Vacuum pump operation state monitoring method based on adaptive multi-domain feature fusion

Abstract

The invention discloses a self-adaptive multi-domain feature fusion vacuum pump operation state monitoring method, and relates to the technical field of vacuum pump monitoring, and the method comprises the following steps: collecting multi-source heterogeneous parameters in the operation of a vacuum pump; performing time domain feature extraction, frequency domain feature extraction and time-frequency domain feature extraction, and performing multi-domain feature fusion to obtain vacuum pump operation multi-domain features; the weight coefficient of the vacuum pump operation multi-domain features is adjusted, and screened multi-domain features are obtained; carrying out weight coefficient optimization; and carrying out secondary adjustment on the weight coefficient of the vacuum pump operation multi-domain feature to obtain a final multi-domain feature, and judging the operation state of the vacuum pump according to the final multi-domain feature. According to the method, through multi-domain feature extraction and fusion and in combination with a weight adaptive optimization technology, the reliability of monitoring the running state of the vacuum pump is improved; and different operation conditions of the vacuum pump can be matched, and generalization is improved.

Description

Technical Field

[0001] The present application relates to the technical field of vacuum pump monitoring, and in particular to a vacuum pump operating status monitoring method based on adaptive multi-domain feature fusion. Background Art

[0002] Vacuum pumps are widely used in various industrial fields, such as semiconductor manufacturing, aerospace, food processing, etc. During operation, vacuum pumps may experience performance degradation, failures, etc. due to long-term operation, wear, pollution, etc. Therefore, real-time monitoring of the operating status of vacuum pumps is of great significance to ensure the normal operation of vacuum pumps and improve production efficiency.

[0003] In the prior art, Chinese patent CN110173422A discloses an online monitoring system, method and terminal device of the Internet of Things for a vacuum pump, including a power module, a monitoring module, a processing module and a communication module; the power module is used to provide power to the monitoring module, the processing module and the communication module; the monitoring module is used to detect first parameter information of the vacuum pump; the processing module is used to read the first parameter information, calculate the second parameter information according to the first parameter information, and determine whether the first parameter information and the second parameter information exceed a preset range. When exceeding the preset range, it is determined that there is an abnormality and the communication module is started to transmit the first parameter information, the second parameter information and / or the abnormal information.

[0004] However, the above-mentioned existing technologies do not take into account the degradation characteristics of the vacuum pump. When constructing the model, it is difficult to match the parameter weights with the actual physical quantities. The discrimination effect is poor when the vacuum pump fault characteristics are sparse. The segmentation and monotonicity of the vacuum pump operation degradation trend are not significant, so the reliability of the vacuum pump operation status monitoring is poor. In addition, the above-mentioned existing technologies are difficult to match different operating conditions of the vacuum pump and have poor generalization. Summary of the invention

[0005] The present application provides a vacuum pump operating status monitoring method with adaptive multi-domain feature fusion, which is used to solve the problems of poor reliability and poor generalization of vacuum pump operating status monitoring in existing vacuum pump monitoring technologies.

[0006] On the one hand, the present application provides a vacuum pump operating status monitoring method based on adaptive multi-domain feature fusion, comprising the following steps: Step 1: Collect multi-source heterogeneous parameters during the operation of the vacuum pump.

[0007] Step 2: Perform time domain feature extraction, frequency domain feature extraction and time-frequency domain feature extraction on the multi-source heterogeneous parameters, and perform multi-domain feature fusion to obtain multi-domain features of vacuum pump operation.

[0008] Step three, using a weight adaptive optimization model to adjust the weight coefficients of the multi-domain characteristics of the vacuum pump operation to obtain the screened multi-domain characteristics.

[0009] Step 4: Optimize the weight coefficient of the weight adaptive optimization model according to the screened multi-domain features and the vacuum pump historical operating status curve library.

[0010] Step five, using the optimized weight adaptive optimization model to perform secondary adjustment on the weight coefficient of the multi-domain characteristics of the vacuum pump operation to obtain the final multi-domain characteristics, and judging the operating status of the vacuum pump according to the final multi-domain characteristics.

[0011] In a possible implementation, in step one, the multi-source heterogeneous parameters include: a vacuum signal, a temperature signal, a current signal, a voltage signal, a cooling water signal, and a vibration signal.

[0012] In a possible implementation manner, in step 1, after obtaining the multi-source heterogeneous parameters, the multi-source heterogeneous parameters are preprocessed.

[0013] The preprocessing includes: outlier removal processing and noise removal processing.

[0014] In a possible implementation, in step 2, the time domain feature extraction uses one or both of statistical extraction and inverse Fourier transform.

[0015] The frequency domain feature extraction adopts one or more of discrete Fourier transform, fast Fourier transform, and power spectral density.

[0016] The time-frequency domain feature extraction adopts one or more of wavelet transform, short-time Fourier transform, and energy spectral density.

[0017] In a possible implementation, step 4 includes: The current operating state curve of the vacuum pump is obtained by fitting the screened multi-domain characteristics and the actual degradation condition of the vacuum pump.

[0018] A curve that is most similar to the current operating state curve of the vacuum pump is selected from a vacuum pump historical operating state curve library as an approximate operating condition curve.

[0019] The weight coefficient of the weight adaptive optimization model is optimized according to the approximate operating condition curve.

[0020] In a possible implementation, in step 4, selecting a curve that is most similar to the current operating state curve of the vacuum pump from a historical operating state curve library of the vacuum pump as an approximate operating condition curve includes: The distance between the current operating state curve of the vacuum pump and each curve in the historical operating state curve library of the vacuum pump is calculated by a distance function, and the curve with the smallest distance to the current operating state curve of the vacuum pump is the approximate operating condition curve.

[0021] In a possible implementation, in step five, after obtaining the final multi-domain feature, the failure probability and remaining life of the vacuum pump are calculated according to the final multi-domain feature, and maintenance recommendations are given.

[0022] The vacuum pump operating status monitoring method of the present application based on adaptive multi-domain feature fusion has the following advantages: Through multi-domain feature extraction and fusion combined with weight adaptive optimization technology, the reliability of vacuum pump operation status monitoring is improved; it can match different operating conditions of the vacuum pump and improve generalization.

[0023] The distance between the current operating state curve of the vacuum pump and each curve in the historical operating state curve library of the vacuum pump is calculated through the distance function. The curve with the smallest distance to the current operating state curve of the vacuum pump is the approximate operating condition curve, which improves the accuracy of matching different operating conditions of the vacuum pump. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 A flowchart of a vacuum pump operating status monitoring method based on adaptive multi-domain feature fusion provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0027] like Figure 1 As shown, the embodiment of the present application provides a vacuum pump operating status monitoring method based on adaptive multi-domain feature fusion, comprising the following steps: Step 1: Collect multi-source heterogeneous parameters during the operation of the vacuum pump.

[0028] Step 2: Perform time domain feature extraction, frequency domain feature extraction and time-frequency domain feature extraction on the multi-source heterogeneous parameters, and perform multi-domain feature fusion to obtain multi-domain features of vacuum pump operation.

[0029] Step three, using a weight adaptive optimization model to adjust the weight coefficients of the multi-domain characteristics of the vacuum pump operation to obtain the screened multi-domain characteristics.

[0030] Step 4: Optimize the weight coefficient of the weight adaptive optimization model according to the screened multi-domain features and the vacuum pump historical operating status curve library.

[0031] Step five, using the optimized weight adaptive optimization model to perform secondary adjustment on the weight coefficient of the multi-domain characteristics of the vacuum pump operation to obtain the final multi-domain characteristics, and judging the operating status of the vacuum pump according to the final multi-domain characteristics.

[0032] Exemplarily, in step one, the multi-source heterogeneous parameters include: a vacuum signal, a temperature signal, a current signal, a voltage signal, a cooling water signal, and a vibration signal.

[0033] Exemplarily, in step 1, after obtaining the multi-source heterogeneous parameters, the multi-source heterogeneous parameters are preprocessed.

[0034] The preprocessing includes: outlier removal processing and noise removal processing.

[0035] Specifically, in this embodiment, both the outlier removal process and the noise removal process adopt the existing data preprocessing technology.

[0036] Exemplarily, in step 2, the time domain feature extraction adopts one or both of statistical extraction and inverse Fourier transform.

[0037] The frequency domain feature extraction adopts one or more of discrete Fourier transform, fast Fourier transform, and power spectral density.

[0038] The time-frequency domain feature extraction adopts one or more of wavelet transform, short-time Fourier transform, and energy spectral density.

[0039] Specifically, in this embodiment, in step 2, time domain feature extraction and frequency domain feature extraction are performed on vacuum degree signal, temperature signal, current signal, voltage signal, cooling water signal, and vibration signal. The time domain feature extraction uses statistical quantity extraction, including statistical features such as mean, maximum value, minimum value, and standard deviation; the frequency domain feature extraction uses discrete Fourier transform. In this embodiment, time-frequency domain feature extraction is also performed on current signal, voltage signal, and vibration signal, and the time-frequency domain feature extraction uses short-time Fourier transform.

[0040] Specifically, in this embodiment, in step three, a weight adaptive optimization model is used to adjust the weight coefficients of the multi-domain characteristics of the vacuum pump operation, so as to screen the multi-domain characteristics.

[0041] Exemplarily, in step three, the weight adaptive optimization model includes steps such as weight initialization, weight update rules, and adaptive mechanisms, wherein weight initialization can assign initial weights to each parameter of the model; weight update rules require the design of weight update algorithms, such as those based on gradient descent algorithms, momentum methods, adaptive learning rate algorithms (such as Adam, RMSprop), etc.; adaptive mechanisms need to dynamically adjust weights according to the performance of the model during training, which may include: regularization techniques, attention mechanisms, meta-learning and other methods.

[0042] Exemplarily, step four includes: The current operating state curve of the vacuum pump is obtained by fitting the screened multi-domain characteristics and the actual degradation condition of the vacuum pump.

[0043] A curve that is most similar to the current operating state curve of the vacuum pump is selected from a vacuum pump historical operating state curve library as an approximate operating condition curve.

[0044] The weight coefficient of the weight adaptive optimization model is optimized according to the approximate operating condition curve.

[0045] Specifically, in this embodiment, in step 4, the approximate operating condition curve is output y, the multi-domain feature of the vacuum pump is input x, and y n It is equal to multiplying each input x by the corresponding weight coefficient a and accumulating them, such as: y n =a 1 *x 1 +a 2 *x 2 +…+a n *x n , where a n represents the weight coefficient of the nth feature in the multi-domain feature of the vacuum pump, x n Represents the nth feature in the multi-domain feature of the vacuum pump. Constantly adjust the weight coefficient a 1 to a n , until y and y n When the error is within a certain range, the calculation stops and the current weight coefficient a 1 to a n is the optimized weight coefficient.

[0046] Exemplarily, in step 4, selecting a curve that is most similar to the current operating state curve of the vacuum pump from a historical operating state curve library of the vacuum pump as an approximate operating condition curve includes: The distance between the current operating state curve of the vacuum pump and each curve in the historical operating state curve library of the vacuum pump is calculated by a distance function, and the curve with the smallest distance to the current operating state curve of the vacuum pump is the approximate operating condition curve.

[0047] Specifically, in this embodiment, in step four, the distance function adopts the Hausdorff distance function. In other possible embodiments, one of the Euclidean distance function, the Manhattan distance function, and the Chebyshev distance function can also be used to calculate the distance between the current operating status curve of the vacuum pump and each curve in the historical operating status curve library of the vacuum pump.

[0048] Exemplarily, in step five, after obtaining the final multi-domain feature, the failure probability and remaining life of the vacuum pump are calculated based on the final multi-domain feature, and maintenance recommendations are given.

[0049] Specifically, in this embodiment, in step five, according to the final multi-domain characteristics of the real-time vacuum pump operation, combined with the historical operation data and historical operation status of the vacuum pump, the current failure probability of the vacuum pump is dynamically evaluated through machine learning and other methods; the final multi-domain characteristics of the vacuum pump operation are used to construct a long-term multi-domain feature set of the vacuum pump operation in time series, and on this basis and in combination with the historical operating conditions of the vacuum pump and deep learning and other methods, the remaining life of the vacuum pump is predicted; in combination with the vacuum pump failure probability and the predicted remaining life, the operating status of the vacuum pump can be determined, and on this basis, in combination with the expert database system, targeted maintenance suggestions are put forward, such as corrective maintenance, improvement maintenance and predictive maintenance.

[0050] The embodiment of the present application improves the reliability of vacuum pump operating status monitoring through multi-domain feature extraction and fusion combined with weight adaptive optimization technology; it can match different operating conditions of the vacuum pump and improve generalization.

[0051] The distance between the current operating state curve of the vacuum pump and each curve in the historical operating state curve library of the vacuum pump is calculated through the distance function. The curve with the smallest distance to the current operating state curve of the vacuum pump is the approximate operating condition curve, which improves the accuracy of matching different operating conditions of the vacuum pump.

[0052] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0053] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A vacuum pump operating status monitoring method based on adaptive multi-domain feature fusion, characterized in that: The following steps are involved: Step 1, collecting multi-source heterogeneous parameters during the operation of the vacuum pump; Step 2, performing time domain feature extraction, frequency domain feature extraction and time-frequency domain feature extraction on the multi-source heterogeneous parameters, and performing multi-domain feature fusion to obtain multi-domain features of vacuum pump operation; Step 3, using a weight adaptive optimization model to adjust the weight coefficient of the multi-domain characteristics of the vacuum pump operation to obtain the screening multi-domain characteristics; Step 4, optimizing the weight coefficient of the weight adaptive optimization model according to the screened multi-domain features and the vacuum pump historical operating status curve library; Step five, using the optimized weight adaptive optimization model to perform secondary adjustment on the weight coefficient of the multi-domain characteristics of the vacuum pump operation to obtain the final multi-domain characteristics, and judging the operating status of the vacuum pump according to the final multi-domain characteristics.

2. The method for monitoring the operating status of a vacuum pump based on adaptive multi-domain feature fusion according to claim 1, characterized in that: In step one, the multi-source heterogeneous parameters include: vacuum signal, temperature signal, current signal, voltage signal, cooling water signal, and vibration signal.

3. The method for monitoring the operating status of a vacuum pump based on adaptive multi-domain feature fusion according to claim 1, characterized in that: In step 1, after obtaining the multi-source heterogeneous parameters, preprocessing the multi-source heterogeneous parameters; The preprocessing includes: outlier removal processing and noise removal processing.

4. The method for monitoring the operating status of a vacuum pump by adaptive multi-domain feature fusion according to claim 1, characterized in that: In step 2, the time domain feature extraction adopts one or both of statistical quantity extraction and inverse Fourier transform; The frequency domain feature extraction adopts one or more of discrete Fourier transform, fast Fourier transform, and power spectral density; The time-frequency domain feature extraction adopts one or more of wavelet transform, short-time Fourier transform, and energy spectral density.

5. The method for monitoring the operating status of a vacuum pump by adaptive multi-domain feature fusion according to claim 1, characterized in that: Step 4 includes: According to the multi-domain characteristics of the screening and the actual degradation condition of the vacuum pump, a current operating state curve of the vacuum pump is obtained by fitting; Selecting a curve that is most similar to the current operating state curve of the vacuum pump from a historical operating state curve library of the vacuum pump as an approximate operating condition curve; The weight coefficient of the weight adaptive optimization model is optimized according to the approximate operating condition curve.

6. The method for monitoring the operating status of a vacuum pump by adaptive multi-domain feature fusion according to claim 5, characterized in that: In step 4, the step of selecting a curve that is most similar to the current operating state curve of the vacuum pump from the historical operating state curve library of the vacuum pump as the approximate operating state curve includes: The distance between the current operating state curve of the vacuum pump and each curve in the historical operating state curve library of the vacuum pump is calculated by a distance function, and the curve with the smallest distance to the current operating state curve of the vacuum pump is the approximate operating condition curve.

7. The method for monitoring the operating status of a vacuum pump by adaptive multi-domain feature fusion according to claim 1, characterized in that: In step five, after obtaining the final multi-domain features, the failure probability and remaining life of the vacuum pump are calculated according to the final multi-domain features, and maintenance recommendations are given.

Images