CN111738455B - Fault diagnosis method and system based on integration domain self-adaptation - Google Patents
Fault diagnosis method and system based on integration domain self-adaptation Download PDFInfo
- Publication number
- CN111738455B CN111738455B CN202010490493.7A CN202010490493A CN111738455B CN 111738455 B CN111738455 B CN 111738455B CN 202010490493 A CN202010490493 A CN 202010490493A CN 111738455 B CN111738455 B CN 111738455B
- Authority
- CN
- China
- Prior art keywords
- domain
- data
- feature
- fault diagnosis
- different
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The utility model provides a fault diagnosis method and system based on integration domain self-adaptation, utilize two feature extractors to project source domain and target domain data to different feature spaces, wherein one feature extractor learns the characteristic based on domain confrontation learning, another feature extractor learns as the loss function with the biggest mean value difference, different feature extractors use different loss functions, obtain different classification results, and integrate the classification result into domain self-adaptation, output and obtain the fault diagnosis result, when there is great difference between two domains, effectively extract the feature expression in the data, greatly improve the performance of fault diagnosis.
Description
Technical Field
The disclosure belongs to the technical field of fault diagnosis, and relates to a fault diagnosis method and system based on integration domain self-adaptation.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Fault diagnosis is the key to ensure the safety of industrial production. In recent years, with the development of machine learning and big data, intelligent fault diagnosis methods based on data driving are rapidly developed. However, although the conventional machine learning method has achieved a certain success in fault diagnosis, the following problems exist in use: 1) they need to be used in conjunction with feature extraction methods, and the selection of features will largely influence the final classification result. In addition, the feature extraction network and the classifier are designed separately, which consumes a lot of time and cannot realize global optimization. 2) Most of the methods belong to shallow structures, and the effective feature representation and the nonlinear mapping relation of a complex system are difficult to learn.
The deep learning algorithm can be used for self-adaptively extracting fault characteristics, has a complex network structure and the like, and is widely applied to mechanical fault diagnosis. Deep learning algorithms such as Deep Belief Networks (DBN) and Convolutional Neural Networks (CNN) are prominent in fault diagnosis and health monitoring. However, due to the complexity of the industrial environment in the field, operating conditions (such as mechanical load and speed) or the device under test may vary. In practical application, because the domains of the training data set and the testing data set are different, the diagnosis precision of the well-trained fault diagnosis model to the testing data set is low.
Although the problem that the domains of the training data set and the testing data set are not matched can be solved to a certain extent by the domain-adaptive fault diagnosis method, if the source domain data and the target domain data have large differences, the method of projecting the data to the same feature space loses some important feature representations, and further the performance of the trained fault diagnosis model is poor in practical application.
Disclosure of Invention
The invention provides a fault diagnosis method and system based on integrated domain self-adaptation to solve the problems, and the method and system effectively extract feature expression in data when a large difference exists between two domains, thereby greatly improving the performance of fault diagnosis.
According to some embodiments, the following technical scheme is adopted in the disclosure:
an integrated domain self-adaptive fault diagnosis method comprises the following steps:
the method comprises the steps that two feature extractors are used for projecting source domain data and target domain data into different feature spaces, one feature extractor learns features based on domain confrontation learning, the other feature extractor learns the maximum mean difference as a loss function, the different feature extractors use different loss functions to obtain different classification results, the classification results are integrated into domain self-adaptation, and the fault diagnosis results are output.
As an alternative embodiment, data preprocessing is performed on the source domain data and the target domain data, specifically including dividing the data into data segments of set length, and normalizing the divided data.
As an alternative embodiment, different labels are defined for the preprocessed source domain and target domain data, and the first feature extractor and the first classification network are used for training in combination with the domain prediction network on the basis of domain confrontation training, so that the first feature extractor and the domain prediction network obtain optimal parameters.
As an alternative embodiment, the training is performed based on the maximum average difference of the radial basis function kernels, using a second feature extractor and a second classification network.
As an alternative embodiment, in the integration step, the two classification results are averaged to obtain the final prediction result.
An integrated domain adaptation-based fault diagnosis system comprising:
the data preprocessing module is configured to divide the acquired source domain data and the acquired target domain data and normalize the data;
the feature processing module comprises two feature extractors and projects the data of a source domain and a target domain into different feature spaces, wherein one feature extractor learns features based on domain confrontation learning, the other feature extractor learns by taking the maximum mean difference as a loss function, and different feature extractors use different loss functions to obtain different classification results;
and the integration module is configured to integrate the classification result into the domain self-adaption and output the obtained fault diagnosis result.
A computer readable storage medium having stored therein a plurality of instructions adapted to be loaded by a processor of a terminal device and to execute a method for integrated domain adaptive based fault diagnosis as described.
A terminal device comprising a processor and a computer readable storage medium, the processor being configured to implement instructions; the computer readable storage medium is used for storing a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the fault diagnosis method based on integration domain self-adaption.
Compared with the prior art, the beneficial effect of this disclosure is:
the present disclosure introduces an efficient structure to model and integrate different feature representations between source domain and target domain data into domain adaptation. When a large difference exists between the two domains, the feature expression in the data is effectively extracted, and the performance of fault diagnosis is greatly improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure and are not to limit the disclosure.
FIG. 1(a) is a prior art domain confrontation training network framework;
FIG. 1(b) is a schematic diagram of a process for predicting test data using a trained model;
FIG. 2 is a network architecture diagram of the present disclosure;
fig. 3 is a flow chart of fault diagnosis performed by the trained network of the present disclosure.
The specific implementation mode is as follows:
the present disclosure is further described with reference to the following drawings and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
In domain confrontation training, source domain data and target domain data are projected into the same subspace, and a feature extractor outputs two domain invariant features. In most cases, domain confrontation training is effective and applicable. However, if there is a large difference between the two domains, some representative features may be lost when only one feature model is considered. Here we introduce an efficient structure to model the different feature representations between source domain and target domain data and integrate them into domain adaptation. In particular, two feature extractors are used to project data into different feature spaces. One of the feature extractors learns features based on domain confrontation learning, the other feature extractor learns the Maximum Mean Difference (MMD) as a loss function, and different feature extractors use different loss functions. In addition, since the two feature extractors use separate classifiers, ensemble learning is used to obtain the final result.
Specifically, a conventional Domain Adaptive Training (DAT) plays an important role in Domain adaptation, and a framework of the Domain adaptive Training is shown in fig. 1 (a). Let training data be XsThe training label is YsTest data is XtThe feature extractors of the training data and the test data are respectively MsAnd MtThe classification network is CsThe domain prediction network is D. In the training process, training data and test data are first input into the feature extractor. Feature M of training datasAs input to the classification network. M is trained using a back propagation algorithm to minimize the loss function shown in equation (1)sAnd CsThe parameter (1).
Wherein N is the number of classes, and when k is ysWhen it is time toCs(Ms(xs) Represents the probability that the classification is correct. Then using MsParameter pair MtInitialization is performed. Training domain prediction network D and feature extractor M using a method of countertrainings、Mt. First, the domain prediction network D aims to distinguish the characteristics of the target domain and the source domain, and connects Ms(Xs) And Mt(Xt) There are two categories. Let the source domain data have a characteristic label of 1,if the feature label of the target domain data is 0, the loss function of the domain prediction network D is defined as:
in the confrontation training, M is settSeen as a generator using the target domain data as input. Expectation of Mt(Xt) The generated data can trick the domain prediction network. In other words, the domain prediction network D will Mt(Xt) The source domain data and the target domain data are projected to the same feature space by being mistaken for the features of the source domain and marked as 1. Therefore, M is trained using a method that minimizes the following loss functiont:
After the domain confrontation training method, the prediction result of the test data can pass through the feature extractor MtAnd a classification network CsObtained as shown in FIG. 1 (b).
The scheme of the disclosure, as shown in fig. 2, includes:
(1) data preprocessing:
data preprocessing includes data segmentation and data normalization. The data division means dividing the vibration signal data into data segments having the same length. For example, a complete segment of vibration signal data contains 48000 data segments, and the data may be divided into 1024,2048 or 4096 data segments by data division. The data segmentation divides the high-frequency vibration signal data into data sections with equal length, so that the subsequent machine learning algorithm can be favorably used for data analysis, and the data segmentation is a general data preprocessing method in the field of machine learning.
After segmentation, the training data set and the test data set are represented asAndwherein n is1And n2Respectively represent XsAnd XtNumber of samples in (1). The data normalization can be expressed as:
where mean (x)s,i) Denotes xs,iAverage value of (a), std (x)t,j) Denotes xs,iStandard deviation of (2). Data normalization is also a common method for data and processing in machine learning.
(2) Training feature representation based on domain confrontation training
In this process, as shown in fig. 2, a feature extractor (feature extractor I), a classifier (classification network I) is trained in combination with the domain prediction network based on the domain confrontation training. Let the training data set beTraining dataset label YsThe test data set isThe feature extractor I is MIThe classification network I is CIThe domain prediction network is D.Andare respectively asAndthe characteristic expression of (1). To train a domain prediction network, given source domain dataLabel 1, target domain dataThe label is 0, and the loss function of the domain prediction network is defined as:
employing counter training to make MIAnd D obtaining optimal parameters, training M by minimizing the loss functionI:
M is then trained using the following classification loss functionIAnd CIThe parameters of (2):
wherein N is the number of failure classes, and when k is ysWhen it is time toIndicating the probability that the source domain classification is correct. After training, the predicted results of the test data are passed through feature extractor MIAnd a classification network CIAnd (4) obtaining.
(3) Maximum Mean Difference (MMD) based feature representation training
Further, MMD-based feature representation and training of classifiers. Let feature extractor II be M as shown in FIG. 2IIThe classifier II is CII。Andare respectively asAndusing MMD based on Radial Basis Function (RBF) kernel, we can write as:
where G (-) denotes an RBF core, which may be denoted as G (x)1,x2)=exp(-||x1-x2||2τ). Due to MIICan be learned adaptive to the bandwidth τ, and τ can be set to an arbitrary value, here set to 1.
Classifier CIIAnd MIISimultaneous training, CIITraining was performed using the following loss function:
the method adopts two classifiers, and each classifier adopts a feature extractor.
(4) The final prediction result is obtained by ensemble learning
The method provided by the scheme uses two classification networks. To obtain the final prediction result, the two classification results need to be averaged. The output of the two classifiers is averaged based on an ensemble learning method. Let the predicted output result of the classification network I be YIThe predicted output result of the classification network II is YIIThen the final predicted resultIs composed of
(5) Example of network architecture implementation
The proposed network architecture is as shown in table I. Batch Normalization (BN) and leakage corrected Linear units (leakage ReLU) are added after each convolutional layer, and BN, leakage ReLU and dropout are added after each fully-connected layer.
Table 1 network architecture for the method presented
The integrated domain adaptive fault diagnosis method introduces an effective structure to model different feature representations between source domain data and target domain data, and integrates the feature representations into the domain adaptation. When a large difference exists between the two domains, the feature expression in the data is effectively extracted, and the performance of fault diagnosis is greatly improved.
As will be appreciated by one skilled in the art, embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present disclosure 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 so forth) having computer-usable program code embodied therein.
The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams 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 flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
Although the present disclosure has been described with reference to specific embodiments, it should be understood that the scope of the present disclosure is not limited thereto, and those skilled in the art will appreciate that various modifications and changes can be made without departing from the spirit and scope of the present disclosure.
Claims (9)
1. A fault diagnosis method based on integration domain self-adaptation is characterized in that: the method comprises the following steps: introducing an effective structure to model different feature expressions between source domain data and target domain data, integrating the feature expressions into domain self-adaptation, and effectively extracting feature expressions in the data; specifically, two classifiers are adopted, two classification networks are used, each classifier adopts a feature extractor, the two feature extractors are used for projecting source domain data and target domain data into different feature spaces, one feature extractor learns features based on domain confrontation learning, the other feature extractor learns by taking the maximum mean difference as a loss function, different feature extractors use different loss functions to obtain different classification results, the classification results are integrated into domain self-adaptation, and fault diagnosis results are output; in the integration step, the two classification results are averaged to obtain a final prediction result.
2. The integrated domain adaptive-based fault diagnosis method as claimed in claim 1, wherein: data preprocessing is carried out on data of a source domain and data of a target domain, and the data preprocessing specifically comprises the steps of dividing the data into data sections with set lengths and normalizing the divided data.
3. The integrated domain adaptive-based fault diagnosis method as claimed in claim 1, wherein: different labels are defined for the preprocessed source domain data and the preprocessed target domain data, and the first feature extractor and the first classification network are used for training in combination with the domain prediction network on the basis of domain confrontation training, so that the first feature extractor and the domain prediction network can obtain optimal parameters.
4. The integrated domain adaptive-based fault diagnosis method as claimed in claim 1, wherein: training is performed based on the maximum average difference of the radial basis function kernels using a second feature extractor and a second classification network.
5. A fault diagnosis system based on integration domain self-adaptation is characterized in that: the method comprises the following steps:
the data preprocessing module is configured to divide the acquired source domain data and the acquired target domain data and normalize the data;
the feature processing module comprises two feature extractors and projects the data of a source domain and a target domain into different feature spaces, wherein one feature extractor learns features based on domain confrontation learning, the other feature extractor learns by taking the maximum mean difference as a loss function, and different feature extractors use different loss functions to obtain different classification results;
the integration module is configured to integrate the classification result into the domain self-adaption and output a fault diagnosis result; in the integration step, the two classification results are averaged to obtain a final prediction result.
6. The integrated domain adaptive-based fault diagnosis system according to claim 5, wherein: different labels are defined for the preprocessed source domain data and the preprocessed target domain data, and the first feature extractor and the first classification network are used for training in combination with the domain prediction network on the basis of domain confrontation training, so that the first feature extractor and the domain prediction network can obtain optimal parameters.
7. The integrated domain adaptive-based fault diagnosis system according to claim 5, wherein: training is performed based on the maximum average difference of the radial basis function kernels using a second feature extractor and a second classification network.
8. A computer-readable storage medium characterized by: a plurality of instructions are stored, wherein the instructions are suitable for being loaded by a processor of the terminal equipment and executing the fault diagnosis method based on the integration domain self-adaption in any one of claims 1-4.
9. A terminal device is characterized in that: the system comprises a processor and a computer readable storage medium, wherein the processor is used for realizing instructions; the computer readable storage medium is used for storing a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the fault diagnosis method based on integration domain self-adaption in any one of claims 1-4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010490493.7A CN111738455B (en) | 2020-06-02 | 2020-06-02 | Fault diagnosis method and system based on integration domain self-adaptation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010490493.7A CN111738455B (en) | 2020-06-02 | 2020-06-02 | Fault diagnosis method and system based on integration domain self-adaptation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111738455A CN111738455A (en) | 2020-10-02 |
CN111738455B true CN111738455B (en) | 2021-05-11 |
Family
ID=72648227
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010490493.7A Active CN111738455B (en) | 2020-06-02 | 2020-06-02 | Fault diagnosis method and system based on integration domain self-adaptation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111738455B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112683532B (en) * | 2020-11-25 | 2022-04-05 | 西安交通大学 | Cross-working condition countermeasure diagnostic method for bearing |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109308318A (en) * | 2018-08-14 | 2019-02-05 | 深圳大学 | Training method, device, equipment and the medium of cross-domain texts sentiment classification model |
CN110210371A (en) * | 2019-05-29 | 2019-09-06 | 华南理工大学 | A kind of aerial hand-written inertia sensing signal creating method based on depth confrontation study |
CN110728377A (en) * | 2019-10-21 | 2020-01-24 | 山东大学 | Intelligent fault diagnosis method and system for electromechanical equipment |
CN110837850A (en) * | 2019-10-23 | 2020-02-25 | 浙江大学 | Unsupervised domain adaptation method based on counterstudy loss function |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10956817B2 (en) * | 2018-04-18 | 2021-03-23 | Element Ai Inc. | Unsupervised domain adaptation with similarity learning for images |
CN109492099B (en) * | 2018-10-28 | 2022-03-15 | 北京工业大学 | Cross-domain text emotion classification method based on domain impedance self-adaption |
CN110309798B (en) * | 2019-07-05 | 2021-05-11 | 中新国际联合研究院 | Face spoofing detection method based on domain self-adaptive learning and domain generalization |
-
2020
- 2020-06-02 CN CN202010490493.7A patent/CN111738455B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109308318A (en) * | 2018-08-14 | 2019-02-05 | 深圳大学 | Training method, device, equipment and the medium of cross-domain texts sentiment classification model |
CN110210371A (en) * | 2019-05-29 | 2019-09-06 | 华南理工大学 | A kind of aerial hand-written inertia sensing signal creating method based on depth confrontation study |
CN110728377A (en) * | 2019-10-21 | 2020-01-24 | 山东大学 | Intelligent fault diagnosis method and system for electromechanical equipment |
CN110837850A (en) * | 2019-10-23 | 2020-02-25 | 浙江大学 | Unsupervised domain adaptation method based on counterstudy loss function |
Also Published As
Publication number | Publication date |
---|---|
CN111738455A (en) | 2020-10-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10187415B2 (en) | Cognitive information security using a behavioral recognition system | |
US20220075946A1 (en) | Perceptual associative memory for a neuro-linguistic behavior recognition system | |
US20240071037A1 (en) | Mapper component for a neuro-linguistic behavior recognition system | |
Mao et al. | Deep Learning of Segment-Level Feature Representation with Multiple Instance Learning for Utterance-Level Speech Emotion Recognition. | |
US20220012422A1 (en) | Lexical analyzer for a neuro-linguistic behavior recognition system | |
CN112784929B (en) | Small sample image classification method and device based on double-element group expansion | |
CN112464233B (en) | RNN-based malicious software detection method on cloud platform | |
CN111653275A (en) | Method and device for constructing voice recognition model based on LSTM-CTC tail convolution and voice recognition method | |
CN112860904B (en) | External knowledge-integrated biomedical relation extraction method | |
Kwon et al. | Multi-scale speaker embedding-based graph attention networks for speaker diarisation | |
CN114048729A (en) | Medical document evaluation method, electronic device, storage medium, and program product | |
CN116337448A (en) | Method, device and storage medium for diagnosing faults of transfer learning bearing based on width multi-scale space-time attention | |
Wang et al. | Contrastive Predictive Coding of Audio with an Adversary. | |
CN111738455B (en) | Fault diagnosis method and system based on integration domain self-adaptation | |
Fonseca et al. | Model-agnostic approaches to handling noisy labels when training sound event classifiers | |
CN113220865B (en) | Text similar vocabulary retrieval method, system, medium and electronic equipment | |
CN113849653A (en) | Text classification method and device | |
US11586874B2 (en) | Anomaly score adjustment across anomaly generators | |
CN111832815A (en) | Scientific research hotspot prediction method and system | |
CN117235286B (en) | Attention-strengthening entity relation extraction model, construction method thereof and storage medium | |
CN114330320A (en) | Entity extraction method, training method of first entity extraction model and related device | |
CN117831566A (en) | Multilingual speech emotion recognition system based on domain countermeasure learning | |
Dongfang | Electronic Music Signal Identification Model Based On Machine Learning Algorithm | |
CN116245148A (en) | Compression method and device for automatic driving model, electronic equipment and storage medium | |
CN112698977A (en) | Server fault positioning method, device, equipment and medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |