Three-phase asynchronous motor fault diagnosis method and device

Abstract

The invention discloses a three-phase asynchronous motor fault diagnosis method. The method comprises the following steps: 1, collecting a vibration signal and a current signal of a three-phase asynchronous motor; 2, the noisy signal is processed, noise and redundant signals are removed, then a fault characteristic signal is obtained, and the fault characteristic signal comprises a vibration characteristic signal and a current characteristic signal; 3, converting the obtained vibration characteristic signal and the current characteristic signal into a GADF data image; 4, according to a certain weight ratio, performing image fusion on the GADF data images obtained after conversion of the vibration characteristic signals and the current characteristic signals to obtain GADF data containing fault characteristics; and 5, constructing a fault diagnosis neural network model with an attention mechanism, and diagnosing the fault of the three-phase asynchronous motor by using the vibration signal and the current signal. The invention further provides a three-phase asynchronous motor fault diagnosis device. Compared with the prior art, the accuracy of motor fault diagnosis can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of motors, and in particular to a method and device for diagnosing faults of a three-phase asynchronous motor. Background Art

[0002] Three-phase asynchronous motors are widely used in the field of industrial machinery. However, many application scenarios (such as climbing frames) mainly use post-disassembly and maintenance to repair motor failures. There is a lack of monitoring of the motor's operation process. Generally, during use, replacement measures are only taken when the motor is completely damaged. During turnover maintenance, the motor is mainly subjected to simple rotation tests. Therefore, it is easy for defective motors to continue to serve.

[0003] Common motor fault monitoring mainly relies on vibration signal analysis, but since the working environment of the motor is usually harsh during operation, the signals collected by the sensor often contain a large number of redundant signals and interference signals, which affects the accuracy of state monitoring. The present invention uses signal processing and image fusion technology to strip redundant information from the signal, image the motor vibration signal and current signal and fuse them, effectively extracting the motor classification fault feature information, thereby improving the accuracy of motor fault diagnosis.

[0004] Common motor fault monitoring mainly relies on vibration signal analysis, but because the working environment of the motor is often harsh during operation, the signals collected by the sensor often contain a large number of redundant signals and interference signals, which affects the accuracy of state monitoring. In addition, after the motor has been running for a period of time, it may have different degrees of explicit or hidden faults in multiple parts. After the relevant signals are collected, different fault signals cross-influence each other, making it difficult to determine the specific location of the fault. The traditional fault diagnosis method cannot effectively extract the characteristic signals of multiple faults, which is not conducive to the analysis of motor faults.

[0005] In view of this, a three-phase asynchronous motor fault diagnosis method and device are proposed. Summary of the invention

[0006] The purpose of the present invention is to provide a three-phase asynchronous motor fault diagnosis method and device to solve the problems pointed out in the background technology.

[0007] The above technical objectives of the present invention are achieved through the following technical solutions:

[0008] A three-phase asynchronous motor fault diagnosis method comprises the following steps:

[0009] Step 1: Collect vibration signal and current signal of three-phase asynchronous motor;

[0010] Step 2: Process the noisy signal, remove the noise and redundant signals, and obtain the fault characteristic signal, which includes the vibration characteristic signal and the current characteristic signal;

[0011] Step 3: Convert the obtained vibration characteristic signal and current characteristic signal into a GADF data image;

[0012] Step 4: Perform image fusion on the GADF images converted from the vibration characteristic signal and the current characteristic signal according to a certain weight ratio to obtain a GADF data image containing fault characteristics;

[0013] Step 5: Construct a fault diagnosis neural network model with an attention mechanism, and use the vibration signal and current signal to diagnose the three-phase asynchronous motor fault.

[0014] In a preferred embodiment, the current signal includes a stator current signal at different speeds and loads, and the vibration signal includes signal data of XYZ axes at different speeds and loads.

[0015] In a preferred embodiment, in step 2, DEGWO-VMD is used to process the noisy signal.

[0016] In a preferred embodiment, the process of processing a noisy signal using DEGWO-VMD includes: taking the minimum residual index as the fitness function, using DEGWO-VMD to decompose the vibration signal and current signal containing strong noise to obtain several IMF components, and using the kurtosis criterion to perform preliminary signal reconstruction on the IMF components with kurtosis values ​​greater than 3 to obtain the preliminary reconstructed vibration signals and current signals.

[0017] In a preferred embodiment, the process of processing a noisy signal using DEGWO-VMD also includes: constructing a Hankel matrix to perform singular value decomposition and reconstruction on the initially reconstructed signal, and obtaining a vibration signal and a current signal after noise reduction processing after removing noise and a large amount of redundant signals.

[0018] In a preferred embodiment, the process of processing the noisy signal using DEGWO-VMD also includes: for the vibration signal and the current signal after denoising, using the minimum envelope kurtosis factor as the fitness function, using DEGWO-VMD to perform a second decomposition on the denoised signal, and reconstructing the IMF component of the main fault characteristic signal through spectral analysis to obtain the corresponding fault characteristic signal.

[0019] In a preferred embodiment, the process of processing the noisy signal using DEGWO-VMD further includes: dividing the processed data samples into a training set and a test set, and performing label processing according to different types of faults.

[0020] In a preferred embodiment, in step three, the obtained vibration characteristic signal and current characteristic signal are converted into a GADF data image using a Gram angle difference field.

[0021] In a preferred embodiment, in step five, a neural network model based on Resnet with an attention mechanism is constructed to analyze and diagnose faults in image data.

[0022] A three-phase asynchronous motor fault diagnosis device, comprising:

[0023] Signal acquisition unit: used to collect vibration signals and current signals of three-phase asynchronous motors;

[0024] Preprocessing unit: used to process the noisy signal, remove the noise and redundant signals to obtain the fault characteristic signal, the fault characteristic signal includes the vibration characteristic signal and the current characteristic signal;

[0025] Image conversion unit: used for converting the obtained vibration characteristic signal and current characteristic signal into a GADF data image;

[0026] Image fusion unit: used to fuse the GADF image converted from the vibration characteristic signal and the current characteristic signal according to a certain weight ratio to obtain a GADF data image containing fault characteristics;

[0027] Fault diagnosis unit: used to construct a fault diagnosis neural network model with an attention mechanism, and use the vibration signal and current signal to diagnose the fault of the three-phase asynchronous motor.

[0028] Compared with the prior art, the present invention uses motor vibration signals and current signals to monitor the state of the motor. Through signal processing and image fusion technology, the redundant information in the signal is stripped off, the motor vibration signal and the current signal are processed and fused in an image form, and the characteristic information of various motor faults is effectively extracted, thereby improving the accuracy of motor fault diagnosis. It has the following beneficial effects:

[0029] 1. Multi-source signal acquisition provides a comprehensive information foundation

[0030] By collecting the vibration signal and current signal of the motor, where the vibration signal covers the three-phase asynchronous motor at different speeds and different torque loads, and includes signal data of the XYZ axes, and the current signal includes the stator current signal at different speeds and loads, etc., this multi-source signal collection method makes the grasp of the motor's operating status more comprehensive and no longer limited to a single signal dimension. It provides rich and diverse basic data information for the subsequent accurate diagnosis of various possible faults of the motor, such as bearing faults, stator faults, rotor faults, etc., avoiding the problem of missing key fault characteristics due to a single signal source.

[0031] 2. Efficient noise reduction and feature extraction improve data quality

[0032] The use of improved grey wolf optimization algorithm to optimize variational mode decomposition (DEGWO-VMD) plays a key role in the entire data processing process. First, the minimum residual index is used as the fitness function to perform the initial decomposition of the vibration and current signals containing strong noise, and then the signal is preliminarily reconstructed according to the kurtosis criterion, and then the Hankel matrix is ​​constructed to further remove the noise and redundant signals. After that, the second decomposition is performed with the minimum envelope kurtosis factor as the fitness function, and the main fault feature signal is obtained through spectrum analysis and reconstruction. After such a secondary decomposition and reconstruction process, a large amount of noise and redundant parts contained in the original signal can be effectively removed, and the fault feature signal can be accurately extracted, which greatly improves the quality of data used for subsequent fault diagnosis, so that the characteristics that truly reflect the fault situation can be highlighted, and the influence of interference factors on the accuracy of diagnosis is reduced.

[0033] 3. Data conversion and fusion to enhance feature presentation

[0034] The Gram angle difference field is used to convert the vibration characteristic signal and the current characteristic signal into a 256x256 GADF two-dimensional data image. This conversion method allows the relatively abstract signal characteristics to be presented in an intuitive image form, which is convenient for subsequent analysis and processing. In addition, the alpha image blending technology is used to fuse the GADF images converted from the vibration characteristic signal and the current characteristic signal according to a certain weight ratio, further integrating the fault characteristics contained in the two signals, making the fault characteristics in the final GADF data clearer and more concentrated, and can be more effectively captured and used by the neural network model in the subsequent fault diagnosis link, providing more recognizable feature data support for accurate diagnosis of motor faults.

[0035] 4. Neural network model ensures diagnostic accuracy and stability

[0036] A neural network model with an attention mechanism based on Resnet is constructed to analyze and diagnose the processed, converted and fused image data. The introduction of the attention mechanism allows the model to focus on the key fault feature area, avoid the interference of irrelevant information, and accurately mine the fault-related content in the data. The Resnet architecture itself has good deep feature extraction capabilities and complex data processing capabilities. The combination of the two makes the model more accurate when diagnosing three-phase asynchronous motor faults, and the convergence of the accuracy curve is better and smoother, which means that with the increase in the number of diagnostic samples and the advancement of the training process, its diagnostic results can quickly and stably approach the accurate state, reducing the fluctuation of the diagnostic results, and can reliably provide accurate fault judgment basis for the operation and maintenance of three-phase asynchronous motors, which helps to timely discover and solve the hidden faults in the operation of the motor, ensure the normal and stable operation of the motor, and reduce the adverse effects such as production delays and equipment damage caused by motor failures.

[0037] In summary, this fault diagnosis method achieves efficient and accurate diagnosis of three-phase asynchronous motor faults through the coordinated cooperation of multiple links, from signal acquisition, data processing to the final fault diagnosis model construction, and has significant technical advantages and good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 The present invention relates to a flow chart of a three-phase asynchronous motor fault diagnosis method.

[0039] Figure 2 The present invention relates to a flowchart of step 2 of a three-phase asynchronous motor fault diagnosis method.

[0040] Figure 3 The present invention relates to a noise reduction effect schematic diagram of a three-phase asynchronous motor fault diagnosis method.

[0041] Figure 4 The invention relates to a three-phase asynchronous motor fault diagnosis method. DETAILED DESCRIPTION

[0042] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0043] This specific embodiment is merely an explanation of the present invention and is not a limitation of the present invention. After reading this specification, those skilled in the art may make non-creative modifications to the present embodiment as needed. However, as long as they are within the scope of the claims of the present invention, they are protected by the patent law.

[0044] Embodiment 1:

[0045] like Figures 1 to 4 As shown:

[0046] A three-phase asynchronous motor fault diagnosis method comprises the following steps:

[0047] Step 1: Collect vibration signal and current signal of three-phase asynchronous motor;

[0048] Step 2: Process the noisy signal, remove the noise and redundant signals, and obtain the fault characteristic signal, which includes the vibration characteristic signal and the current characteristic signal;

[0049] Step 3: Convert the obtained vibration characteristic signal and current characteristic signal into a GADF data image;

[0050] Step 4: Perform image fusion on the GADF images converted from the vibration characteristic signal and the current characteristic signal according to a certain weight ratio to obtain a GADF data image containing fault characteristics;

[0051] Step 5: Construct a fault diagnosis neural network model with an attention mechanism, and use the vibration signal and current signal to diagnose the three-phase asynchronous motor fault.

[0052] In this embodiment, the bearing fault of the three-phase asynchronous motor includes a bearing fault, a stator fault, a rotor fault, etc.

[0053] Furthermore, the current signal includes a stator current signal at different speeds and loads, and the vibration signal includes signal data of XYZ axes at different speeds and loads.

[0054] Furthermore, in step 2, DEGWO-VMD is used to process the noisy signal.

[0055] Furthermore, the process of using DEGWO-VMD to process noisy signals includes: taking the minimum residual index as the fitness function, using DEGWO-VMD to decompose the vibration signal and current signal containing strong noise to obtain several IMF components, and using the kurtosis criterion to perform preliminary signal reconstruction on the IMF components with kurtosis values ​​greater than 3 to obtain the preliminary reconstructed vibration signals and current signals.

[0056] Furthermore, the process of using DEGWO-VMD to process the noisy signal also includes: constructing a Hankel matrix to perform singular value decomposition and reconstruction on the preliminary reconstructed signal, removing noise and a large number of redundant signals to obtain the vibration signal and current signal after noise reduction processing. The noise reduction effect is as follows: Figure 3 .

[0057] Furthermore, the process of using DEGWO-VMD to process the noisy signal also includes: for the vibration signal and current signal after denoising, using the minimum envelope kurtosis factor as the fitness function, using DEGWO-VMD to perform a second decomposition on the denoised signal, and reconstructing the IMF component of the main fault characteristic signal through spectrum analysis to obtain the corresponding fault characteristic signal.

[0058] Furthermore, the process of processing the noisy signal using DEGWO-VMD also includes: dividing the processed data samples into a training set and a test set, and labeling them according to different types of faults. In this embodiment, the rolling bearing fault can be divided into a normal working condition, with a label of 0; an inner ring fault, with a label of 1; an outer ring fault, with a label of 2; and a ball fault, with a label of 3.

[0059] Furthermore, in the step three, the fault signal data is normalized, and the obtained vibration characteristic signal and current characteristic signal are converted into a 256x256 GADF data image using the Gram angle difference field.

[0060] Furthermore, in step 5, a neural network model with an attention mechanism based on Resnet is constructed to analyze and diagnose faults of image data, thereby improving the accuracy of fault diagnosis of three-phase asynchronous motors. In this embodiment, the accuracy of rolling bearing fault analysis is as high as 99.21%. Figure 4 .

[0061] Furthermore, the alpha image blending technology is used to fuse the GADF images converted from the vibration characteristic signal and the current characteristic signal according to a certain weight ratio to obtain a GADF data image containing fault characteristics. In this embodiment, the rolling bearing fault is fused according to a weight ratio of 0.6, that is, 60% of the vibration data image and 40% of the current data image to obtain a GADF data image containing fault characteristics.

[0062] In this embodiment, the basic process of the DEGWO-VMD algorithm is as follows:

[0063] 1) Parameter initialization, including:

[0064] Set the size of the gray wolf population to 40;

[0065] It is determined that the number of VMD parameters that need to be optimized is 3;

[0066] Set the maximum number of iterations of the algorithm to 30;

[0067] Set the scaling factor of the differential evolution algorithm to 0.5;

[0068] Set the Bernoulli mapping parameter to 0.5;

[0069] VMD parameter range: Determine k 0 =6, sampling frequency f 0 =10kHz, and the global optimization range of k is selected as [3, 12].

[0070] The global optimization range of the penalty item weight value is [5000, 100000]

[0071] 2) Initialize the population

[0072] In order to solve the problem that the GWO algorithm falls into the local optimal solution, the Bernoulli chaotic map is introduced during the algorithm initialization to improve the population initialization method. The specific steps are as follows:

[0073] Randomly generate N initial values ​​x i ∈(0,1).

[0074] The chaotic sequence is iteratively generated using the provided Bernoulli mapping formula to obtain N chaotic values.

[0075] Mapping these chaotic values ​​to the search space of VMD parameters, we obtain N D-dimensional vectors as the initial gray wolf population X = [X 1 ,X 2 ,...,X n ]. Note that the chaotic sequence needs to be scaled and offset according to the range of VMD parameters.

[0076] 3) Iterative optimization (t = 1 to T)

[0077] Fitness evaluation, for each gray wolf individual X i , use its corresponding VMD parameters to perform VMD decomposition, and calculate its fitness value fitness according to the selected evaluation index i The smaller the fitness value, the better the decomposition result.

[0078] Update α, β, δ, find the three individuals with the smallest fitness values ​​in the current population, and assign them to α, β, δ respectively.

[0079] The differential evolution algorithm (DE) is integrated into the standard gray wolf algorithm. The mutation and selection operators of the differential algorithm can effectively maintain the diversity of the population and prevent the algorithm from converging prematurely. The mutation operation generates new individuals, and the selection operation retains better individuals, thereby preventing the population from falling into the local optimum prematurely. The crossover strategy selects binomial crossover, and the mutation strategy is:

[0080] V(i,:)=bestX+F*(X(r(1),:)-X(r(2),:));

[0081] Where V(i,:) represents the mutation vector of the i-th individual, bestX is the position vector of the best individual in the current population. F is the scaling factor, set to 0.5, which controls the strength of the mutation. X(r(1),:) and X(r(2),:) are the position vectors of two different individuals randomly selected from the population.

[0082] The gray wolf position is updated according to the standard GWO algorithm, using α, β, δ to guide each individual X i Move towards the optimal solution.

[0083] The individual U generated by differential evolution i With the original individual X i Compare and select individuals with smaller fitness values ​​as the next generation of individuals.

[0084] Boundary processing is performed to ensure that the location of the updated gray wolf individuals is still within the valid range of the VMD parameters.

[0085] 4) Termination conditions:

[0086] When the termination condition of the maximum number of iterations T is reached, the algorithm terminates.

[0087] Embodiment 2:

[0088] A three-phase asynchronous motor fault diagnosis device, comprising:

[0089] Signal acquisition unit: used to collect vibration signals and current signals of three-phase asynchronous motors;

[0090] Preprocessing unit: used to process the noisy signal, remove the noise and redundant signals to obtain the fault characteristic signal, the fault characteristic signal includes the vibration characteristic signal and the current characteristic signal;

[0091] Image conversion unit: used for converting the obtained vibration characteristic signal and current characteristic signal into a GADF data image;

[0092] Image fusion unit: used to fuse the GADF data image converted from the vibration characteristic signal and the current characteristic signal according to a certain weight ratio to obtain a GADF data image containing fault characteristics;

[0093] Fault diagnosis unit: used to construct a fault diagnosis neural network model with an attention mechanism, and use the vibration signal and current signal to diagnose the fault of the three-phase asynchronous motor.

[0094] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "include..." or "comprise..." do not exclude the existence of other elements in the process, method, article or terminal device including the elements. In addition, in this article, "greater than", "less than", "exceed" and the like are understood to exclude the number itself; "above", "below", "within" and the like are understood to include the number itself.

[0095] The above description of the embodiments is to facilitate the understanding and use of the present invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to the embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments. Improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the protection scope of the present invention.

Claims

1. A three-phase asynchronous motor fault diagnosis method, characterized in that: The steps include: Step 1: Collect vibration signal and current signal of three-phase asynchronous motor; Step 2: Process the noisy signal, remove the noise and redundant signals, and obtain the fault characteristic signal, which includes the vibration characteristic signal and the current characteristic signal; Step 3: Convert the obtained vibration characteristic signal and current characteristic signal into a GADF data image; Step 4: Perform image fusion on the GADF images converted from the vibration characteristic signal and the current characteristic signal according to a certain weight ratio to obtain a GADF data image containing fault characteristics; Step 5: Construct a fault diagnosis neural network model with an attention mechanism, and use the vibration signal and current signal to diagnose the three-phase asynchronous motor fault.

2. A three-phase asynchronous motor fault diagnosis method according to claim 1, characterized in that: The current signal includes a stator current signal at different speeds and loads, and the vibration signal includes signal data of XYZ axes at different speeds and loads.

3. A three-phase asynchronous motor fault diagnosis method according to claim 1, characterized in that: In step 2, DEGWO-VMD is used to process the noisy signal.

4. A three-phase asynchronous motor fault diagnosis method according to claim 3, characterized in that: The process of using DEGWO-VMD to process noisy signals includes: taking the minimum residual index as the fitness function, using DEGWO-VMD to decompose the vibration signal and current signal containing strong noise to obtain several IMF components, and using the kurtosis criterion to perform preliminary signal reconstruction on the IMF components with kurtosis values ​​greater than 3 to obtain the preliminary reconstructed vibration signals and current signals.

5. A three-phase asynchronous motor fault diagnosis method according to claim 4, characterized in that: The process of processing noisy signals using DEGWO-VMD also includes: constructing a Hankel matrix to perform singular value decomposition and reconstruction on the initially reconstructed signal, and obtaining a vibration signal and a current signal after noise reduction after removing noise and a large number of redundant signals.

6. A three-phase asynchronous motor fault diagnosis method according to claim 5, characterized in that: The process of using DEGWO-VMD to process noisy signals also includes: for the vibration signal and current signal after denoising, using the minimum envelope kurtosis factor as the fitness function, using DEGWO-VMD to perform a second decomposition of the denoised signal, and reconstructing the IMF component of the main fault characteristic signal through spectrum analysis to obtain the corresponding fault characteristic signal.

7. A three-phase asynchronous motor fault diagnosis method according to claim 6, characterized in that: The process of processing noisy signals using DEGWO-VMD also includes: dividing the processed data samples into a training set and a test set, and performing label processing according to different types of faults.

8. A three-phase asynchronous motor fault diagnosis method according to claim 1, characterized in that: In the step three, the obtained vibration characteristic signal and current characteristic signal are converted into a GADF data image using the Gram angle difference field.

9. A three-phase asynchronous motor fault diagnosis method according to claim 1, characterized in that: In the step five, a neural network model with an attention mechanism based on Resnet is constructed to analyze and diagnose faults in image data.

10. A three-phase asynchronous motor fault diagnosis device, characterized in that: include: Signal acquisition unit: used to collect vibration signals and current signals of three-phase asynchronous motors; Preprocessing unit: used to process the noisy signal, remove the noise and redundant signals to obtain the fault characteristic signal, the fault characteristic signal includes the vibration characteristic signal and the current characteristic signal; Image conversion unit: used for converting the obtained vibration characteristic signal and current characteristic signal into a GADF data image; Image fusion unit: used to fuse the GADF image converted from the vibration characteristic signal and the current characteristic signal according to a certain weight ratio to obtain a GADF data image containing fault characteristics; Fault diagnosis unit: used to construct a fault diagnosis neural network model with an attention mechanism, and use the vibration signal and current signal to diagnose the fault of the three-phase asynchronous motor.

Images