by a power sensor.

[0087] In the above-mentioned special turning and milling compound motor 100 with hydraulic brake, the power feature extraction module 120 is used to arrange the power values ​​of the special turning and milling compound motor at the plurality of predetermined time points as input vectors according to the time dimension and then pass through the multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector. Since the power value of the special turning and milling compound motor has different pattern characteristics under different time period spans, that is, the power of the special turning and milling compound motor will also change with different working scenes, and the power value has different change characteristics in different time ranges in terms of time sequence. Therefore, after arranging the power values ​​of the special turning and milling compound motor at the plurality of predetermined time points as input vectors according to the time dimension, feature extraction is performed in the multi-scale neighborhood feature extraction module to extract the multi-scale neighborhood correlation features of the power of the special turning and milling compound motor under different time spans, thereby obtaining a multi-scale power feature vector. Here, the power values ​​of the special turning and milling compound motor at the plurality of predetermined time points are arranged as input vectors according to the time dimension in order to structure the data for the convenience of model calculation.

[0088] Figure 4 FIG. 1 is a block diagram of a power feature extraction module in a milling and turning hybrid motor with hydraulic brake according to an embodiment of the present application. Figure 4 As shown, the power feature extraction module 120 includes: a first scale unit 121, used to input the input vector into the first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first-scale power feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel of a first length; a second scale unit 122, used to input the input vector into the second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second-scale power feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and a multi-scale cascade unit 123, used to cascade the first-scale power feature vector and the second-scale power feature vector to obtain the multi-scale power feature vector.

[0089] Specifically, in the embodiment of the present application, the first scale unit 121 is further used to: use the first convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the input vector according to the following formula to obtain the first scale power feature vector; wherein the formula is:

[0090]

[0091] Wherein, a is the width of the first convolution kernel in the x direction, F(a) is the first convolution kernel parameter vector, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents the input vector; the second scale unit 122 is further used to: use the second convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the input vector according to the following formula to obtain the second scale power feature vector; wherein the formula is:

[0092]

[0093] Among them, b is the width of the second convolution kernel in the x direction, F(b) is the second convolution kernel parameter vector, G(xb) is the local vector matrix operated with the convolution kernel function, m is the size of the second convolution kernel, and X represents the input vector.

[0094] In the above-mentioned milling-turning motor 100 with hydraulic brake, the power feature enhancement module 130 is used to perform feature-level data enhancement on the multi-scale power feature vector based on the Gaussian density map to obtain a power feature matrix. Considering that the power value of the milling-turning motor is volatile within the predetermined time period, in order to improve the accuracy of abnormal detection of the working state of the milling-turning motor, it is necessary to perform feature-level data enhancement on the power dynamic characteristics of the milling-turning motor in a high-dimensional feature space.

[0095] It should be understood that as a learning target of the neural network model, the Gaussian density map can represent the joint distribution of a single eigenvalue of the feature distribution due to its probability density when multiple eigenvalues ​​constitute an overall distribution, that is, the feature distribution is used as a prior distribution to obtain the probability density of each prior distribution position due to the correlation of other prior distribution positions as a posterior distribution, thereby more accurately describing the feature distribution in a higher dimension. Therefore, in the technical solution of the present application, the dynamic multi-scale neighborhood correlation characteristics of the power of the milling-turning composite motor can be enhanced by the prior distribution of the power value, that is, the Gaussian distribution. Specifically, a Gaussian density map of the multi-scale power feature vector is constructed, and the Gaussian density map is Gaussian discretized so as not to generate information loss when the data features are augmented, thereby obtaining a power feature matrix.

[0096] Specifically, in the embodiment of the present application, the power feature enhancement module 130 includes: a Gaussian density map construction unit, which is used to: construct a self-Gaussian density map of the multi-scale power feature vector according to the following formula; wherein the formula is:

[0097]

[0098] Wherein, μ represents the multi-scale power feature vector, and the value of each position of σ represents the variance between the eigenvalues ​​of the corresponding two positions in the multi-scale power feature vector; and a Gaussian discretization unit is used to perform Gaussian discretization on the Gaussian distribution of each position in the self-Gaussian density map to obtain a power feature matrix.

[0099] In the above-mentioned milling and turning compound motor 100 with hydraulic brake, the vibration generation module 140 is used to pass the power feature matrix through the vibration generator based on the adversarial generation network to obtain the generated vibration signal. It is worth mentioning that here, the vibration generator based on the adversarial generation network includes a generator model and a discriminator model. The generator model of the vibration generator can use the dynamic characteristics of the power value of the milling and turning compound motor in the time series dimension to generate a generated vibration signal that characterizes the working characteristics of the milling and turning compound motor; and the discriminator model of the vibration generator can make the generated vibration signal and the feature information extracted from the reference vibration signal similar by "cheating" the discriminator, so that the generated vibration signal can characterize the vibration signal with the working characteristics of the milling and turning compound motor as the excitation. That is, the discriminator is used to compare and calculate the discriminator loss function value between the generated vibration signal and the real vibration signal, and use the discriminator loss function value as the loss function value to update the neural network parameters of the vibration generator based on the adversarial generation network.

[0100] In the above-mentioned hydraulic brake lathe-milling compound special motor 100, the twin detection module 150 is used to pass the vibration signal and the generated vibration signal through a twin network model including a first image encoder and a second image encoder to obtain a detection vibration feature vector and a generated vibration feature vector, and the first image encoder and the second image encoder have the same network structure. That is, the first image encoder of the twin network model is used to perform image encoding on the waveform of the vibration signal to extract the feature distribution information of the local implicit features of the waveform of the vibration signal in the high-dimensional space, thereby obtaining the detection vibration feature vector; and the second image encoder of the twin network model is used to encode the generated vibration signal to extract the local implicit feature distribution information of the generated vibration signal, thereby obtaining the generated vibration feature vector.

[0101] In particular, here, the first image encoder and the second image encoder have the same network structure. It should be understood that here, using the twin network model to perform convolution encoding of the same network structure on the vibration signal and the generated vibration signal respectively can extract feature information whose feature difference in the source domain image is not obvious, so as to focus more on the vibration hidden feature information of the turning-milling compound motor in the source domain images of the two.

[0102] In the above-mentioned milling and turning motor 100 with hydraulic brake, the difference module 160 is used to calculate the difference feature vector between the detected vibration feature vector and the generated vibration feature vector, and perform distribution transfer optimization for the differential feature vector oriented to the bounded closed domain to obtain the optimized difference feature vector. That is, the difference feature vector between the detected vibration feature vector and the generated vibration feature vector is calculated to represent the difference feature information between the vibration hidden feature and the power hidden feature of the milling and turning motor.

[0103] Specifically, in the embodiment of the present application, the difference module 160 is further used to calculate the difference feature vector between the detected vibration feature vector and the generated vibration feature vector using the following formula; wherein the formula is:

[0104]

[0105] Among them, V d is the differential eigenvector, V 1 is the detected vibration feature vector, V 2 Generate a vibration eigenvector for the It represents the subtraction of the elements at corresponding positions of the detected vibration feature vector and the generated vibration feature vector.

[0106] In particular, in the technical solution of the present application, when the multi-scale power feature vector is subjected to feature-level data enhancement based on the Gaussian density map to obtain the power feature matrix, due to the randomness during Gaussian discretization, some abnormal distributions that deviate from the natural state will be introduced into the power feature matrix. These abnormal distributions will be amplified after passing through the vibration generator based on the adversarial generative network and the second image encoder, thereby causing abnormalities in the characteristic distribution of the generated vibration feature vector, so that the differential feature vector obtained by calculating the difference between the detected vibration feature vector and the generated vibration feature vector has inductive divergence when passing through the classifier, affecting the training speed of the classifier and the accuracy of the classification results. Therefore, preferably, the differential feature vector is subjected to distribution transfer optimization oriented to the class bounded closed domain.

[0107] Specifically, in the embodiment of the present application, the difference module 160 is further used to: perform distribution transfer optimization for the differential feature vector oriented to a class bounded closed domain according to the following formula to obtain an optimized differential feature vector; wherein the formula is:

[0108]

[0109] Among them, v iis the i-th eigenvalue of the differential eigenvector, L is the length of the differential eigenvector, and log represents the logarithm with base 2.

[0110] Here, the class-bounded closed domain-oriented distribution transfer optimization targets the possible inductive divergence of the high-dimensional feature distribution represented by the differential feature vector when it is transferred to the target domain of the classification problem. The feature distribution is converged toward the bounded closed domain of the feature set through structured information constraints based on conditional classification, so that the feature distribution of the differential feature vector is transferred to a range with stable structured boundaries in the target domain, thereby improving the stability of the inductive iteration of the classification solution, that is, improving the training speed of the classifier and the accuracy of the classification results.

[0111] In the above-mentioned milling-turning motor 100 with hydraulic brake, the working condition monitoring result generation module 170 is used to pass the optimized differential feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the working condition of the milling-turning motor is normal. Here, the optimized differential feature vector is used as a classification feature vector to be classified by a classifier. That is, the classifier is used to divide and determine the class boundaries of the high-dimensional data manifold of the optimized differential feature vector to obtain a classification result for indicating whether the working condition of the milling-turning motor is normal.

[0112] Specifically, in an embodiment of the present application, the operating condition monitoring result generation module 170 includes: a fully connected encoding unit, used to use the fully connected layer of the classifier to fully connect the optimized differential feature vector to obtain a coded classification feature vector; and a classification result generation unit, used to input the coded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0113] In the above-mentioned milling machine motor 100 with hydraulic brake, the hydraulic brake control module 180 is used to determine whether to generate a hydraulic emergency brake command based on the classification result. That is, in response to the classification result that the working condition of the milling machine motor is abnormal, a hydraulic emergency brake command is generated. In this way, the working state of the milling machine motor can be accurately detected in real time to ensure the working performance of the milling machine motor while ensuring its operating safety.

[0114] In summary, the special turning and milling motor 100 with hydraulic brake according to the embodiment of the present application is explained, which adopts artificial intelligence monitoring technology based on deep learning to extract the multi-scale dynamic feature information of the power signal of the special turning and milling motor in the time dimension, and obtains the generated vibration signal with the dynamic feature of the power, and inputs it and the actually collected vibration signal into the twin network with the same network structure respectively to mine the source domain image features of the two, and further extracts the difference feature information of the source domain image by the difference comparison of the obtained detection vibration feature and the generated vibration feature, and uses it to detect the abnormal working condition of the special turning and milling motor. In this way, when the working condition of the special turning and milling motor is detected to be abnormal, a hydraulic emergency braking command is generated to ensure the working performance of the special turning and milling motor while ensuring the safety of operation.

[0115] Exemplary Methods

[0116] Figure 5 Flow chart of the control method of the milling machine special motor with hydraulic brake according to the embodiment of the present application. Figure 5 As shown, according to the control method of the special turning and milling compound motor with hydraulic brake according to the embodiment of the present application, the steps include: S110, obtaining the power value of the special turning and milling compound motor at multiple predetermined time points within a predetermined time period and the vibration signal of the special turning and milling compound motor in the predetermined time period; S120, arranging the power values ​​of the special turning and milling compound motor at the multiple predetermined time points as input vectors according to the time dimension and then passing them through a multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector; S130, performing feature-level data enhancement on the multi-scale power feature vector based on a Gaussian density map to obtain a power feature matrix; S140, passing the power feature matrix through a vibration generator based on an adversarial generative network to obtain a generated vibration signal; S150, The vibration signal and the generated vibration signal are obtained by a twin network model including a first image encoder and a second image encoder to obtain a detected vibration feature vector and a generated vibration feature vector, wherein the first image encoder and the second image encoder have the same network structure; S160, calculating the differential feature vector between the detected vibration feature vector and the generated vibration feature vector, and performing distribution transfer optimization for a class bounded closed domain on the differential feature vector to obtain an optimized differential feature vector; S170, passing the optimized differential feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the working condition of the turning-milling compound special motor is normal; and, S180, determining whether to generate a hydraulic emergency braking instruction based on the classification result.

[0117] In one example, in the above-mentioned control method of the special turning-milling compound motor with hydraulic brake, the power values ​​of the special turning-milling compound motor at the multiple predetermined time points are arranged as input vectors according to the time dimension and then passed through a multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector, including: inputting the input vector into the first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first-scale power feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel of a first length; inputting the input vector into the second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second-scale power feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and, cascading the first-scale power feature vector and the second-scale power feature vector to obtain the multi-scale power feature vector.

[0118] In one example, in the control method of the above-mentioned hydraulic brake milling composite special motor, the multi-scale power feature vector is enhanced at the feature level based on the Gaussian density map to obtain a power feature matrix, including: constructing a self-Gaussian density map of the multi-scale power feature vector according to the following formula; wherein the formula is:

[0119]

[0120] Wherein, μ represents the multi-scale power feature vector, and the value of each position of σ represents the variance between the eigenvalues ​​of the corresponding two positions in the multi-scale power feature vector; and the Gaussian distribution of each position in the self-Gaussian density map is Gaussian discretized to obtain a power feature matrix.

[0121] In one example, in the control method of the above-mentioned turning-milling hybrid motor with hydraulic brake, the vibration generator based on the adversarial generative network includes a discriminator and a generator.

[0122] In one example, in the above-mentioned control method of the special turning-milling compound motor with hydraulic brake, the first image encoder and the second image encoder are convolutional neural network models as filters.

[0123] In one example, in the control method of the above-mentioned hydraulic brake milling composite special motor, the calculation of the differential feature vector between the detected vibration feature vector and the generated vibration feature vector, and the distribution transfer optimization of the differential feature vector oriented to the bounded closed domain to obtain the optimized differential feature vector, includes: calculating the differential feature vector between the detected vibration feature vector and the generated vibration feature vector by the following formula; wherein the formula is:

[0124]

[0125] Among them, V d is the differential eigenvector, V 1 is the detected vibration feature vector, V 2 Generate a vibration eigenvector for the It represents the subtraction of the elements at corresponding positions of the detected vibration feature vector and the generated vibration feature vector.

[0126] In one example, in the control method of the above-mentioned hydraulic brake milling composite special motor, the calculation of the differential feature vector between the detected vibration feature vector and the generated vibration feature vector, and the distribution transfer optimization of the differential feature vector oriented to a class bounded closed domain to obtain an optimized differential feature vector, also includes: the distribution transfer optimization of the differential feature vector oriented to a class bounded closed domain to obtain an optimized differential feature vector according to the following formula; wherein the formula is:

[0127]

[0128] Among them, v i is the i-th eigenvalue of the differential eigenvector, L is the length of the differential eigenvector, and log represents the logarithm with base 2.

[0129] In one example, in the control method of the above-mentioned special turning-milling compound motor with hydraulic brake, the optimized differential feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the working condition of the special turning-milling compound motor is normal, including: using the fully connected layer of the classifier to fully connect encode the optimized differential feature vector to obtain an encoded classification feature vector; and, inputting the encoded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0130] In summary, the control method of the special turning and milling compound motor with hydraulic brake according to the embodiment of the present application is explained, which adopts artificial intelligence monitoring technology based on deep learning to extract the multi-scale dynamic feature information of the power signal of the special turning and milling compound motor in the time dimension, and uses the dynamic characteristics of the power to obtain the generated vibration signal, which is respectively input into the twin network with the same network structure as the actual collected vibration signal to mine the source domain image features of the two, and further extracts the difference feature information of the source domain image by the difference comparison of the obtained detection vibration feature and the generated vibration feature, and uses this to detect the abnormal working condition of the special turning and milling compound motor. In this way, a hydraulic emergency braking command is generated when the working condition of the special turning and milling compound motor is detected to ensure the safety of operation while ensuring the working performance of the special turning and milling compound motor.

Claims

1. A special motor for turning and milling with hydraulic brake, characterized in that: include: A working condition monitoring module, used for acquiring power values ​​of the turning-milling compound motor at a plurality of predetermined time points within a predetermined time period and a vibration signal of the turning-milling compound motor within the predetermined time period; A power feature extraction module, used for arranging the power values ​​of the turning-milling compound special motor at the plurality of predetermined time points according to the time dimension as an input vector and then passing the vector through a multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector; A power feature enhancement module, used for performing feature-level data enhancement on the multi-scale power feature vector based on a Gaussian density map to obtain a power feature matrix; A vibration generation module, used for passing the power feature matrix through a vibration generator based on an adversarial generation network to generate a vibration signal; A twin detection module, configured to pass the vibration signal and the generated vibration signal through a twin network model including a first image encoder and a second image encoder to obtain a detected vibration feature vector and a generated vibration feature vector, wherein the first image encoder and the second image encoder have the same network structure; A difference module, used for calculating a difference feature vector between the detected vibration feature vector and the generated vibration feature vector, and performing a distribution transfer optimization oriented to a quasi-bounded closed domain on the difference feature vector to obtain an optimized difference feature vector; A working condition monitoring result generating module, used for passing the optimized differential feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the working condition of the turning-milling compound special motor is normal; as well as a hydraulic brake control module, configured to determine whether to generate a hydraulic emergency brake instruction based on the classification result; Wherein, the power feature enhancement module includes: A Gaussian density map construction unit is used to: construct a self-Gaussian density map of the multi-scale power feature vector according to the following formula; Wherein, the formula is: Wherein, μ represents the multi-scale power feature vector, and the value of each position of σ represents the variance between the feature values ​​of corresponding two positions in the multi-scale power feature vector; and A Gaussian discretization unit, used for performing Gaussian discretization on the Gaussian distribution of each position in the self-Gaussian density map to obtain a power feature matrix; Wherein, the differential module is also used for: The differential feature vector is subjected to distribution transfer optimization oriented to a class-bounded closed domain using the following formula to obtain an optimized differential feature vector; Wherein, the formula is: Among them, v i is the i-th eigenvalue of the differential eigenvector, L is the length of the differential eigenvector, and log represents the logarithm with base 2.

2. The special motor for turning and milling with hydraulic brake according to claim 1 is characterized in that: The power feature extraction module comprises: A first scale unit, configured to input the input vector into a first convolutional layer of the multi-scale neighborhood feature extraction module to obtain a first scale power feature vector, wherein the first convolutional layer has a first one-dimensional convolutional kernel of a first length; a second scaling unit, configured to input the input vector into a second convolutional layer of the multi-scale neighborhood feature extraction module to obtain a second-scale power feature vector, wherein the second convolutional layer has a second one-dimensional convolutional kernel of a second length, and the first length is different from the second length; and A multi-scale cascading unit is used to cascade the first-scale power feature vector and the second-scale power feature vector to obtain the multi-scale power feature vector.

3. The special motor for turning and milling with hydraulic brake according to claim 2 is characterized in that: The first scale unit is further used to: use the first convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the input vector according to the following formula to obtain the first scale power feature vector; Wherein, the formula is: Wherein, a is the width of the first convolution kernel in the x direction, F(a) is the first convolution kernel parameter vector, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents the input vector; The second scale unit is further used to: use the second convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the input vector according to the following formula to obtain the second scale power feature vector; Wherein, the formula is: Among them, b is the width of the second convolution kernel in the x direction, F(b) is the second convolution kernel parameter vector, G(xb) is the local vector matrix operated with the convolution kernel function, m is the size of the second convolution kernel, and X represents the input vector.

4. The special motor for turning and milling with hydraulic brake according to claim 3 is characterized in that: The vibration generator based on the adversarial generative network includes a discriminator and a generator.

5. The special motor for turning and milling with hydraulic brake according to claim 4 is characterized in that: The first image encoder and the second image encoder are convolutional neural network models as filters.

6. The special motor for turning and milling with hydraulic brake according to claim 5, characterized in that: The differential module is further used for: The differential feature vector between the detected vibration feature vector and the generated vibration feature vector is calculated by the following formula; Wherein, the formula is: Among them, V d is the differential feature vector, V1 is the detected vibration feature vector, V2 is the generated vibration feature vector, It represents the subtraction of the elements at corresponding positions of the detected vibration feature vector and the generated vibration feature vector.

7. The special turning and milling motor with hydraulic brake according to claim 6 is characterized in that: The working condition monitoring result generating module comprises: a fully connected encoding unit, configured to perform fully connected encoding on the optimized differential feature vector using the fully connected layer of the classifier to obtain an encoded classification feature vector; and The classification result generating unit is used to input the encoded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

8. A control method for a special turning and milling motor with hydraulic brake, used for the special turning and milling motor with hydraulic brake according to claim 1, characterized in that: include: Acquire power values ​​of the turning-milling compound motor at a plurality of predetermined time points within a predetermined time period and a vibration signal of the turning-milling compound motor in the predetermined time period; The power values ​​of the milling-turning motor at the plurality of predetermined time points are arranged as input vectors according to the time dimension and then passed through a multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector; Performing feature-level data enhancement on the multi-scale power feature vector based on a Gaussian density map to obtain a power feature matrix; Passing the power feature matrix through a vibration generator based on a generative adversarial network to generate a vibration signal; Passing the vibration signal and the generated vibration signal through a twin network model including a first image encoder and a second image encoder to obtain a detected vibration feature vector and a generated vibration feature vector, wherein the first image encoder and the second image encoder have the same network structure; calculating a differential eigenvector between the detected vibration eigenvector and the generated vibration eigenvector; Passing the differential feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the working condition of the milling-turning motor is normal; as well as Based on the classification result, it is determined whether to generate a hydraulic emergency brake command.

Images