CN112464559B - Data fusion method and system for fuel cell engine fault prediction - Google Patents
Data fusion method and system for fuel cell engine fault prediction Download PDFInfo
- Publication number
- CN112464559B CN112464559B CN202011332973.7A CN202011332973A CN112464559B CN 112464559 B CN112464559 B CN 112464559B CN 202011332973 A CN202011332973 A CN 202011332973A CN 112464559 B CN112464559 B CN 112464559B
- Authority
- CN
- China
- Prior art keywords
- sample
- data
- fusion
- fault
- fault sample
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/30—Hydrogen technology
- Y02E60/50—Fuel cells
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Fuel Cell (AREA)
Abstract
The invention discloses a data fusion method and a system for predicting faults of a fuel cell engine, wherein the method comprises the following steps of S1, obtaining an actually measured fault sample through a test; s2, obtaining a simulated fault sample; s3, performing feature level fusion on the actual measurement fault sample simulation fault sample to obtain a first fusion sample; s4, carrying out data level fusion on the first fusion sample and the actually measured fault sample to obtain a second fusion sample; and S5, the second fusion sample and the actually measured fault sample are used as training samples of a fuel cell engine fault prediction model. The invention can solve the problem of insufficient experimental data of the fault prediction of the fuel cell engine.
Description
Technical Field
The invention relates to the technical field of fault diagnosis of fuel cell automobile engines, in particular to a data fusion method and a data fusion system for fault prediction of fuel cell engines.
Background
With the continuous development of fuel cell automobile technology, fuel cell automobiles have begun to gradually go to practical use. However, in the prior art, since the fuel cell automobile has not been put into practical use in a large amount, there is still little concern about the failure diagnosis of the fuel cell automobile engine.
And less research is needed for predicting the failure of the fuel cell engine. The difficulty in predicting the failure of the fuel cell engine is mainly that: since fuel cell automobiles have not been widely used, the failure data of the fuel cell engine is too small, and it is difficult for the existing prediction model to predict based on the too small data amount. In addition, even through a large-scale test, it is difficult to obtain a sufficient amount of data. This is because, in a fuel cell engine, the probability of occurrence of failure is small, and it is common that actual experimental data can be obtained only after long-term operation, in the case of part degradation, or the like. In this case, to obtain sufficient data, it takes years for data to accumulate to be able to be studied accordingly.
Therefore, how to develop a study for predicting the failure of a fuel cell engine is one of the important problems to be solved in the art under the condition of insufficient experimental data.
Disclosure of Invention
The invention aims to provide a data fusion method for fuel cell engine fault prediction, which aims to solve the defects in the prior art and can solve the problem of insufficient experimental data of the fuel cell engine fault prediction.
The invention provides a data fusion method for predicting faults of a fuel cell engine, which comprises the following steps of,
s1, obtaining an actual measurement fault sample through a test;
s2, obtaining a simulated fault sample;
s3, performing feature level fusion on the actual measurement fault sample simulation fault sample to obtain a first fusion sample;
s4, carrying out data level fusion on the first fusion sample and the actually measured fault sample to obtain a second fusion sample;
and S5, the second fusion sample and the actually measured fault sample are used as training samples of a fuel cell engine fault prediction model.
The data fusion method for fuel cell engine fault prediction as described above, wherein, optionally, step S2 includes,
s21, building a simulation model, and acquiring multi-source information from the simulation model; wherein the multi-source information includes vibration data, temperature data, battery data, and pressure data;
s22, extracting characteristics of the multi-source information to obtain a first characteristic signal;
s23, performing feature enhancement processing on the first feature signal to obtain a second feature signal;
s24, performing deep fusion on the second characteristic signals by using a deep convolutional neural network to obtain a simulated fault sample.
The data fusion method for fuel cell engine fault prediction as described above, wherein, optionally, step S3 includes,
s31, preprocessing an actual measurement fault sample and a simulation fault sample;
s32, constructing an antagonistic neural network, and setting a generator and a discriminator;
s33, training the discriminator by taking the actually measured fault sample as training data;
s34, training the generator by taking the simulated fault sample as training data;
s35, generating a first data sample by using the trained generator; inputting the first data sample into a discriminator, and identifying the data in the first data sample as true by the discriminator as data in a first fusion sample;
s36, repeating the step S35 until the data volume of the first fusion sample reaches a set value.
The data fusion method for fuel cell engine fault prediction as described above, wherein, optionally, the generator and the arbiter both employ 6-layer fully connected neural networks;
the hidden layer node number of the generator is 32-64-128-256, and the hidden layer node number of the discriminator is 256-128-64-32;
the activation functions of the generator and the arbiter are SELU functions.
The data fusion method for fuel cell engine fault prediction as described above, wherein, optionally, in step S32, the antagonistic neural network model constructed is a waserstein GAN network model.
The data fusion method for fuel cell engine failure prediction as described above, wherein, optionally, step S31 includes,
s311, carrying out noise reduction treatment on the actually measured fault sample and the simulated fault sample;
s312, carrying out A/D conversion on the simulated fault sample and the actually measured fault sample after noise reduction processing according to a unified rule so as to convert data in the actually measured fault sample of the fault sample into digital signals;
s313, extracting feature vectors of the actual measurement fault sample and the simulation fault sample;
the invention provides a data fusion system for predicting failure of a fuel cell engine, which comprises the following steps of,
the data acquisition module is used for acquiring actual measurement fault data so as to acquire an actual measurement fault sample;
the simulation module is used for obtaining a simulated fault sample of the combustion battery engine through simulation;
the first data fusion module is electrically connected with the data acquisition module and the simulation module and is used for carrying out feature level fusion on the actually measured fault sample and the simulated fault sample so as to obtain a first fusion sample;
and the second data fusion module is electrically connected with the first data fusion module and the data acquisition module and is used for carrying out data fusion on the first fusion sample and the actually-measured fault sample so as to obtain a second fusion sample.
The data fusion system for fuel cell engine fault prediction as described above, wherein optionally the data acquisition module comprises a vibration sensor, a temperature sensor, a pressure sensor, and a battery data collector.
A data fusion system for fuel cell engine fault prediction as described above, wherein optionally the second data fusion module comprises an antagonistic neural network model;
the countermeasure neural network model comprises a generator and a discriminator; the generator and the discriminator are 6-layer fully-connected neural networks, the number of hidden layer nodes of the generator is 32-64-128-256, and the number of hidden layer nodes of the discriminator is 256-128-64-32; the activation functions of the generator and the discriminator are SELU functions;
the generator is trained by taking the simulated fault sample as training data; the discriminator is trained by using the actually measured fault sample as training data.
A data fusion system for fuel cell engine fault prediction as described above, wherein optionally the antagonistic neural network model is a wasperstein GAN network model.
Compared with the prior art, the method has the advantages that the actually measured fault samples obtained through the test are fused with the simulated fault samples obtained through the model simulation, so that a sufficient number of second fused samples are obtained. Therefore, training samples of the fuel cell engine fault prediction model can be ensured, and the research of the fuel cell engine fault prediction can be smoothly performed under the current condition that the number of actually measured fault samples is small.
When fusion is carried out on the actually measured fault sample and the simulated fault sample, fusion is carried out on the actually measured fault sample and the simulated fault sample to obtain a first fusion sample; carrying out data fusion on the first fusion sample and the actually measured fault sample to obtain a second fusion sample; therefore, the actually measured fault samples are repeatedly applied, when the actually measured fault sample quantity is far smaller than the simulated fault sample quantity, the proportion of the actually measured fault samples in fusion can be increased, and the second fusion sample can approach the vacuum fault samples as accurately as possible.
When the first fusion sample is obtained, the fusion between the actually measured fault sample and the simulated fault sample is the feature level fusion, and the data level fusion is carried out between the first fusion sample and the actually measured fault sample. Thus, the proportion of the actual measurement fault samples during fusion can be improved, and the fusion result caused by simply recycling the actual measurement fault samples can be prevented from excessively approaching the actual measurement fault samples.
Drawings
FIG. 1 is a flow chart of the steps of embodiment 1 of the present invention;
FIG. 2 is a flowchart showing the steps in step S2 of the present invention;
FIG. 3 is a flowchart showing the steps in step S3 of the present invention;
FIG. 4 is a flowchart showing the steps in step S31 of the present invention;
FIG. 5 is a schematic diagram showing the data fusion process in example 1 of the present invention;
fig. 6 is a block diagram of the structure of embodiment 2 of the present invention.
Detailed Description
The embodiments described below by referring to the drawings are illustrative only and are not to be construed as limiting the invention.
Example 1
Referring to fig. 1 to 5, the present invention provides a data fusion method for predicting failure of a fuel cell engine, which includes the following steps,
s1, obtaining an actual measurement fault sample through a test; in the specific implementation, the actually measured fault sample is obtained through a test or fault data generated in the actual application process of the fuel cell engine. In the prior art, since the application of the fuel cell engine is still in an early stage, and the fuel cell steam engine relates to the electrochemical field, the kinetic field, the thermodynamic field and the hydrodynamic field, the structure thereof is extremely complex. In the prior art, the accumulated fault data is difficult to support the development of the research work for predicting the fault of the fuel cell.
S2, obtaining a simulated fault sample. Specifically, based on a system dynamics domain, an electrochemical domain, a hydrodynamic domain, a thermal dynamic domain and other physical domains, a corresponding mathematical model is established to deeply study an operation mechanism and a failure mechanism of the fuel cell engine in a normal operation state, the influence of factors such as different operation modes, load disturbance, operation parameter change, structural vibration and the like of the system on internal and external state variables of the fuel cell is fully considered, chemical durability and degradation mechanisms of key parts such as bipolar plates, catalysts and membrane electrodes are studied, the influence of internal and external key state parameters of the fuel cell is analyzed, and the failure evolution rule of the fuel cell engine is analyzed to build a high-fidelity model of the fuel cell engine by integrating the normal operation mechanism, the failure mechanism and the degradation mechanism. The applicant has already described in detail in the patent 2020109599767 of the prior application and will not be described here again.
Specifically, step S2 includes:
s21, building a simulation model, and acquiring multi-source information from the simulation model; wherein the multi-source information includes vibration data, temperature data, battery data, and pressure data;
s22, extracting characteristics of the multi-source information to obtain a first characteristic signal;
s23, performing feature enhancement processing on the first feature signal to obtain a second feature signal;
s24, performing deep fusion on the second characteristic signals by using a deep convolutional neural network to obtain a simulated fault sample.
That is, in step S2, the simulated fault samples are generated by running the simulation model, and since the simulated fault samples are not final samples that can be used directly, in the subsequent steps, regeneration of the antagonistic neural network and screening using the discriminators are still required, in which case the generation of the simulated data is based on the simulation model, rather than being generated randomly. This can effectively improve the efficiency against the neural network.
Through the steps S22 to S24, the multi-source information is extracted with features, and then is subjected to deep fusion after the feature extraction, so that noise and redundant data in the multi-source information can be effectively and rapidly removed. That is, the fusion performed in steps S21 to S24 is feature level fusion, and can give the feature information necessary for decision to the maximum extent.
S3, performing feature level fusion on the actual measurement fault sample simulation fault sample to obtain a first fusion sample;
in particular, the present step includes,
s31, preprocessing an actual measurement fault sample and a simulation fault sample; specifically, the preprocessing of the actually measured fault sample and the simulated fault sample includes:
s311, carrying out noise reduction treatment on the actually measured fault sample and the simulated fault sample; in this way, noise can be effectively reduced to reduce the influence of noise on data fusion.
S312, carrying out A/D conversion on the simulated fault sample and the actually measured fault sample after noise reduction processing according to a unified rule so as to convert data in the actually measured fault sample of the fault sample into digital signals;
and S313, extracting the eigenvectors of the actual measurement fault sample and the simulated fault sample. The information after data preprocessing has more efficient characteristic information, and then characteristic extraction is carried out according to a certain rule, so that the information which finally needs to be subjected to fusion operation is obtained. That is, through steps S311 to S313, feature level fusion is performed on the actually measured failure sample and the simulated failure sample.
S32, constructing an antagonistic neural network, and setting a generator and a discriminator; the constructed antagonistic neural network model is a Wasserstein GAN network model. In the specific implementation, the generator and the discriminator are all 6 layers of fully-connected neural networks; the hidden layer node number of the generator is 32-64-128-256, and the hidden layer node number of the discriminator is 256-128-64-32; the activation functions of the generator and the arbiter are SELU functions. In this way, it can be ensured that the constructed antagonistic neural network can generate the first data sample quickly and efficiently.
In specific implementation, S33, the actual measurement fault sample is used as training data to train the arbiter.
S34, training the generator by taking the simulated fault sample as training data; in the training of the generator, the actual measurement fault sample and the simulation fault sample may be used together as training data to train the generator.
S35, generating a first data sample by using the trained generator; inputting the first data sample into a discriminator, and identifying the data in the first data sample as true by the discriminator as data in a first fusion sample; because the generator is trained based on measured and simulated fault samples, the generated fault samples have a higher accuracy than randomly generated samples. In this step, when the generator generates the first data sample and the arbiter performs the first data sample discrimination, the generator and the arbiter perform the joint training simultaneously, that is, the generator acquires the discrimination result of the arbiter and uses the discrimination result as a primary training process of the generator, and at the same time, the arbiter also uses the generation result of the generator and the discrimination result of the arbiter as a primary training process of the arbiter.
S36, repeating the step S35 until the data volume of the first fusion sample reaches a set value. By repeating step S35, a large number of data are obtained to meet the data size of the first fusion sample.
S4, carrying out data level fusion on the first fusion sample and the actually measured fault sample to obtain a second fusion sample; in the implementation, the first fusion sample and the actually measured fault sample are not further processed before the data-level fusion, and the data sources are directly associated and then fused. Therefore, the original data characteristics can be reserved to the greatest extent, and more detail information can be provided. In the step, as the first fusion sample is data subjected to primary feature level fusion, the data is extracted once, and noise caused by the data generated by the simulation model can be removed; and for the actually measured fault sample, the corresponding feature extraction is not performed, and all information of the actually measured fault sample is reserved.
Therefore, the reliability and the accuracy of the actually measured fault sample are far greater than those of the simulated fault sample, and in the two fusion processes, the feature level fusion and the data level fusion can be respectively carried out to ensure that: for the simulated fault sample, good information compression is achieved, required characteristic information can be given to the greatest extent while real-time performance can be guaranteed, meanwhile, data loss in the actually measured fault sample is prevented through data level fusion, and fusion accuracy is guaranteed.
And S5, the second fusion sample and the actually measured fault sample are jointly used as training samples of a fuel cell engine fault prediction model.
In the implementation, the second fusion sample may be a primary data-level fusion with the actually measured fault sample, or may be a training sample of the fuel cell engine fault prediction model after directly combining the second fusion sample with the implementation fault sample.
Compared with the prior art, the method has the advantages that the actually measured fault samples obtained through the test are fused with the simulated fault samples obtained through the model simulation, so that a sufficient number of second fused samples are obtained. Therefore, training samples of the fuel cell engine fault prediction model can be ensured, and the research of the fuel cell engine fault prediction can be smoothly performed under the current condition that the number of actually measured fault samples is small.
When fusion is carried out on the actually measured fault sample and the simulated fault sample, fusion is carried out on the actually measured fault sample and the simulated fault sample to obtain a first fusion sample; carrying out data fusion on the first fusion sample and the actually measured fault sample to obtain a second fusion sample; therefore, the actually measured fault samples are repeatedly applied, when the actually measured fault sample quantity is far smaller than the simulated fault sample quantity, the proportion of the actually measured fault samples in fusion can be increased, and the second fusion sample can approach the vacuum fault samples as accurately as possible.
When the first fusion sample is obtained, the fusion between the actually measured fault sample and the simulated fault sample is the feature level fusion, and the data level fusion is carried out between the first fusion sample and the actually measured fault sample. Thus, the proportion of the actual measurement fault samples during fusion can be improved, and the fusion result caused by simply recycling the actual measurement fault samples can be prevented from excessively approaching the actual measurement fault samples.
In the case of example 2,
referring to fig. 6, the present embodiment proposes a data fusion system for fuel cell engine failure prediction, which includes,
the data acquisition module is used for acquiring actual measurement fault data so as to acquire an actual measurement fault sample;
the simulation module is used for obtaining a simulated fault sample of the combustion battery engine through simulation;
the first data fusion module is electrically connected with the data acquisition module and the simulation module and is used for carrying out feature level fusion on the actually measured fault sample and the simulated fault sample so as to obtain a first fusion sample;
and the second data fusion module is electrically connected with the first data fusion module and the data acquisition module and is used for carrying out data fusion on the first fusion sample and the actually-measured fault sample so as to obtain a second fusion sample.
Specifically, the data acquisition module comprises a vibration sensor, a temperature sensor, a pressure sensor and a battery data acquisition device.
More specifically, the second data fusion module includes an antagonistic neural network model;
the countermeasure neural network model comprises a generator and a discriminator; the generator and the discriminator are 6-layer fully-connected neural networks, the number of hidden layer nodes of the generator is 32-64-128-256, and the number of hidden layer nodes of the discriminator is 256-128-64-32; the activation functions of the generator and the discriminator are SELU functions;
the generator is trained by taking the simulated fault sample as training data; the discriminator is trained by using the actually measured fault sample as training data.
In practice, the antagonistic neural network model is the Wasserstein GAN network model.
Note that in this embodiment, the same method as in embodiment 1 is used, and the technical effect described in embodiment 1 can be achieved. Therefore, the description is not repeated in this embodiment.
While the foregoing is directed to embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow.
Claims (3)
1. A data fusion method for fuel cell engine fault prediction, characterized by: comprises the following steps of the method,
s1, obtaining an actual measurement fault sample through a test;
s2, obtaining a simulated fault sample;
s3, performing feature level fusion on the actual measurement fault sample simulation fault sample to obtain a first fusion sample;
s4, carrying out data level fusion on the first fusion sample and the actually measured fault sample to obtain a second fusion sample;
s5, the second fusion sample and the actually measured fault sample are jointly used as training samples of a fuel cell engine fault prediction model;
the step S2 includes the steps of,
s21, building a simulation model, and acquiring multi-source information from the simulation model; wherein the multi-source information includes vibration data, temperature data, battery data, and pressure data; the model used in this step is: establishing a corresponding mathematical model based on a system dynamics domain, an electrochemical domain, a hydrodynamic domain and a thermal dynamic domain multi-physical domain, deeply researching an operation mechanism and a failure mechanism of the fuel cell engine in a normal operation state, considering the influence of a system on internal and external state variables of the fuel cell by different operation modes, load disturbance, operation parameter changes and structural vibration factors, researching chemical durability and degradation mechanisms of a key component bipolar plate, a catalyst and a membrane electrode, analyzing the influence of internal and external key state parameters of the fuel cell, and analyzing the failure evolution rule of the fuel cell engine by combining the normal operation mechanism, the failure mechanism and the degradation mechanism to establish a high-fidelity model of the fuel cell engine;
s22, extracting characteristics of the multi-source information to obtain a first characteristic signal;
s23, performing feature enhancement processing on the first feature signal to obtain a second feature signal;
s24, performing deep fusion on the second characteristic signals by using a deep convolutional neural network to obtain simulated fault samples;
the step S3 includes the steps of,
s31, preprocessing an actual measurement fault sample and a simulation fault sample;
s32, constructing an antagonistic neural network, and setting a generator and a discriminator;
s33, training the discriminator by taking the actually measured fault sample as training data;
s34, training the generator by taking the simulated fault sample as training data;
s35, generating a first data sample by using the trained generator; inputting the first data sample into a discriminator, and identifying the data in the first data sample as true by the discriminator as data in a first fusion sample;
s36, repeating the step S35 until the data volume of the first fusion sample reaches a set value;
the step S31 includes the steps of,
s311, carrying out noise reduction treatment on the actually measured fault sample and the simulated fault sample;
s312, carrying out A/D conversion on the simulated fault sample and the actually measured fault sample after noise reduction processing according to a unified rule so as to convert data in the actually measured fault sample of the fault sample into digital signals;
s313, extracting feature vectors of the actual measurement fault sample and the simulation fault sample;
a system for using the method comprises:
the data acquisition module is used for acquiring actual measurement fault data so as to acquire an actual measurement fault sample;
the simulation module is used for obtaining a simulated fault sample of the combustion battery engine through simulation;
the first data fusion module is electrically connected with the data acquisition module and the simulation module and is used for carrying out feature level fusion on the actually measured fault sample and the simulated fault sample so as to obtain a first fusion sample;
the second data fusion module is electrically connected with the first data fusion module and the data acquisition module and is used for carrying out data fusion on the first fusion sample and the actually measured fault sample so as to obtain a second fusion sample;
the data acquisition module comprises a vibration sensor, a temperature sensor, a pressure sensor and a battery data acquisition device;
the second data fusion module includes an antagonistic neural network model;
the countermeasure neural network model comprises a generator and a discriminator; the generator and the discriminator are 6-layer fully-connected neural networks, the number of hidden layer nodes of the generator is 32-64-128-256, and the number of hidden layer nodes of the discriminator is 256-128-64-32; the activation functions of the generator and the discriminator are SELU functions;
the generator is trained by taking the simulated fault sample as training data; the discriminator is trained by taking the actually measured fault sample as training data;
the antagonistic neural network model is a Wasserstein GAN network model.
2. The data fusion method for fuel cell engine fault prediction according to claim 1, wherein: the generator and the discriminator are all 6 layers of fully-connected neural networks;
the hidden layer node number of the generator is 32-64-128-256, and the hidden layer node number of the discriminator is 256-128-64-32;
the activation functions of the generator and the arbiter are SELU functions.
3. The data fusion method for fuel cell engine fault prediction according to claim 2, wherein: in step S32, the constructed antagonistic neural network model is the wasperstein GAN network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011332973.7A CN112464559B (en) | 2020-11-24 | 2020-11-24 | Data fusion method and system for fuel cell engine fault prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011332973.7A CN112464559B (en) | 2020-11-24 | 2020-11-24 | Data fusion method and system for fuel cell engine fault prediction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112464559A CN112464559A (en) | 2021-03-09 |
| CN112464559B true CN112464559B (en) | 2023-05-16 |
Family
ID=74799833
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011332973.7A Active CN112464559B (en) | 2020-11-24 | 2020-11-24 | Data fusion method and system for fuel cell engine fault prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112464559B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114267844B (en) * | 2021-11-09 | 2024-02-27 | 深圳市氢瑞燃料电池科技有限公司 | System and method for producing fuel cell polar plate |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414601A (en) * | 2019-07-30 | 2019-11-05 | 南京工业大学 | Photovoltaic module fault diagnosis method, system and equipment based on deep convolution countermeasure network |
| CN111947928A (en) * | 2020-08-10 | 2020-11-17 | 山东大学 | Multi-source information fusion bearing fault prediction system and method |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295714B (en) * | 2016-08-22 | 2020-01-21 | 中国科学院电子学研究所 | Multi-source remote sensing image fusion method based on deep learning |
| CN110780146B (en) * | 2019-12-10 | 2021-04-27 | 武汉大学 | Transformer fault identification and positioning diagnosis method based on multi-stage transfer learning |
| CN111400930B (en) * | 2020-04-09 | 2022-04-15 | 武汉大学 | Power equipment small sample fault diagnosis method and system based on virtual and real twin space |
| CN111652297B (en) * | 2020-05-25 | 2021-05-25 | 哈尔滨市科佳通用机电股份有限公司 | Fault picture generation method for image detection model training |
| CN111898660B (en) * | 2020-07-17 | 2022-10-04 | 武汉大学 | Hydrological simulation method for average fusion of multi-source data based on Bayesian mode |
-
2020
- 2020-11-24 CN CN202011332973.7A patent/CN112464559B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414601A (en) * | 2019-07-30 | 2019-11-05 | 南京工业大学 | Photovoltaic module fault diagnosis method, system and equipment based on deep convolution countermeasure network |
| CN111947928A (en) * | 2020-08-10 | 2020-11-17 | 山东大学 | Multi-source information fusion bearing fault prediction system and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112464559A (en) | 2021-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109726452B (en) | Online proton exchange membrane fuel cell fault diagnosis method based on impedance spectrum | |
| CN113870260B (en) | Welding defect real-time detection method and system based on high-frequency time sequence data | |
| CN110781592B (en) | Method and system for predicting residual life of equipment | |
| CN109000930B (en) | Turbine engine performance degradation evaluation method based on stacking denoising autoencoder | |
| CN110596506A (en) | Converter fault diagnosis method based on time convolution network | |
| CN112883497B (en) | Space valve reliability assessment method based on multi-source information fusion | |
| CN112541233A (en) | Rotary machine fault diagnosis method based on improved convolutional neural network | |
| CN112464559B (en) | Data fusion method and system for fuel cell engine fault prediction | |
| CN113866642A (en) | Lithium ion battery fault diagnosis method based on gradient lifting tree | |
| CN112666479A (en) | Battery service life prediction method based on charging cycle fusion | |
| CN115098962A (en) | Method for predicting residual life of mechanical equipment in degradation state based on hidden half Markov model | |
| Jouin et al. | PHM of Proton-Exchange Membrane Fuel Cells-A review. | |
| CN113765909A (en) | Big data detection method and system for coping with intelligent education data wind control | |
| CN110633314A (en) | Internet of vehicles data processing method and device | |
| CN108613820A (en) | A kind of online allophone monitoring algorithm for GIS bulk mechanicals defect diagonsis and positioning | |
| CN117056865A (en) | Method and device for diagnosing operation faults of machine pump equipment based on feature fusion | |
| CN111858351A (en) | Deep learning inference engine test method based on differential evaluation | |
| CN116741304A (en) | Diesel engine original exhaust nitrogen oxide prediction method based on Internet of vehicles big data | |
| CN116910574A (en) | Mechanical equipment diagnosis method and device based on meta-learning and time convolution network | |
| CN115456013A (en) | Wind turbine generator converter power module fault diagnosis method based on operation data | |
| CN115356599A (en) | Multi-mode urban power grid fault diagnosis method and system | |
| CN115267586A (en) | Lithium battery SOH evaluation method | |
| CN114858455A (en) | Rolling bearing fault diagnosis method and system based on improved GAN-OSNet | |
| Hou et al. | A vehicle alarm network for high-temperature fault diagnosis of electric vehicles | |
| CN113887452A (en) | Fault diagnosis optimization method based on correlation matrix |
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 |