CN115358277A - Bearing fault diagnosis method, device, equipment and readable storage medium - Google Patents
Bearing fault diagnosis method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN115358277A CN115358277A CN202211224305.1A CN202211224305A CN115358277A CN 115358277 A CN115358277 A CN 115358277A CN 202211224305 A CN202211224305 A CN 202211224305A CN 115358277 A CN115358277 A CN 115358277A
- Authority
- CN
- China
- Prior art keywords
- bearing
- model
- training
- data
- signal
- 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.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 91
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000012549 training Methods 0.000 claims abstract description 132
- 238000013145 classification model Methods 0.000 claims abstract description 76
- 239000011159 matrix material Substances 0.000 claims abstract description 36
- 230000009466 transformation Effects 0.000 claims abstract description 35
- 230000008569 process Effects 0.000 claims abstract description 28
- 238000012545 processing Methods 0.000 claims abstract description 27
- 238000003062 neural network model Methods 0.000 claims description 33
- 238000002405 diagnostic procedure Methods 0.000 claims 2
- 238000006243 chemical reaction Methods 0.000 abstract description 6
- 230000008859 change Effects 0.000 description 10
- 238000004891 communication Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 9
- 230000000737 periodic effect Effects 0.000 description 5
- 238000000605 extraction Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000013526 transfer learning Methods 0.000 description 2
- 230000001131 transforming effect Effects 0.000 description 2
- GNFTZDOKVXKIBK-UHFFFAOYSA-N 3-(2-methoxyethoxy)benzohydrazide Chemical compound COCCOC1=CC=CC(C(=O)NN)=C1 GNFTZDOKVXKIBK-UHFFFAOYSA-N 0.000 description 1
- FGUUSXIOTUKUDN-IBGZPJMESA-N C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 Chemical compound C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 FGUUSXIOTUKUDN-IBGZPJMESA-N 0.000 description 1
- YTAHJIFKAKIKAV-XNMGPUDCSA-N [(1R)-3-morpholin-4-yl-1-phenylpropyl] N-[(3S)-2-oxo-5-phenyl-1,3-dihydro-1,4-benzodiazepin-3-yl]carbamate Chemical compound O=C1[C@H](N=C(C2=C(N1)C=CC=C2)C1=CC=CC=C1)NC(O[C@H](CCN1CCOCC1)C1=CC=CC=C1)=O YTAHJIFKAKIKAV-XNMGPUDCSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 229910052918 calcium silicate Inorganic materials 0.000 description 1
- 235000012241 calcium silicate Nutrition 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Acoustics & Sound (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a bearing fault diagnosis method, a device, equipment and a readable storage medium, wherein the method comprises the following steps: acquiring signal data of a bearing in the running process; performing matrix transformation processing on the signal data to obtain target data; inputting the target data to a diagnosis model to obtain a classification result; determining the fault type of the bearing based on the classification result; the diagnostic model is obtained by performing iterative training on an image classification model based on a bearing signal training data set; the bearing signal training data set is obtained by performing matrix transformation processing on the signal data. The method and the device have the advantages that the signal data of the bearing in the operation process are converted, so that the target data are obtained, the target data subjected to matrix conversion are used as training samples, the diagnostic model is obtained through training, the classification and recognition accuracy of the trained model is improved, and the diagnosis accuracy of the diagnostic model on the bearing fault is high.
Description
Technical Field
The present disclosure relates to the field of signals, and in particular, to a bearing fault diagnosis method, device, apparatus, and readable storage medium.
Background
During the bearing operation, a corresponding operation signal during operation is generated, and whether the bearing has a fault or not is determined by detecting the signal change.
The collected signal data of the bearing in the operation process is usually a one-dimensional signal in a time domain, but the image classification model cannot accurately identify the significant features of the one-dimensional signal, so that the signal data is used as a training sample for training the image classification model, and the classification precision of the trained model is low.
Therefore, the accuracy of fault diagnosis of the bearing signal using the model in which the signal data is used as the training sample is low.
Disclosure of Invention
In view of the above, the present application provides a bearing fault diagnosis method, apparatus, device and readable storage medium, which aim to improve the accuracy of fault diagnosis of a bearing generated by an image classification model.
In order to achieve the above object, the present application provides a bearing fault diagnosis method, including the steps of:
acquiring signal data of a bearing in the running process;
performing matrix transformation processing on the signal data to obtain target data;
inputting the target data to a diagnosis model to obtain a classification result;
determining the fault type of the bearing based on the classification result; the diagnostic model is obtained by performing iterative training on an image classification model based on a bearing signal training data set; the bearing signal training data set is obtained by performing matrix transformation processing on the signal data.
Illustratively, the matrix transforming the signal data to obtain the target data includes:
converting the signal data to a two-dimensional matrix of a time-frequency domain;
acquiring time domain information and frequency domain information in the two-dimensional matrix;
and integrating the time domain information and the frequency domain information to obtain target data.
Illustratively, before converting the signal data into a two-dimensional matrix form, the method includes:
detecting a frequency of the signal data;
and selecting a Gaussian window with the length and the frequency resolution both suitable for the frequency to extract information on different frequency bands and extract frequency components of the signal data at each time point in a preset time length, wherein the information on the frequency bands and the frequency components are time domain information and frequency domain information.
Illustratively, the integrating the time domain information and the frequency domain information to obtain the target data includes:
establishing a time-frequency domain with time, frequency and amplitude as coordinate axes based on the time domain information and the frequency domain information, and drawing image information in the time-frequency domain;
and selecting the projection of the image information in a plane formed by time and frequency coordinate axes to obtain target data.
Illustratively, the taking the target data as a training sample, outputting the training sample to the neural network model, and training the neural network model to obtain a diagnostic model includes:
acquiring training samples when different faults occur to a bearing;
inputting the training sample to an image classification model, and obtaining a diagnosis model after the image classification model finishes training; the image classification model is a neural network model for image classification.
Exemplarily, the training sample is input to an image classification model, and a diagnosis model is obtained after the image classification model is trained; the image classification model is a neural network model for image classification, and comprises the following steps:
inputting the training samples to an image classification model, and classifying the training samples to obtain training classification labels;
calculating the gradient of the image classification model based on the training classification label and a preset real label corresponding to the training sample;
determining whether the image classification model meets a preset iterative training end condition or not based on the gradient;
if so, taking the image classification model as a diagnosis model;
if not, continuing to carry out iterative training on the image classification model until the image classification model meets a preset iterative training end condition.
For example, the inputting the training sample to an image classification model, and after the image classification model is trained to obtain a diagnosis model, the method includes:
obtaining training samples of bearings of other models;
and taking the diagnosis model as an initial model, and training the initial model by using the training samples of the bearings of other models to obtain a target model for diagnosing the bearings of different models.
Illustratively, to achieve the above object, the present application also provides a bearing fault diagnosis apparatus, comprising:
the acquisition module is used for acquiring signal data of the bearing in the operation process;
the processing module is used for carrying out matrix transformation processing on the signal data to obtain target data;
the input module is used for inputting the target data to a diagnosis model to obtain a classification result;
the determining module is used for determining the fault type of the bearing based on the classification result; the diagnostic model is obtained by performing iterative training on an image classification model based on a bearing signal training data set; the bearing signal training data set is obtained by performing matrix transformation processing on the signal data.
Illustratively, to achieve the above object, the present application also provides a bearing fault diagnosis apparatus, comprising: a memory, a processor and a bearing fault diagnosis program stored on the memory and executable on the processor, the bearing fault diagnosis program being configured to implement the steps of the bearing fault diagnosis method as described above.
Illustratively, to achieve the above object, the present application further provides a computer readable storage medium having stored thereon a bearing fault diagnosis program, which when executed by a processor, implements the steps of the bearing fault diagnosis method as described above.
In the prior art, an image classification model cannot directly perform feature extraction and accurate identification on signal data generated when a bearing operates, so that the signal data is directly used as a training sample, and compared with the situation that a diagnosis model obtained by training the image classification model has low accuracy rate of bearing fault diagnosis, in the application, the signal data of the bearing in the operation process is obtained, the signal data is subjected to matrix transformation to obtain target data, the target data is input into the diagnosis model to obtain a classification result, and whether the bearing has a fault type is determined according to the classification result, wherein the used diagnosis model is obtained by performing iterative training on the image classification model based on a bearing signal training data set, meanwhile, the data in the bearing signal training data set is obtained by performing matrix transformation on the signal data of the bearing in the operation process, namely, the signal data which cannot directly extract features of the original image classification model is converted into data capable of being used as the training sample, so that the accuracy of the trained image classification model on the feature classification of the signal data is improved, and the bearing fault diagnosis rate can be accurately diagnosed by performing matrix transformation on the signal data of the diagnosis model.
Drawings
FIG. 1 is a schematic flow chart diagram of a first embodiment of a bearing fault diagnosis method of the present application;
FIG. 2 is a schematic flow chart of a second embodiment of the bearing fault diagnosis method of the present application;
FIG. 3 is a schematic diagram of a vibration signal collected when a fault occurs in an inner race of a bearing gearbox according to the present application;
FIG. 4 is a schematic illustration of one-dimensional signal data of the present application;
FIG. 5 is a schematic diagram of the conversion of one-dimensional signal data to two-dimensional image data according to the present application;
fig. 6 is a schematic structural diagram of a hardware operating environment according to an embodiment of the present application.
The implementation, functional features and advantages of the object of the present application will be further explained with reference to the embodiments, and with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.
The application provides a bearing fault diagnosis method, and with reference to fig. 1, fig. 1 is a schematic flow chart of a first embodiment of the bearing fault diagnosis method.
While a logical order is shown in the flow chart, in some cases, the steps shown or described may be performed in an order different than that shown or described herein. For convenience of description, the following omits to perform the steps of the subject description bearing fault diagnosis method, which includes:
step S110: acquiring signal data of a bearing in the running process;
the bearing generates different operation signals in the rotation process, for example, the operation signals are different when the bearing runs stably and when the bearing breaks down, and meanwhile, the fault signals generated when the bearing runs are different according to different fault types (for example, faults such as bearing outer ring breakage, bearing ball breakage and the like) generated by the bearing.
Referring to fig. 2, fig. 2 is a schematic diagram of a vibration signal acquired when a fault occurs in an inner ring of a bearing gearbox, and it can be seen that the fault in the inner ring of the bearing causes a periodic impact event to occur in the bearing, and the periodic impact event appears as a plurality of peaks and small tails on a time domain signal.
Wherein the abscissa of fig. 2 is the detection period, and the ordinate of fig. 2 is the amplitude of the signal.
The method for acquiring the signal data of the bearing in the running process mainly comprises two categories: a normal operation signal and a fault operation signal. Wherein the fault operation signal has different signal characteristics according to different fault types.
Step S120: performing matrix transformation processing on the signal data to obtain target data;
because the normally acquired images of the signals are all in a one-dimensional signal state in a time domain, only the one-dimensional form of the bearing signals is aimed at, and the distribution of the bearing signals in a frequency domain is not concerned, thereby causing information waste.
The form of the one-dimensional signal in the time domain is the variation of the signal amplitude in a period of time, and the distribution in the frequency domain is the specific value of the signal amplitude in a certain frequency.
In addition, the one-dimensional signal has noise and other conditions in the time domain, the direct extraction of the characteristics of the one-dimensional signal can cause the characteristics to be affected by the noise, and the problem does not exist in the frequency domain because most of the noise is distributed in the whole frequency band, but the noise amplitude of a certain frequency band is small, so that the time domain and the frequency domain are divided, and meanwhile, the information of the signal in the time domain and the frequency domain is considered, and the low-noise and accurate signal characteristic information can be obtained.
Therefore, when the signal data is transformed, the S transform (Stockwell transform) is used, thereby avoiding the problem of poor frequency resolution of the prior STFT (short-time Fourier transform). In addition, in the S transformation process, windows with different lengths are added to the low-frequency signals and the high-frequency signals respectively, so that different information on different frequency bands can be better extracted through the S transformation. After S transformation, the frequency components of the signal at each time point in a certain time can be clearly seen, and the information on the frequency band and the frequency components are time domain information and frequency domain information.
For example, when the signal data is subjected to the transformation processing, in addition to the S transformation, the one-dimensional signal can be converted into a two-dimensional image by using a method such as a grammite angular field or a markov transition field.
In addition, when the signal data is subjected to transformation processing, the signal data is converted from a one-dimensional signal state to a two-dimensional image state, so that more general image recognition neural network models can be used for the signals, and the design pressure of algorithm developers is reduced.
Therefore, after the signal data are transformed, the target data are obtained, wherein the target data are two-dimensional image data, that is, in this step, the process of converting the signal data is completed, and a one-dimensional signal which cannot accurately capture the signal characteristics of the original neural network model is converted into image information more suitable for the neural network model, so that the signal data of the bearing during operation can be identified and extracted by the neural network model, and the transformed target data are sample data which can be used for establishing the neural network model.
Step S130: and inputting the target data to a diagnosis model to obtain a classification result.
The target data is data which is subjected to transformation processing and can be identified or extracted by the neural network model, so that the target data is used as a training sample to train the neural network model, and the diagnosis model is obtained.
The training sample of the diagnosis model is signal data of the bearing during operation, wherein the signal data comprises a normal operation signal and a fault operation signal, and therefore the trained neural network model can be used for judging whether the operation state of the bearing has a fault.
For example, after the target data is input into the diagnostic model, the diagnostic model diagnoses the data, the diagnostic model performs a screening or classifying operation on the signal data, classifies the signals of the bearing in the normal operation state into one class, and classifies the signals of the bearing in the fault operation state into one class, so as to obtain a classification result.
Exemplarily, the taking the target data as a training sample, outputting the training sample to the neural network model, and training the neural network model to obtain a diagnostic model includes:
a, step a: acquiring training samples when different faults occur to a bearing;
when the target data is used as a training sample, different signals generated by different faults of the bearing are considered, the target data when the bearing has different faults needs to be stored and used as the training sample, so that the types of the training sample are increased, and the work content which can be completed by the neural network model is increased when the training sample is used.
For example, when target data of different faults of the bearing is used as a training sample, the neural network model learns different types of fault operation signals, and comprehensively grasps the fault type of the bearing, for example, a fault operation signal when the bearing inner ring is cracked, a fault operation signal when the bearing balls are broken, a fault operation signal when the bearing inner ring raceway is dropped, and the like are input.
Step b: inputting the training sample to an image classification model, and obtaining a diagnosis model after the image classification model finishes training; the image classification model is a neural network model for image classification.
Since the target data is two-dimensional image data, when outputting the training samples to the neural network model, the selected neural network model is a model for image classification, for example, resNet (residual neural network) or densnet (dense connection convolutional network) in a CNN (convolutional neural network) model series.
Exemplarily, the training sample is input to an image classification model, and a diagnosis model is obtained after the image classification model is trained; the image classification model is a neural network model for image classification, and comprises the following steps:
step c: inputting the training samples to an image classification model, and classifying the training samples to obtain training classification labels;
when classifying the training samples, the data type in the training samples converted from the signal data of the bearing in the operation process is considered, the data comprises the data of the bearing in normal operation and the operation data of the bearing in fault operation, and meanwhile, when the bearing is in fault, various faults (bearing outer ring cracking and the like) may occur in the bearing, and different faults generate data results of different types.
Thus, after the classification process, two levels of classification labels are obtained.
Step d: calculating the gradient of the image classification model based on the training classification label and a preset real label corresponding to the training sample;
the preset real label corresponds to different types of data of normal operation of the bearing, fault operation of the bearing and fault of the bearing.
After classification, the gradient of the model is calculated to find the optimal weights and biases.
Step e: determining whether the image classification model meets a preset iterative training end condition or not based on the gradient;
in the classification model, after the gradient is calculated, the size of the current gradient value is determined, the minimum value of the gradient is taken as the optimal weight point, meanwhile, the continuous iterative calculation process of model training and the classification precision condition of the model are considered, the preset iterative training end condition is set, and under the condition that the gradient value of the classification model is not the minimum value but meets the classification precision, the content of overlarge iterative training is avoided.
Step f: if so, taking the image classification model as a diagnosis model;
step g: and if not, continuing to carry out iterative training on the image classification model until the image classification model meets a preset iterative training end condition.
If the gradient meets the preset iterative training end condition, the diagnostic model is obtained, if the gradient does not meet the preset iterative training end condition, the image classification model needs to be subjected to iterative training continuously, the classification weight of the image classification model is changed, the current gradient is calculated until the gradient meets the preset iterative training end condition, and the model meeting the condition at last is used as the diagnostic model.
And the finally obtained diagnosis model meets two classification conditions, normal operation and fault operation of the bearing are obtained by classification, data in the fault operation of the bearing are further classified, and classification is carried out according to the fault type.
Meanwhile, when the bearing operation data is classified according to the fault, the classification point is considered, and the characteristic of the data during the operation of the bearing is taken as a classification judgment basis, so that the diagnostic model is used for extracting the characteristic of the operation data of the bearing.
For example, the inputting the training sample to an image classification model, and obtaining a diagnostic model after the image classification model is trained, includes:
step h: obtaining training samples of bearings of other models;
after the diagnostic model is obtained, the types of the bearings are different, for example, the sizes of the bearings are different (the large bearings or the oversize bearings are used in large equipment, the size range of the large bearings is 200-430mm, and the size range of the oversize bearings is more than 440 mm), the types of the bearings are different (ball bearings, deep groove ball bearings, cylindrical roller bearings, thrust bearings and the like), and the signal states and the signal characteristics of the bearings in different types are different during operation.
Therefore, after a diagnosis model suitable for one type of bearing is obtained, the diagnosis model is converted into a diagnosis model suitable for other types of bearings in a transfer learning mode, so that training samples of other types of bearings are obtained, and the data content of the training samples is two-dimensional image data after transformation, but not one-dimensional signal data.
Step i: and taking the diagnosis model as an initial model, and training the initial model by using the training samples of the bearings of other models to obtain a target model for diagnosing the bearings of different models.
The transfer learning process is to use a trained existing model as a basis of a new neural network model, use a new training sample (the training sample and a training sample used by the existing model are approximate samples, for example, use a diagnostic model as a diagnostic model suitable for training bearings of other models), output the new training sample to the existing model, and perform secondary training on the existing model to obtain a new model.
Step S140: determining the fault type of the bearing based on the classification result; the diagnostic model is obtained by performing iterative training on an image classification model based on a bearing signal training data set; the bearing signal training data set is obtained by performing matrix transformation processing on the signal data.
After the diagnosis model is obtained, the diagnosis model is used to diagnose the signals of the bearing, and the signals of the bearing need to be pre-processed, that is, the original one-dimensional signals to be diagnosed are converted into two-dimensional signal images, so that the diagnosis model can perform related processing on the two-dimensional images.
Therefore, after the diagnosis model is obtained, a transformation module (the transformation module is provided with S-transformation operation logic, for example, a one-dimensional signal is input, and two-dimensional image data is output through transformation module logic operation) is arranged to serve as a module for preprocessing the bearing signal, the module is directly communicated with the neural network model, relevant workers only need to input the bearing signal into the transformation module, the S-transformation of the one-dimensional signal of the bearing is realized through a logic operation rule added in the transformation module, image data is obtained, and the image data is directly input into the diagnosis model, so that the diagnosis efficiency is improved.
The diagnostic model is a neural network model for image classification, and is particularly excellent in functions of extracting and recognizing image features, and therefore specific applications of the diagnostic model include diagnosing whether an operation signal of a bearing is normal or faulty, and extracting feature information in a converted bearing signal image.
The characteristic information includes an amplitude characteristic of the bearing signal (for example, a peak characteristic or a peak tailing characteristic of the bearing signal within a certain time, the characteristic is mainly represented by amplitude change, amplitude height, amplitude change frequency and the like of the bearing signal), an overall periodic change of the bearing signal and the like (for example, when the bearing fails, a bearing inner ring cracks, and referring to a signal image of fig. 2, the signal has a periodic peak change).
When the operation signals of the bearing are diagnosed by using the diagnosis model, on one hand, the health state of the operation signals of the bearing is diagnosed, and on the other hand, the motion state of the bearing is diagnosed, wherein the health state comprises the normal state and the abnormal state (fault state) of the bearing, and the motion state comprises the signal state (normal operation signals, fault operation signals and the bearing fault type to which the fault operation signals belong) when the bearing operates.
When a bearing signal is diagnosed, firstly, an operation signal of a bearing is input into a neural network model; secondly, screening out data of the bearing with faults; thirdly, executing a judging process after screening out the data with faults of the bearing, and judging the specific fault type of the data; and fourthly, outputting a diagnosis result, wherein the diagnosis result comprises a part of normal operation signals and a part of fault operation signals of the input bearing operation signals.
Therefore, the neural network model comprises at least two intermediate layers, wherein one intermediate layer classifies the normal operation signal and the fault operation signal, the other intermediate layer judges the fault operation signal, and the fault type of the bearing represented by the signal is diagnosed.
In the prior art, an image classification model cannot directly perform feature extraction and accurate identification on signal data generated when a bearing operates, so that the signal data is directly used as a training sample, and compared with the situation that a diagnosis model obtained by training the image classification model has low accuracy rate of bearing fault diagnosis, in the application, the signal data of the bearing in the operation process is obtained, the signal data is subjected to matrix transformation to obtain target data, the target data is input into the diagnosis model to obtain a classification result, and whether the bearing has a fault type is determined according to the classification result, wherein the used diagnosis model is obtained by performing iterative training on the image classification model based on a bearing signal training data set, meanwhile, the data in the bearing signal training data set is obtained by performing matrix transformation on the signal data of the bearing in the operation process, namely, the signal data which cannot directly extract features of the original image classification model is converted into data capable of being used as the training sample, so that the accuracy of the trained image classification model on the feature classification of the signal data is improved, and the bearing fault diagnosis rate can be accurately diagnosed by performing matrix transformation on the signal data of the diagnosis model.
Exemplarily, referring to fig. 3, fig. 3 is a schematic flowchart of a second embodiment of the bearing fault diagnosis method of the present application, and the second embodiment is proposed based on the above first embodiment of the bearing fault diagnosis method of the present application, and the method further includes:
illustratively, the matrix transforming the signal data to obtain the target data includes:
step S210: converting the signal data to a two-dimensional matrix of a time-frequency domain;
when the signal data are subjected to transformation processing, the original one-dimensional form of the signal data is converted into two-dimensional image data which can be accurately extracted or identified by a neural network model.
The one-dimensional signal data in the one-dimensional form is amplitude characteristics of a bearing signal in a certain time period, that is, the signal data of the bearing in the operation process is data in a coordinate system with the abscissa axis as time and the ordinate axis as amplitude.
Referring to fig. 4 and 5, fig. 4 is a schematic diagram of one-dimensional signal data, fig. 5 is a schematic diagram of converting the one-dimensional signal data into two-dimensional image data, where the two-dimensional image data is two-dimensional image data obtained by processing signal data in a time domain (in a time and amplitude coordinate system) and converting the signal data into a time-frequency domain (in a three-axis coordinate system of time, frequency and amplitude), where each point in the time-frequency domain represents a component amplitude of the signal in a short neighborhood of a certain time and at a certain frequency, that is, the time, frequency and amplitude characteristics of the signal are simultaneously represented.
Fig. 4 shows signal data that has not been subjected to transform processing, where an abscissa axis of a coordinate system where the data is located is time, an ordinate axis is amplitude, and is specifically represented as time domain information, that is, amplitude variation characteristics in a time domain, and fig. 5 shows signal data that has been subjected to transform processing, where an abscissa axis of a coordinate system where the data is located is time, an ordinate axis is frequency, and is specifically represented as time-frequency domain information, that is, meaning that any point in an image represents component amplitude at a certain frequency in a short neighborhood of the signal at a certain time.
When signal data is converted, the current signal data is one-dimensional signal data, and when the current signal data is converted into two-dimensional image data, it is necessary to first separate and convert information contained in the one-dimensional signal data, and convert signal data in which original information is overlapped (time, frequency, and amplitude are superimposed in the one-dimensional signal data) into a two-dimensional matrix by using a main operation step of S conversion.
Illustratively, before converting the signal data into a two-dimensional matrix form, the method includes:
step j: detecting a frequency of the signal data;
before converting the signal data, the overall frequency of the signal data needs to be detected, and corresponding conversion strategies are made according to different frequency sections and different frequency sizes.
The signal data is data in the coordinate axes of time and amplitude, and therefore, the change condition of the overall frequency is detected directly on the basis of the acquired signal data, and the frequency of a periodic peak generated in the whole signal data, the tailing frequency after the peak of the signal data is detected, and the like are detected.
Step k: and selecting a Gaussian window with the length and the frequency resolution both suitable for the frequency to extract information on different frequency bands and extract frequency components of the signal data at each time point in a preset duration, wherein the information on the frequency bands and the frequency components are time domain information and frequency domain information.
In the process of converting signal data, a Gaussian window is selected and is transformed according to a window frame selection range, wherein the window is used for intercepting and selecting a section of continuous signals to extract frequency components of different frequency bands or at each time point in a preset time length, and information on the frequency bands and the frequency components are time domain information and frequency domain information.
When selecting the Gaussian window, the selection range of the window is considered, the selection range is at least a section of complete frequency section, and meanwhile, the Gaussian window needs to select different frequency resolutions according to different frequency sizes of different frequency sections.
Step S220: acquiring time domain information and frequency domain information in the two-dimensional matrix;
the two-dimensional matrix mainly comprises two aspects: time domain information and frequency domain information.
The time domain information is data content of the signal in a coordinate system of time and amplitude, that is, for a change situation of the amplitude in a time domain coordinate.
The frequency domain information is data content of the signal in a coordinate system of frequency and amplitude, that is, for a change situation of the amplitude in the frequency domain coordinate.
That is, when one-dimensional signal data is transformed into two-dimensional image data, information on the frequency domain is added, so that the neural network model directly diagnoses time domain information and frequency domain information.
Step S230: and integrating the time domain information and the frequency domain information to obtain target data.
After signal data are transformed, the problems of feature overlapping and unobvious feature information in pure one-dimensional signal data are solved, a neural network model cannot be accurately identified, meanwhile, noise interference usually exists in the one-dimensional signal data, so that the feature information is hidden, and usable signal segments cannot be directly extracted from the one-dimensional signal data.
Therefore, the time domain information and the frequency domain information are integrated to obtain target data capable of being directly extracted by the neural network model.
Illustratively, the integrating the time domain information and the frequency domain information to obtain the target data includes:
step l: establishing a time-frequency domain with time, frequency and amplitude as coordinate axes based on the time domain information and the frequency domain information, and drawing image information in the time-frequency domain;
and when integrating the time domain information and the frequency domain information, establishing a three-dimensional coordinate system, and drawing image information under the coordinate system related to time, frequency and amplitude.
Namely, the image with noise of the original one-dimensional signal data is converted into two-dimensional image data, the noise is reduced, and the image information obviously represents the characteristics of the rotation signal.
The image information is a three-dimensional image in a three-dimensional coordinate system. The image information shows the change characteristics of the rotation signal of the bearing in a time-frequency domain.
Step m: and selecting the projection of the image information in a plane formed by time and frequency coordinate axes to obtain target data.
The current image information is a three-dimensional image, that is, an image in a three-dimensional space, the image is detected at different observation angles, different detection effects are generated, and the problem of abnormality in the extracted information is caused by directly extracting feature information from the image, so that the image cannot be directly subjected to feature extraction work through a neural network model for analyzing the image.
Therefore, after the current three-dimensional image information is obtained, the projection effect of the selected image in a certain plane is taken as target data, the target data comprises information of at least two coordinate axes, and simultaneously, variables of at least one coordinate axis are hidden in the curve change of projection, so that the information in three-way coordinate axes comprising time, frequency and amplitude is contained.
The characteristics of the image information in the time-frequency domain are considered, namely when the one-dimensional signal data are converted into the two-dimensional image data, the signals are converted from the original time domain to the time-frequency domain, and the noise interference is reduced, so that a plane formed by a time coordinate axis and a frequency coordinate axis in a three-dimensional coordinate system is selected as a projection plane, and the projection of the image information on the projection plane is selected as target data.
In this embodiment, the obtained signal data is converted, that is, the one-dimensional signal data is converted into two-dimensional image data, so that the original signal data in the time domain is converted into image data in the time-frequency domain, a group of characteristic information of the original signal data in the time domain and the frequency domain is obtained through conversion, the image information in the three-dimensional coordinate system is drawn by using a three-dimensional coordinate system with time, frequency and amplitude as coordinate axes, and an image of the image information in a plane formed by the time coordinate axis and the frequency coordinate axis is selected as target data.
In addition, this application still provides a bearing fault diagnosis device, a bearing fault diagnosis device includes:
the acquisition module is used for acquiring signal data of the bearing in the operation process;
the processing module is used for carrying out matrix transformation processing on the signal data to obtain target data;
the input module is used for inputting the target data to a diagnosis model to obtain a classification result;
the determining module is used for determining the fault type of the bearing based on the classification result; the diagnostic model is obtained by performing iterative training on an image classification model based on a bearing signal training data set; the bearing signal training data set is obtained by performing matrix transformation processing on the signal data.
Illustratively, the processing module includes:
the conversion submodule is used for converting the signal data to a two-dimensional matrix of a time-frequency domain;
the first acquisition submodule is used for acquiring time domain information and frequency domain information in the two-dimensional matrix;
and the integration submodule is used for integrating the time domain information and the frequency domain information to obtain target data.
Illustratively, the transformation module includes:
a detection unit for detecting a frequency of the signal data;
the first selecting unit is used for selecting a Gaussian window with the length and the frequency resolution both suitable for the frequency to extract information on different frequency bands and extract frequency components of the signal data at each time point in a preset duration, wherein the information on the frequency bands and the frequency components are time domain information and frequency domain information.
Illustratively, the synthesis submodule includes:
the establishing unit is used for establishing a time-frequency domain with time, frequency and amplitude as coordinate axes based on the time domain information and the frequency domain information, and drawing image information under the time-frequency domain;
and the second selecting unit is used for selecting the projection of the image information in a plane formed by time and frequency coordinate axes to obtain target data.
Illustratively, the input module includes:
the second obtaining submodule is used for obtaining training samples when different faults occur to the bearing;
the input submodule is used for inputting the training sample to an image classification model and obtaining a diagnosis model after the image classification model finishes training; the image classification model is a neural network model for image classification;
the third acquisition submodule is used for acquiring training samples of bearings of other models;
and the training submodule is used for taking the diagnostic model as an initial model, training the initial model by using the training samples of the bearings of other models, and obtaining target models for diagnosing the bearings of different models.
Illustratively, the input submodule includes:
the input unit is used for inputting the training samples to an image classification model and classifying the training samples to obtain training classification labels;
the calculating unit is used for calculating the gradient of the image classification model based on the training classification label and a preset real label corresponding to the training sample;
the determining unit is used for determining whether the image classification model meets a preset iteration training end condition or not based on the gradient;
a first judgment unit, configured to take the image classification model as a diagnosis model if the first judgment unit is satisfied;
and the second judging unit is used for continuing to carry out iterative training on the image classification model if the image classification model is not met until the image classification model meets a preset iterative training end condition.
The specific implementation of the bearing fault diagnosis device of the present application is substantially the same as that of each embodiment of the bearing fault diagnosis method, and is not described herein again.
In addition, this application still provides a bearing fault diagnosis equipment. As shown in fig. 6, fig. 6 is a schematic structural diagram of a hardware operating environment according to an embodiment of the present application.
For example, fig. 6 is a schematic structural diagram of a hardware operating environment of the bearing fault diagnosis device.
As shown in fig. 6, the bearing fault diagnosis apparatus may include a processor 601, a communication interface 602, a memory 603, and a communication bus 604, wherein the processor 601, the communication interface 602, and the memory 603 complete communication with each other through the communication bus 604, and the memory 603 stores a computer program; the processor 601 is configured to implement the steps of the bearing fault diagnosis method when executing the program stored in the memory 603.
The communication bus 604 mentioned in the above bearing fault diagnosis device may be a Peripheral Component Interconnect (PCI) bus, an Extended Industrial Standard Architecture (EISA) bus, or the like. The communication bus 604 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.
The communication interface 602 is used for communication between the above-described bearing failure diagnosis apparatus and other apparatuses.
The Memory 603 may include a Random Access Memory (RMD) and may also include a Non-Volatile Memory (NM), such as at least one disk Memory. Optionally, the memory 603 may also be at least one storage device located remotely from the processor 601.
The Processor 601 may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also Digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components.
The specific implementation of the bearing fault diagnosis device of the present application is substantially the same as that of each embodiment of the bearing fault diagnosis method, and is not described herein again.
Furthermore, an embodiment of the present application also provides a computer-readable storage medium, on which a bearing fault diagnosis program is stored, and the bearing fault diagnosis program, when executed by a processor, implements the steps of the bearing fault diagnosis method as described above.
The specific implementation of the computer-readable storage medium of the present application is substantially the same as that of the embodiments of the bearing fault diagnosis method described above, and is not described herein again.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system 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, article, or system. Without further limitation, an element defined by the phrases "comprising one of 8230; \8230;" 8230; "does not exclude the presence of additional like elements in a process, method, article, or system that comprises the element.
The above-mentioned serial numbers of the embodiments of the present application are merely for description and do not represent the merits of the embodiments.
Through the description of the foregoing embodiments, it is clear to those skilled in the art that the method of the foregoing embodiments may be implemented by software plus a necessary general hardware platform, and certainly may also be implemented by hardware, but in many cases, the former is a better implementation. Based on such understanding, the technical solutions of the present application or portions thereof contributing to the prior art may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) as described above and includes several instructions for enabling a terminal device (which may be a mobile phone, a computer, a server, or a network device) to execute the method described in the embodiments of the present application.
The above description is only a preferred embodiment of the present application, and not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings of the present application, or which are directly or indirectly applied to other related technical fields, are included in the scope of the present application.
Claims (10)
1. A bearing fault diagnosis method is characterized by comprising the following steps:
acquiring signal data of a bearing in the running process;
performing matrix transformation processing on the signal data to obtain target data;
inputting the target data to a diagnosis model to obtain a classification result;
determining the fault type of the bearing based on the classification result; the diagnostic model is obtained by performing iterative training on an image classification model based on a bearing signal training data set; the bearing signal training data set is obtained by performing matrix transformation processing on the signal data.
2. The bearing fault diagnosis method according to claim 1, wherein the matrix-transforming the signal data to obtain target data comprises:
converting the signal data to a two-dimensional matrix of a time-frequency domain;
acquiring time domain information and frequency domain information in the two-dimensional matrix;
and integrating the time domain information and the frequency domain information to obtain target data.
3. A bearing fault diagnostic method as claimed in claim 2 wherein said converting said signal data to a two-dimensional matrix form comprises:
detecting a frequency of the signal data;
and selecting a Gaussian window with the length and the frequency resolution both suitable for the frequency to extract information on different frequency bands and extract frequency components of the signal data at each time point in a preset duration, wherein the information on the frequency bands and the frequency components are time domain information and frequency domain information.
4. The bearing fault diagnosis method according to claim 2, wherein said integrating the time domain information and the frequency domain information to obtain target data comprises:
establishing a time-frequency domain with time, frequency and amplitude as coordinate axes based on the time domain information and the frequency domain information, and drawing image information in the time-frequency domain;
and selecting the projection of the image information in a plane formed by time and frequency coordinate axes to obtain target data.
5. The bearing fault diagnostic method of claim 1, wherein said inputting said target data prior to a diagnostic model comprises:
acquiring training samples when different faults occur to the bearing;
inputting the training sample to an image classification model, and obtaining a diagnosis model after the image classification model finishes training; the image classification model is a neural network model for image classification.
6. The bearing fault diagnosis method according to claim 5, wherein the training samples are input to an image classification model, and after the image classification model is trained, a diagnosis model is obtained; the image classification model is a neural network model for image classification, and comprises the following steps:
inputting the training samples to an image classification model, and classifying the training samples to obtain training classification labels;
calculating the gradient of the image classification model based on the training classification label and a preset real label corresponding to the training sample;
determining whether the image classification model meets a preset iteration training end condition or not based on the gradient;
if so, taking the image classification model as a diagnosis model;
if not, continuing to carry out iterative training on the image classification model until the image classification model meets a preset iterative training end condition.
7. The method for diagnosing a bearing fault of claim 1, wherein the inputting the training samples into an image classification model, after the image classification model is trained and a diagnostic model is obtained, comprises:
obtaining training samples of bearings of other models;
and taking the diagnosis model as an initial model, and training the initial model by using the training samples of the bearings of other models to obtain a target model for diagnosing the bearings of different models.
8. A bearing fault diagnosis apparatus characterized by comprising:
the acquisition module is used for acquiring signal data of the bearing in the operation process;
the processing module is used for carrying out matrix transformation processing on the signal data to obtain target data;
the input module is used for inputting the target data to a diagnosis model to obtain a classification result;
the determining module is used for determining the fault type of the bearing based on the classification result; the diagnostic model is obtained by performing iterative training on an image classification model based on a bearing signal training data set; the bearing signal training data set is obtained by performing matrix transformation processing on the signal data.
9. A bearing fault diagnosis apparatus characterized by comprising: a memory, a processor, and a bearing fault diagnosis program stored on the memory and executable on the processor, the bearing fault diagnosis program configured to implement the steps of the bearing fault diagnosis method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that a bearing fault diagnosis program is stored thereon, which when executed by a processor implements the steps of the bearing fault diagnosis method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211224305.1A CN115358277A (en) | 2022-10-09 | 2022-10-09 | Bearing fault diagnosis method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211224305.1A CN115358277A (en) | 2022-10-09 | 2022-10-09 | Bearing fault diagnosis method, device, equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115358277A true CN115358277A (en) | 2022-11-18 |
Family
ID=84008675
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211224305.1A Pending CN115358277A (en) | 2022-10-09 | 2022-10-09 | Bearing fault diagnosis method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115358277A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109632309A (en) * | 2019-01-17 | 2019-04-16 | 燕山大学 | Based on the rolling bearing fault intelligent diagnosing method for improving S-transformation and deep learning |
| CN111076934A (en) * | 2019-12-24 | 2020-04-28 | 江苏大学 | Method for diagnosing potential fault of bearing based on S transformation |
| CN112304611A (en) * | 2020-09-14 | 2021-02-02 | 江苏大学 | Deep learning-based bearing fault diagnosis method |
| CN113567131A (en) * | 2021-07-29 | 2021-10-29 | 沈阳建筑大学 | Bearing fault diagnosis method based on S transformation and miniature convolution neural network model |
-
2022
- 2022-10-09 CN CN202211224305.1A patent/CN115358277A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109632309A (en) * | 2019-01-17 | 2019-04-16 | 燕山大学 | Based on the rolling bearing fault intelligent diagnosing method for improving S-transformation and deep learning |
| CN111076934A (en) * | 2019-12-24 | 2020-04-28 | 江苏大学 | Method for diagnosing potential fault of bearing based on S transformation |
| CN112304611A (en) * | 2020-09-14 | 2021-02-02 | 江苏大学 | Deep learning-based bearing fault diagnosis method |
| CN113567131A (en) * | 2021-07-29 | 2021-10-29 | 沈阳建筑大学 | Bearing fault diagnosis method based on S transformation and miniature convolution neural network model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20230351177A1 (en) | Enhanced discriminate feature learning deep residual cnn for multi-task rotating machinery fault diagnosis with information fusion | |
| CN109858352B (en) | Fault diagnosis method based on compressed sensing and improved multi-scale network | |
| CN112534470A (en) | System and method for image-based inspection of target objects | |
| US11435719B2 (en) | System and method for identifying manufacturing defects | |
| CN112465049A (en) | Method and device for generating anomaly detection model and method and device for detecting anomaly event | |
| KR102416474B1 (en) | Fault diagnosis apparatus and method based on machine-learning | |
| CN112729835B (en) | Multi-domain semi-supervised fault diagnosis method and device for axial plunger pump bearing | |
| CN114170184A (en) | Product image anomaly detection method and device based on embedded feature vector | |
| CN110263730B (en) | Image recognition method and device, electronic equipment and storage medium | |
| CN114862838A (en) | Unsupervised learning-based defect detection method and equipment | |
| CN110553839A (en) | Single and composite fault diagnosis method, equipment and system for gearbox | |
| Wang et al. | Construction of the efficient attention prototypical net based on the time–frequency characterization of vibration signals under noisy small sample | |
| CN114861718B (en) | Bearing fault diagnosis method and system based on improved depth residual error algorithm | |
| CN113865872A (en) | Bearing fault diagnosis method based on wavelet packet reconstruction imaging and CNN | |
| CN116595313A (en) | Fault diagnosis method and system based on multi-scale residual error attention network | |
| CN112434717B (en) | Model training method and device | |
| CN110750876B (en) | Bearing data model training and using method | |
| CN115358277A (en) | Bearing fault diagnosis method, device, equipment and readable storage medium | |
| CN116625654A (en) | Vibration signal fault diagnosis method, device, equipment and storage medium | |
| CN116399592A (en) | Bearing fault diagnosis method based on channel attention dual-path feature extraction | |
| CN116380466A (en) | Rolling bearing intelligent fault diagnosis method and system based on enhanced event visual data | |
| CN116541771A (en) | Unbalanced sample bearing fault diagnosis method based on multi-scale feature fusion | |
| Hao et al. | New fusion features convolutional neural network with high generalization ability on rolling bearing fault diagnosis | |
| CN113792770A (en) | Zero-sample rolling bearing fault diagnosis method and system based on attribute description | |
| CN111753981A (en) | Reuse of neural network training for higher dimensional data |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20221118 |