CN118052154A - Electric drive system fault prediction method for few-sample data enhancement and migration - Google Patents
Electric drive system fault prediction method for few-sample data enhancement and migration Download PDFInfo
- Publication number
- CN118052154A CN118052154A CN202410452105.4A CN202410452105A CN118052154A CN 118052154 A CN118052154 A CN 118052154A CN 202410452105 A CN202410452105 A CN 202410452105A CN 118052154 A CN118052154 A CN 118052154A
- Authority
- CN
- China
- Prior art keywords
- fault
- data
- drive system
- electric drive
- migration
- 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
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000013508 migration Methods 0.000 title claims abstract description 32
- 230000005012 migration Effects 0.000 title claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 8
- 239000013598 vector Substances 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000003860 storage Methods 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
Landscapes
- Control Of Electric Motors In General (AREA)
Abstract
The invention provides a failure prediction method of an electric drive system with less sample data enhancement and migration, which comprises the following steps: acquiring operation data of the electric drive system in a normal working state, and pre-training an LSTM network model by using the operation data; collecting fault data of the electric drive system in a fault state, performing characteristic tag calibration operation on the fault data, and performing data enhancement processing on the fault data with few samples through a time sliding window; performing migration learning on the pre-trained LSTM network model to obtain a fault prediction model, and training the fault prediction model by using the enhanced fault data; and performing fault prediction by using the trained fault prediction model, and outputting a predicted fault curve. The invention has the beneficial effects that: less data can be used for training with high efficiency, and potential faults of the motor assembly can be accurately identified and predicted relative to a traditional model.
Description
Technical Field
The invention belongs to the field of fault test, and particularly relates to a fault prediction method of an electric drive system with less sample data enhancement and migration.
Background
The occurrence of various types of faults in the drive motor system and the electric drive assembly contained therein during the test process can lead to project stagnation, and particularly, the occurrence of the project development stage can greatly affect production. The electric drive assembly fault prediction model is used as a machine learning-based fault prediction method for identifying and predicting possible faults in advance, and is usually driven by data, so that the risk of sudden faults is reduced.
However, in the actual testing process, the sample size of the fault data is relatively small, the accuracy obtained by training the deep learning model is poor, the actual application requirement cannot be met, and how to train a robust model under the effective data condition is still a difficult problem to be solved.
Disclosure of Invention
In view of the foregoing, the present invention is directed to a method for predicting failure of an electric drive system with less sample data enhancement and migration, so as to solve at least one of the above-mentioned problems.
In order to achieve the above purpose, the technical scheme of the invention is realized as follows:
the first aspect of the present invention provides a method for predicting failure of an electric drive system with little sample data enhancement and migration, comprising:
acquiring operation data of the electric drive system in a normal working state, and pre-training an LSTM network model by using the operation data;
collecting fault data of the electric drive system in a fault state, performing characteristic tag calibration operation on the fault data, and performing data enhancement processing on the fault data with few samples through a time sliding window;
Performing migration learning on the pre-trained LSTM network model to obtain a fault prediction model, and training the fault prediction model by using the enhanced fault data;
and performing fault prediction by using the trained fault prediction model, and outputting a predicted fault curve.
Further, the process of collecting the operation data of the electric drive system in the normal working state comprises the following steps: selecting a plurality of driving devices with different powers in an electric driving system as prototypes, and collecting operation data of the prototypes under the same working condition and normal working state; the operation data comprise output current, temperature and shell vibration data;
and carrying out normalization pretreatment after unifying time scales on the acquired data.
Further, the LSTM network model comprises an encoder and a decoder, a repetition vector layer is arranged between the encoder and the decoder, and a time distribution connecting layer is arranged at the output end of the decoder;
the encoder calculates activation vectors of the forgetting gate, the input gate, the output gate and the cell state, and carries out forward propagation through the LSTM unit to obtain a hidden state of each time step, and the hidden state is used as a context to represent the cell state;
The decoder receives the hidden state corresponding to the last time step as an initial state, calculates the hidden state and the cell state of the decoder using the same LSTM unit as the encoder, and generates a reconstructed output sequence.
Further, the process of performing the feature tag calibration operation on the fault data includes: setting tag points in fault data by taking fault occurrence time as a reference, and calculating the absolute value of the difference between the time of each tag point and the time of fault occurrence to obtain the residual time;
And inputting the residual time into the LSTM network model as a characteristic label, and performing time point correlation on fault data input into the LSTM network model.
Further, after collecting fault data of the electric drive system in a fault state, executing normalization preprocessing operation identical to that of operation data on the fault data;
The process of carrying out data enhancement processing on the few sample fault data through the time sliding window comprises the following steps: and inputting the preprocessed fault data, setting the window size and the moving step length of the time sliding window, controlling the time sliding window to extract a plurality of samples in a sequence of the fault data according to the window size and the step length, and adding the context relation of the time sequence in the samples.
Further, the process of performing migration learning on the pre-trained LSTM network model to obtain a fault prediction model includes:
And reserving an encoder part in the pre-trained LSTM network model, adding a plurality of LSTM layers and a regression output layer on the basis of the encoder part, and obtaining a fault prediction model by using Relu functions as activation functions.
Further, after the predicted fault curve is output, the predicted fault curve is compared with the real fault curve, and the effectiveness of the model is evaluated.
A second aspect of the present invention provides an electronic device comprising a processor and a memory communicatively coupled to the processor for storing instructions executable by the processor, characterized by: the processor is configured to perform a method for predicting failure of an electric drive system with reduced sample data enhancement and migration as described in any one of the first aspects above.
A third aspect of the present invention provides a server, characterized in that: comprising at least one processor, and a memory communicatively coupled to the processor, the memory storing instructions executable by the at least one processor to cause the at least one processor to perform a method of low sample data enhancement and migration electric drive system failure prediction as described in any of the first aspects.
A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, characterized in that: the computer program, when executed by a processor, implements a method for failure prediction of an electric drive system for small sample data enhancement and migration as described in any one of the first aspects.
Compared with the prior art, the electric drive system fault prediction method for enhancing and migrating the few-sample data has the following beneficial effects:
the potential faults of the motor assembly can be accurately identified and predicted relative to the traditional model;
By accurately predicting potential failure of the electric drive assembly, it helps to improve the reliability and performance of the overall drive system, which is particularly important to ensure safe operation and meet high standard performance requirements.
The method provided by the invention uses a large amount of normal operation data as a source domain to train a model, and can provide references for fault prediction in other component fields.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention. In the drawings:
FIG. 1 is a flow chart of a method for predicting failure of an electric drive system with little sample data enhancement and migration according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an LSTM network model according to an embodiment of the present invention;
Fig. 3 is a schematic diagram of a fault prediction curve and a real fault life curve of a model machine according to three LSTM models in an embodiment of the present invention.
Detailed Description
It should be noted that, without conflict, the embodiments of the present invention and features of the embodiments may be combined with each other.
The invention will be described in detail below with reference to the drawings in connection with embodiments.
Embodiment one:
as shown in fig. 1: a method of low sample data enhancement and migration electric drive system failure prediction comprising:
s1, acquiring operation data of an electric drive system in a normal working state, and pre-training an LSTM network model by using the operation data;
S2, collecting fault data of the electric drive system in a fault state, performing feature tag calibration operation on the fault data, and performing data enhancement processing on the fault data of a few samples through a time sliding window;
S3, performing migration learning on the pre-trained LSTM network model to obtain a fault prediction model, and training the fault prediction model by using the enhanced fault data;
s4, performing fault prediction by using the trained fault prediction model, and outputting a predicted fault curve.
The principle of pre-training the LSTM network model by using the operation data and the training target are shown as the following formula:
the encoder hidden state is:
;
the decoder hidden state is:
;
the self-encoder loss function is:
;
wherein, Is the normal working data of the source domain,/>Is the hidden state of the encoder t-1 instant,Is the hidden state at the moment of the decoder t-1.
The objective function of the encoder is:
;
For each of the source domains Calculating the square/>, of the reconstruction errors of the encoder and decoder in the self-encoder by LSTMAnd minimizes it.
The process of collecting the operation data of the electric drive system in the normal working state comprises the following steps:
Selecting 5 driving devices with different powers in an electric driving system as prototypes, and collecting operation data of the 5 prototypes under the same working condition and normal working state; the operation data comprise output current, temperature and shell vibration data; the electric signals, the temperature signals and the mechanical signals of the electric drive system are used as characteristic data, so that the running state of the system can be comprehensively reflected;
The acquired data is subjected to normalization pretreatment after the unification of time scales so as to ensure that the acquired data has consistent time standard during analysis; the formula for processing the data by the normalization pretreatment method (Min-Max-scaler) is as follows: ;
Where X is the original record point of a feature of the prototype, min (X) is the minimum value in the dataset, max (X) is the maximum value in the dataset, and X norm is the value after normalization.
The LSTM network model comprises an encoder and a decoder, wherein a repetition vector layer is arranged between the encoder and the decoder, and a time distribution connecting layer is arranged at the output end of the decoder;
Encoder computing forget door Input door/>Output door/>Cell status/>Forward propagation through LSTM cells to obtain hidden states/>, per time stepAnd representing the cell state with the hidden state as a context;
The decoder receives the hidden state corresponding to the last time step as the initial state, calculates the hidden state and the cell state of the decoder by using the same LSTM unit as the encoder, and generates a reconstructed output sequence 。
The specific process comprises the following steps: because the invention relates to unsupervised training, a Autoencoder network structure is used for being embedded into an LSTM network, a normal data set is learned, important features are extracted, and normal input data of different prototypes are characterized;
Assume that a sequence is input as Wherein/>Is an input vector at time step t, the encoder propagates forward through LSTM units, and calculates hidden state/>, of each time step:
The arithmetic mathematical interpretation formulas of the automatic encoder and decoder are as follows:
;
;
;
;
;
;
wherein, ,/>,/>Activation vectors of forget gate, input gate and output gate, respectively,/>The sigmoid function is represented as a function,Representing the hadamard product, W and b are the weight matrix and the bias vector. Final state of encoder/>Is used as a context to represent cell status/>。
The decoder section is used to reconstruct the input sequence X. It receives the final hidden state from the encoderAs an initial state, and attempting to reconstruct the original input X;
Using LSTM cells similar to the encoder, but with the initial concealment state and cell state initialized by the context provided by the encoder, the concealment state of the decoder is t for each time step And cell status/>Is updated to generate a reconstructed output sequence/>, using hidden states of the decoder, similar to the way the encoder is updatedThe specific formula is as follows:
;
;
The final goal is to make the reconstructed sequence as close to the original acquisition data X as possible;
specifically, the structure of the encoder selects two LSTM layers, the first LSTM layer selects 64 units, the second LSTM layer selects 32 units, and the activation function selects Relu functions;
Adding Repeatvector layers between the encoder and decoder, repeating the last output of the encoder multiple times to maintain the input for each step of the decoder;
The decoder part structure selects two LSTM layers, the first LSTM layer selects 32 units, the second LSTM layer selects 64 units, the activation function also selects Relu functions, and finally the TimeDistributed connection layer is used as an output layer to complete the construction of the self-coding network.
In the process of pre-training the LSTM network model by using the operation data, an Adam optimizer is used by the optimizer, an adaptive method is used by the learning rate, the beta selection is 0.04, the epoch selection is 50, and the batch-size selection is 32; using the Mean Square Error (MSE) as a training error function, the resulting parametric weights provide for transfer learning.
The process of performing the characteristic tag calibration operation on the fault data comprises the following steps: setting tag points in fault data by taking fault occurrence time as a reference, and calculating the absolute value of the difference between the time of each tag point and the time of fault occurrence to obtain the residual time; and inputting the residual time into the LSTM network model as a characteristic label, and performing time point correlation on fault data input into the LSTM network model.
After collecting fault data of the electric drive system in a fault state, executing normalization preprocessing operation identical to that of operation data on the fault data; the process of carrying out data enhancement processing on the few sample fault data through the time sliding window comprises the following steps: and inputting the preprocessed fault data, setting the window size and the moving step length of the time sliding window, controlling the time sliding window to extract a plurality of samples in a sequence of the fault data according to the window size and the step length, and adding the context relation of the time sequence in the samples.
The concrete setting mode of the time sliding window method is as follows:
Let the original input be Where N is the total length of the sequence, defining a window size W and a step size S, M samples can be extracted,/>The definition of each sample is as follows: ; wherein i=1, 2, …, M; in this embodiment, the window size W is set to 3, the step size S of the window movement is set to 2, and the extracted data maintains the context of the time sequence.
The process for obtaining the fault prediction model by performing migration learning on the pre-trained LSTM network model comprises the following steps:
And reserving an encoder part in the pre-trained LSTM network model, ensuring that the encoder part is completely the same as a new model on a shared layer architecture, correctly matching weights, adding two LSTM layers on the basis of the encoder part, wherein the number of first LSTM layer selection nodes is 16, the number of second LSTM layer selection nodes is 64, adding a regression output layer, and using Relu functions as an activation function to obtain a fault prediction model.
The target domain deep learning model iteration update target of the migration learning of the encoder is as follows:
;
Wherein the adaptive learning rate is defined as:
; where β is a super parameter that adjusts the learning rate.
The model was trained using Adam optimizer as the gradient descent method, MSE as the error function, and epoch iterated 1000 times.
After the predicted fault curve is output, the predicted fault curve is compared with the real fault curve, and the effectiveness of the model is evaluated.
Embodiment two:
An electronic device comprising a processor and a memory communicatively coupled to the processor for storing processor-executable instructions, characterized in that: the processor is configured to perform a method for predicting failure of an electric drive system with reduced sample data enhancement and migration as in any of the above embodiments.
Embodiment III:
a server, characterized by: the system comprises at least one processor and a memory communicatively connected to the processor, wherein the memory stores instructions executable by the at least one processor to cause the at least one processor to perform a method of low sample data enhancement and migration electric drive system failure prediction as described in any of the embodiments.
Embodiment four:
a computer-readable storage medium storing a computer program, characterized in that: the computer program, when executed by a processor, implements a method for failure prediction of an electric drive system of small sample data enhancement and migration as described in any of the embodiments.
Those of ordinary skill in the art will appreciate that the elements and method steps of each example described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the elements and steps of each example have been described generally in terms of functionality in the foregoing description to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the several embodiments provided in the present application, it should be understood that the disclosed methods and systems may be implemented in other ways. For example, the above-described division of units is merely a logical function division, and there may be another division manner when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted or not performed. The units may or may not be physically separate, and components shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiment of the present application.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; 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 or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention, and are intended to be included within the scope of the appended claims and description.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.
Claims (10)
1. A method for predicting failure of an electric drive system with reduced sample data enhancement and migration, comprising:
acquiring operation data of the electric drive system in a normal working state, and pre-training an LSTM network model by using the operation data;
collecting fault data of the electric drive system in a fault state, performing characteristic tag calibration operation on the fault data, and performing data enhancement processing on the fault data with few samples through a time sliding window;
Performing migration learning on the pre-trained LSTM network model to obtain a fault prediction model, and training the fault prediction model by using the enhanced fault data;
and performing fault prediction by using the trained fault prediction model, and outputting a predicted fault curve.
2. The method for predicting failure of an electric drive system with reduced sample data enhancement and migration of claim 1, wherein:
The process of collecting the operation data of the electric drive system in the normal working state comprises the following steps: selecting a plurality of driving devices with different powers in an electric driving system as prototypes, and collecting operation data of the prototypes under the same working condition and normal working state; the operation data comprise output current, temperature and shell vibration data;
and carrying out normalization pretreatment after unifying time scales on the acquired data.
3. The method for predicting failure of an electric drive system with reduced sample data enhancement and migration of claim 1, wherein:
The LSTM network model comprises an encoder and a decoder, wherein a repetition vector layer is arranged between the encoder and the decoder, and a time distribution connecting layer is arranged at the output end of the decoder;
the encoder calculates activation vectors of the forgetting gate, the input gate, the output gate and the cell state, and carries out forward propagation through the LSTM unit to obtain a hidden state of each time step, and the hidden state is used as a context to represent the cell state;
The decoder receives the hidden state corresponding to the last time step as an initial state, calculates the hidden state and the cell state of the decoder using the same LSTM unit as the encoder, and generates a reconstructed output sequence.
4. The method for predicting failure of an electric drive system with reduced sample data enhancement and migration of claim 1, wherein:
the process of performing the characteristic tag calibration operation on the fault data comprises the following steps: setting tag points in fault data by taking fault occurrence time as a reference, and calculating the absolute value of the difference between the time of each tag point and the time of fault occurrence to obtain the residual time;
And inputting the residual time into the LSTM network model as a characteristic label, and performing time point correlation on fault data input into the LSTM network model.
5. The method for predicting failure of an electric drive system with reduced sample data enhancement and migration of claim 1, wherein:
after collecting fault data of the electric drive system in a fault state, executing normalization preprocessing operation identical to that of operation data on the fault data;
The process of carrying out data enhancement processing on the few sample fault data through the time sliding window comprises the following steps: and inputting the preprocessed fault data, setting the window size and the moving step length of the time sliding window, controlling the time sliding window to extract a plurality of samples in a sequence of the fault data according to the window size and the step length, and adding the context relation of the time sequence in the samples.
6. The method for predicting failure of an electric drive system with reduced sample data enhancement and migration of claim 1, wherein:
the process for obtaining the fault prediction model by performing migration learning on the pre-trained LSTM network model comprises the following steps:
And reserving an encoder part in the pre-trained LSTM network model, adding a plurality of LSTM layers and a regression output layer on the basis of the encoder part, and obtaining a fault prediction model by using Relu functions as activation functions.
7. The method for predicting failure of an electric drive system with reduced sample data enhancement and migration of claim 1, wherein:
after the predicted fault curve is output, the predicted fault curve is compared with the real fault curve, and the effectiveness of the model is evaluated.
8. An electronic device comprising a processor and a memory communicatively coupled to the processor for storing processor-executable instructions, characterized in that: the processor is configured to perform a method for predicting failure of an electric drive system with reduced sample data enhancement and migration as set forth in any one of claims 1-7.
9. A server, characterized by: comprising at least one processor, and a memory communicatively coupled to the processor, the memory storing instructions executable by the at least one processor to cause the at least one processor to perform a method of low sample data enhancement and migration electric drive system failure prediction as claimed in any one of claims 1-7.
10. A computer-readable storage medium storing a computer program, characterized in that: the computer program, when executed by a processor, implements a method for low sample data enhancement and migration of electric drive system failure prediction as claimed in any one of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410452105.4A CN118052154A (en) | 2024-04-16 | 2024-04-16 | Electric drive system fault prediction method for few-sample data enhancement and migration |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410452105.4A CN118052154A (en) | 2024-04-16 | 2024-04-16 | Electric drive system fault prediction method for few-sample data enhancement and migration |
Publications (1)
Publication Number | Publication Date |
---|---|
CN118052154A true CN118052154A (en) | 2024-05-17 |
Family
ID=91048694
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410452105.4A Pending CN118052154A (en) | 2024-04-16 | 2024-04-16 | Electric drive system fault prediction method for few-sample data enhancement and migration |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN118052154A (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112434602A (en) * | 2020-11-23 | 2021-03-02 | 西安交通大学 | Fault diagnosis method based on migratable common feature space mining |
CN113469013A (en) * | 2021-06-28 | 2021-10-01 | 江苏大学 | Motor fault prediction method and system based on transfer learning and time sequence |
CN115859791A (en) * | 2022-11-23 | 2023-03-28 | 北京无线电计量测试研究所 | Method and device for estimating residual life of aero-engine |
WO2023071217A1 (en) * | 2021-10-27 | 2023-05-04 | 中国华能集团清洁能源技术研究院有限公司 | Multi-working-condition process industrial fault detection and diagnosis method based on deep transfer learning |
CN117150243A (en) * | 2023-10-27 | 2023-12-01 | 湘江实验室 | Fault isolation and estimation method based on fault influence decoupling network |
-
2024
- 2024-04-16 CN CN202410452105.4A patent/CN118052154A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112434602A (en) * | 2020-11-23 | 2021-03-02 | 西安交通大学 | Fault diagnosis method based on migratable common feature space mining |
CN113469013A (en) * | 2021-06-28 | 2021-10-01 | 江苏大学 | Motor fault prediction method and system based on transfer learning and time sequence |
WO2023071217A1 (en) * | 2021-10-27 | 2023-05-04 | 中国华能集团清洁能源技术研究院有限公司 | Multi-working-condition process industrial fault detection and diagnosis method based on deep transfer learning |
CN115859791A (en) * | 2022-11-23 | 2023-03-28 | 北京无线电计量测试研究所 | Method and device for estimating residual life of aero-engine |
CN117150243A (en) * | 2023-10-27 | 2023-12-01 | 湘江实验室 | Fault isolation and estimation method based on fault influence decoupling network |
Non-Patent Citations (3)
Title |
---|
付松等: "基于迁移学习的民航发动机小样本故障诊断", 计算机集成制造系统, vol. 27, no. 12, 31 December 2021 (2021-12-31), pages 3450 - 3461 * |
尹杰等: "基于迁移学习和降噪自编码器-长短时间记忆的锂离子电池剩余寿命预测", 电工技术学报, vol. 39, no. 1, 31 January 2024 (2024-01-31), pages 289 - 302 * |
王梦蝶: "基于深度学习的核电厂一回路典型故障的诊断方法研究", 中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑, no. 1, 15 January 2024 (2024-01-15), pages 27 - 29 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Sangiorgio et al. | Robustness of LSTM neural networks for multi-step forecasting of chaotic time series | |
Mattey et al. | A novel sequential method to train physics informed neural networks for Allen Cahn and Cahn Hilliard equations | |
US20200272905A1 (en) | Artificial neural network compression via iterative hybrid reinforcement learning approach | |
Jouin et al. | Particle filter-based prognostics: Review, discussion and perspectives | |
CN110909926A (en) | TCN-LSTM-based solar photovoltaic power generation prediction method | |
CN110956260A (en) | System and method for neural architecture search | |
CN111832216A (en) | Rolling bearing residual service life prediction method based on EEMD-MCNN-GRU | |
EP3380948B1 (en) | Environmental monitoring systems, methods and media | |
Ayodeji et al. | Causal augmented ConvNet: A temporal memory dilated convolution model for long-sequence time series prediction | |
CN114266278B (en) | Dual-attention network-based equipment residual service life prediction method | |
CN112213771A (en) | Seismic wave impedance inversion method and device | |
Zhao et al. | Probabilistic remaining useful life prediction based on deep convolutional neural network | |
Koeppe et al. | Explainable artificial intelligence for mechanics: physics-explaining neural networks for constitutive models | |
Mo et al. | Few-shot RUL estimation based on model-agnostic meta-learning | |
CN113487019A (en) | Circuit fault diagnosis method and device, computer equipment and storage medium | |
CN113646776A (en) | Using hysteresis gradients in machine learning model training | |
CN113971423A (en) | Method, system and computer readable medium for training a new neural network | |
CN116075836A (en) | Quantum hardware generative modeling | |
Juesas et al. | Ascertainment-adjusted parameter estimation approach to improve robustness against misspecification of health monitoring methods | |
CN118052154A (en) | Electric drive system fault prediction method for few-sample data enhancement and migration | |
CN116484868A (en) | Cross-domain named entity recognition method and device based on diffusion model generation | |
CN114720129A (en) | Rolling bearing residual life prediction method and system based on bidirectional GRU | |
Hu et al. | Neural-PDE: a RNN based neural network for solving time dependent PDEs | |
Zembra et al. | Use of physics-augmented neural networks for unsupervised learning of material constitutive relations-Comparison of the NN-Euclid and NN-mCRE methods | |
Kocijan et al. | System identification with GP models |
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 |