CN117648600A - Fault diagnosis method, fault diagnosis device, electronic apparatus, and medium - Google Patents
Fault diagnosis method, fault diagnosis device, electronic apparatus, and medium Download PDFInfo
- Publication number
- CN117648600A CN117648600A CN202311386862.8A CN202311386862A CN117648600A CN 117648600 A CN117648600 A CN 117648600A CN 202311386862 A CN202311386862 A CN 202311386862A CN 117648600 A CN117648600 A CN 117648600A
- Authority
- CN
- China
- Prior art keywords
- domain
- target
- fault
- source domain
- fault diagnosis
- 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 173
- 238000000034 method Methods 0.000 title claims abstract description 77
- 238000012549 training Methods 0.000 claims abstract description 56
- 238000005457 optimization Methods 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 16
- 238000003860 storage Methods 0.000 claims description 9
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 description 7
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000013024 troubleshooting Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0475—Generative networks
-
- 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
- G06N3/094—Adversarial learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The invention provides a fault diagnosis method, a fault diagnosis device, an electronic apparatus and a medium. The method comprises the following steps: acquiring target domain data to be diagnosed; inputting the target domain data to be diagnosed into a target fault diagnosis model to obtain a fault type of the target domain data to be diagnosed, wherein the target fault diagnosis model is obtained by training based on a source domain normal sample, a source domain fault sample and a target domain normal sample, the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and the source domain fault sample and the target domain normal sample are used for optimizing target domain characteristics in the target fault diagnosis model to generate a network. According to the fault diagnosis method provided by the invention, fault diagnosis can be carried out through the target diagnosis model obtained under the condition of no target domain fault sample, and the fault type of the target domain data can be rapidly and accurately diagnosed.
Description
Technical Field
The present invention relates to the field of computer technologies, and in particular, to a fault diagnosis method, a fault diagnosis device, an electronic apparatus, and a medium.
Background
In the fields of modern industrial manufacturing, electronic equipment, transportation and the like, fault diagnosis is a key element for ensuring normal operation and production efficiency of equipment. Conventional fault diagnosis methods typically rely on training and learning of marker data for target domain fault samples that are readily available and contain information about known faults. However, in practical applications, it is sometimes difficult to obtain adequate samples of target domain faults, especially for some new or rare fault conditions.
Aiming at the condition of no target domain fault sample, the traditional fault diagnosis method has the following problems:
1. lack of marking data: the non-target domain fault samples often lack marking data and are difficult to train in a traditional supervised learning method.
2. The universality is not enough: the feature distribution of the target domain fault samples is greatly different from that of the source domain samples, and the traditional model may be poor in performance in the target domain.
3. Failure diagnosis accuracy declines: due to the lack of sufficient target domain sample information, conventional approaches have difficulty in accurately diagnosing new or rare faults.
Disclosure of Invention
The invention provides a fault diagnosis method, a fault diagnosis device, electronic equipment and a medium, which are used for solving the problem that in the prior art, no target domain fault sample exists and the fault diagnosis application is limited.
The invention provides a fault diagnosis method, which comprises the following steps:
acquiring target domain data to be diagnosed;
inputting the target domain data to be diagnosed into a target fault diagnosis model to obtain a fault type of the target domain data to be diagnosed, wherein the target fault diagnosis model is obtained by training based on a source domain normal sample, a source domain fault sample and a target domain normal sample, the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and the source domain fault sample and the target domain normal sample are used for optimizing target domain characteristics in the target fault diagnosis model to generate a network.
In some embodiments, before the target domain data to be diagnosed is input to a target fault diagnosis model to obtain a fault type of the target domain data to be diagnosed, the method further includes:
inputting the source domain normal sample to a source domain feature generation network to generate a source domain normal sample feature;
inputting the target domain normal sample to a target domain feature generation network to generate target domain normal sample features;
and performing countermeasure training on the source domain normal sample characteristics and the target domain normal sample characteristics through a domain discriminator, and updating the target domain characteristic generation network and the domain discriminator.
In some embodiments, after the updating the target domain feature generation network, the method further comprises:
executing at least one iterative optimization process until the mixed domain classification loss obtained in the last iterative optimization process reaches a target loss value;
based on network parameters obtained in the last iterative optimization process, obtaining an optimized target domain feature generation network, wherein the optimized target domain feature generation network meets the condition that the difference value between a first inter-class distance and a second inter-class distance is smaller than a target difference value, the first inter-class distance is the inter-class distance between the normal sample of the target domain and the source domain fault sample, and the second inter-class distance is the inter-class distance between the normal sample of the source domain and the source domain fault sample;
generating a network based on the optimized target domain characteristics, and determining the target fault diagnosis model;
wherein the iterative optimization process comprises:
generating a network based on the target domain characteristics, and updating the target domain normal sample characteristics;
inputting source domain fault sample characteristics and target domain normal sample characteristics into a source domain fault diagnosis model, and calculating mixed domain classification loss, wherein the source domain fault sample characteristics are obtained after the source domain fault samples are input into a source domain characteristic generation network, and the source domain fault diagnosis model is obtained based on the source domain fault samples and the source domain normal samples in a training mode;
and adjusting network parameters of the target domain feature generation network based on the mixed domain classification loss.
In some embodiments, the source domain fault diagnosis model is constructed by:
constructing the source domain feature generation network based on a convolutional neural network;
constructing a fault classification network based on the full connection layer;
and constructing the source domain fault diagnosis model based on the fault classification network and the source domain feature generation network.
In some embodiments, before the inputting the source domain fault sample feature and the target domain normal sample feature to a source domain fault diagnosis model, the method further comprises:
inputting the source domain fault sample characteristics and the source domain normal sample characteristics into the diagnosis classification network, and determining fault types corresponding to the source domain fault sample characteristics and the source domain normal sample characteristics;
calculating a source domain classification loss based on the fault type corresponding to the source domain fault sample characteristics and the fault type corresponding to the source domain normal sample characteristics;
and training the source domain fault diagnosis model based on the source domain classification loss.
In some embodiments, said training the source domain normal sample feature and the target domain normal sample feature against by a domain arbiter, updating the target domain feature generation network and the domain arbiter, comprises:
performing countermeasure training on the source domain normal sample characteristics and the target domain normal sample characteristics through a domain discriminator, and calculating domain discrimination loss;
updating the target domain feature generation network and the domain arbiter based on the domain discrimination loss.
The invention also provides a fault diagnosis device, comprising:
the acquisition module is used for acquiring target domain data to be diagnosed;
the diagnosis module is used for inputting the target domain data to be diagnosed into a target fault diagnosis model to obtain the fault type of the target domain data to be diagnosed, the target fault diagnosis model is obtained by training based on a source domain normal sample, a source domain fault sample and a target domain normal sample, the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and the source domain fault sample and the target domain normal sample are used for optimizing a target domain feature generation network in the target fault diagnosis model.
The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the fault diagnosis method as described above when executing the program.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a fault diagnosis method as described in any of the above.
The invention also provides a computer program product comprising a computer program which, when executed by a processor, implements a fault diagnosis method as described in any of the above.
According to the fault diagnosis method, the fault diagnosis device, the electronic equipment and the medium, the target fault diagnosis model is obtained by training the source domain fault sample and the target domain normal sample under the condition that no target domain fault sample exists, so that efficient and accurate fault diagnosis of target domain data to be diagnosed can be realized; will help to increase the versatility and practicality of the fault diagnosis technique, providing a more reliable solution for equipment maintenance and troubleshooting in various fields, thereby increasing production efficiency and equipment reliability.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a fault diagnosis method provided by the invention;
FIG. 2 is a schematic diagram of a fault diagnosis model construction and training architecture of the fault diagnosis method provided by the invention;
fig. 3 is a schematic structural view of a fault diagnosis apparatus provided by the present invention;
fig. 4 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The fault diagnosis method, the fault diagnosis apparatus, the electronic device, and the medium of the present invention are described below with reference to fig. 1 to 4.
The main body of the fault diagnosis method provided by the invention can be an electronic device, a component in the electronic device, an integrated circuit or a chip. The electronic device may be a mobile electronic device or a non-mobile electronic device. By way of example, the mobile electronic device may be a cell phone, tablet computer, notebook computer, palm computer, vehicle mounted electronic device, wearable device, ultra-mobile personal computer (ultra-mobile personal computer, UMPC), netbook or personal digital assistant (personal digital assistant, PDA), etc., and the non-mobile electronic device may be a server, network attached storage (Network Attached Storage, NAS), personal computer (personal computer, PC), television (TV), teller machine or self-service machine, etc., without limitation of the present invention.
The following describes the technical scheme of the present invention in detail by taking a computer to execute the fault diagnosis method provided by the present invention as an example.
Fig. 1 is a schematic flow chart of a fault diagnosis method provided by the invention. Referring to fig. 1, the fault diagnosis method provided by the present invention may include: step 110 and step 120.
Step 110, obtaining target domain data to be diagnosed;
step 120, inputting target domain data to be diagnosed into a target fault diagnosis model to obtain a fault type of the target domain data to be diagnosed, wherein the target fault diagnosis model is obtained by training a source domain normal sample, a source domain fault sample and a target domain normal sample, the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and the source domain fault sample and the target domain normal sample are used for optimizing a target domain characteristic generation network in the target fault diagnosis model.
In the existing fault diagnosis technology, the target domain fault samples are not easy to obtain and mark, but for the case of no target domain fault samples, the application of the conventional method is limited due to lack of sufficient marking data.
According to the invention, an unsupervised feature learning mechanism is introduced, a target fault diagnosis model is trained through a source domain fault sample and a target domain normal sample, and implicit feature representation of the source domain fault sample and the target domain normal sample is learned, so that the diagnosis capability of the target fault diagnosis model on target domain data is enhanced. After training is completed, the fault type of the target domain data to be diagnosed, which is output by the target fault diagnosis model, can be obtained by carrying out feature extraction and identification on the target domain data to be diagnosed.
It will be appreciated that the fault type may be determined according to the actual application scenario of the source domain data or the target domain data, and is not specifically limited herein.
According to the fault diagnosis method provided by the invention, the target fault diagnosis model is obtained by training the source domain fault sample and the target domain normal sample under the condition of no target domain fault sample, so that the high-efficiency and accurate fault diagnosis of the target domain data to be diagnosed can be realized; will help to increase the versatility and practicality of the fault diagnosis technique, providing a more reliable solution for equipment maintenance and troubleshooting in various fields, thereby increasing production efficiency and equipment reliability.
In actual execution, the invention utilizes advanced technology of the countermeasure generation network (Generative Adversarial Network, GAN) to virtually generate the target domain feature distribution meeting the diagnosis distribution space distance in the source domain feature space and the target domain feature space, thereby realizing efficient and accurate diagnosis on the target domain fault sample.
As shown in fig. 2, the fault diagnosis method provided by the invention under the condition of no-target-domain fault sample based on the countermeasure generation network comprises the following processes:
first, data preparation is performed.
Source domain failure samples and target domain failure-free samples are collected. The target domain fault-free sample is the target domain normal sample.
In actual execution, after the source domain time sequence data are collected, marked, segmented and preprocessed, the source domain time sequence data are divided into a source domain fault sample data set and a source domain normal sample data set, and a source domain fault sample can be obtained from the source domain fault sample data set, so that the accuracy of the labels of the source domain fault sample data is ensured. If there is no fault sample data of the target domain and part of fault sample data of the target domain, the fault sample data of the target domain and part of fault sample data of the target domain are also collected together.
And secondly, constructing a source domain fault diagnosis model.
In some embodiments, the source domain fault diagnosis model is constructed by:
constructing a source domain feature generation network based on a convolutional neural network;
constructing a fault classification network based on the full connection layer;
and constructing a source domain fault diagnosis model based on the fault classification network and the source domain feature generation network.
In practical implementation, a convolutional neural network is adopted to construct a source domain feature generation network, and a fault classification network is constructed by a relatively simple fully-connected network.
The output end of the source domain feature generation network is connected with the input end of the fault classification network, and the fault classification network generally adopts a full-connection network with 2-3 layers.
By training the model structure, the fault type can be ensured to be basically linearly separable in the source domain feature space.
Thirdly, constructing an countermeasure generation network.
The architecture of GAN is employed to build the generator and domain arbiter. In actual execution, a source domain feature generator and a target domain feature generator are established for the source domain and the target domain, respectively.
Specifically, the source domain feature generator and the target domain feature generator can be constructed by adopting a multi-layer 1-dimensional convolutional neural network with the same structure, so that the target domain normal sample and the source domain normal sample are similar as much as possible in feature space.
The source domain feature generator generates a network for the source domain features in the source domain fault diagnosis model of the above embodiment. The target domain feature generator is a target domain feature generation network, and the target domain feature generation network can map the target domain normal sample to the same feature space as the source domain. The domain arbiter is responsible for determining whether the generated sample is a real sample or a virtual sample.
And fourthly, training a source domain fault diagnosis model.
In some embodiments, before inputting the source domain fault sample feature and the target domain normal sample feature into the source domain fault diagnosis model, the fault diagnosis method further comprises:
inputting the source domain fault sample characteristics and the source domain normal sample characteristics into a diagnosis classification network, and determining fault types corresponding to the source domain fault sample characteristics and fault types corresponding to the source domain normal sample characteristics;
calculating the classification loss of the source domain based on the fault type corresponding to the source domain fault sample characteristics and the fault type corresponding to the source domain normal sample characteristics;
based on the source domain classification loss, a source domain fault diagnosis model is trained.
In actual execution, the source domain fault sample characteristics and the source domain normal sample characteristics are input into a diagnosis classification network, implicit characteristic representations in a source domain are automatically learned, fault types corresponding to the source domain fault sample characteristics and fault types corresponding to the source domain normal sample characteristics are determined, then source domain classification losses are calculated, and a trained source domain fault diagnosis model is obtained under the condition that the source domain classification losses meet training conditions, so that the diagnosis capability of the source domain fault diagnosis model on the source domain normal samples and the source domain fault samples can be enhanced.
Fifth, the source domain normal sample generates the countermeasure training of the network for the target domain characteristics.
In some embodiments, prior to step 120, the fault diagnosis method further comprises:
inputting the source domain normal sample into a source domain feature generation network to generate source domain normal sample features;
inputting the target domain normal sample into a target domain feature generation network to generate target domain normal sample features;
and performing countermeasure training on the source domain normal sample characteristics and the target domain normal sample characteristics through a domain discriminator, and updating the target domain characteristics to generate a network and the domain discriminator.
In actual execution, the source domain normal sample and the target domain normal sample are respectively input into a source domain feature generation network and a target domain feature generation network to obtain a source domain normal sample feature and a target domain normal sample feature.
And through interaction of the source domain feature generation network, the target domain feature generation network and the domain discriminator, performing countermeasure training, and updating parameters of the target domain feature generation network and parameters of the domain discriminator until the target domain feature generation network can output a result meeting the requirement.
In some embodiments, the source domain normal sample feature and the target domain normal sample feature are counter-trained by a domain arbiter, and the target domain feature generation network and the domain arbiter are updated, comprising:
performing countermeasure training on the source domain normal sample characteristics and the target domain normal sample characteristics through a domain discriminator, and calculating domain discrimination loss;
based on the domain discrimination loss, the target domain feature generation network and the domain arbiter are updated.
In each round of training, the target domain feature generation network may generate more realistic source domain normal sample features, while the domain arbiter strives to distinguish between real and virtual samples and calculate domain discrimination loss. The target domain feature generation network and the domain discriminator gradually improve the respective performances through iterative optimization domain discrimination loss. Through the training, the target domain normal sample is similar to the source domain normal sample as much as possible in the characteristic space.
Sixth, training the source domain fault sample to generate a network for the target domain features.
In some embodiments, after updating the target domain feature generation network, the fault diagnosis method further comprises:
executing at least one iterative optimization process until the mixed domain classification loss obtained in the last iterative optimization process reaches a target loss value;
based on network parameters obtained in the last iterative optimization process, obtaining an optimized target domain feature generation network, wherein the optimized target domain feature generation network meets the condition that the difference value between a first inter-class distance and a second inter-class distance is smaller than a target difference value, the first inter-class distance is the inter-class distance between a target domain normal sample and a source domain fault sample, and the second inter-class distance is the inter-class distance between the source domain normal sample and the source domain fault sample;
generating a network based on the optimized target domain characteristics, and determining a target fault diagnosis model;
the iterative optimization process comprises the following steps:
generating a network based on the target domain characteristics, and updating the target domain normal sample characteristics;
the method comprises the steps of inputting source domain fault sample characteristics and target domain normal sample characteristics into a source domain fault diagnosis model, and calculating mixed domain classification loss, wherein the source domain fault sample characteristics are obtained after source domain fault samples are input into a source domain characteristic generation network, and the source domain fault diagnosis model is obtained based on source domain fault samples and source domain normal sample training;
based on the mixed domain classification loss, network parameters of the target domain feature generation network are adjusted.
In actual execution, the source domain fault sample characteristics generated after the source domain fault samples pass through the source domain characteristic generating network and the target domain normal sample characteristics generated after the target domain normal samples pass through the target domain characteristic generating network are input into a diagnosis classification network in a trained source domain fault diagnosis model, and the mixed domain classification loss is calculated.
The invention aims at minimizing the classification loss of the mixed domain and optimizes the characteristics of the target domain to generate a network.
It can be understood that the objective of this step is to make the sample generated by the target domain feature generating network more realistic, and adjust the network parameters of the target domain feature generating network through at least one iterative optimization process, so that the recalculated mixed domain classification loss gradually reaches the target loss value, and at the same time, the difference between the second type of the source domain normal sample and the source domain fault sample and the first type of the target domain normal sample and the target domain fault sample is reduced.
Through the round of learning, when the difference value between the first inter-class distance and the second inter-class distance is smaller than the target difference value, the first inter-class distance representing the normal sample of the target domain and the fault sample of the target domain is similar to the source domain feature space. The target difference value and the target loss value may be set according to practical situations, and are not specifically limited herein.
And seventh, extracting and diagnosing target domain features.
And performing feature extraction on target domain data to be diagnosed by adopting the trained target domain feature generation network to obtain target domain data features. And then, diagnosing the extracted target domain data characteristics by using the trained source domain fault diagnosis model, so as to accurately identify the fault type of the target domain data to be diagnosed.
In actual implementation, the target fault diagnosis model provided by the invention can be determined by the trained diagnosis classification network and the trained target domain feature generation network in the source domain fault diagnosis model. The target fault diagnosis model in the invention can better adapt to the characteristic distribution of the target domain data, and realize accurate prediction and diagnosis of unknown faults.
And eighth step, model evaluation and optimization.
And evaluating the target fault diagnosis model, and comparing the performance difference of the target fault diagnosis model with that of the traditional method. And according to the evaluation result, optimizing and adjusting the model to achieve higher accuracy and robustness.
And ninth, practical application.
And applying the optimized target fault diagnosis model to an actual scene, and performing diagnosis and monitoring on the fault data of the target domain so as to ensure efficient fault early warning and normal operation of equipment.
The invention utilizes the advanced technology of GAN to virtually generate the target domain feature distribution meeting the space distance of the diagnosis distribution in the source domain feature space and the target domain feature space, thereby realizing the efficient and accurate diagnosis of the target domain fault sample. According to the method, an unsupervised feature learning mechanism is introduced, and implicit feature representation of the target domain fault sample can be automatically learned under the condition that no marked data exists, so that the diagnosis capability of the target fault diagnosis model on the target domain data is enhanced.
According to the invention, through the cooperation of the GAN generator and the arbiter, the network is generated by training the target domain features from the marked source domain normal sample feature space and the source domain fault sample feature space. The target domain feature generation network trains the target domain feature generation network to generate fault features similar to the target domain by learning the feature distribution of the source domain samples, and meanwhile, the domain discriminator is responsible for judging whether the generated samples are real or not. Through iterative optimization, the target domain feature generator gradually improves the fidelity of the generated samples, and the domain discriminator more accurately distinguishes the real samples from the virtual samples.
In order to further improve the performance of the target fault diagnosis model, the invention also introduces an unsupervised feature learning mechanism, and enhances the diagnosis capability of the target fault diagnosis model on the target domain sample by learning the implicit feature representation of the source domain fault sample and the target domain normal sample. After training, the fault type of the target domain data can be rapidly and accurately diagnosed by the target fault diagnosis model through feature extraction and identification of the fault sample without the target domain.
The invention builds the target fault diagnosis model under the condition of no target domain fault sample based on the countermeasure generation network, provides a fault diagnosis method of the target fault diagnosis model, has wide application prospect in the fault diagnosis field, can be applied to a plurality of fields such as industrial manufacture, electronic equipment, transportation and the like, and provides an effective solution for further development and promotion of the fault diagnosis technology.
The fault diagnosis apparatus provided by the present invention will be described below, and the fault diagnosis apparatus described below and the fault diagnosis method described above may be referred to correspondingly to each other.
Fig. 3 is a schematic structural diagram of a fault diagnosis apparatus provided by the present invention. Referring to fig. 3, the fault diagnosis apparatus provided by the present invention may include: an acquisition module 310 and a diagnostic module 320.
An acquisition module 310, configured to acquire target domain data to be diagnosed;
the diagnostic module 320 is configured to input the target domain data to be diagnosed to a target fault diagnosis model, so as to obtain a fault type of the target domain data to be diagnosed, where the target fault diagnosis model is obtained by training based on a source domain normal sample, a source domain fault sample, and a target domain normal sample, where the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and where the source domain fault sample and the target domain normal sample are used for optimizing a target domain feature generating network in the target fault diagnosis model.
According to the fault diagnosis device provided by the invention, the target fault diagnosis model is obtained by training the source domain fault sample and the target domain normal sample under the condition of no target domain fault sample, so that the high-efficiency and accurate fault diagnosis of the target domain data to be diagnosed can be realized; will help to increase the versatility and practicality of the fault diagnosis technique, providing a more reliable solution for equipment maintenance and troubleshooting in various fields, thereby increasing production efficiency and equipment reliability.
In some embodiments, the apparatus further comprises:
the first training module is used for inputting the source domain normal sample into a source domain characteristic generating network to generate a source domain normal sample characteristic before the target domain data to be diagnosed is input into a target fault diagnosis model to obtain the fault type of the target domain data to be diagnosed;
inputting the target domain normal sample to a target domain feature generation network to generate target domain normal sample features;
and performing countermeasure training on the source domain normal sample characteristics and the target domain normal sample characteristics through a domain discriminator, and updating the target domain characteristic generation network and the domain discriminator.
In some embodiments, the apparatus further comprises:
the model optimization module is used for executing at least one iterative optimization process after the target domain feature generation network is updated until the mixed domain classification loss obtained in the last iterative optimization process reaches a target loss value;
based on network parameters obtained in the last iterative optimization process, obtaining an optimized target domain feature generation network, wherein the optimized target domain feature generation network meets the condition that the difference value between a first inter-class distance and a second inter-class distance is smaller than a target difference value, the first inter-class distance is the inter-class distance between the normal sample of the target domain and the source domain fault sample, and the second inter-class distance is the inter-class distance between the normal sample of the source domain and the source domain fault sample;
generating a network based on the optimized target domain characteristics, and determining the target fault diagnosis model;
wherein the iterative optimization process comprises:
generating a network based on the target domain characteristics, and updating the target domain normal sample characteristics;
inputting source domain fault sample characteristics and target domain normal sample characteristics into a source domain fault diagnosis model, and calculating mixed domain classification loss, wherein the source domain fault sample characteristics are obtained after the source domain fault samples are input into a source domain characteristic generation network, and the source domain fault diagnosis model is obtained based on the source domain fault samples and the source domain normal samples in a training mode;
and adjusting network parameters of the target domain feature generation network based on the mixed domain classification loss.
In some embodiments, the source domain fault diagnosis model is constructed by:
constructing the source domain feature generation network based on a convolutional neural network;
constructing a fault classification network based on the full connection layer;
and constructing the source domain fault diagnosis model based on the fault classification network and the source domain feature generation network.
In some embodiments, the apparatus further comprises:
the second training module is configured to input the source domain fault sample feature and the source domain normal sample feature to the diagnostic classification network before the source domain fault sample feature and the target domain normal sample feature are input to a source domain fault diagnosis model, and determine a fault type corresponding to the source domain fault sample feature and a fault type corresponding to the source domain normal sample feature;
calculating a source domain classification loss based on the fault type corresponding to the source domain fault sample characteristics and the fault type corresponding to the source domain normal sample characteristics;
and training the source domain fault diagnosis model based on the source domain classification loss.
In some embodiments, the first training module is specifically configured to:
performing countermeasure training on the source domain normal sample characteristics and the target domain normal sample characteristics through a domain discriminator, and calculating domain discrimination loss;
updating the target domain feature generation network and the domain arbiter based on the domain discrimination loss.
Fig. 4 illustrates a physical schematic diagram of an electronic device, as shown in fig. 4, which may include: processor 410, communication interface (Communications Interface) 420, memory 430 and communication bus 440, wherein processor 410, communication interface 420 and memory 430 communicate with each other via communication bus 440. The processor 410 may invoke logic instructions in the memory 430 to perform a fault diagnosis method comprising:
acquiring target domain data to be diagnosed;
inputting the target domain data to be diagnosed into a target fault diagnosis model to obtain a fault type of the target domain data to be diagnosed, wherein the target fault diagnosis model is obtained by training based on a source domain normal sample, a source domain fault sample and a target domain normal sample, the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and the source domain fault sample and the target domain normal sample are used for optimizing target domain characteristics in the target fault diagnosis model to generate a network.
Further, the logic instructions in the memory 430 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, the computer program product comprising a computer program, the computer program being storable on a non-transitory computer readable storage medium, the computer program, when executed by a processor, being capable of performing the fault diagnosis method provided by the above methods, the method comprising:
acquiring target domain data to be diagnosed;
inputting the target domain data to be diagnosed into a target fault diagnosis model to obtain a fault type of the target domain data to be diagnosed, wherein the target fault diagnosis model is obtained by training based on a source domain normal sample, a source domain fault sample and a target domain normal sample, the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and the source domain fault sample and the target domain normal sample are used for optimizing target domain characteristics in the target fault diagnosis model to generate a network.
In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the fault diagnosis method provided by the above methods, the method comprising:
acquiring target domain data to be diagnosed;
inputting the target domain data to be diagnosed into a target fault diagnosis model to obtain a fault type of the target domain data to be diagnosed, wherein the target fault diagnosis model is obtained by training based on a source domain normal sample, a source domain fault sample and a target domain normal sample, the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and the source domain fault sample and the target domain normal sample are used for optimizing target domain characteristics in the target fault diagnosis model to generate a network.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A fault diagnosis method, characterized by comprising:
acquiring target domain data to be diagnosed;
inputting the target domain data to be diagnosed into a target fault diagnosis model to obtain a fault type of the target domain data to be diagnosed, wherein the target fault diagnosis model is obtained by training based on a source domain normal sample, a source domain fault sample and a target domain normal sample, the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and the source domain fault sample and the target domain normal sample are used for optimizing target domain characteristics in the target fault diagnosis model to generate a network.
2. The fault diagnosis method according to claim 1, wherein before the target domain data to be diagnosed is input to a target fault diagnosis model to obtain a fault type of the target domain data to be diagnosed, the method further comprises:
inputting the source domain normal sample to a source domain feature generation network to generate a source domain normal sample feature;
inputting the target domain normal sample to a target domain feature generation network to generate target domain normal sample features;
and performing countermeasure training on the source domain normal sample characteristics and the target domain normal sample characteristics through a domain discriminator, and updating the target domain characteristic generation network and the domain discriminator.
3. The fault diagnosis method according to claim 2, wherein after the updating the target domain feature generation network, the method further comprises:
executing at least one iterative optimization process until the mixed domain classification loss obtained in the last iterative optimization process reaches a target loss value;
based on network parameters obtained in the last iterative optimization process, obtaining an optimized target domain feature generation network, wherein the optimized target domain feature generation network meets the condition that the difference value between a first inter-class distance and a second inter-class distance is smaller than a target difference value, the first inter-class distance is the inter-class distance between the normal sample of the target domain and the source domain fault sample, and the second inter-class distance is the inter-class distance between the normal sample of the source domain and the source domain fault sample;
generating a network based on the optimized target domain characteristics, and determining the target fault diagnosis model;
wherein the iterative optimization process comprises:
generating a network based on the target domain characteristics, and updating the target domain normal sample characteristics;
inputting source domain fault sample characteristics and target domain normal sample characteristics into a source domain fault diagnosis model, and calculating mixed domain classification loss, wherein the source domain fault sample characteristics are obtained after the source domain fault samples are input into a source domain characteristic generation network, and the source domain fault diagnosis model is obtained based on the source domain fault samples and the source domain normal samples in a training mode;
and adjusting network parameters of the target domain feature generation network based on the mixed domain classification loss.
4. A fault diagnosis method according to claim 3, wherein the source domain fault diagnosis model is constructed by:
constructing the source domain feature generation network based on a convolutional neural network;
constructing a fault classification network based on the full connection layer;
and constructing the source domain fault diagnosis model based on the fault classification network and the source domain feature generation network.
5. The fault diagnosis method according to claim 4, wherein before the source domain fault sample feature and the target domain normal sample feature are input to a source domain fault diagnosis model, the method further comprises:
inputting the source domain fault sample characteristics and the source domain normal sample characteristics into the diagnosis classification network, and determining fault types corresponding to the source domain fault sample characteristics and the source domain normal sample characteristics;
calculating a source domain classification loss based on the fault type corresponding to the source domain fault sample characteristics and the fault type corresponding to the source domain normal sample characteristics;
and training the source domain fault diagnosis model based on the source domain classification loss.
6. The fault diagnosis method according to claim 2, wherein the training the source domain normal sample feature and the target domain normal sample feature against by a domain discriminator, updating the target domain feature generation network and the domain discriminator, comprises:
performing countermeasure training on the source domain normal sample characteristics and the target domain normal sample characteristics through a domain discriminator, and calculating domain discrimination loss;
updating the target domain feature generation network and the domain arbiter based on the domain discrimination loss.
7. A fault diagnosis apparatus characterized by comprising:
the acquisition module is used for acquiring target domain data to be diagnosed;
the diagnosis module is used for inputting the target domain data to be diagnosed into a target fault diagnosis model to obtain the fault type of the target domain data to be diagnosed, the target fault diagnosis model is obtained by training based on a source domain normal sample, a source domain fault sample and a target domain normal sample, the source domain normal sample and the target domain normal sample are used for performing countermeasure training, and the source domain fault sample and the target domain normal sample are used for optimizing a target domain feature generation network in the target fault diagnosis model.
8. 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 fault diagnosis method of any one of claims 1 to 6 when the program is executed by the processor.
9. A non-transitory computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements the fault diagnosis method according to any one of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the fault diagnosis method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311386862.8A CN117648600A (en) | 2023-10-24 | 2023-10-24 | Fault diagnosis method, fault diagnosis device, electronic apparatus, and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311386862.8A CN117648600A (en) | 2023-10-24 | 2023-10-24 | Fault diagnosis method, fault diagnosis device, electronic apparatus, and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117648600A true CN117648600A (en) | 2024-03-05 |
Family
ID=90042304
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311386862.8A Pending CN117648600A (en) | 2023-10-24 | 2023-10-24 | Fault diagnosis method, fault diagnosis device, electronic apparatus, and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117648600A (en) |
-
2023
- 2023-10-24 CN CN202311386862.8A patent/CN117648600A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111914873B (en) | Two-stage cloud server unsupervised anomaly prediction method | |
| CN109302410A (en) | A kind of internal user anomaly detection method, system and computer storage medium | |
| CN112257263B (en) | Equipment residual life prediction system based on self-attention mechanism | |
| CN110298374B (en) | Driving track energy consumption analysis method and device based on deep learning | |
| Shajihan et al. | CNN based data anomaly detection using multi-channel imagery for structural health monitoring | |
| CN115184054B (en) | Mechanical equipment semi-supervised fault detection and analysis method, device, terminal and medium | |
| CN116610092A (en) | Method and system for vehicle analysis | |
| CN115791174B (en) | Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium | |
| CN114418189A (en) | Water quality grade prediction method, system, terminal device and storage medium | |
| CN116523177A (en) | Vehicle energy consumption prediction method and device integrating mechanism and deep learning model | |
| CN117036843A (en) | Target detection model training method, target detection method and device | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| CN117061318A (en) | Fault diagnosis method, device, equipment, medium and product | |
| CN110135428B (en) | Image segmentation processing method and device | |
| Duan et al. | Serial transfer learning (STL) theory for processing data insufficiency: Fault diagnosis of transformer windings | |
| Zhang et al. | Sonar image quality evaluation using deep neural network | |
| CN117407665A (en) | Retired battery time sequence data missing value filling method based on generation countermeasure network | |
| CN116543538B (en) | Internet of things fire-fighting electrical early warning method and early warning system | |
| CN117632659A (en) | Log exception processing method, device, equipment and medium | |
| Chen et al. | Application of Data‐Driven Iterative Learning Algorithm in Transmission Line Defect Detection | |
| CN115865458B (en) | Network attack behavior detection method, system and terminal based on LSTM and GAT algorithm | |
| CN116630989A (en) | Visual fault detection method and system for intelligent ammeter, electronic equipment and storage medium | |
| CN117648600A (en) | Fault diagnosis method, fault diagnosis device, electronic apparatus, and medium | |
| CN115174421B (en) | Network fault prediction method and device based on self-supervision unwrapping hypergraph attention | |
| CN116359738A (en) | Method, device, equipment and storage medium for monitoring health state of battery |
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 |