CN116128882A - Motor bearing fault diagnosis method, equipment and medium based on unbalanced data set - Google Patents
Motor bearing fault diagnosis method, equipment and medium based on unbalanced data set Download PDFInfo
- Publication number
- CN116128882A CN116128882A CN202310416034.8A CN202310416034A CN116128882A CN 116128882 A CN116128882 A CN 116128882A CN 202310416034 A CN202310416034 A CN 202310416034A CN 116128882 A CN116128882 A CN 116128882A
- Authority
- CN
- China
- Prior art keywords
- fault diagnosis
- data set
- motor bearing
- gradient
- positive
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Abstract
The invention relates to the field of fault diagnosis and discloses a motor bearing fault diagnosis method, equipment and medium based on an unbalanced data set. The method comprises the following steps: obtaining vibration data of a motor bearing to form an unbalanced vibration data set; converting vibration data in the data set into a two-dimensional image by utilizing continuous wavelet transformation to form an unbalanced image data set; randomly dividing the image data set into a training sample data set and a test sample data set; and inputting the sample data in the training sample data set into a fault diagnosis model to be trained based on the loss balance of dynamic gradient accumulation so as to train the fault diagnosis model to be trained, inputting the sample data in the test sample data set into the trained fault diagnosis model so as to verify the trained fault diagnosis model, and performing motor bearing fault diagnosis by using the verified fault diagnosis model. And fault diagnosis of the motor bearing is realized.
Description
Technical Field
The invention relates to the field of equipment fault diagnosis, in particular to a motor bearing fault diagnosis method, equipment and medium based on an unbalanced data set.
Background
Since the overall structure of the machine tends to be complicated, and each component has a certain probability of failing, the probability of the machine and the system failing is greatly increased. For example, due to long-term operation of the equipment, the bearings in the machine may be easily damaged during operation due to overload, fatigue, corrosion and the like, and thus, the equipment may be abnormal or stopped, and even casualties may be caused. Therefore, the method has important significance in fault diagnosis of core components or equipment in the mechanical system.
Three main methods of equipment fault diagnosis are available, one is a method based on an analysis model, the method passes through enough sensor information and needs to accurately establish a mathematical model, and typical methods of the method comprise a state estimation method, a parameter estimation method and the like; the other is a method based on experience knowledge, which is suitable for a system to build an accurate mathematical model because a sensor cannot obtain enough information, and typical methods of the method include a signed directed graph, an expert system and the like; the last one is based on industrial big data driving method, and the data is more easily collected and analyzed due to the coming of the times of the Internet of things and the information physical system, so that the method is receiving more attention and importance.
Industrial big data refers to various data generated by various links in the whole life cycle of the whole industrial product and the general name of related technologies and applications in a typical intelligent manufacturing mode in the industrial field. Industrial big data has characteristics of mass property, multi-source isomerism, unbalance and the like. For example, in the case of fault diagnosis of industrial big data equipment, since equipment is in a normal state most of the time in the industrial production process, the frequency of occurrence of faults is very small, so that most of data obtained by monitoring equipment sensors are normal state information, and abnormal data is very small, thus unbalanced data exists. The number of samples of a certain class (minority class) in unbalanced data is far smaller than that of samples of other classes, the existing unbalanced data brings greater difficulty to preprocessing, classifying, feature extraction and data mining of the data, and the existing feature selection and sampling algorithm is mainly based on the maximum classifying accuracy, so that the classification of the majority class data is facilitated, and the effective recognition of the minority class samples is greatly restricted. Therefore, there is a need for solving the problem of data imbalance in fault diagnosis based on industrial big data.
In view of this, the present invention has been made.
Disclosure of Invention
In order to solve the technical problems, the invention provides a motor bearing fault diagnosis method, equipment and medium based on an unbalanced data set, which solve the problem of unbalanced data in the fault diagnosis of industrial big data and realize the fault diagnosis of a motor bearing.
In a first aspect, an embodiment of the present invention provides a motor bearing fault diagnosis method based on an unbalanced data set, the method including:
s1, vibration data of a motor bearing are obtained, wherein the vibration data comprise vibration data of the motor bearing in a normal state and vibration data of the motor bearing in a fault state, the vibration data of the motor bearing in the normal state and the vibration data of the motor bearing in the fault state form an unbalanced vibration data set, and the vibration data of the motor bearing in the normal state is more than the vibration data of the motor bearing in the fault state;
s2, converting vibration data in the data set into a two-dimensional image by utilizing continuous wavelet transformation to form an unbalanced image data set; randomly dividing the image data set into a training sample data set and a test sample data set;
s3, inputting sample data in the training sample data set into a fault diagnosis model to be trained based on dynamic gradient accumulation loss balance so as to train the fault diagnosis model to be trained and obtain a trained fault diagnosis model;
s4, inputting sample data in the test sample data set into the trained fault diagnosis model to verify the trained fault diagnosis model;
s5, performing motor bearing fault diagnosis by using the verified fault diagnosis model.
In a second aspect, an embodiment of the present invention provides an electronic device, including:
a processor and a memory;
the processor is configured to execute the steps of the motor bearing fault diagnosis method based on the unbalance data set according to any one of the embodiments by calling a program or instructions stored in the memory.
In a third aspect, embodiments of the present invention provide a computer-readable storage medium storing a program or instructions that cause a computer to perform the steps of the motor bearing fault diagnosis method based on an imbalance data set of any of the embodiments.
The embodiment of the application provides a motor bearing fault diagnosis method based on an unbalanced data set, which solves the problem of data unbalance in fault diagnosis of industrial big data by using a fault diagnosis model based on dynamic gradient accumulation and loss balance, and realizes the fault diagnosis of a motor bearing.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a motor bearing fault diagnosis method based on an unbalanced data set according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the invention, are within the scope of the invention.
Since the overall structure of the machine tends to be complicated, and each component (e.g., motor bearing) has a certain probability of failing, the probability of the machine failing is greatly increased. Mechanical equipment is easily damaged during operation, and thus, the equipment can be abnormal or shut down. Therefore, the method has important significance in fault diagnosis of core parts in mechanical equipment. Meanwhile, in the data acquisition process, the frequency of the mechanical equipment failure is low, so that the acquired data set has serious unbalance (the unbalance refers to that the positive sample number and the negative sample number are far different), and the difficulty of equipment failure diagnosis is increased. In order to solve the problem, the technical scheme of the application is provided.
Fig. 1 is a schematic flow chart of a motor bearing fault diagnosis method based on an unbalanced data set according to an embodiment of the present invention, as shown in fig. 1, where the motor bearing fault diagnosis method based on an unbalanced data set includes the following steps:
s1, vibration data of a motor bearing are obtained, the vibration data comprise vibration data of the motor bearing in a normal state and vibration data of the motor bearing in a fault state, the vibration data of the motor bearing in the normal state and the vibration data of the motor bearing in the fault state form an unbalanced vibration data set, and the vibration data of the motor bearing in the normal state is more than the vibration data of the motor bearing in the fault state.
S2, converting vibration data in the data set into a two-dimensional image by utilizing continuous wavelet transformation to form an unbalanced image data set; and randomly dividing the image dataset into a training sample dataset and a test sample dataset.
Wherein the vibration data is a one-dimensional vibration signal, the image is a two-dimensional matrix, and the two-dimensional image can carry more information to represent more complex structural distribution than the one-dimensional vibration signal. Conventional one-dimensional signal analysis, such as time domain analysis or frequency domain analysis, is often faced with the difficulty of acquiring the natural modes of the mechanical failure state. Therefore, the present invention converts one-dimensional vibration data into a two-dimensional image.
S3, inputting sample data in the training sample data set into a fault diagnosis model to be trained based on dynamic gradient accumulation loss balance so as to train the fault diagnosis model to be trained, and obtaining a trained fault diagnosis model.
Specifically, step S3 includes the following substeps S31-S36:
s31, sample data in the training sample data set are input to the fault diagnosis model to be trained in batches, wherein the sample data obtained based on vibration data of the motor bearing in a normal state is a negative sample, and the sample data obtained based on vibration data of the motor bearing in a fault state is a positive sample.
In training of the deep learning model, because the data volume of the sample data is huge, the model cannot process all data at one time, so the sample data enters the model in batches (batch), the sample data of each batch plays a role in model training by providing gradients for the model, positive samples provide positive gradients, negative samples provide negative gradients, the gradients can be accumulated in the calculation of loss, the differences between the predicted result and the actual result are calculated and then propagated forward, the network weight is adjusted, the next batch of data is welcome, and the process can be iterated until the model converges. The related parameters involved include a total number of training iterations epoch and a batch size, for example, a total of 1000 samples, and a batch size of 10, so that in one epoch, it is guaranteed that all samples participate in the training of the one epoch after 1000/10=100 iterations, and the training of the one epoch is called a complete iteration process. The value of epoch is set by the programmer, assuming that the value of epoch is set to 200, a total of 200 x 100 iterations are required.
S32, when the ratio of the positive sample number to the negative sample number in the sample data of the current batch is inconsistent with the ratio of the positive sample number to the negative sample number in the training sample data set, multiplying the ratio of the accumulated positive gradient and the accumulated negative gradient accumulated until the current time by a dynamic weight, and obtaining the processed positive-negative gradient ratio, wherein the dynamic weight is the ratio of the negative sample number to the positive sample number in the sample data of the current batch.
Optionally, when the ratio of the positive sample number to the negative sample number in the sample data of the current batch is greater than 1, it is determined that the ratio of the positive sample number to the negative sample number in the sample data of the current batch is inconsistent with the ratio of the positive sample number to the negative sample number in the training sample data set.
Assume that when the batch size is 10, there are 10 samples in total in the sample data of the current batch, where the ratio of the positive number of samples to the negative number of samples is 9:1, and the training sample data set has 10 positive samples and 90 negative samples, which are 100 sample data. The ratio of positive and negative samples was 1:9. At this time, the ratio of the positive number of samples to the negative number of samples in the sample data of the current batch is not considered to be identical to the ratio of the positive number of samples to the negative number of samples in the training sample data set. Or, in other words, when the positive number of samples is greater than the negative number of samples in the sample data of the current batch, the ratio of the positive number of samples to the negative number of samples in the sample data of the current batch is considered to be inconsistent with the ratio of the positive number of samples to the negative number of samples in the training sample data set.
Due to the randomness of the data distribution, there may be a situation that the fault data (i.e., the positive samples) are excessively concentrated locally, which causes an imbalance situation with respect to the whole data set (i.e., the ratio of the positive sample number to the negative sample number in the sample data of the current batch is inconsistent with the ratio of the positive sample number to the negative sample number in the training sample data set), and causes an inaccurate weight calculated by the cumulative gradient. In order to solve the problem, a technical scheme of the application is provided. The method is characterized in that a dynamic weight is added on the basis of statistics of accumulated gradients, so that the problem that the calculated weight of the accumulated gradients is inaccurate due to the fact that positive samples are locally and excessively concentrated is effectively solved, and the processing precision and the robustness of the model can be guaranteed.
Illustratively, the dynamic weights are labeled asD batch Dynamic weight markingD batch Can be determined by the following expression (1):
wherein, the liquid crystal display device comprises a liquid crystal display device,D batch representing the dynamic weight of the model,
representing the negative number of samples in the sample data of the current batch, i.e. the number of sample data obtained based on the vibration data of the motor bearing in a normal state, +.>
Representing the number of samples of the current batchAccording to the positive number of samples, i.e. the number of sample data obtained based on vibration data of the motor bearing in a faulty state,/->
Representing the positive-negative gradient ratio after the treatment, i.e. the dynamic cumulative gradient for single lot weighting,/->
Represents the cumulative positive gradient accumulated until now, < >>
Representing the cumulative negative gradient that has been accumulated until now.
Wherein the accumulated positive gradient accumulated up to the present
And cumulative negative gradient up to the current cumulative +.>
Can be determined by the following expression (2):
wherein, the liquid crystal display device comprises a liquid crystal display device,N batch indicating the number of all the batches that have been calculated,
representing positive gradient result values in iterations of a single batch calculated, +.>
Representing the negative gradient result values in the iterations of the single batch calculated.
Further, positive gradient result values in single batch iterations
And negative gradient result value +.>
Is determined by the following expression (3): />
Wherein, the liquid crystal display device comprises a liquid crystal display device,N batchsize indicating the batch size, i.e. the number of samples a batch contains,
representing the probability that the ith sample is predicted as failure data F, +.>
Indicating one-hot encoding of the ith sample for the failure data F.
In the dynamic training process, it is assumed that 792 pieces of training data can be divided into 40 batches in a batch size of 20, each 1 batch being an iteration, each 40 batches being a complete iteration process. If in 20 samples of 1 batch: the positive number of samples is less than the negative number of samples, and the positive and negative sample gradients are weighted directly by the previously accumulated gradients. If the number of positive samples is more than the number of negative samples, the previous accumulated gradient is adjusted through the dynamic weight, and then the positive and negative sample gradients are weighted. The dynamic weight reduces the negative influence of batches in the positive sample set on the whole gradient accumulation result, maintains the relationship of small positive sample gradient and large negative sample gradient, continuously weights the positive sample gradient in the model training process, and reduces the negative sample gradient, so that the model pays attention to fault data serving as the positive sample better.
S33, respectively obtaining positive gradient weight and negative gradient weight based on the processed positive and negative gradient ratio.
Illustratively, the positive and negative gradient weights are determined based on the following expression (4):
wherein, the liquid crystal display device comprises a liquid crystal display device,
representing the positive-negative gradient ratio after said treatment, < >>
Representing the positive gradient weight, +.>
Representing the weight of the negative gradient,xas independent variables, μ and γ are super parameters for adjusting independent variable range, mapping functionf（x) The aim of (2) is to control the calculation result to take a value between 0 and 1.
S34, carrying out weighting processing on the positive gradient result value in the iteration of the current batch based on the positive gradient weight to obtain a final positive gradient value in the iteration of the current batch, and carrying out weighting processing on the negative gradient result value in the iteration of the current batch based on the negative gradient weight to obtain a final negative gradient value in the iteration of the current batch, wherein the process of inputting sample data of a batch into the model is an iteration process of the model.
Wherein the final positive gradient value in the iteration of the current batch and the final negative gradient value in the iteration of the current batch are determined by the following expression (5):
wherein in expression (5)
Positive gradient result value (determined by expression (3) above) in iteration representing the current lot,>
negative gradient result value (determined by expression (3) above) in iteration representing the current lot,>
representing the final positive gradient value in the iteration of the current batch,/->
Representing the final negative gradient value in the iteration of the current batch.
And S35, continuing to accumulate the final positive gradient value and the accumulated positive gradient, and continuing to accumulate the final negative gradient value and the accumulated negative gradient to obtain a gradient result value in the iteration of the current batch.
Specifically, the final positive gradient value and the final negative gradient value are substituted into the above expression (2) to perform accumulation processing, specifically
Substitution of +.2 in expression (2)>
，/>
Substitution of +.2 in expression (2)>
Gradient result values in iterations of the current batch are obtained. The gradient result value in the iteration of the current batch is in particular the accumulated positive gradient +.>
And cumulative negative gradient accumulated up to the current lot +.>
。
And S36, carrying out loss calculation based on the gradient result to obtain a loss value, and transmitting the loss value forward so that the model adjusts parameters based on the loss value, and ending the current iteration process.
S4, inputting sample data in the test sample data set into the trained fault diagnosis model to verify the trained fault diagnosis model.
S5, performing motor bearing fault diagnosis by using the verified fault diagnosis model.
According to the technical scheme, the training process of the model is improved, so that the model obtained through training of the positive and negative samples based on unbalanced distribution has higher detection precision, and the motor bearing fault diagnosis precision is improved. The problem of unbalanced data in the fault diagnosis of industrial big data is solved, and the fault diagnosis of the motor bearing is realized.
Further, the fault diagnosis model comprises a space-time sequence prediction module ConvLSTM, a convolution block attention module CBAM and a full connection layer;
the step S5 comprises the following steps:
s51, inputting a two-dimensional image obtained based on vibration data of a motor bearing to be diagnosed to ConvLSTM, so as to extract a first feature comprising space information by convolution calculation;
s52, processing the first characteristic through a channel attention mechanism of the CBAM to obtain a second characteristic.
When fault prediction and diagnosis are carried out, vibration data at different moments have different degrees of influence on a prediction result, when fault diagnosis is carried out, convLSTM normally carries out convolution on a hidden state to obtain a prediction output, and the convolution kernel parameters learned in the mode are fixed and cannot be changed along with the change of input. To solve this problem, the channel attention mechanism is added on the basis of ConvLSTM, and the distribution of the attention weights input to the network is automatically completed, so that the network focuses more on the critical information, and the diagnosis accuracy of the model is improved.
And S53, processing the second characteristic through a spatial attention mechanism of the CBAM to obtain a third characteristic.
The CBAM attention mechanism is used for focusing on partial information of auxiliary judgment in the image, and can suppress other useless information, so that the classification efficiency is greatly improved. The CBAM attention mechanism can effectively solve the problem that important features are lost and the fault diagnosis accuracy is low due to the fact that the traditional convolutional neural network model ignores channel attention and space attention.
S54, inputting the third characteristic to ConvLSTM to obtain an output result.
S55, inputting the output result to a full-connection layer, and completing fault classification by using a Softmax activation function.
Fig. 2 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. As shown in fig. 2, electronic device 400 includes one or more processors 401 and memory 402.
The processor 401 may be a Central Processing Unit (CPU) or other form of processing unit having data processing capabilities and/or instruction execution capabilities and may control other components in the electronic device 400 to perform desired functions.
In one example, the electronic device 400 may further include: an input device 403 and an output device 404, which are interconnected by a bus system and/or other forms of connection mechanisms (not shown). The input device 403 may include, for example, a keyboard, a mouse, and the like. The output device 404 may output various information to the outside, including early warning prompt information, braking force, etc. The output device 404 may include, for example, a display, speakers, a printer, and a communication network and remote output devices connected thereto, etc.
Of course, only some of the components of the electronic device 400 that are relevant to the present invention are shown in fig. 2 for simplicity, components such as buses, input/output interfaces, etc. are omitted. In addition, electronic device 400 may include any other suitable components depending on the particular application.
In addition to the methods and apparatus described above, embodiments of the invention may also be a computer program product comprising computer program instructions which, when executed by a processor, cause the processor to perform the steps of the motor bearing fault diagnosis method based on an imbalance data set provided by any of the embodiments of the invention.
The computer program product may write program code for performing operations of embodiments of the present invention in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server.
Furthermore, embodiments of the present invention may also be a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, cause the processor to perform the steps of the motor bearing fault diagnosis method based on an unbalanced data set provided by any of the embodiments of the present invention.
The computer readable storage medium may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of the present application. As used in this specification, the terms "a," "an," "the," and/or "the" are not intended to be limiting, but rather are to be construed as covering the singular and the plural, unless the context clearly dictates otherwise. The terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method or apparatus comprising such elements.
It should also be noted that the positional or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. are based on the positional or positional relationship shown in the drawings, are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or element in question must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention. Unless specifically stated or limited otherwise, the terms "mounted," "connected," and the like are to be construed broadly and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.
Claims (5)
1. The motor bearing fault diagnosis method based on the unbalanced data set is characterized by comprising the following steps of:
s1, vibration data of a motor bearing are obtained, wherein the vibration data comprise vibration data of the motor bearing in a normal state and vibration data of the motor bearing in a fault state, the vibration data of the motor bearing in the normal state and the vibration data of the motor bearing in the fault state form an unbalanced vibration data set, and the vibration data of the motor bearing in the normal state is more than the vibration data of the motor bearing in the fault state;
s2, converting vibration data in the data set into a two-dimensional image by utilizing continuous wavelet transformation to form an unbalanced image data set; randomly dividing the image data set into a training sample data set and a test sample data set;
s3, inputting sample data in the training sample data set into a fault diagnosis model to be trained based on dynamic gradient accumulation loss balance so as to train the fault diagnosis model to be trained and obtain a trained fault diagnosis model;
s4, inputting sample data in the test sample data set into the trained fault diagnosis model to verify the trained fault diagnosis model;
s5, performing motor bearing fault diagnosis by using the verified fault diagnosis model.
2. The motor bearing failure diagnosis method based on an unbalance data set according to claim 1, wherein S3 comprises:
s31, sample data in the training sample data set are input to the fault diagnosis model to be trained in batches, wherein the sample data obtained based on vibration data of the motor bearing in a normal state is a negative sample, and the sample data obtained based on vibration data of the motor bearing in a fault state is a positive sample;
s32, when the ratio of the positive sample number to the negative sample number in the sample data of the current batch is inconsistent with the ratio of the positive sample number to the negative sample number in the training sample data set, multiplying the ratio of the accumulated positive gradient and the accumulated negative gradient accumulated until the current time by a dynamic weight to obtain a processed positive-negative gradient ratio, wherein the dynamic weight is the ratio of the negative sample number to the positive sample number in the sample data of the current batch;
s33, respectively obtaining a positive gradient weight and a negative gradient weight based on the processed positive and negative gradient ratio;
s34, carrying out weighting processing on the positive gradient result value in the iteration of the current batch based on the positive gradient weight to obtain a final positive gradient value in the iteration of the current batch, and carrying out weighting processing on the negative gradient result value in the iteration of the current batch based on the negative gradient weight to obtain a final negative gradient value in the iteration of the current batch, wherein the process of inputting sample data of a batch into the model is an iteration process of the model;
s35, continuing to accumulate the final positive gradient value and the accumulated positive gradient, and continuing to accumulate the final negative gradient value and the accumulated negative gradient to obtain a gradient result value in the iteration of the current batch;
and S36, carrying out loss calculation based on the gradient result to obtain a loss value, and transmitting the loss value forward so that the model adjusts parameters based on the loss value, and ending the current iteration process.
3. The motor bearing fault diagnosis method based on the unbalanced data set according to claim 1, wherein the fault diagnosis model comprises a spatio-temporal sequence prediction module ConvLSTM, a convolution block attention module CBAM and a fully connected layer;
the step S5 comprises the following steps:
s51, inputting a two-dimensional image obtained based on vibration data of a motor bearing to be diagnosed to ConvLSTM, so as to extract a first feature comprising space information by convolution calculation;
s52, processing the first characteristic through a channel attention mechanism of the CBAM to obtain a second characteristic;
s53, processing the second characteristic through a spatial attention mechanism of the CBAM to obtain a third characteristic;
s54, inputting the third feature to ConvLSTM to obtain an output result;
s55, inputting the output result to a full-connection layer, and completing fault classification by using a Softmax activation function.
4. An electronic device, the electronic device comprising:
a processor and a memory;
the processor is configured to execute the steps of the motor bearing failure diagnosis method based on the unbalance data set as claimed in any one of claims 1 to 3 by calling a program or instructions stored in the memory.
5. A computer-readable storage medium storing a program or instructions that cause a computer to execute the steps of the unbalance dataset-based motor bearing fault diagnosis method according to any one of claims 1 to 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310416034.8A CN116128882B (en) | 2023-04-19 | 2023-04-19 | Motor bearing fault diagnosis method, equipment and medium based on unbalanced data set |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310416034.8A CN116128882B (en) | 2023-04-19 | 2023-04-19 | Motor bearing fault diagnosis method, equipment and medium based on unbalanced data set |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116128882A true CN116128882A (en) | 2023-05-16 |
| CN116128882B CN116128882B (en) | 2023-06-23 |
Family
ID=86295977
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310416034.8A Active CN116128882B (en) | 2023-04-19 | 2023-04-19 | Motor bearing fault diagnosis method, equipment and medium based on unbalanced data set |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116128882B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112861982A (en) * | 2021-02-24 | 2021-05-28 | 佛山市南海区广工大数控装备协同创新研究院 | Long-tail target detection method based on gradient average |
| CN113887136A (en) * | 2021-10-08 | 2022-01-04 | 东北大学 | Improved GAN and ResNet based electric vehicle motor bearing fault diagnosis method |
| CN114048787A (en) * | 2022-01-11 | 2022-02-15 | 杭州云智声智能科技有限公司 | Method and system for intelligently diagnosing bearing fault in real time based on Attention CNN model |
| CN114387483A (en) * | 2022-01-06 | 2022-04-22 | 上海商汤智能科技有限公司 | Target detection method, model training method, device, equipment and storage medium |
| CN114563186A (en) * | 2022-01-21 | 2022-05-31 | 苏州大学 | Rolling bearing unbalance fault diagnosis method and system |
| CN114707534A (en) * | 2022-03-07 | 2022-07-05 | 华北电力大学（保定） | Rolling bearing fault diagnosis method under small sample data |
-
2023
- 2023-04-19 CN CN202310416034.8A patent/CN116128882B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112861982A (en) * | 2021-02-24 | 2021-05-28 | 佛山市南海区广工大数控装备协同创新研究院 | Long-tail target detection method based on gradient average |
| CN113887136A (en) * | 2021-10-08 | 2022-01-04 | 东北大学 | Improved GAN and ResNet based electric vehicle motor bearing fault diagnosis method |
| CN114387483A (en) * | 2022-01-06 | 2022-04-22 | 上海商汤智能科技有限公司 | Target detection method, model training method, device, equipment and storage medium |
| CN114048787A (en) * | 2022-01-11 | 2022-02-15 | 杭州云智声智能科技有限公司 | Method and system for intelligently diagnosing bearing fault in real time based on Attention CNN model |
| CN114563186A (en) * | 2022-01-21 | 2022-05-31 | 苏州大学 | Rolling bearing unbalance fault diagnosis method and system |
| CN114707534A (en) * | 2022-03-07 | 2022-07-05 | 华北电力大学（保定） | Rolling bearing fault diagnosis method under small sample data |
Non-Patent Citations (5)
| Title |
|---|
| QIANWEN CUI 等: ""Intelligent Fault Quantitative Identification via the Improved Deep Deterministic Policy Gradient (DDPG) Algorithm Accompanied With Imbalanced Sample"", 《 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》 * |
| YAOXIANG YU 等: ""PCWGAN-GP: A New Method for Imbalanced Fault Diagnosis of Machines"", 《 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》 * |
| YIJUN HUANG 等: ""A Fast bearing Fault diagnosis method based on lightweight Neural Network RepVGG"", 《AISS》 * |
| 包萍;刘运节;: "不均衡数据集下基于生成对抗网络的改进深度模型故障识别研究", 电子测量与仪器学报, no. 03 * |
| 王骁贤;张保华;陆思良;: "基于连续小波变换和卷积神经网络的无刷直流电机故障诊断", 机械与电子, no. 06 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116128882B (en) | 2023-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xia et al. | Multi-stage fault diagnosis framework for rolling bearing based on OHF Elman AdaBoost-Bagging algorithm | |
| CN112766342A (en) | Abnormity detection method for electrical equipment | |
| Chang et al. | Heterogeneous bi-directional recurrent neural network combining fusion health indicator for predictive analytics of rotating machinery | |
| CN116560895B (en) | Fault diagnosis method for mechanical equipment | |
| Ayodeji et al. | Causal augmented ConvNet: A temporal memory dilated convolution model for long-sequence time series prediction | |
| CN112906282B (en) | Method and system for identifying working mode parameters of Sanger neural network parallel principal component extraction | |
| CN111784061B (en) | Training method, device and equipment for power grid engineering cost prediction model | |
| CN112766140A (en) | Transformer fault identification method based on kernel function extreme learning machine | |
| Saravanakumar et al. | Hierarchical symbolic analysis and particle swarm optimization based fault diagnosis model for rotating machineries with deep neural networks | |
| Guo et al. | A Hybrid Prognosis Scheme for Rolling Bearings Based on a Novel Health Indicator and Nonlinear Wiener Process | |
| Orăşan et al. | Benchmarking tensorflow lite quantization algorithms for deep neural networks | |
| CN116128882B (en) | Motor bearing fault diagnosis method, equipment and medium based on unbalanced data set | |
| CN115824261B (en) | Control moment gyroscope fault detection method and device and related storage medium | |
| CN116702005A (en) | Neural network-based data anomaly diagnosis method and electronic equipment | |
| CN109187898B (en) | Soft measurement method and device for ammonia nitrogen content of water in aquaculture environment | |
| CN113624466B (en) | Method, device, equipment and storage medium for diagnosing turbine rotor faults | |
| CN116739154A (en) | Fault prediction method and related equipment thereof | |
| CN116484209A (en) | Dam structure monitoring abnormality identification method and device | |
| US7840391B2 (en) | Model-diversity technique for improved proactive fault monitoring | |
| Godói et al. | Denoising convolutional autoencoder configuration for condition monitoring of rotating machines | |
| CN117493980B (en) | Bearing fault diagnosis method integrating feature extraction and sequencing causal discovery | |
| Yu et al. | Fault diagnosis of wearable temperature sensors based on multi-scale feature extraction | |
| CN115712726B (en) | Emotion analysis method, device and equipment based on double word embedding | |
| CN113151842B (en) | Method and device for determining conversion efficiency of wind-solar complementary water electrolysis hydrogen production | |
| CN117407797B (en) | Equipment fault diagnosis method and model construction method based on incremental learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |