Preventive maintenance diagnosis method for motor equipment based on multi-parameter verification
By denoising the motor operating parameters and converting them into two-dimensional images, combined with texture and geometric feature extraction, the problem of low efficiency and accuracy in motor early warning fault diagnosis is solved, and efficient and accurate fault identification is achieved.
Patent Information
- Application Number
- CN202411216181.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-02
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-09-02
AI Technical Summary
Existing technologies for motor early warning fault diagnosis have low efficiency and accuracy, and it is difficult to effectively extract the features of motor operating parameters.
The motor's operating parameters are monitored by sensors, and after noise reduction, they are converted into two-dimensional images. Combined with texture and geometric feature extraction, fault type identification is performed using models such as support vector machines.
It achieves efficient and accurate motor early warning fault diagnosis, simplifies the data acquisition process, reduces computational complexity, and improves diagnostic efficiency and accuracy.
Smart Images

Figure CN118731689B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of motor equipment maintenance, and in particular to a motor equipment preventive maintenance diagnosis method based on multi-parameter verification. Background Art
[0002] The motor is an important electromechanical device that can convert electrical energy into mechanical energy and vice versa. It plays a very important role in many industrial fields such as electric transmission, transportation, servo control, etc. Therefore, it is very important to ensure the safety and reliability of the motor operation.
[0003] In recent years, with the development of intelligent technology, technologies such as intelligent monitoring and early warning have been widely used in routine electromechanical equipment maintenance operations. Through early warning fault diagnosis, potential electromechanical equipment failures can be identified and repaired in advance, ensuring the safety and reliability of system operation.
[0004] In the existing technology, deep learning technology is usually used to improve the accuracy of early warning fault diagnosis, reduce labor costs, thereby simplifying the diagnosis process and avoiding economic losses. However, the motor operation parameter signal often contains multiple signal dimensions such as three-phase current, three-phase voltage, stator current, rotor current, motor speed, etc., which makes it difficult to extract the characteristics of the motor operation parameters, and the efficiency and accuracy of early warning fault diagnosis are low.
[0005] Therefore, how to accurately and efficiently perform motor early warning fault diagnosis is a technical problem that needs to be solved urgently. Summary of the invention
[0006] The present application provides a preventive maintenance diagnosis method for motor equipment based on multi-parameter verification, which is used to solve the technical problem of low efficiency and accuracy of early warning fault diagnosis of motor equipment in the prior art.
[0007] In a first aspect, the present application provides a preventive maintenance diagnosis method for motor equipment based on multi-parameter verification, comprising:
[0008] Monitor the operating parameters of the motor to be diagnosed by a sensor to obtain a real-time monitoring signal, wherein the real-time monitoring signal is any one of a current signal, a temperature signal, and a vibration signal;
[0009] The real-time monitoring signal is de-noised to obtain the target monitoring signal, and the abnormal data detection method is used to determine whether the target monitoring signal is abnormal. If it is judged to be abnormal, the early warning fault type identification process is entered;
[0010] A one-dimensional time series signal with a preset time series length is obtained from the target monitoring signal, and the one-dimensional time series signal is converted into a two-dimensional image using a Gram angle field;
[0011] Performing feature extraction on the two-dimensional image, wherein the feature extraction includes texture feature extraction and geometric feature extraction;
[0012] Determine the statistical characteristics of the motor operating parameters based on the target monitoring signal, and splice the statistical characteristics of the target monitoring signal with the texture feature vector and the geometric feature vector of the two-dimensional image to obtain the target feature vector;
[0013] The target feature vector is input into the pre-trained target model to determine the early warning fault type of the motor.
[0014] Furthermore, before performing real-time early warning fault diagnosis, it also includes obtaining multiple groups of sample signals characterizing motor operating parameters and the types corresponding to the sample signals, determining the target feature vector corresponding to the sample signal, and using the target feature vector as input and the type as output, and inputting them into the target model for training.
[0015] In a second aspect, the present application provides a preventive maintenance diagnostic device for motor equipment based on multi-parameter verification, the device comprising a data acquisition module, an abnormality detection module, an image generation module, an image feature extraction module, a target feature generation module and a warning fault identification module.
[0016] A data acquisition module, used to monitor the operating parameters of the motor to be diagnosed through sensors to obtain a real-time monitoring signal, wherein the real-time monitoring signal is any one of a current signal, a temperature signal, and a vibration signal;
[0017] The anomaly detection module is used to perform denoising on the real-time monitoring signal to obtain the target monitoring signal, and use the abnormal data detection method to determine whether the target monitoring signal is abnormal. If it is judged to be abnormal, it will be transferred to the image generation module for data processing;
[0018] An image generation module is used to obtain a one-dimensional time series signal with a preset time series length in the target monitoring signal, and convert the one-dimensional time series signal into a two-dimensional image using a Gram angle field;
[0019] An image feature extraction module, used for performing feature extraction on the two-dimensional image, wherein the feature extraction includes texture feature extraction and geometric feature extraction;
[0020] A target feature generation module is used to determine the statistical features of the motor operating parameters based on the target monitoring signal, and to splice the statistical features of the target monitoring signal with the texture feature vector and the geometric feature vector of the two-dimensional image to obtain a target feature vector;
[0021] A warning fault identification module is used to input the target feature vector into a pre-trained target model to determine the type of warning fault of the motor;
[0022] Furthermore, the device also includes a model training module;
[0023] The model training module is used to obtain multiple groups of sample signals representing motor operating parameters and the types corresponding to the sample signals, determine the target feature vector corresponding to the sample signal, and use the target feature vector as input and the type as output to input into the target model for training.
[0024] Compared with the prior art, this application has the following beneficial effects:
[0025] 1. The present application realizes data dimensionality upgrading by converting the motor operating parameter signal into a two-dimensional image. High-dimensional features often contain more data features. The deep-level data features can be extracted through the two-dimensional image corresponding to the motor operating parameter signal, and further combined with the statistical characteristics of the operating parameter signal to obtain a more accurate feature representation, which can more effectively determine the warning fault type of the motor. While simplifying the data collection process, it realizes accurate fault diagnosis and identification relying only on a single operating parameter indicator, thereby improving the efficiency and accuracy of motor warning fault diagnosis.
[0026] 2. This application designs a feature extraction method for texture feature vectors and geometric feature vectors of two-dimensional images, which not only reduces the input feature dimension of the target model and reduces the computational complexity of the model, but also, compared with traditional image feature extraction methods such as CNN convolutional neural networks, the feature extraction method of this application greatly simplifies the feature extraction process. While ensuring the accuracy of feature extraction, it effectively reduces the computational amount of image feature extraction, further improving the efficiency of early warning fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0028] Figure 1 A flowchart of a method for preventive maintenance diagnosis of motor equipment based on multi-parameter verification provided in an embodiment of the present application;
[0029] Figure 2 A schematic diagram of the convolution kernel EM value provided in an embodiment of the present application;
[0030] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION
[0031] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.
[0032] Usually, a working motor will generate various characteristic signals, such as current, temperature, vibration and other physical quantities, which can be used as the representation of motor performance and operating conditions. Motor equipment warning faults can be divided into three types, namely electrical warning faults, mechanical warning faults and magnetic warning faults. Among them, mechanical faults include bearing warning faults, eccentricity warning faults and other types of mechanical warning faults.
[0033] Figure 1 The present invention provides a flow chart of a method for preventive maintenance diagnosis of motor equipment based on multi-parameter verification. Figure 1 As shown, the method includes:
[0034] S110: Monitor the operating parameters of the motor to be diagnosed through sensors to obtain a real-time monitoring signal, where the real-time monitoring signal is any one of a current signal, a temperature signal, and a vibration signal.
[0035] Wherein, the current signal of the motor may be at least one of a load current, a phase current, and a line current;
[0036] The temperature signal may be at least one of the motor housing temperature, the bearing temperature, and the fan temperature;
[0037] The vibration amount signal may be at least one of a base bearing vibration amount, a drive end bearing vibration amount, a rotor vibration amount, and a stator vibration amount.
[0038] S120: Perform denoising on the real-time monitoring signal to obtain a target monitoring signal, and use an abnormal data detection method to determine whether the target monitoring signal is abnormal. If it is determined that an abnormality occurs, proceed to step S130.
[0039] There is usually a certain amount of noise interference in the real-time monitoring signals collected by sensors. The presence of noise may introduce interference in the analysis and processing of the monitoring signals, resulting in deviations or errors in the analysis results. Therefore, in order to facilitate the subsequent analysis and processing of the monitoring signals, it is necessary to denoise the real-time monitoring signals to obtain the target monitoring signals.
[0040] The denoising method may adopt conventional signal filtering methods, such as wavelet transform, Gaussian filtering, median filtering, etc.
[0041] In this step, the abnormal data detection method can use the Shewhart control chart in conjunction with the CUCUM control chart and / or the EWMA control chart.
[0042] Among them, the Shewhart control chart is a Shewhart control chart, which is based on the statistical 3σ principle and is effective in detecting larger drifts and outliers; the CUCUM (Cumulative Sum) control chart is a cumulative sum control chart, which is based on the sequential probability ratio test and is effective in detecting small and medium drifts; the EWMA (Exponentially Weighted Moving-Average) control chart is an exponentially weighted moving average control chart, a time-weighted control chart, which is effective in detecting small and medium drifts.
[0043] Since the Shewhart control chart is a good complement to the CUCUM control chart or the EWMA control chart, and can ensure that all process drifts are effectively detected, the Shewhart control chart is preferably used in conjunction with the CUCUM control chart and / or the EWMA control chart. In practical applications, when at least one of the control charts generates an abnormal alarm during abnormality detection, it is considered that an abnormality has occurred.
[0044] Exemplarily, when the target monitoring signal is current, and the target monitoring signal includes load current, phase current, and line current, data anomaly detection is performed on the load current signal, phase current signal, and line current signal, respectively. When any of the current signals is abnormal, it is determined that the monitoring signal is abnormal.
[0045] It can be seen that in the real-time early warning fault diagnosis process, since only abnormal signals will be transferred to the early warning fault type identification process, the amount of data that the model needs to process is greatly reduced, making the early warning fault diagnosis process more efficient.
[0046] S130: Acquire a one-dimensional time series signal with a preset time series length in the target monitoring signal, and convert the one-dimensional time series signal into a two-dimensional image using a Gram angular field.
[0047] Among them, the one-dimensional time series signal is represented by X= , n is the preset time series length, x i The instantaneous value of the target monitoring signal corresponding to each time point. Of course, the one-dimensional time series signal includes the abnormal time point data determined in the above step S120.
[0048] In this embodiment, the two-dimensional image may be a GADF image or a GASF image; the target monitoring signal may be any one of current, temperature, and vibration.
[0049] For example, when the target monitoring signal is current and the image is a GADF image, the target monitoring signal includes at least one of load current, phase current, and line current. Accordingly, the GADF image includes at least one of load current GADF image, phase current GADF image, and line current GADF image.
[0050] S140: performing feature extraction on the two-dimensional image, wherein the feature extraction includes texture feature extraction and geometric feature extraction.
[0051] Generally, texture features focus on information such as the surface structure and appearance of an image, while geometric features focus on information such as the shape, size, and position of an image. The key information of a two-dimensional image can be effectively represented by texture features and geometric features.
[0052] It can be understood that for each two-dimensional image, a texture feature vector and a geometric feature vector can be obtained.
[0053] S150: Determine statistical features of motor operating parameters based on the target monitoring signal, and concatenate the statistical features of the target monitoring signal with the texture feature vector and the geometric feature vector of the two-dimensional image to obtain a target feature vector.
[0054] Among them, when multiple target monitoring signals are included, the statistical characteristics of different target monitoring signals need to be determined separately. For example, when the target monitoring signal is current, and the target monitoring signal includes load current, phase current, and line current, the statistical characteristics of the load current signal, phase current signal, and line current signal are extracted separately. When calculating the statistical characteristics, it is necessary to first perform maximum and minimum normalization processing on different target monitoring signals separately. The statistical characteristics can be taken as mean, standard deviation, variance, signal period, etc.
[0055] Of course, different target monitoring signals correspond to different two-dimensional images, that is, multiple two-dimensional images, which correspond to multiple texture feature vectors and multiple geometric feature vectors. Therefore, when performing feature splicing, it is necessary to splice together the multiple texture feature vectors and multiple geometric feature vectors and multiple statistical features corresponding to the multiple target monitoring signals to obtain the target feature vector.
[0056] S160: Input the target feature vector into a pre-trained target model to determine the warning fault type of the motor.
[0057] Among them, the target model can adopt machine learning models such as support vector machine SVM, convolutional neural network CNN, multi-layer perceptron MLP, etc.
[0058] It can be seen that in this embodiment, data dimensionality is increased by converting the motor operating parameter signal into a two-dimensional image, and high-dimensional features often contain more data features. The deep data features can be extracted through the two-dimensional image corresponding to the motor operating parameter signal, and the statistical characteristics of the operating parameter signal are further combined to obtain a more accurate feature representation, which can more effectively determine the warning fault type of the motor and simplify the data collection process. At the same time, accurate warning fault diagnosis can be achieved by relying only on a single operating parameter indicator (such as current, temperature or vibration), thereby improving the efficiency and accuracy of motor warning fault identification.
[0059] In this embodiment, the step S130 specifically includes:
[0060] For one-dimensional time series signal X= Perform maximum and minimum normalization and map to [-1, 1] to obtain the normalized one-dimensional time series signal, which is expressed as:
[0061] =
[0062] The normalized one-dimensional time series signal is transformed into an angle and radius The polar coordinate form of is expressed as:
[0063]
[0064] Based on the data after polar coordinate transformation, GASF image or GADF image is extracted according to two different inner product definitions of Gram angle field (GAF).
[0065] The expression of the GADF image is:
[0066] = · ·
[0067] The expression of the GASF image is:
[0068] = ·
[0069] As a preferred example, step S140 in this embodiment includes:
[0070] S140-1: Obtain a preset Gabor function, determine a convolution kernel by using the preset Gabor function, slide the convolution kernel on the two-dimensional image like a window, obtain a convolution result at each position, sum the absolute values of the convolution results at each position, and obtain a value M.
[0071] S140 - 2 : Change the parameters of the Gabor function, repeat step S140 - 1, and obtain a plurality of numerical values M. The plurality of numerical values M form a vector, namely, a texture feature vector.
[0072] Among them, the preset Gabor function is:
[0073]
[0074] Where x and y are the position coordinates, are the wavelength, phase shift, standard deviation, and aspect ratio, respectively.
[0075] For example, a group The combination of parameters corresponds to a Gabor function, and for multiple groups , which corresponds to multiple Gabor functions. The number of different Gabor functions can be defined as 2K, which corresponds to 2K convolution kernels.
[0076] Among them, the convolution kernel RM determined based on the Gabor function can be set to 16*16 dimensions, and the calculation formula of the convolution kernel RM is as follows:
[0077]
[0078] in, The value range of [ 1 , 16 ] ,and They are all integers, and RM(i, j) represents the element in the i-th row and j-th column of the convolution kernel RM.
[0079] After obtaining the convolution kernel RM, the convolution kernel RM and the two-dimensional image matrix AM are convolved, and the RM convolution kernel is slid on the two-dimensional image matrix AM like a window, and a convolution result is obtained at each position.
[0080] For example, when the dimension of the two-dimensional image matrix AM is 128*128, the two-dimensional image is exactly divided into 8*8=64 windows. After the convolution calculation, 64 convolution results can be obtained. The absolute values of the 64 convolution results are summed to obtain the value M.
[0081] Then, change the parameters in the Gabor function The value of is taken, and the convolution kernel RM is redetermined based on the Gabor function. The redetermined convolution kernel RM and the two-dimensional image matrix AM are convolved, and the absolute values of the convolution results are summed to obtain another value M.
[0082] The above 2K convolution kernels RM are respectively used to perform convolution calculation on the two-dimensional image matrix AM to obtain 2K numerical values M, which form a vector, namely, a texture feature vector. Wherein, K is a positive integer, and preferably, K can be 32.
[0083] S140-3: Calculate the gradient of the two-dimensional image and perform binarization processing to obtain an edge feature matrix of the two-dimensional image.
[0084] S140-4: Obtain a preset straight line function and a circular function, determine the convolution kernel by the preset straight line function and the circular function respectively, slide the convolution kernel on the edge feature matrix like a window, obtain a convolution result at each position, sum the absolute values of the convolution results at each position, and obtain the value H.
[0085] S140-5: Change the parameters of the straight line function and the circular function respectively, repeat step S140-4, and obtain multiple numerical values H. The multiple numerical values H form a vector, namely, a geometric feature vector.
[0086] It can be understood that the process of extracting geometric feature vectors is essentially to detect the correlation between the two-dimensional image and the geometric function. Taking the straight line function y = x as an example, if there are more oblique line elements in the image, the obtained H value will be larger, otherwise it will be smaller.
[0087] Exemplarily, the gradient of the two-dimensional image matrix AM is first calculated. The gradient is represented by a matrix TM. TM is a 128*128-dimensional matrix. The calculation formula of TM is as follows:
[0088] TM i , j = abs ( AM i , j - AM i -1, j )+ abs ( AM i , j - AM i , j -1 if i , j ∈ [2,128] 0 if i =1 or j = 1
[0089] in, All are integers, abs means taking the absolute value function, Represents the element in the i-th row and j-th column of the matrix TM.
[0090] Furthermore, the matrix TM is binarized to obtain the edge feature matrix BM, as shown in the following formula:
[0091]
[0092] Wherein, T represents a binarization threshold, and the threshold can be set to 0.5 based on experience, or can be adaptively set based on the edge extraction effect.
[0093] It can be understood that what is stored in the two-dimensional image matrix AM is the image, while what is stored in the edge feature matrix BM is the lines.
[0094] Now use BM to calculate the geometric eigenvectors.
[0095] Define a geometric function. For example, there are 2K geometric functions, K of which are linear functions and the other K are circular functions. The formula of the linear function is:
[0096]
[0097] in, The value range is from -90° to 90°, and the values are taken at equal intervals, with a total of K values, that is, K straight line functions are defined.
[0098] The formula for the circular function is:
[0099]
[0100] The value range of r is [0, ], take values at equal intervals, take K values in total, that is, define K circular functions. Wherein, K is a positive integer, and preferably, K can be 32.
[0101] The convolution kernel EM determined based on the above geometric function can be represented by a 16*16 dimensional matrix. For example, when When , the straight line function is: , at this time the value of the convolution kernel EM is Figure 2 As shown, the black square indicates that the EM value at that position is 1, and the white square indicates that the EM value at that position is 0. In short, the elements in the matrix EM are composed of 0 and 1, and the shape defined by the position of the non-zero element value distribution in the EM matrix is determined by the geometric function.
[0102] Use the convolution kernel EM to perform window sliding convolution on the edge feature matrix BM, similar to the above texture feature extraction method, to obtain 2K convolution results, sum the absolute values of the 2K convolution results to obtain the value G.
[0103] The convolution kernel EM determined by the above 2K geometric functions (ie, K straight line functions and K circular functions) is respectively used to perform convolution calculation on the edge feature matrix BM to obtain 2K numerical values G. The 2K numerical values G form a vector, namely, the geometric feature vector.
[0104] In this embodiment, the texture feature vector and geometric feature vector of the two-dimensional image can be extracted through the above step S140, thereby reducing the input feature dimension of the target model and reducing the computational complexity of the model; and, compared with traditional image feature extraction methods such as CNN convolutional neural networks, the above feature extraction method simplifies the feature extraction process, while ensuring the accuracy of feature extraction, effectively reducing the computational amount of image feature extraction, thereby further improving the efficiency of early warning fault diagnosis.
[0105] Furthermore, before step S110, a training step of the target model is also included:
[0106] S100: Acquire multiple groups of sample signals representing motor operating parameters and types corresponding to the sample signals, determine a target feature vector corresponding to the sample signal, and use the target feature vector as input and the type as output to input into a target model for training.
[0107] Among them, the multiple groups of sample signals are target monitoring signals after denoising, and the data processing method for determining the target feature vector based on the multiple groups of sample signals is the same as the above steps S130-S150, which will not be repeated here.
[0108] The following takes the support vector machine model SVM as an example. This process can be completed using Python's machine learning library:
[0109] First, the prepared data is divided into two parts: a training set and a test set. The data includes target feature vectors corresponding to multiple groups of sample signals and types corresponding to the sample signals, of which the training set accounts for 70% and the test set accounts for 30%. Assuming that the training set is named x_train and y_train, and the test set is x_test and y_test, then each row in x_train and x_test represents a target feature vector, and each row in y_train and y_test represents a warning fault category corresponding to each target feature vector, with a dimension of 1.
[0110] Secondly, create and train the model: import the sklearn related library, initialize an SVM model, the model name is model, and the kernel function selects the linear function:
[0111] model = svm.SVC(kernel='linear');
[0112] Train the model:
[0113] model.fit(x_train, y_train);
[0114] Use the test set to evaluate the model accuracy, and the accuracy is named score:
[0115] score = accuracy_score(y_test, y_pred)).
[0116] The trained target model is obtained through multiple rounds of iterative updates. In practical applications, the target feature vector is input into the target model to output the corresponding early warning fault diagnosis type.
[0117] In another embodiment of the present application, a preventive maintenance diagnostic device for motor equipment based on multi-parameter verification is provided, which includes a data acquisition module, an abnormality detection module, an image generation module, an image feature extraction module, a target feature generation module and a warning fault identification module.
[0118] A data acquisition module, used to monitor the operating parameters of the motor to be diagnosed through sensors to obtain a real-time monitoring signal, wherein the real-time monitoring signal is any one of a current signal, a temperature signal, and a vibration signal;
[0119] The anomaly detection module is used to perform denoising on the real-time monitoring signal to obtain the target monitoring signal, and use the abnormal data detection method to determine whether the target monitoring signal is abnormal. If it is judged to be abnormal, it will be transferred to the image generation module for data processing;
[0120] An image generation module is used to obtain a one-dimensional time series signal with a preset time series length in the target monitoring signal, and convert the one-dimensional time series signal into a two-dimensional image using a Gram angle field;
[0121] An image feature extraction module, used for performing feature extraction on the two-dimensional image, wherein the feature extraction includes texture feature extraction and geometric feature extraction;
[0122] A target feature generation module is used to determine the statistical features of the motor operating parameters based on the target monitoring signal, and to splice the statistical features of the target monitoring signal with the texture feature vector and the geometric feature vector of the two-dimensional image to obtain a target feature vector;
[0123] The early warning fault identification module is used to input the target feature vector into the pre-trained target model to determine the early warning fault type of the motor.
[0124] Furthermore, the device also includes a model training module;
[0125] The model training module is used to obtain multiple groups of sample signals representing motor operating parameters and the types corresponding to the sample signals, determine the target feature vector corresponding to the sample signal, and use the target feature vector as input and the type as output to input into the target model for training.
[0126] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.
[0127] It should be further noted that, although the various steps in the flowchart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowchart may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
[0128] It should be understood that the above-mentioned device embodiments are only illustrative, and the device of the present application can also be implemented in other ways. For example, the division of units / modules in the above-mentioned embodiments is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units, modules or components can be combined, or can be integrated into another system, or some features can be ignored or not executed.
[0129] In addition, unless otherwise specified, each functional unit / module in each embodiment of the present application may be integrated into one unit / module, each unit / module may exist physically separately, or two or more units / modules may be integrated together. The above-mentioned integrated unit / module may be implemented in the form of hardware or in the form of a software program module.
[0130] In the above embodiments, the description of each embodiment has its own emphasis. For the part not described in detail in a certain embodiment, please refer to the relevant description of other embodiments. The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0131] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.
[0132] It should be understood that the present application is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof.
Claims
1. A preventive maintenance diagnosis method for motor equipment based on multi-parameter verification, characterized in that: The method comprises: Step S110: monitoring the operating parameters of the motor to be diagnosed through sensors to obtain real-time monitoring signals; Step S120: performing denoising processing on the real-time monitoring signal to obtain a target monitoring signal, and using an abnormal data detection method to determine whether the target monitoring signal is abnormal. If it is determined that an abnormality occurs, proceeding to step S130; Step S130: obtaining a one-dimensional time series signal with a preset time series length in the target monitoring signal, and converting the one-dimensional time series signal into a two-dimensional image using a Gram angle field; Step S140: performing feature extraction on the two-dimensional image, wherein the feature extraction includes texture feature extraction and geometric feature extraction; Step S150: determining the statistical characteristics of the motor operating parameters based on the target monitoring signal, and splicing the statistical characteristics of the target monitoring signal with the texture feature vector and the geometric feature vector of the two-dimensional image to obtain a target feature vector; Step S160: inputting the target feature vector into a pre-trained target model to determine the early warning fault type of the motor; Wherein, the real-time monitoring signal is any one of a current signal, a temperature signal, and a vibration signal; The step of performing feature extraction on the two-dimensional image, wherein the feature extraction includes texture feature extraction and geometric feature extraction, comprises: Step S140-1: obtaining a preset Gabor function, determining a convolution kernel RM by using the preset Gabor function, sliding the convolution kernel RM on the two-dimensional image like a window, obtaining a convolution result at each position, summing the absolute values of the convolution results at each position, and obtaining a value M; Step S140-2: changing the parameters of the Gabor function, repeating step S140-1, and obtaining a plurality of numerical values M, wherein the plurality of numerical values M form a vector, namely, a texture feature vector; The feature extraction of the two-dimensional image includes texture feature extraction and geometric feature extraction, and further includes: The preset Gabor function is: Where x and y are the position coordinates, are wavelength, phase shift, standard deviation, and aspect ratio, respectively; The feature extraction of the two-dimensional image includes texture feature extraction and geometric feature extraction, and further includes: Step S140-3: Calculate the gradient of the two-dimensional image and perform binarization processing to obtain an edge feature matrix of the two-dimensional image; Step S140-4: obtaining a preset straight line function and a preset circular function, determining a convolution kernel EM by using the preset straight line function and the preset circular function respectively, sliding the convolution kernel EM on the edge feature matrix like a window, obtaining a convolution result at each position, summing the absolute values of the convolution results at each position, and obtaining a value H; Step S140-5: respectively change the parameters of the straight line function and the circular function, repeat step S140-4, and obtain multiple numerical values H. The multiple numerical values H form a vector, namely, a geometric feature vector.
2. The preventive maintenance diagnosis method for motor equipment based on multi-parameter verification according to claim 1 is characterized in that: The abnormal data detection method uses a Shewhart control chart in conjunction with a CUCUM control chart; Alternatively, the abnormal data detection method uses a Shewhart control chart in combination with an EWMA control chart.
3. The preventive maintenance diagnosis method for motor equipment based on multi-parameter verification according to claim 1 is characterized in that: The step of obtaining a one-dimensional time series signal of a preset time series length in the target monitoring signal and converting the one-dimensional time series signal into a two-dimensional image by using a Gram angle field includes: The two-dimensional image is a GADF image or a GASF image, and the target monitoring signal is any one of current, temperature, and vibration.
4. The preventive maintenance diagnosis method for motor equipment based on multi-parameter verification according to claim 1 is characterized in that: The convolution kernel EM is determined by a preset straight line function and a preset circular function respectively, including: The elements in the convolution kernel EM are composed of 0 and 1, and the shape defined by the positions where the non-zero element values are distributed in the convolution kernel EM is determined by a preset straight line function or a preset circular function.
5. The preventive maintenance diagnosis method for motor equipment based on multi-parameter verification according to claim 1 is characterized in that: The statistical feature includes at least one of a mean, a standard deviation, a variance, and a signal period.
6. The preventive maintenance diagnosis method for motor equipment based on multi-parameter verification according to claim 1 is characterized in that: Before step S110, the method further includes: Step S100: Acquire multiple groups of sample signals representing motor operating parameters and types corresponding to the sample signals, determine the target feature vector corresponding to the sample signal, and use the target feature vector as input and the type as output to input into the target model for training.
7. A preventive maintenance diagnostic device for motor equipment based on multi-parameter verification, used to execute any of the methods of claims 1-6, characterized in that: The device includes a data acquisition module, an anomaly detection module, an image generation module, an image feature extraction module, a target feature generation module and a warning fault identification module; A data acquisition module, used to monitor the operating parameters of the motor to be diagnosed through sensors to obtain a real-time monitoring signal, wherein the real-time monitoring signal is any one of a current signal, a temperature signal, and a vibration signal; An abnormality detection module is used to perform denoising on the real-time monitoring signal to obtain a target monitoring signal, and use an abnormal data detection method to determine whether the target monitoring signal is abnormal. If it is determined to be abnormal, the target monitoring signal is transferred to the image generation module for data processing; An image generation module is used to obtain a one-dimensional time series signal with a preset time series length in the target monitoring signal, and convert the one-dimensional time series signal into a two-dimensional image using a Gram angle field; An image feature extraction module, used for performing feature extraction on the two-dimensional image, wherein the feature extraction includes texture feature extraction and geometric feature extraction; A target feature generation module, used for determining the statistical features of the motor operating parameters based on the target monitoring signal, and splicing the statistical features of the target monitoring signal with the texture feature vector and the geometric feature vector of the two-dimensional image to obtain a target feature vector; The early warning fault identification module is used to input the target feature vector into a pre-trained target model to determine the early warning fault type of the motor.
Citation Information
Cited By
A pattern recognition method based on signal image conversion and large model inference
CN123016077A
