CN118132986A - Equipment fault diagnosis method and device and electronic equipment - Google Patents
Equipment fault diagnosis method and device and electronic equipment Download PDFInfo
- Publication number
- CN118132986A CN118132986A CN202410247878.9A CN202410247878A CN118132986A CN 118132986 A CN118132986 A CN 118132986A CN 202410247878 A CN202410247878 A CN 202410247878A CN 118132986 A CN118132986 A CN 118132986A
- Authority
- CN
- China
- Prior art keywords
- features
- local
- frequency domain
- time domain
- vibration 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 86
- 238000000034 method Methods 0.000 title claims abstract description 58
- 230000001419 dependent effect Effects 0.000 claims abstract description 47
- 238000001228 spectrum Methods 0.000 claims abstract description 43
- 230000002159 abnormal effect Effects 0.000 claims abstract description 26
- 238000012545 processing Methods 0.000 claims description 27
- 238000013527 convolutional neural network Methods 0.000 claims description 22
- 238000004590 computer program Methods 0.000 claims description 19
- 238000000605 extraction Methods 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 10
- 238000004140 cleaning Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 7
- 230000003595 spectral effect Effects 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 15
- 238000012549 training Methods 0.000 description 15
- 238000010586 diagram Methods 0.000 description 13
- 230000008569 process Effects 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 4
- 238000005096 rolling process Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000007405 data analysis Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The present application relates to the field of fault diagnosis technologies, and in particular, to a device fault diagnosis method and apparatus, and an electronic device. The method comprises the following steps: obtaining data to be detected, wherein the data to be detected comprises a time domain vibration signal and a frequency domain vibration signal of equipment to be detected. And extracting the spatial characteristics of the time domain vibration signal and the frequency domain vibration signal to obtain a first local time domain characteristic and a first local frequency domain characteristic, splicing the first local time domain characteristic and the first local frequency domain characteristic, and inputting the spliced first local time domain characteristic and the spliced first local frequency domain characteristic into a hierarchical network to obtain a device state diagnosis result. If the equipment state diagnosis result shows that the equipment is abnormal, extracting time sequence features of the first local time domain features and the first local frequency domain features to obtain time sequence dependent features and frequency spectrum semantic features, splicing the time sequence dependent features and the frequency spectrum semantic features, and then inputting the spliced time sequence dependent features and frequency spectrum semantic features into a hierarchical network to determine an equipment fault structure. By the aid of the scheme, the interpretability and the expandability of equipment fault diagnosis can be improved.
Description
Technical Field
The present application relates to the field of fault diagnosis technologies, and in particular, to a device fault diagnosis method and apparatus, and an electronic device.
Background
As natural language processing technology (GENERATIVE PRETRAINED transducer, GPT) of pre-trained models grows, the use of GPT in fault diagnosis scenarios has become a trend. However, there is a problem in that the interpretability is poor and the expandability is poor.
Disclosure of Invention
The embodiment of the application provides a device fault diagnosis method and device and electronic equipment, which are used for improving the interpretability and expandability of device fault diagnosis.
In a first aspect, an embodiment of the present application provides an apparatus fault diagnosis method, where the method includes:
obtaining data to be detected, wherein the data to be detected comprises a time domain vibration signal and a frequency domain vibration signal of equipment to be detected;
Extracting spatial features of a time domain vibration signal and a frequency domain vibration signal to obtain a first local time domain feature and a first local frequency domain feature, splicing the first local time domain feature and the first local frequency domain feature, inputting the first local time domain feature and the first local frequency domain feature into a hierarchical network to obtain a device state diagnosis result, wherein the device state diagnosis result is used for indicating whether the device is abnormal or not, the hierarchical network is trained through a preset self-supervision learning task, the self-supervision learning task is used for limiting the hierarchical network to generate a pseudo tag by utilizing the internal structure and the attribute of sample data, and adjusting based on the difference between the pseudo tag and a prediction result;
if the equipment state diagnosis result shows that the equipment is abnormal, extracting time sequence features of the first local time domain features and the first local frequency domain features to obtain time sequence dependent features and frequency spectrum semantic features, splicing the time sequence dependent features and the frequency spectrum semantic features, and then inputting the spliced time sequence dependent features and frequency spectrum semantic features into a hierarchical network to determine an equipment fault structure.
Optionally, before extracting the time sequence features of the local time domain features and the local frequency domain features to obtain the time sequence dependent features and the spectrum semantic features, the method further includes:
If the equipment state diagnosis result shows that the equipment is abnormal, extracting the spatial characteristics of the first local time domain characteristics and the first local frequency domain characteristics to obtain second local time domain characteristics and second local frequency domain characteristics, splicing the second local time domain characteristics and the second local frequency domain characteristics, and inputting the spliced second local time domain characteristics and the spliced second local frequency domain characteristics into a hierarchical network to determine the equipment type of the equipment fault structure.
Optionally, the extracting spatial features of the time domain vibration signal and the frequency domain vibration signal to obtain a first local time domain feature and a first local frequency domain feature specifically includes:
preprocessing the time domain vibration signal and the frequency domain vibration signal to obtain a standardized vibration signal and a standardized frequency domain vibration signal, wherein the preprocessing at least comprises data cleaning processing and normalization processing;
and inputting the normalized time domain vibration signal and the normalized frequency domain vibration signal into a convolutional neural network encoder for spatial feature extraction to obtain a first local time domain feature and a first local frequency domain feature.
Optionally, extracting the time sequence features of the first local time domain feature and the first local frequency domain feature to obtain the time sequence dependent feature and the spectrum semantic feature specifically includes:
And inputting the first local time domain features and the first local frequency domain features into a transducer encoder for time sequence feature extraction to obtain time sequence dependent features and frequency spectrum semantic features.
Optionally, before extracting the time sequence features of the local time domain features and the local frequency domain features to obtain the time sequence dependent features and the spectrum semantic features, the method further includes:
And respectively carrying out random discarding operation on the first local time domain features and the first local frequency domain features.
Optionally, the data to be detected further includes a data tag, where the data tag is used to indicate a device status, or a fault class, or a fault structure.
In a second aspect, an embodiment of the present application provides an apparatus fault diagnosis device, including:
The receiving and transmitting module is used for obtaining data to be detected, wherein the data to be detected comprises a time domain vibration signal and a frequency domain vibration signal of equipment to be detected;
The processing module is used for extracting spatial characteristics of the time domain vibration signal and the frequency domain vibration signal to obtain a first local time domain characteristic and a first local frequency domain characteristic, inputting the first local time domain characteristic and the first local frequency domain characteristic into the hierarchical network after being spliced to obtain a device state diagnosis result, wherein the device state diagnosis result is used for indicating whether the device is abnormal or not, the hierarchical network is trained through a preset self-supervision learning task, the self-supervision learning task is used for limiting the hierarchical network to generate a pseudo tag by utilizing the internal structure and the attribute of sample data, and the pseudo tag is adjusted based on the difference between the pseudo tag and a prediction result;
And the processing module is also used for extracting the time sequence features of the first local time domain features and the first local frequency domain features to obtain time sequence dependent features and spectrum semantic features if the equipment state diagnosis result indicates equipment abnormality, splicing the time sequence dependent features and the spectrum semantic features, and then inputting the spliced time sequence dependent features and spectrum semantic features into a hierarchical network to determine an equipment fault structure.
In a third aspect, an embodiment of the present application further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the computer program, when executed by the processor, causes the processor to implement any one of the device fault diagnosis methods of the first aspect.
In a fourth aspect, an embodiment of the present application further provides a computer readable storage medium, in which a computer program is stored, where the computer program, when executed by a processor, implements any one of the device fault diagnosis methods of the first aspect.
In a fifth aspect, embodiments of the present application also provide a computer program product comprising a computer program that is executed by a processor to implement a device fault diagnosis method as in any of the first aspects above.
The technical effects caused by any implementation manner of the second aspect to the fifth aspect may refer to the technical effects caused by the corresponding implementation manner of the first aspect, and are not described herein.
Drawings
FIG. 1 is a flow chart of a method for diagnosing equipment faults, which is provided by the embodiment of the application;
FIG. 2 is a schematic diagram of a data tag according to an embodiment of the present application;
FIG. 3 is a schematic diagram of equipment fault diagnosis according to an embodiment of the present application;
FIG. 4 is a flowchart of another method for diagnosing equipment failure according to an embodiment of the present application;
FIG. 5 is a schematic diagram of an apparatus for diagnosing a device failure according to an embodiment of the present application;
fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In the following, some terms in the embodiments of the present application are explained for easy understanding by those skilled in the art.
1. "And/or", describes an association relationship of an association object, and indicates that there may be three relationships, for example, a and/or B, and may indicate: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship.
2. The time domain signal and the frequency domain signal of the vibration signal have the same points and differences in the following aspects: the same points: both the time domain signal and the frequency domain signal are ways of describing the signal characteristics, and can represent the physical characteristics of the vibration signal. Both the time domain signal and the frequency domain signal can be used to analyze the frequency component, amplitude magnitude, phase information, etc. of the vibration signal. The difference is: the time domain signal and the frequency domain signal are represented differently. The time domain signal is a law of variation representing the vibration signal in the time domain, and the frequency domain signal is a frequency component representing the vibration signal in the frequency domain. The time domain signal and the frequency domain signal are analyzed in different manners. In the time domain, we analyze the amplitude and time history of the signal, while in the frequency domain, we analyze the frequency content, amplitude magnitude, phase information, etc. of the signal. The application range of the time domain signal and the frequency domain signal is different. The analysis method of the time domain signal is more suitable for processing non-stationary signals, namely signals with obvious change along with time; the analysis method of the frequency domain signal is more suitable for processing stable signals, namely signals with relatively smooth change along with time.
3. In neural networks Dropout is a regularization technique that prevents overfitting by randomly closing a portion of the neurons in the network.
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be described in further detail below with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
The application scenario described in the embodiment of the present application is for more clearly describing the technical solution of the embodiment of the present application, and does not constitute a limitation on the technical solution provided by the embodiment of the present application, and as a person of ordinary skill in the art can know that the technical solution provided by the embodiment of the present application is applicable to similar technical problems as the new application scenario appears. In the description of the present application, unless otherwise indicated, the meaning of "a plurality" is two or more.
As natural language processing technology (GENERATIVE PRETRAINED transducer, GPT) of pre-trained models grows, the use of GPT in fault diagnosis scenarios has become a trend. But has problems of poor interpretability and poor scalability.
In order to solve the problems, the embodiment of the application provides a device fault diagnosis method and device and electronic equipment. For example, data to be detected is obtained, the data to be detected including a time domain vibration signal and a frequency domain vibration signal of the device to be detected. Extracting spatial features of a time domain vibration signal and a frequency domain vibration signal to obtain a first local time domain feature and a first local frequency domain feature, splicing the first local time domain feature and the first local frequency domain feature, inputting the first local time domain feature and the first local frequency domain feature into a hierarchical network to obtain a device state diagnosis result, wherein the device state diagnosis result is used for indicating whether the device is abnormal or not, the hierarchical network is trained through a preset self-supervision learning task, the self-supervision learning task is used for limiting the hierarchical network to generate a pseudo tag by utilizing the internal structure and the attribute of sample data, and adjusting based on the difference between the pseudo tag and a prediction result. If the equipment state diagnosis result shows that the equipment is abnormal, extracting time sequence features of the first local time domain features and the first local frequency domain features to obtain time sequence dependent features and frequency spectrum semantic features, splicing the time sequence dependent features and the frequency spectrum semantic features, and then inputting the spliced time sequence dependent features and frequency spectrum semantic features into a hierarchical network to determine an equipment fault structure.
Therefore, the self-supervision learning task is performed for pre-training, so that the representation of the time domain signal and the frequency domain signal is extracted rapidly, and the cross-equipment support is realized. When the downstream task is executed, the fault reasoning process can be explained and illustrated in detail by combining the hierarchical network, so that a user can more clearly determine the equipment state diagnosis result and the fault structure, the equipment state diagnosis result and the fault structure are easier to understand and accept, and the interpretability and the expandability of equipment fault diagnosis are improved. Meanwhile, the application is not only suitable for fault diagnosis of single equipment or scenes, but also can be easily expanded to other equipment and scenes, and has strong generalization capability.
As shown in fig. 1, a flowchart of an apparatus fault diagnosis method is provided in an embodiment of the present application. Specifically, the method comprises the following steps.
Step S101, obtaining data to be detected.
The data to be detected comprises a time domain vibration signal and a frequency domain vibration signal of the equipment to be detected.
For example, the data to be detected may be from public laboratory datasets of various universities, as well as on-site datasets proprietary to chemical plants, energy stations, and the like. For another example, the data to be detected may be from multiple rotating machinery of multiple plants. Specifically, a vibration sensor can be additionally arranged on each device to collect vibration signals in the running process of the device. For example, the vibration sensor is a vibration velocity sensor, a vibration acceleration sensor, or a vibration displacement sensor.
Optionally, after the data to be detected is obtained, the time domain vibration signal and the frequency domain vibration signal may be preprocessed, so as to obtain a standardized vibration signal and a standardized frequency domain vibration signal. And inputting the normalized time domain vibration signal and the normalized frequency domain vibration signal into a convolutional neural network encoder for spatial feature extraction to obtain a first local time domain feature and a first local frequency domain feature. The pretreatment at least comprises data cleaning treatment and normalization treatment.
For example, the time domain vibration signal and the frequency domain vibration signal are subjected to data cleaning, data alignment and normalization to obtain a normalized time domain vibration signal and a normalized frequency domain vibration signal.
In the method, the data can be adjusted to a unified specification or standard by the mode of carrying out data standardization on the time domain vibration signals and the frequency domain vibration signals, and unit influence among different variables can be eliminated, so that the numerical values of all indexes are in the same order of magnitude, and comprehensive analysis and comparison are convenient to carry out, so that the accuracy and the reliability of subsequent analysis are ensured.
Step S102, extracting spatial features of the time domain vibration signals and the frequency domain vibration signals to obtain first local time domain features and first local frequency domain features, splicing the first local time domain features and the first local frequency domain features, and inputting the spliced first local time domain features and the spliced first local frequency domain features into a hierarchical network to obtain a device state diagnosis result.
Wherein, the device status diagnosis result is used for indicating whether the device is abnormal. The hierarchical network is trained through preset self-supervision learning tasks. The self-supervised learning task is used for limiting the hierarchical network to generate a pseudo tag by utilizing the internal structure and the attribute of the sample data, and adjustment is performed based on the difference between the pseudo tag and the prediction result.
It will be appreciated that the application is not particularly limited as to the manner in which the local time domain features are spliced with the local frequency domain features. For example, the local time domain features are head-to-tail with the local frequency domain features.
In an alternative embodiment, after the data to be detected is acquired, the time domain vibration signal and the frequency domain vibration signal may be preprocessed, so as to obtain a normalized time domain vibration signal and a normalized frequency domain vibration signal. The pretreatment at least comprises data cleaning treatment and normalization treatment.
For example, the time domain vibration signal and the frequency domain vibration signal are subjected to data cleaning processing, data alignment and normalization to obtain a normalized time domain vibration signal and a normalized frequency domain vibration signal.
In the method, the data can be adjusted to a unified specification or standard by the mode of carrying out data standardization on the time domain vibration signals and the frequency domain vibration signals, and unit influence among different variables can be eliminated, so that the numerical values of all indexes are in the same order of magnitude, and comprehensive analysis and comparison are convenient to carry out, so that the accuracy and the reliability of subsequent analysis are ensured.
After the normalized time domain vibration signal and the normalized frequency domain vibration signal are obtained, the normalized time domain vibration signal and the normalized frequency domain vibration signal can be input into a convolutional neural network encoder to perform spatial feature extraction, so as to obtain a first local time domain feature and a first local frequency domain feature. And then the first local time domain features and the first local frequency domain features are spliced and then input into a hierarchical network to obtain a device state diagnosis result.
How to train the hierarchical network is described as follows:
In one possible scenario, the training of the hierarchical network may be supervised training. That is, during the training process, the sample data is data tagged. The data tags may be obtained by manual labeling or using other supervised learning methods. The hierarchical network computes a prediction result of the sample data. And taking the difference between the prediction result and the data tag as a comparison loss function, and adjusting the hierarchical network based on the comparison loss function until the hierarchical network converges or reaches the preset iteration times.
Wherein the contrast loss function may satisfy the following formula:
Wherein, C is the category number, p i is the data tag, and q i is the prediction result.
In the method, by using the sample data with the data tag, the hierarchical network can learn the mapping relation between the input sample data and the output tag, so that accurate prediction or classification can be performed when the data to be detected is faced. Meanwhile, the training of the hierarchical network can be fine-tuned and optimized for specific tasks. The diagnosis effect on different types of faults can be remarkably improved.
In another possible scenario, the training of the hierarchical network may be unsupervised training. Sample data is input into a hierarchical network, and the hierarchical network calculates a prediction result of the sample data. And taking the difference between the prediction result and the pseudo tag as a comparison loss function, and adjusting the hierarchical network based on the comparison loss function until the hierarchical network converges or reaches the preset iteration times.
In the method, by using unlabeled sample data, self-supervised learning can be pre-trained with a large amount of unlabeled data, thereby learning the inherent structure and features of the data without manual labeling. This greatly expands the amount of data available for training, improving the generalization ability of the model. The inherent structure and relationships of the data are captured, thereby helping to address various machine learning tasks. The dependence on the marking data is reduced. Meanwhile, through self-supervision learning pre-training, richer data representation can be learned in downstream tasks, so that understanding and generalization capability of the model to the tasks are improved.
It is understood that a hierarchical network may be composed of a fully connected layer of two neurons and a layer of nonlinear activation functions (relus).
Optionally, embodiments of the present application may include a multi-layered hierarchical network. A contrast loss function may be employed at the loss of each layer. And the weight lost by each layer is dynamically changed by the downstream task demand. It will be appreciated that the sum of the weights of the multi-layered hierarchical network is 1. And the weights from the first hierarchical network to the nth hierarchical network are decremented. The loss weight of each layer of hierarchical network can be preset by a person skilled in the art, and can be changed based on a specific application scene.
For example, assume that a three-layer hierarchical network may be included. The weight of the loss of the first layered network is 0.5. The weight of the loss of the second layered network is 0.3. The weight of the loss of the first layered network is 0.2.
Step S103, if the equipment state diagnosis result shows that the equipment is abnormal, extracting time sequence features of the first local time domain features and the first local frequency domain features to obtain time sequence dependent features and frequency spectrum semantic features, splicing the time sequence dependent features and the frequency spectrum semantic features, and then inputting the spliced time sequence dependent features and frequency spectrum semantic features into a hierarchical network to determine an equipment fault structure.
Optionally, after determining that the device state diagnosis result is abnormal, before extracting the time sequence features of the first local time domain feature and the first local frequency domain feature to obtain the time sequence dependent feature and the spectrum semantic feature, a random discarding operation (such as a Dropout operation) may be performed on the first local time domain feature and the first local frequency domain feature respectively. And inputting the first local time domain features and the first local frequency domain features into a transform encoder for time sequence feature extraction to obtain time sequence dependent features and frequency spectrum semantic features. And the time sequence dependent features and the spectrum semantic features are spliced and then input into a hierarchical network to determine a device fault structure.
In the method, the first local time domain features and the first local frequency domain features are respectively subjected to random discarding operation, so that overfitting can be prevented. Meanwhile, since the random discarding operation is realized by randomly closing a part of neurons in the network, the generalization capability of the model can be increased, so that the model is better suitable for new data, and a smaller learning rate can be used in training. The calculation amount can be reduced, and the subsequent training process can be accelerated.
Optionally, when the downstream task performs fault diagnosis, a downstream task fine tuning technology may be used to adjust the fault diagnosis basic model to obtain a plurality of fault diagnosis models corresponding to different types of downstream tasks. In one possible scenario, sample data for training a hierarchical network may be tagged with data. Then the training of the hierarchical network is supervised training at this time. The data to be detected may also include a data tag for indicating a status of the device, or a fault category, or a fault structure.
As shown in fig. 2, an embodiment of the present application provides a schematic diagram of a data tag. For example, the data tag used to represent the status of the device may be normal, abnormal. The data labels for indicating the fault categories may be rotor faults, bearing faults, gearbox faults.
For example, if the data tag for indicating the failure category is a rotor failure, the corresponding data tag for indicating the failure structure may include unbalance, poor centering, rub-and-bump. For another example, if the data tag for indicating the failure category is a bearing failure, the corresponding data tag for indicating the failure structure may include an inner ring, an outer ring, rolling elements, and the like. For another example, if the data tag used to represent the fault category is a gearbox fault, the corresponding data tag used to represent the fault structure may include tooth breakage, tooth surface wear, tooth root cracking.
It can be appreciated that in order to adapt to the hierarchical network, the requirements of user data analysis are met, and the user experience is improved. The data tag may be set to change accordingly before entering a different hierarchical network. For example, assume a three-layer hierarchical network. Upon entering the first tier hierarchical network, a data tag for indicating the status of the device may be employed. Upon entering the second tier hierarchical network, a data tag for indicating the failure category may be employed. Upon entering the second tier hierarchical network, a data tag for indicating the failure category may be employed. For another example, assume a three-layer hierarchical network. Upon entering the first tier hierarchical network, a data tag for indicating the status of the device may be employed. Upon entering the second tier hierarchical network, a data tag for indicating the failure category may be employed. Upon entering the second tier hierarchical network, a data tag for representing the failure structure may be employed.
Optionally, in order to facilitate the subsequent data analysis of the user, the fault state diagnosis result and the fault diagnosis result may be displayed after the fault diagnosis result is determined. The device status diagnosis result, the fault category diagnosis result, and the fault structure diagnosis result may also be displayed.
For example, it is possible to display: anomaly-rotor failure-poor centering. For another example, it is possible to display: normal. For another example, it is possible to display: abnormal-bearing failure-rolling bodies. For another example, it is possible to display: abnormality-gearbox failure-tooth breakage.
It can be appreciated that to enhance the user experience, the interpretability of the fault diagnosis is enhanced. The application can extract the time domain vibration signal and the space characteristic of the frequency domain vibration signal for many times. And inputting the corresponding spatial characteristics into the hierarchical network for fault diagnosis. Meanwhile, in order to prevent the overfitting, before extracting the time sequence features of the time domain vibration signals and the frequency domain vibration signals, random discarding operation is required to be carried out on the spatial features of the time domain vibration signals and the inter-feature of the frequency domain vibration signals respectively.
For example, a convolutional neural network encoder may be comprised of a three-layer convolutional neural network. The first local time domain features and the first local frequency domain features can be spliced and then input into a first layered network to obtain a device state diagnosis result. The first local time domain feature is obtained by inputting a time domain vibration signal into a first layer convolutional neural network to extract spatial features. The first local frequency domain feature is obtained by inputting a frequency domain vibration signal into a first layer convolutional neural network to extract spatial features. If the equipment state diagnosis result shows that the equipment is abnormal, the first local time domain feature and the first local frequency domain feature are input into a second layer convolutional neural network to perform space feature extraction to obtain a second local time domain feature and a second local frequency domain feature, the second local time domain feature and the second local frequency domain feature are spliced and then input into a second layer hierarchical network, and the equipment type of the equipment fault structure is determined. And then, the second local time domain feature and the second local frequency domain feature are input into a third layer convolutional neural network to perform spatial feature extraction to obtain a third local time domain feature and a third local frequency domain feature, and Dropout operation is performed on the third local time domain feature and the third local frequency domain feature respectively. And inputting the third local time domain feature and the third local frequency domain feature after the Dropout operation into a transducer encoder for time sequence feature extraction to obtain time sequence dependent features and frequency spectrum semantic features. And the time sequence dependent features and the spectrum semantic features are spliced and then input into a hierarchical network to determine a device fault structure.
The embodiment of fig. 1 is illustrated below:
As shown in fig. 3, an alternative device fault diagnosis schematic is provided in an embodiment of the present application. Wherein, Representing the time domain vibration signal. /(I)Representing the frequency domain vibration signal. B1 represents a first-tier hierarchical network. B2 represents a layer-two hierarchical network. B3 represents a third layered network. C is a convolutional neural network. /(I)And a first layer convolutional neural network corresponding to the time domain vibration signal. /(I)And a second layer convolutional neural network corresponding to the time domain vibration signal. /(I)And a third layer convolutional neural network corresponding to the time domain vibration signal. /(I)And a first layer convolutional neural network corresponding to the frequency domain vibration signal. /(I)And a second layer convolutional neural network corresponding to the frequency domain vibration signal. /(I)And a third layer convolutional neural network corresponding to the frequency domain vibration signal.
It will be appreciated that the fault diagnosis system shown in fig. 3 may maintain maximum consistency during the pre-training process.
By combining time-domain signalsAnd frequency domain signal/>And respectively inputting the first layer convolutional neural network to obtain a first local time domain feature and a first local frequency domain feature. And the first local time domain features and the first local frequency domain features are spliced and then input into the first layered network B1. The first layered network B1 determines the device status diagnosis result.
In one possible case, the device state diagnosis result is normal, and the fault diagnosis is stopped. In another possible case, if the device state diagnosis result is abnormal, the first local time domain feature and the first local frequency domain feature are respectively input into the second layer convolutional neural network to obtain a second local time domain feature and a second local frequency domain feature. And splicing the second local time domain features and the second local frequency domain features, and inputting the spliced second local time domain features and the second local frequency domain features into a second layered network B2. The second hierarchical network B2 determines the device class of the device failure structure. For example, the device class diagnostic result may be a rotor fault. For another example, the device class diagnostic result may be a bearing failure. For another example, the equipment category diagnostic result may be a gearbox fault.
After the equipment type of the equipment fault structure is determined, the second local time domain feature and the second local frequency domain feature are respectively input into a third layer convolutional neural network to obtain a third local time domain feature and a third local frequency domain feature. Characterizing a third local time domainThird local frequency domain feature/>Dropout operation is respectively carried out, and fourth local time domain characteristics/>, after the Dropout operation, are obtainedFourth local frequency domain feature/>The fourth local time domain feature/>, after the Dropout operation, is respectively carried outFourth local frequency domain feature/>Input to a transducer encoder to obtain a timing dependent feature/>Spectral semantic features/>Splice timing dependent features/>Enhanced timing dependent features/>Spliced Spectrum semantic features/>Enhanced spectrum semantic features are obtainedSpliced and enhanced time sequence dependency characteristics/>Enhanced spectral semantic features/>And then, the fault structure of the equipment is input into the third layered network B3. For example, assuming that the first layered network determines that the device status diagnosis result is abnormal, the second layered network determines that the failure category result is a rotor failure, and the third layered network determines that the device failure structure may be a centering failure. For another example, assuming that the first layered network determines that the device state diagnosis result is abnormal, the second layered network determines that the failure type is a bearing failure, and the third layered network determines that the device failure structure may be a rolling body.
After determining the device fault structure, the fault state diagnosis result, the fault category and the fault structure may be sent to the terminal and displayed. For example, display: anomaly-rotor failure-poor centering. For another example, it is possible to display: normal. For another example, it is possible to display: abnormal-bearing failure-rolling bodies.
As shown in fig. 4, an embodiment of the present application provides an exemplary flow chart for device fault diagnosis. The method can comprise the following steps:
step S401, obtaining data to be detected;
step S402, preprocessing the time domain vibration signal and the frequency domain vibration signal to obtain a standardized vibration signal and a standardized frequency domain vibration signal;
Step S403, inputting the normalized time domain vibration signal and the normalized frequency domain vibration signal into a convolutional neural network encoder for spatial feature extraction to obtain a first local time domain feature and a first local frequency domain feature;
Step S404, the first local time domain features and the first local frequency domain features are spliced and then input into a hierarchical network to obtain a device state diagnosis result;
Step S405, if the equipment state diagnosis result indicates that the equipment is abnormal, extracting the spatial characteristics of the first local time domain characteristics and the first local frequency domain characteristics to obtain second local time domain characteristics and second local frequency domain characteristics, and inputting the spliced second local time domain characteristics and second local frequency domain characteristics into a hierarchical network to determine the equipment category of the equipment fault structure;
step S406, extracting the spatial features of the second local time domain feature and the second local frequency domain feature to obtain a third local time domain feature and a third local frequency domain feature;
Step S407, carrying out Dropout operation on the third local time domain feature and the third local frequency domain feature respectively;
Step S408, inputting the third local time domain feature and the third local frequency domain after the Dropout operation into a transducer encoder to obtain a time sequence dependent feature and a frequency spectrum semantic feature;
Step S409, the time sequence dependent features and the spectrum semantic features are spliced and then input into a hierarchical network to determine a device fault structure;
step S410, sending the device status diagnosis result, the device class of the device failure structure, and the device failure structure to the terminal, so that the terminal displays the device status diagnosis result, the device class of the device failure structure, and the device failure structure.
Fig. 5 is a schematic structural diagram of an apparatus for diagnosing a device fault according to an embodiment of the present application, as shown in fig. 5, the apparatus includes: transceiver module 501, processing module 502.
The transceiver module 501 is configured to obtain data to be detected, where the data to be detected includes a time domain vibration signal and a frequency domain vibration signal of a device to be detected;
The processing module 502 is configured to extract spatial features of the time domain vibration signal and the frequency domain vibration signal to obtain a first local time domain feature and a first local frequency domain feature, splice the first local time domain feature and the first local frequency domain feature, and input the spliced first local time domain feature and the spliced first local frequency domain feature into the hierarchical network to obtain a device state diagnosis result, where the device state diagnosis result is used to indicate whether the device is abnormal, the hierarchical network is trained through a preset self-supervision learning task, the self-supervision learning task is used to define the hierarchical network to generate a pseudo tag by using an internal structure and an attribute of sample data, and adjust the pseudo tag based on a difference between the pseudo tag and a prediction result;
The processing module 502 is further configured to extract a time sequence feature of the first local time domain feature and the first local frequency domain feature to obtain a time sequence dependent feature and a spectrum semantic feature if the device state diagnosis result indicates that the device is abnormal, splice the time sequence dependent feature and the spectrum semantic feature, and input the spliced time sequence dependent feature and the spectrum semantic feature into the hierarchical network to determine a device fault structure.
Optionally, before extracting the time-domain features and the time-domain features to obtain the time-domain dependent features and the spectrum semantic features, the processing module 502 is further configured to:
If the equipment state diagnosis result shows that the equipment is abnormal, extracting the spatial characteristics of the first local time domain characteristics and the first local frequency domain characteristics to obtain second local time domain characteristics and second local frequency domain characteristics, splicing the second local time domain characteristics and the second local frequency domain characteristics, and inputting the spliced second local time domain characteristics and the spliced second local frequency domain characteristics into a hierarchical network to determine the equipment type of the equipment fault structure.
Optionally, the above extracting the spatial features of the time domain vibration signal and the frequency domain vibration signal to obtain the first local time domain feature and the first local frequency domain feature, and the processing module 502 is specifically configured to:
preprocessing the time domain vibration signal and the frequency domain vibration signal to obtain a standardized vibration signal and a standardized frequency domain vibration signal, wherein the preprocessing at least comprises data cleaning processing and normalization processing;
and inputting the normalized time domain vibration signal and the normalized frequency domain vibration signal into a convolutional neural network encoder for spatial feature extraction to obtain a first local time domain feature and a first local frequency domain feature.
Optionally, extracting the time sequence features of the first local time domain feature and the first local frequency domain feature to obtain a time sequence dependent feature and a spectrum semantic feature, where the processing module 502 is specifically configured to:
And inputting the first local time domain features and the first local frequency domain features into a transducer encoder for time sequence feature extraction to obtain time sequence dependent features and frequency spectrum semantic features.
Optionally, before extracting the time-domain local features and the frequency-domain local features to obtain the time-domain dependent features and the spectrum semantic features, the processing module 502 is further configured to:
And respectively carrying out random discarding operation on the first local time domain features and the first local frequency domain features.
Optionally, the data to be detected further includes a data tag, where the data tag is used to indicate a device status, or a fault class, or a fault structure.
Based on the same technical conception, the embodiment of the application also provides electronic equipment, which can realize the functions of the equipment fault diagnosis device.
Fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
At least one processor 601, and a memory 602 connected to the at least one processor 601, a specific connection medium between the processor 601 and the memory 602 is not limited in the embodiment of the present application, and in fig. 6, the processor 601 and the memory 602 are connected through a bus 600 as an example. Bus 600 is shown in bold lines in fig. 6, and the manner in which the other components are connected is illustrated schematically and not by way of limitation. The bus 600 may be divided into an address bus, a data bus, a control bus, etc., and is represented by only one thick line in fig. 6 for convenience of representation, but does not represent only one bus or one type of bus. Alternatively, the processor 601 may be referred to as a controller, and the names are not limited.
In the embodiment of the present application, the memory 602 stores instructions executable by the at least one processor 601, and the at least one processor 601 may perform a device fault diagnosis method as described above by executing the instructions stored in the memory 602. The processor 601 may implement the functions of the respective modules in the apparatus shown in fig. 4.
The processor 601 is a control center of the device, and various interfaces and lines can be used to connect various parts of the whole control device, and through running or executing instructions stored in the memory 602 and calling data stored in the memory 602, various functions of the device and processing data can be performed, so that the device can be monitored as a whole.
In one possible design, processor 601 may include one or more processing units, and processor 601 may integrate an application processor that primarily processes operating systems, driver interfaces, application programs, and the like, and a modem processor that primarily processes wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 601. In some embodiments, processor 601 and memory 602 may be implemented on the same chip, or they may be implemented separately on separate chips in some embodiments.
The processor 601 may be a general purpose processor such as a Central Processing Unit (CPU), digital signal processor, application specific integrated circuit, field programmable gate array or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, which may implement or perform the methods, steps and logic blocks disclosed in embodiments of the application. The general purpose processor may be a microprocessor or any conventional processor or the like. The steps of an equipment fault diagnosis method disclosed in connection with the embodiment of the application can be directly embodied as the execution of a hardware processor or the execution of the combination of hardware and software modules in the processor.
The memory 602 is a non-volatile computer readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs, and modules. The Memory 602 may include at least one type of storage medium, which may include, for example, flash Memory, hard disk, multimedia card, card Memory, random access Memory (Random Access Memory, RAM), static random access Memory (StaticRandom Access Memory, SRAM), programmable Read Only Memory (Programmable Read OnlyMemory, PROM), read Only Memory (ROM), charged erasable programmable Read Only Memory (ELECTRICALLY ERASABLE PROGRAMMABLE READ-Only Memory, EEPROM), magnetic Memory, magnetic disk, optical disk, and the like. Memory 602 is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited to such. The memory 602 in embodiments of the present application may also be circuitry or any other device capable of performing storage functions for storing program instructions and/or data.
By programming the processor 601, the code corresponding to one of the device fault diagnosis methods described in the foregoing embodiments can be cured into the chip, so that the chip can execute one of the device fault diagnosis methods of the embodiment shown in fig. 2 at the time of operation. How to design and program the processor 601 is a well-known technique for those skilled in the art, and will not be described in detail herein.
It should be noted that, the above-mentioned power-on electronic device provided in the embodiment of the present application can implement all the method steps implemented in the above-mentioned method embodiment, and can achieve the same technical effects, and specific details of the same parts and beneficial effects as those of the method embodiment in the present embodiment are not described herein.
The embodiment of the application also provides a computer readable storage medium, wherein the computer readable storage medium stores computer executable instructions for causing a computer to execute the equipment fault diagnosis method in the embodiment.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present application without departing from the spirit or scope of the application. Thus, it is intended that the present application also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (10)
1. A method of diagnosing a device fault, the method comprising:
Obtaining data to be detected, wherein the data to be detected comprises a time domain vibration signal and a frequency domain vibration signal of equipment to be detected;
Extracting spatial features of the time domain vibration signal and the frequency domain vibration signal to obtain a first local time domain feature and a first local frequency domain feature, splicing the first local time domain feature and the first local frequency domain feature, and inputting the first local time domain feature and the first local frequency domain feature into a hierarchical network to obtain a device state diagnosis result, wherein the device state diagnosis result is used for indicating whether the device is abnormal or not, the hierarchical network is trained through a preset self-supervision learning task, the self-supervision learning task is used for limiting the hierarchical network to generate a pseudo tag by utilizing the internal structure and the attribute of sample data, and the self-supervision learning task is adjusted based on the difference between the pseudo tag and a prediction result;
and if the equipment state diagnosis result shows that the equipment is abnormal, extracting the time sequence features of the first local time domain features and the first local frequency domain features to obtain time sequence dependent features and spectrum semantic features, splicing the time sequence dependent features and the spectrum semantic features, and then inputting the spliced time sequence dependent features and spectrum semantic features into the hierarchical network to determine an equipment fault structure.
2. The method of claim 1, wherein before extracting the time-domain features and the time-domain features to obtain time-domain dependent features and spectral semantic features, the method further comprises:
And if the equipment state diagnosis result shows that the equipment is abnormal, extracting the spatial characteristics of the first local time domain characteristics and the first local frequency domain characteristics to obtain second local time domain characteristics and second local frequency domain characteristics, splicing the second local time domain characteristics and the second local frequency domain characteristics, and inputting the spliced second local time domain characteristics and second local frequency domain characteristics into the hierarchical network to determine the equipment category of the equipment fault structure.
3. The method according to claim 1, wherein the extracting spatial features of the time domain vibration signal and the frequency domain vibration signal to obtain a first local time domain feature and a first local frequency domain feature specifically includes:
preprocessing the time domain vibration signal and the frequency domain vibration signal to obtain a standardized vibration signal and a standardized frequency domain vibration signal, wherein the preprocessing at least comprises data cleaning processing and normalization processing;
and inputting the normalized time domain vibration signal and the normalized frequency domain vibration signal into a convolutional neural network encoder for spatial feature extraction to obtain the first local time domain feature and the first local frequency domain feature.
4. The method according to claim 1, wherein the extracting the time-series features of the first local time-domain features and the first local frequency-domain features results in time-series dependent features and spectrum semantic features, specifically comprising:
and inputting the first local time domain features and the first local frequency domain features into a transducer encoder to extract time sequence features, so as to obtain the time sequence dependent features and the frequency spectrum semantic features.
5. The method of claim 1, wherein before extracting the time-domain features and the time-domain features to obtain time-domain dependent features and spectral semantic features, the method further comprises:
and respectively carrying out random discarding operation on the first local time domain features and the first local frequency domain features.
6. The method according to claim 1 or 2, wherein the data to be detected further comprises a data tag for indicating a device status, or a fault class, or a fault structure.
7. An apparatus failure diagnosis device, comprising:
the receiving and transmitting module is used for obtaining data to be detected, wherein the data to be detected comprises a time domain vibration signal and a frequency domain vibration signal of equipment to be detected;
The processing module is used for extracting spatial characteristics of the time domain vibration signal and the frequency domain vibration signal to obtain a first local time domain characteristic and a first local frequency domain characteristic, inputting the first local time domain characteristic and the first local frequency domain characteristic into a hierarchical network after being spliced to obtain a device state diagnosis result, wherein the device state diagnosis result is used for indicating whether the device is abnormal or not, the hierarchical network is trained through a preset self-supervision learning task, the self-supervision learning task is used for limiting the hierarchical network to generate a pseudo tag by utilizing the internal structure and the attribute of sample data, and adjusting based on the difference between the pseudo tag and a prediction result;
and the processing module is further configured to extract the time sequence features of the first local time domain feature and the first local frequency domain feature to obtain a time sequence dependent feature and a spectrum semantic feature if the device state diagnosis result indicates that the device is abnormal, splice the time sequence dependent feature and the spectrum semantic feature, and then input the spliced time sequence dependent feature and the spectrum semantic feature into the hierarchical network to determine a device fault structure.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable by the processor, characterized in that the processor implements the steps of the method according to any one of claims 1-6 when executing the computer program.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any one of claims 1-6.
10. A computer program product, characterized in that the computer program product, when called by a computer, causes the computer to perform the steps of the method according to any of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410247878.9A CN118132986A (en) | 2024-03-05 | 2024-03-05 | Equipment fault diagnosis method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410247878.9A CN118132986A (en) | 2024-03-05 | 2024-03-05 | Equipment fault diagnosis method and device and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118132986A true CN118132986A (en) | 2024-06-04 |
Family
ID=91237136
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410247878.9A Pending CN118132986A (en) | 2024-03-05 | 2024-03-05 | Equipment fault diagnosis method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118132986A (en) |
-
2024
- 2024-03-05 CN CN202410247878.9A patent/CN118132986A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Unsupervised fault diagnosis of rolling bearings using a deep neural network based on generative adversarial networks | |
| Zhang et al. | A hybrid attention improved ResNet based fault diagnosis method of wind turbines gearbox | |
| Principi et al. | Unsupervised electric motor fault detection by using deep autoencoders | |
| Meng et al. | Data segmentation and augmentation methods based on raw data using deep neural networks approach for rotating machinery fault diagnosis | |
| CN113379123A (en) | Fault prediction method, device, server and storage medium based on digital twin | |
| CN111340238A (en) | Fault diagnosis method, device, equipment and storage medium of industrial system | |
| Ayodeji et al. | Causal augmented ConvNet: A temporal memory dilated convolution model for long-sequence time series prediction | |
| Gao et al. | ASM1D-GAN: An intelligent fault diagnosis method based on assembled 1D convolutional neural network and generative adversarial networks | |
| CN113837000A (en) | Small sample fault diagnosis method based on task sequencing meta-learning | |
| CN107942956A (en) | Information processor, information processing method, message handling program and recording medium | |
| Islam et al. | Motor bearing fault diagnosis using deep convolutional neural networks with 2d analysis of vibration signal | |
| US20220334573A1 (en) | Sensor-agnostic mechanical machine fault identification | |
| Gültekin et al. | A novel deep learning approach for intelligent fault diagnosis applications based on time-frequency images | |
| CN110837852A (en) | Fault diagnosis method and device for rolling mill gearbox and terminal equipment | |
| CN109254577A (en) | A kind of intelligence manufacture procedure fault classification method and device based on deep learning | |
| CN116012681A (en) | Method and system for diagnosing motor faults of pipeline robot based on sound vibration signal fusion | |
| Xu et al. | Automatic roller bearings fault diagnosis using DSAE in deep learning and CFS algorithm | |
| CN115758225A (en) | Fault prediction method and device based on multi-mode data fusion and storage medium | |
| Shi et al. | DecouplingNet: A stable knowledge distillation decoupling net for fault detection of rotating machines under varying speeds | |
| Du et al. | RUL prediction based on GAM–CNN for rotating machinery | |
| Lu et al. | Rotating Machinery Fault Diagnosis Under Multiple Working Conditions via A Time Series Transformer Enhanced by Convolutional Neural Network | |
| Wang et al. | A multiscale convolution neural network for bearing fault diagnosis based on frequency division denoising under complex noise conditions | |
| CN117351273A (en) | Causal knowledge-guided power equipment partial discharge fault diagnosis method | |
| CN118132986A (en) | Equipment fault diagnosis method and device and electronic equipment | |
| CN115278757A (en) | Method and device for detecting abnormal data and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination |