CN116127383A - Fault detection method and device, electronic equipment and storage medium - Google Patents
Fault detection method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN116127383A CN116127383A CN202211275239.0A CN202211275239A CN116127383A CN 116127383 A CN116127383 A CN 116127383A CN 202211275239 A CN202211275239 A CN 202211275239A CN 116127383 A CN116127383 A CN 116127383A
- Authority
- CN
- China
- Prior art keywords
- layer
- operation data
- fault detection
- convolution
- data
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 111
- 238000003860 storage Methods 0.000 title claims abstract description 17
- 238000000605 extraction Methods 0.000 claims abstract description 53
- 230000007246 mechanism Effects 0.000 claims abstract description 34
- 238000000034 method Methods 0.000 claims abstract description 29
- 230000005540 biological transmission Effects 0.000 claims abstract description 13
- 239000011159 matrix material Substances 0.000 claims description 37
- 238000012549 training Methods 0.000 claims description 35
- 238000011176 pooling Methods 0.000 claims description 25
- 238000012360 testing method Methods 0.000 claims description 24
- 230000004927 fusion Effects 0.000 claims description 19
- 238000004364 calculation method Methods 0.000 claims description 14
- 239000013598 vector Substances 0.000 claims description 13
- 230000003321 amplification Effects 0.000 claims description 5
- 238000007477 logistic regression Methods 0.000 claims description 5
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 5
- 238000012512 characterization method Methods 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 4
- 230000002401 inhibitory effect Effects 0.000 claims description 4
- 238000003672 processing method Methods 0.000 claims description 4
- 238000004891 communication Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 8
- 238000013527 convolutional neural network Methods 0.000 description 7
- 238000012545 processing Methods 0.000 description 6
- 238000009826 distribution Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000005484 gravity Effects 0.000 description 3
- 238000009776 industrial production Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000005012 migration Effects 0.000 description 2
- 238000013508 migration Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013501 data transformation Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000005469 granulation Methods 0.000 description 1
- 230000003179 granulation Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Debugging And Monitoring (AREA)
- Accessory Devices And Overall Control Thereof (AREA)
- Control Or Security For Electrophotography (AREA)
Abstract
The application provides a fault detection method, a fault detection device, electronic equipment and a storage medium, wherein the method comprises the steps of obtaining operation data corresponding to a plurality of detection points of target equipment; inputting a plurality of operation data into a pre-trained fault detection model, wherein the fault detection model at least comprises an input layer, a multi-head attention mechanism layer, a plurality of feature extraction layers, a full connection layer and an output layer according to the transmission sequence of the operation data; calculating the weight of each operation data through the multi-head attention mechanism layer, and fusing a plurality of operation data based on the weight; and the plurality of feature extraction layers and the full connection layer are used for carrying out feature extraction and classification on the fused operation data, and the fault probability of the target equipment is output through the output layer, so that the technical problem of excessive model parameters of the fault detection model in the prior art is solved, and the complexity of the fault detection model is reduced.
Description
Technical Field
The present disclosure relates to the field of fault detection technologies, and in particular, to a fault detection method, a fault detection device, an electronic device, and a storage medium.
Background
With the continuous progress of modern science and the rapid development of economy, the social production level is greatly improved, the industrial production field becomes the productivity main force, key equipment of the industrial production field mainly comprises rotating machinery such as motors, engines, transmission shafts and the like, and if the industrial production field fails but cannot be found and treated in time, economic loss is caused, and even casualties are caused.
The rotary mechanical equipment also develops towards comprehensive complicacy, the relevance coupling between equipment modules is stronger and stronger, the fault diagnosis of single equipment based on fault mechanism analysis is not suitable for complex large-scale machinery any more, and the fault detection model adopted in the prior art has the defect of too many calculation parameters when the fault diagnosis of the equipment is carried out, so that the complexity of the fault detection model is increased, and the calculation engineering quantity of the fault detection model is huge.
Disclosure of Invention
In view of the foregoing, an object of the present application is to provide a fault detection method, a fault detection device, an electronic device and a storage medium, so as to overcome all or part of the defects in the prior art.
Based on the above object, the present application provides a fault detection method, including: acquiring operation data corresponding to a plurality of detection points of target equipment; inputting a plurality of operation data into a pre-trained fault detection model, wherein the fault detection model at least comprises an input layer, a multi-head attention mechanism layer, a plurality of feature extraction layers, a full connection layer and an output layer according to the transmission sequence of the operation data; calculating the weight of each operation data through the multi-head attention mechanism layer, and fusing a plurality of operation data based on the weight; and performing feature extraction and classification on the fused operation data through a plurality of feature extraction layers and the full connection layer, and outputting the fault probability of the target equipment through the output layer.
Optionally, the fault detection model is a WDCNN model, and the feature extraction layer sequentially includes a convolution layer and a pooling layer.
Optionally, calculating, by the multi-head attention mechanism layer, a weight of each of the operation data, and fusing a plurality of operation data based on the weights, including: multiplying the operation data with the corresponding weight value to obtain a product value; combining a plurality of the product values into a product value matrix; and performing linear transformation operation on the product value matrix to obtain fused operation data.
Optionally, according to the transmission sequence of the operation data, the first feature extraction layer includes a wide convolution layer and a pooling layer, the other feature extraction layers include a narrow convolution layer and a pooling layer, feature extraction and classification are performed on the operation data after fusion by the plurality of feature extraction layers and the full connection layer, and fault probability of the target device is output via the output layer, including: short-time feature extraction operation is carried out on the fused operation data through the wide convolution layer so as to obtain a feature map, wherein the short-time feature extraction is operation of inhibiting high-frequency noise from extracting middle-low frequency signals; transmitting the feature images in other feature extraction layers in sequence to obtain a target feature image, wherein a convolution operation is performed on the current feature image through each narrow convolution layer, and a feature amplification operation is performed on the current feature image through each pooling layer; and performing logistic regression calculation on the target feature map through the full connection layer to obtain the fault probability of the target equipment, and outputting the fault probability through the output layer.
Optionally, combining a plurality of the product values into a product value matrix includes: the product value matrix is obtained by the following calculation formula:
U 1 ＝QK T ,(batch,n Q ,n K )
U 3 ＝U 2 .masked_fill(mask,-∞)
A＝softmax(U 3 )
output＝AV
wherein ,U1 Is matrix one, U 2 U as matrix two 3 The three-dimensional vector data processing method comprises the steps of taking a matrix three, n as the number of characterization values in a vector, d as the dimension of data of the vector, Q as a query fault characteristic parameter value, K as a key value of operation data, V as the operation data, A as a weight matrix and output as the product value matrix.
Optionally, the method further comprises: performing a convolution operation on the current feature map by each narrow convolution layer to obtain an intermediate feature map, obtaining the intermediate feature map by the following formula,
wherein ,/>
and bL j Respectively representing the weight and offset of the jth convolution kernel in the L-layer narrow convolution layer, wherein J is the number of convolution kernels of the L-layer narrow convolution layer, s is the step length, k is the size of the convolution kernels of the L-layer narrow convolution layer, Q is the number of convolution kernels of the L-1-layer narrow convolution layer, and%>
Intermediate feature map of size M of the (q) th output of the L-1 th narrow convolution layer,/th narrow convolution layer>
And (3) representing a j-th intermediate feature map with the size i output by the L-th narrow convolution layer, wherein i is the size of the intermediate feature map, and p is the padding size.
Optionally, the pre-training method of the fault detection model includes: acquiring historical fault data of equipment; generating training samples using an countermeasure network based on the historical fault data; randomly dividing the training sample into training set data and test set data according to a preset proportion; initializing the weight of the fault detection model, and performing iterative training on the fault detection model by adopting the training set data; and testing the fault detection model subjected to iterative training based on the test set data to obtain test precision, and obtaining the fault detection model subjected to training when the test precision is greater than a preset precision.
Based on the same inventive concept, the application also provides a fault detection device, which comprises an acquisition module configured to acquire operation data corresponding to a plurality of detection points of target equipment; the input module is configured to input a plurality of operation data into a pre-trained fault detection model, and the fault detection model at least comprises an input layer, a multi-head attention mechanism layer, a plurality of convolution layers, a plurality of pooling layers, a full connection layer and an output layer according to the transmission sequence of the operation data; the fusion module is configured to calculate the weight of each piece of operation data through the multi-head attention mechanism layer, and fuse a plurality of pieces of operation data based on the weight; and the output module is configured to extract and classify the characteristics of the fused operation data through a plurality of convolution layers, a plurality of pooling layers and the full connection layer, and output the fault probability of the target equipment through the output layer.
Based on the same inventive concept, the application also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable by the processor, the processor implementing the method as described above when executing the computer program.
Based on the same inventive concept, the present application also provides a non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the method as described above.
From the above, it can be seen that the fault detection method, device, electronic device and storage medium provided by the present application achieve multiple sources of detection data by acquiring the operation data corresponding to multiple detection points of the target device, and reduce inaccuracy caused by single detection data; inputting a plurality of operation data into a pre-trained fault detection model, wherein the fault detection model comprises a multi-head attention mechanism layer, the multi-head attention mechanism layer is used for calculating the weight of each operation data, and the plurality of operation data are fused based on the weight, so that the fusion of the multi-source operation data is realized; and the plurality of feature extraction layers and the full connection layer are used for carrying out feature extraction and classification on the fused operation data, so that the problem of excessive model parameters of the fault detection model is reduced, and the complexity of the fault detection model is reduced.
Drawings
In order to more clearly illustrate the technical solutions of the present application or related art, the drawings that are required to be used in the description of the embodiments or related art will be briefly described below, and it is apparent that the drawings in the following description are only embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort to those of ordinary skill in the art.
Fig. 1 is a schematic flow chart of a fault detection method according to an embodiment of the present application;
FIG. 2 is a schematic diagram of a multi-headed attention mechanism according to an embodiment of the present application;
fig. 3 is a schematic hierarchical structure of a WDCNN model according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of a fault detection device according to an embodiment of the present application;
fig. 5 is a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present application.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail below with reference to the accompanying drawings.
It should be noted that unless otherwise defined, technical or scientific terms used in the embodiments of the present application should be given the ordinary meaning as understood by one of ordinary skill in the art to which the present application belongs. The word "comprising" or "comprises", and the like, means that elements or items preceding the word are included in the element or item listed after the word and equivalents thereof, but does not exclude other elements or items. The terms "connected" or "connected," and the like, are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect.
As described in the background section, the existing fault detection model includes a multi-scale convolutional neural network model, which is a model for performing multi-source information fault detection decision fusion, and includes a plurality of parallel 1DCNN branch networks, each branch samples an input signal in different scales through an average pooling layer, and takes a time sequence after sampling as an input of each branch, and then fuses features extracted by each branch to perform final fault recognition. The method mainly comprises the steps of performing coarse granulation processing on input, namely performing smooth filtering and downsampling operation on an original signal, extracting abstract features from each scale signal through two alternating convolution layers and pooling layers, performing feature learning processes on different scale signals in parallel, cascading and connecting finally obtained feature vectors, judging through a full-connection layer and a softmax layer, and outputting corresponding probability information. However, the multi-scale convolutional neural network model has the following disadvantages: 1) Decision fusion is carried out after data feature extraction, the feature extraction also needs to depend on a plurality of parallel networks, the calculated parameter quantity is increased, and the model is more complex; 2) Aiming at the same input time sequence, coarse-grained splitting is carried out on the time sequence, other sensor information is not utilized, a data set needs to be constructed if the information is required to be comprehensively acquired, and the engineering quantity is huge; 3) The distribution condition of the original signals is different for different equipment or different working conditions, but a large number of parameters of the multi-scale convolutional neural network model are unchanged, and the characteristic extraction by using the model is possibly not adapted, so the model does not have good mobility.
In view of this, the embodiment of the present application proposes a fault detection method, referring to fig. 1, including the following steps:
In the step, a sensor is respectively arranged at each of a plurality of detection points of the target equipment for collecting operation data of the target equipment, wherein the operation data comprise but are not limited to vibration signals of the target equipment and operation time signals of the target equipment, the sensors facilitate data collection and storage, the sensors are deployed at the plurality of detection points, the state detection of the target equipment is realized by utilizing the plurality of sensors at the same time, comprehensive operation information is obtained, the problem of singleness of the operation data collected in the prior art is solved, the problems of the target equipment are more comprehensively reflected, and the accuracy of fault detection is improved.
In the step, the input layer of the fault detection model comprises at least one channel, each channel is mutually independent, operation data from different detection points are transmitted, and the response sensitivity degree of the different detection points to faults is different and the correlation between the different detection points to the same faults is considered, so that the fault detection model introduces a multi-head attention mechanism layer, the independent distribution of the weight of the operation data with larger correlation with the faults of target equipment is realized, and the precision of the fault detection model is improved.
The multi-head attention mechanism layer is used for learning different behaviors based on the same attention mechanism when the same query, key and value set are given, combining different behaviors as knowledge, capturing various range dependency relations in a sequence, for example, taking kesixi data storage as an example, detecting the fault of the inner ring of the driving end, wherein the data of the 3 different position sensors comprise a base end, the driving end and a fan end, the number of channels established at the moment is 3, and the multi-head attention is set to be 3-head attention.
And step 103, calculating the weight of each operation data through the multi-head attention mechanism layer, and fusing a plurality of operation data based on the weight.
In the step, the multi-head attention mechanism layer can perform autonomous learning, the sensitivity degree and the dependence relation of each channel information in the input layer on faults are obtained, different weight values are given to each channel, the data before feature extraction is fused through the weights, the autonomous learning mechanism of the multi-head attention mechanism layer has the advantages of no need of manual setting, reduction of parameters of a fault detection model, and improvement of the learning efficiency and the fault detection efficiency of the fault detection model.
And 104, performing feature extraction and classification on the fused operation data through a plurality of feature extraction layers and the full connection layer, and outputting the fault probability of the target equipment through the output layer.
In the step, the feature extraction layer of the fault detection model has the functions of transforming the fused operation data and highlighting representative features, and the required features and the amplified features are extracted through data analysis and transformation; each node of the full-connection layer of the fault detection model is connected with all nodes of the upper layer, features extracted from the front edge are integrated, and accuracy of the fault detection model is improved.
According to the scheme, the multi-source data fusion fault detection method for end-to-end processing is realized, the self-adaptive information fusion is realized through the fault detection model with the multi-head attention mechanism layer aiming at the operation data of equipment acquired by different sensors, the operation by a professional is not needed, the parameter quantity of the fault detection model is reduced, compared with the parallel feature extraction of a multi-scale convolutional neural network model, the calculation quantity of the model is reduced, and the fusion of the fault detection model at a data level is reduced in complexity of the model compared with the fusion at a feature level.
In some embodiments, in step 102, the fault detection model includes at least an input layer, a multi-head attention mechanism layer, a plurality of feature extraction layers, a full connection layer, and an output layer according to the transmission order of the operation data, and specifically includes:
in step 1021, the fault detection model is a WDCNN model, and the feature extraction layer sequentially includes a convolution layer and a pooling layer.
In the scheme, the WDCNN model is a first-layer wide convolution network model, the first-layer convolution layer is a wide convolution layer, required abnormal fluctuation is intercepted and captured into a convolution window, the function of the WDCNN model is similar to that of short-time Fourier transform, and the purpose of the WDCNN model is to extract short-time characteristics.
In some embodiments, in step 103, the calculating, by the multi-head attention mechanism layer, the weight of each of the operation data, and fusing the plurality of operation data based on the weight specifically includes:
step 1031, multiplying the operation data with the corresponding weight value to obtain a product value.
Step 1032, combining a plurality of said product values into a product value matrix.
And 1033, performing linear transformation operation on the product value matrix to obtain fused operation data.
In the above scheme, the attention value calculation is performed on the operation data, the weight is determined, and the outputs of multiple heads are coupled according to the weight to obtain the product value, wherein the multiple heads can obtain multiple outputs. As shown in fig. 2, fig. 2 is a schematic diagram of a multi-head attention mechanism in the embodiment of the present application, the values of the operation data Q, K, V are respectively input to the Linear layer to obtain a plurality of weight values, the operation data are multiplied by the weight values in the SDA layer to obtain product values, the plurality of outputs are coupled through the concat layer of the multi-head attention mechanism layer to obtain a product value matrix, and the plurality of product value matrices are converted into the fused operation data through the Linear layer of the multi-head attention mechanism layer according to specific gravity, so that adaptive multi-source information fusion can be realized, specific gravity parameters are determined through autonomous learning, manual setting is not required, and model parameters are reduced.
In some embodiments, in step 102, according to the transmission order of the operation data, the first feature extraction layer includes a wide convolution layer and a pooling layer, the other feature extraction layers include a narrow convolution layer and a pooling layer, feature extraction and classification are performed on the fused operation data through a plurality of feature extraction layers and the full connection layer, and the fault probability of the target device is output through the output layer, which specifically includes:
and step 1022, performing short-time feature extraction operation on the fused operation data through the wide convolution layer to obtain a feature map, wherein the short-time feature extraction is an operation of suppressing low-frequency signals in high-frequency noise extraction.
Step 1023, transmitting the feature map in other feature extraction layers in turn to obtain a target feature map, wherein a convolution operation is performed on the current feature map through each narrow convolution layer, and a feature amplification operation is performed on the current feature map through each pooling layer.
Step 1024, performing logistic regression calculation on the target feature map through the full connection layer to obtain the fault probability of the target device, and outputting the fault probability through the output layer.
In the above scheme, as shown in fig. 3, fig. 3 is a schematic diagram of a hierarchical structure of a WDCNN model in an embodiment of the present application, where fused operation data is input into a wide convolution layer, short-time features are extracted by using a wide window, and feature amplification is performed by using a maximum pooling effect of a pooling layer, so that a data dimension of a feature map is reduced; and then slowly analyzing deep semantic features of the feature map by using a narrow convolution layer, inhibiting high-frequency noise from extracting medium-low frequency signals, integrating the feature map at a full connection layer after passing through a plurality of narrow convolution layers and a maximum pooling layer, and outputting a corresponding fault probability result through logistic regression calculation.
In some embodiments, in step 103, combining a plurality of the product values into a product value matrix specifically includes:
step 1034, obtaining the product value matrix by the following calculation formula:
U 1 ＝QK T ,(batch,n Q ,n K )
U 3 ＝U 2 .masked_fill(mask,-∞)
A＝softmax(U 3 )
output＝AV
wherein ,U1 Is matrix one, U 2 U as matrix two 3 The three-dimensional vector data processing method comprises the steps of taking a matrix three, n as the number of characterization values in a vector, d as the dimension of data of the vector, Q as a query fault characteristic parameter value, K as a key value of operation data, V as the operation data, A as a weight matrix and output as the product value matrix.
In the scheme, the product value matrix is calculated through a specific formula, and the abstract numerical value is embodied, so that the fault detection model has more fault detection accuracy.
In some embodiments, step 102 specifically includes:
step 1025, performing convolution operation on the current feature map by each narrow convolution layer to obtain an intermediate feature map, obtaining the intermediate feature map by the following formula,
wherein ,/>
and />
Respectively representing the weight and offset of the jth convolution kernel in the L-layer narrow convolution layer, wherein J is the number of convolution kernels of the L-layer narrow convolution layer, s is the step length, k is the size of the convolution kernels of the L-layer narrow convolution layer, Q is the number of convolution kernels of the L-1-layer narrow convolution layer, and%>
Intermediate feature map of size M of the (q) th output of the L-1 th narrow convolution layer,/th narrow convolution layer>
And (3) representing a j-th intermediate feature map with the size i output by the L-th narrow convolution layer, wherein i is the size of the intermediate feature map, and p is the padding size.
In the scheme, the middle characteristic diagram of the narrow convolution layer is determined through a specific formula, abstract numerical values are embodied, so that the fault detection model has more fault detection accuracy, and it is required to be noted that the wide convolution layer in the first characteristic extraction layer is calculated through the following formula,
wherein ,/>
An intermediate feature map with a size x, which represents the output of the convolution layer of the first wide convolution layer, y is the number of convolution kernels of the first wide convolution layer, z is the size of the convolution kernels of the first wide convolution layer, 1 represents the first wide convolution layer, and T is the runtime signal data of the target device>
and />
The weight and bias of the first convolution kernel in the first wide convolution layer are represented, respectively.
In some embodiments, the pre-training method of the fault detection model specifically includes:
step 1051, historical failure data for the device is obtained.
At step 1052, training samples are generated using the countermeasure network based on the historical fault data.
Step 1053, dividing the training sample into training set data and testing set data according to a preset proportion.
Step 1054, initializing the weight of the fault detection model, and performing iterative training on the fault detection model by adopting the training set data.
And 1055, testing the fault detection model subjected to iterative training based on the test set data to obtain test precision, and obtaining the fault detection model subjected to training when the test precision is greater than a preset precision.
In the above scheme, the data set is constructed through the historical fault data of the equipment, and in the case that the equipment fault information occupies a relatively small amount, the data set with normal proportion is constructed through generating the countermeasure network, so that the following steps of 8:2, performing iterative training on the model by using a training sample, performing parameter adjustment, testing the model by using test set data to obtain test precision, and obtaining a trained fault detection model under the condition that the test precision is more than or equal to preset precision.
In another embodiment provided herein, data-related information from multiple sensors is combined for coordinated optimization and comprehensive processing. The multi-channel characteristic of the convolutional neural network model is utilized, the weight of the information with larger relevance is automatically distributed by combining a multi-head attention mechanism, the sensitivity degree and the dependence relationship of each channel information in the input layer to faults are automatically learned, the channels sensitive to the faults are highlighted while the multi-sensor is cooperatively identified, and the learning efficiency of the model is improved. And extracting short-time features of the fusion information according to specific gravity by using a WDCNN model, and carrying out batch normalization by using an AdaBN algorithm to increase generalization of the network so as to cope with different distribution conditions of the target domain under different working conditions. The WDCNN model performs model migration by using the AdaBN algorithm, has generalization compared with a multi-scale convolutional neural network model, and can achieve cross-domain migration by replacing the mean and variance of the target domain with the source domain, for example, the distribution of data of monitored equipment may be different under different loads or different rotating speeds, and the generalization capability of the model is particularly important when dealing with the situation.
It should be noted that, the method of the embodiments of the present application may be performed by a single device, for example, a computer or a server. The method of the embodiment can also be applied to a distributed scene, and is completed by mutually matching a plurality of devices. In the case of such a distributed scenario, one of the devices may perform only one or more steps of the methods of embodiments of the present application, and the devices may interact with each other to complete the methods.
It should be noted that some embodiments of the present application are described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments described above and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
Based on the same inventive concept, the application also provides a fault detection device corresponding to the method of any embodiment.
Referring to fig. 4, the fault detection apparatus includes:
the acquisition module 10 is configured to acquire operation data corresponding to a plurality of detection points of the target device.
The input module 20 is configured to input a plurality of operation data into a pre-trained fault detection model, wherein the fault detection model at least comprises an input layer, a multi-head attention mechanism layer, a plurality of convolution layers, a plurality of pooling layers, a full connection layer and an output layer according to the transmission sequence of the operation data.
A fusion module 30 configured to calculate a weight of each of the operation data by the multi-head attention mechanism layer, and fuse a plurality of operation data based on the weights.
An output module 40, configured to perform feature extraction and classification on the fused operation data through a plurality of convolution layers, a plurality of pooling layers, and the full connection layer, and output the failure probability of the target device via the output layer.
According to the device provided by the application, the multi-source information fusion fault detection device for end-to-end processing can be realized, the self-adaptive information fusion can be realized through the fault detection model with the multi-head attention mechanism layer aiming at the equipment operation data acquired by different sensors, the operation by a professional is not needed, the parameter of the fault detection model is reduced, the parallel feature extraction of the multi-scale convolutional neural network model is realized, the calculation amount of the model is reduced, and the complexity of the model is reduced in the data level fusion compared with the feature and fusion.
In some embodiments, the fault detection model is a WDCNN model and the feature extraction layer comprises a convolution layer and a pooling layer in order.
In some embodiments, the fusion module 30 is further configured to: multiplying the operation data with the corresponding weight value to obtain a product value; combining a plurality of the product values into a product value matrix; and performing linear transformation operation on the product value matrix to obtain fused operation data.
In some embodiments, the output module 40 is further configured to: short-time feature extraction operation is carried out on the fused operation data through the wide convolution layer so as to obtain a feature map, wherein the short-time feature extraction is operation of inhibiting high-frequency noise from extracting middle-low frequency signals; transmitting the feature images in other feature extraction layers in sequence to obtain a target feature image, wherein a convolution operation is performed on the current feature image through each narrow convolution layer, and a feature amplification operation is performed on the current feature image through each pooling layer; and performing logistic regression calculation on the target feature map through the full connection layer to obtain the fault probability of the target equipment, and outputting the fault probability through the output layer.
In some embodiments, the fusion module 30 includes a merge unit configured to: the product value matrix is obtained by the following calculation formula:
U 1 ＝QK T ,(batch,n Q ,n K )
U 3 ＝U 2 .masked_fill(mask,-∞)
A＝softmax(U 3 )
output＝AV
wherein ,U1 Is matrix one, U 2 U as matrix two 3 The three-dimensional vector data processing method comprises the steps of taking a matrix three, n as the number of characterization values in a vector, d as the dimension of data of the vector, Q as a query fault characteristic parameter value, K as a key value of operation data, V as the operation data, A as a weight matrix and output as the product value matrix.
In some embodiments, the output module 40 includes an operating unit configured to:
wherein ,/>
and bL j Respectively representing the weight and offset of the jth convolution kernel in the L-layer narrow convolution layer, wherein J is the number of convolution kernels of the L-layer narrow convolution layer, s is the step length, k is the size of the convolution kernels of the L-layer narrow convolution layer, Q is the number of convolution kernels of the L-1-layer narrow convolution layer, and%>
Intermediate feature map of size M of the (q) th output of the L-1 th narrow convolution layer,/th narrow convolution layer>
And (3) representing a j-th intermediate feature map with the size i output by the L-th narrow convolution layer, wherein i is the size of the intermediate feature map, and p is the padding size.
In some embodiments, further comprising a training module 50, the training module 50 further configured to: acquiring historical fault data of equipment; generating training samples using an countermeasure network based on the historical fault data; randomly dividing the training sample into training set data and test set data according to a preset proportion; initializing the weight of the fault detection model, and performing iterative training on the fault detection model by adopting the training set data; and testing the fault detection model subjected to iterative training based on the test set data to obtain test precision, and obtaining the fault detection model subjected to training when the test precision is greater than a preset precision.
For convenience of description, the above devices are described as being functionally divided into various modules, respectively. Of course, the functions of each module may be implemented in the same piece or pieces of software and/or hardware when implementing the present application.
The device of the foregoing embodiment is configured to implement the corresponding fault detection method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiment, which is not described herein.
Based on the same inventive concept, the application also provides an electronic device corresponding to the method of any embodiment, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor implements the fault detection method of any embodiment when executing the program.
Fig. 5 shows a more specific hardware architecture of an electronic device according to this embodiment, where the device may include: a processor 1010, a memory 1020, an input/output interface 1030, a communication interface 1040, and a bus 1050. Wherein processor 1010, memory 1020, input/output interface 1030, and communication interface 1040 implement communication connections therebetween within the device via a bus 1050.
The processor 1010 may be implemented by a general-purpose CPU (Central Processing Unit ), microprocessor, application specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, etc. for executing relevant programs to implement the technical solutions provided in the embodiments of the present disclosure.
The Memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory ), static storage device, dynamic storage device, or the like. Memory 1020 may store an operating system and other application programs, and when the embodiments of the present specification are implemented in software or firmware, the associated program code is stored in memory 1020 and executed by processor 1010.
The input/output interface 1030 is used to connect with an input/output module for inputting and outputting information. The input/output module may be configured as a component in a device (not shown) or may be external to the device to provide corresponding functionality. Wherein the input devices may include a keyboard, mouse, touch screen, microphone, various types of sensors, etc., and the output devices may include a display, speaker, vibrator, indicator lights, etc.
It should be noted that although the above-described device only shows processor 1010, memory 1020, input/output interface 1030, communication interface 1040, and bus 1050, in an implementation, the device may include other components necessary to achieve proper operation. Furthermore, it will be understood by those skilled in the art that the above-described apparatus may include only the components necessary to implement the embodiments of the present description, and not all the components shown in the drawings.
The electronic device of the foregoing embodiment is configured to implement the corresponding fault detection method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiment, which is not described herein.
Based on the same inventive concept, corresponding to any of the above-described embodiments of the method, the present application further provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the fault detection method according to any of the above-described embodiments.
The computer readable media of the present embodiments, including both permanent and non-permanent, removable and non-removable media, may be used to implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device.
The storage medium of the foregoing embodiments stores computer instructions for causing the computer to perform the fault detection method according to any one of the foregoing embodiments, and has the advantages of the corresponding method embodiments, which are not described herein.
Those of ordinary skill in the art will appreciate that: the discussion of any of the embodiments above is merely exemplary and is not intended to suggest that the scope of the application (including the claims) is limited to these examples; the technical features of the above embodiments or in the different embodiments may also be combined within the idea of the present application, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of brevity.
Additionally, well-known power/ground connections to Integrated Circuit (IC) chips and other components may or may not be shown within the provided figures, in order to simplify the illustration and discussion, and so as not to obscure the embodiments of the present application. Furthermore, the devices may be shown in block diagram form in order to avoid obscuring the embodiments of the present application, and this also takes into account the fact that specifics with respect to implementation of such block diagram devices are highly dependent upon the platform on which the embodiments of the present application are to be implemented (i.e., such specifics should be well within purview of one skilled in the art). Where specific details (e.g., circuits) are set forth in order to describe example embodiments of the application, it should be apparent to one skilled in the art that embodiments of the application can be practiced without, or with variation of, these specific details. Accordingly, the description is to be regarded as illustrative in nature and not as restrictive.
While the present application has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of those embodiments will be apparent to those skilled in the art in light of the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.
The present embodiments are intended to embrace all such alternatives, modifications and variances which fall within the broad scope of the appended claims. Accordingly, any omissions, modifications, equivalents, improvements and/or the like which are within the spirit and principles of the embodiments are intended to be included within the scope of the present application.
Claims (10)
1. A fault detection method, comprising:
acquiring operation data corresponding to a plurality of detection points of target equipment;
inputting a plurality of operation data into a pre-trained fault detection model, wherein the fault detection model at least comprises an input layer, a multi-head attention mechanism layer, a plurality of feature extraction layers, a full connection layer and an output layer according to the transmission sequence of the operation data;
calculating the weight of each operation data through the multi-head attention mechanism layer, and fusing a plurality of operation data based on the weight;
and performing feature extraction and classification on the fused operation data through a plurality of feature extraction layers and the full connection layer, and outputting the fault probability of the target equipment through the output layer.
2. The method of claim 1, wherein the fault detection model is a WDCNN model and the feature extraction layer comprises a convolutional layer and a pooling layer in that order.
3. The method of claim 1, wherein calculating, by the multi-headed attention mechanism layer, a weight for each of the operational data, fusing a plurality of operational data based on the weights, comprises:
multiplying the operation data with the corresponding weight value to obtain a product value;
combining a plurality of the product values into a product value matrix;
and performing linear transformation operation on the product value matrix to obtain fused operation data.
4. The method of claim 1, wherein the first feature extraction layer includes a wide convolution layer and a pooling layer in a transmission order of the operation data, the other feature extraction layers include a narrow convolution layer and a pooling layer, the feature extraction and classification of the fused operation data by the plurality of feature extraction layers, the full connection layer, and the output of the failure probability of the target device via the output layer include:
short-time feature extraction operation is carried out on the fused operation data through the wide convolution layer so as to obtain a feature map, wherein the short-time feature extraction is operation of inhibiting high-frequency noise from extracting middle-low frequency signals;
transmitting the feature images in other feature extraction layers in sequence to obtain a target feature image, wherein a convolution operation is performed on the current feature image through each narrow convolution layer, and a feature amplification operation is performed on the current feature image through each pooling layer;
and performing logistic regression calculation on the target feature map through the full connection layer to obtain the fault probability of the target equipment, and outputting the fault probability through the output layer.
5. A method according to claim 3, wherein combining a plurality of said product values into a product value matrix comprises:
the product value matrix is obtained by the following calculation formula:
U 1 ＝QK T ,(batch,n Q ,n K )
U 3 ＝U 2 .masked_fill(mask,-∞)
A＝softmax(U 3 )
output＝AV
wherein ,U1 Is matrix one, U 2 U as matrix two 3 The three-dimensional vector data processing method comprises the steps of taking a matrix three, n as the number of characterization values in a vector, d as the dimension of data of the vector, Q as a query fault characteristic parameter value, K as a key value of operation data, V as the operation data, A as a weight matrix and output as the product value matrix.
6. The method as recited in claim 4, further comprising: performing a convolution operation on the current feature map by each narrow convolution layer to obtain an intermediate feature map, obtaining the intermediate feature map by the following formula,
wherein ,
and bL j Respectively representing the weight and offset of the jth convolution kernel in the L-layer narrow convolution layer, wherein J is the number of convolution kernels of the L-layer narrow convolution layer, s is the step length, k is the size of the convolution kernels of the L-layer narrow convolution layer, Q is the number of convolution kernels of the L-1-layer narrow convolution layer, and%>
Represents layer L-1The q-th middle feature map of size M of the narrow convolution layer output, +.>
And (3) representing a j-th intermediate feature map with the size i output by the L-th narrow convolution layer, wherein i is the size of the intermediate feature map, and p is the padding size.
7. The method of claim 1, wherein the pre-training method of the fault detection model comprises:
acquiring historical fault data of equipment;
generating training samples using an countermeasure network based on the historical fault data;
randomly dividing the training sample into training set data and test set data according to a preset proportion;
initializing the weight of the fault detection model, and performing iterative training on the fault detection model by adopting the training set data;
and testing the fault detection model subjected to iterative training based on the test set data to obtain test precision, and obtaining the fault detection model subjected to training when the test precision is greater than a preset precision.
8. A fault detection device, comprising:
the acquisition module is configured to acquire operation data corresponding to a plurality of detection points of the target equipment;
the input module is configured to input a plurality of operation data into a pre-trained fault detection model, and the fault detection model at least comprises an input layer, a multi-head attention mechanism layer, a plurality of convolution layers, a plurality of pooling layers, a full connection layer and an output layer according to the transmission sequence of the operation data;
the fusion module is configured to calculate the weight of each piece of operation data through the multi-head attention mechanism layer, and fuse a plurality of pieces of operation data based on the weight;
and the output module is configured to extract and classify the characteristics of the fused operation data through a plurality of convolution layers, a plurality of pooling layers and the full connection layer, and output the fault probability of the target equipment through the output layer.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of claims 1 to 7 when the program is executed by the processor.
10. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211275239.0A CN116127383A (en) | 2022-10-18 | 2022-10-18 | Fault detection method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211275239.0A CN116127383A (en) | 2022-10-18 | 2022-10-18 | Fault detection method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116127383A true CN116127383A (en) | 2023-05-16 |
Family
ID=86298062
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211275239.0A Pending CN116127383A (en) | 2022-10-18 | 2022-10-18 | Fault detection method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116127383A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116665711A (en) * | 2023-07-26 | 2023-08-29 | 中国南方电网有限责任公司超高压输电公司广州局 | Gas-insulated switchgear on-line monitoring method and device and computer equipment |
| CN116879761A (en) * | 2023-09-06 | 2023-10-13 | 杭州宇谷科技股份有限公司 | Multi-mode-based battery internal short circuit detection method, system, device and medium |
| CN117036732A (en) * | 2023-08-07 | 2023-11-10 | 深圳市同鑫科技有限公司 | Electromechanical equipment detection system, method and equipment based on fusion model |
| CN117074628A (en) * | 2023-10-17 | 2023-11-17 | 山东鑫建检测技术有限公司 | Multi-sensor air quality detection equipment fault positioning method |
| CN117081666A (en) * | 2023-09-25 | 2023-11-17 | 腾讯科技（深圳）有限公司 | Fault prediction method, device, electronic equipment, storage medium and program product |
-
2022
- 2022-10-18 CN CN202211275239.0A patent/CN116127383A/en active Pending
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116665711A (en) * | 2023-07-26 | 2023-08-29 | 中国南方电网有限责任公司超高压输电公司广州局 | Gas-insulated switchgear on-line monitoring method and device and computer equipment |
| CN116665711B (en) * | 2023-07-26 | 2024-01-12 | 中国南方电网有限责任公司超高压输电公司广州局 | Gas-insulated switchgear on-line monitoring method and device and computer equipment |
| CN117036732A (en) * | 2023-08-07 | 2023-11-10 | 深圳市同鑫科技有限公司 | Electromechanical equipment detection system, method and equipment based on fusion model |
| CN117036732B (en) * | 2023-08-07 | 2024-04-16 | 深圳市同鑫科技有限公司 | Electromechanical equipment detection system, method and equipment based on fusion model |
| CN116879761A (en) * | 2023-09-06 | 2023-10-13 | 杭州宇谷科技股份有限公司 | Multi-mode-based battery internal short circuit detection method, system, device and medium |
| CN117081666A (en) * | 2023-09-25 | 2023-11-17 | 腾讯科技（深圳）有限公司 | Fault prediction method, device, electronic equipment, storage medium and program product |
| CN117081666B (en) * | 2023-09-25 | 2024-01-09 | 腾讯科技（深圳）有限公司 | Fault prediction method, device, electronic equipment, storage medium and program product |
| CN117074628A (en) * | 2023-10-17 | 2023-11-17 | 山东鑫建检测技术有限公司 | Multi-sensor air quality detection equipment fault positioning method |
| CN117074628B (en) * | 2023-10-17 | 2024-01-09 | 山东鑫建检测技术有限公司 | Multi-sensor air quality detection equipment fault positioning method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116127383A (en) | Fault detection method and device, electronic equipment and storage medium | |
| Surendran et al. | Deep learning based intelligent industrial fault diagnosis model. | |
| Jin et al. | New domain adaptation method in shallow and deep layers of the CNN for bearing fault diagnosis under different working conditions | |
| Pandya et al. | Fault diagnosis of rolling element bearing by using multinomial logistic regression and wavelet packet transform | |
| US11715190B2 (en) | Inspection system, image discrimination system, discrimination system, discriminator generation system, and learning data generation device | |
| JP2020095258A (en) | Noise data artificial intelligence apparatus and pre-conditioning method for identifying source of problematic noise | |
| Xing et al. | Intelligent fault diagnosis of rolling bearing based on novel CNN model considering data imbalance | |
| CN117104377B (en) | Intelligent management system and method for electric bicycle | |
| Islam et al. | Motor bearing fault diagnosis using deep convolutional neural networks with 2d analysis of vibration signal | |
| CN117592332B (en) | Digital twinning-based gearbox model high-fidelity method, system and storage medium | |
| Shi et al. | Intelligent fault diagnosis of rolling mills based on dual attention-guided deep learning method under imbalanced data conditions | |
| CN112541524A (en) | BP-Adaboost multi-source information motor fault diagnosis method based on attention mechanism improvement | |
| Xue et al. | A novel framework for motor bearing fault diagnosis based on multi-transformation domain and multi-source data | |
| Yang et al. | A novel multiple feature-based engine knock detection system using sparse Bayesian extreme learning machine | |
| Yan et al. | Rotating machinery fault diagnosis based on a novel lightweight convolutional neural network | |
| Saha et al. | Enhancing bearing fault diagnosis using transfer learning and random forest classification: A comparative study on variable working conditions | |
| CN117237359B (en) | Conveyor belt tearing detection method and device, storage medium and electronic equipment | |
| CN118410375A (en) | Deep integration method of information technology and operation technology | |
| Hao et al. | Research on Fault Diagnosis Method Based on Improved CNN | |
| Bui-Ngoc et al. | Structural health monitoring using handcrafted features and convolution neural network | |
| CN116721097A (en) | Bearing fault diagnosis method and device and electronic equipment | |
| Sonmez et al. | A new deep learning model combining CNN for engine fault diagnosis | |
| Zhang et al. | The combination model of CNN and GCN for machine fault diagnosis | |
| CN115659223A (en) | Rolling bearing fault diagnosis method, device, equipment and medium based on multiple algorithms | |
| Baihaqi et al. | A Comparison Support Vector Machine, Logistic Regression And Naï ve Bayes For Classification Sentimen Analisys user Mobile App |
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 |