CN116244640A - Unbalanced bearing fault diagnosis method and system based on federal learning - Google Patents

Unbalanced bearing fault diagnosis method and system based on federal learning Download PDF

Info

Publication number
CN116244640A
CN116244640A CN202310194574.6A CN202310194574A CN116244640A CN 116244640 A CN116244640 A CN 116244640A CN 202310194574 A CN202310194574 A CN 202310194574A CN 116244640 A CN116244640 A CN 116244640A
Authority
CN
China
Prior art keywords
data
local
classifier
user
enhancer
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
Application number
CN202310194574.6A
Other languages
Chinese (zh)
Inventor
李骏
张玮
梅镇
韦康
邵雨蒙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Science and Technology
Original Assignee
Nanjing University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing University of Science and Technology filed Critical Nanjing University of Science and Technology
Priority to CN202310194574.6A priority Critical patent/CN116244640A/en
Publication of CN116244640A publication Critical patent/CN116244640A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/04—Bearings
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
    • 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
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention discloses a federal learning-based unbalanced bearing fault diagnosis method and a federal learning-based unbalanced bearing fault diagnosis system, wherein the method operates on a plurality ofThe method comprises the steps that a local user node and a server aggregation node are connected; each user preprocesses the unbalanced bearing fault data set, pretrains the local classifier, sends the data to the server to be aggregated into a global classifier, and sends the data to the user to continuously pretrain, and continuously pretrains N 1 A wheel; each user receives the pre-trained global classifier as a local classifier, jointly trains the local classifier and the data enhancer, sends the local classifier and the data enhancer into an aggregation node to carry out weighted sum aggregation to obtain the global classifier and the data enhancer, and then broadcasts the global classifier and the data enhancer to the participating local user nodes to continue weighted sum aggregation for N times 2 A wheel; after training, each user uses the obtained local classifier and the data enhancer to detect bearing faults of the test data. The invention solves the unbalanced problem of bearing fault diagnosis under the condition of federal learning, and can obtain accurate fault diagnosis results.

Description

Unbalanced bearing fault diagnosis method and system based on federal learning
Technical Field
The invention belongs to the technical field of combination of artificial intelligence and edge calculation, and particularly relates to a federal learning-based unbalanced bearing fault diagnosis method and system.
Background
The abnormal monitoring of key parts of intelligent factory equipment has important significance for safe and economic operation, and timely and accurate fault diagnosis is an effective means for avoiding huge economic loss and unnecessary casualties. Most of the data driven diagnostic methods are assumed to be performed with extensive and balanced training, however this assumption is often impractical in engineering.
In industry, training samples under different health conditions are often unbalanced, such as rotating mechanical systems, which mostly work under health conditions with very short running times under fault conditions, considering costs and safety. The results show that there are enough samples taken to represent a normal condition, but the corresponding fault data is far less than the healthy samples.
Complex industrial data often has the following problems: 1) In mass data, the characteristic information related to a certain fault, namely class imbalance, is lacking; 2) The data on each device is not independent and co-distributed (non-IID), i.e., the distribution is unbalanced, and the local private data on each device may come from different situations and thus be different from the overall distribution of the data. These problems often result in a decrease in the accuracy of fault diagnosis, severely limiting the learning ability of the deep learning algorithm. Furthermore, researchers in the relevant industry can only analyze and mine datasets that belong to their own institutions. If the amount of data owned by a single organization is not very large and the similarity is high and the diversity is insufficient, machine learning on such a dataset may eventually lead to a model with poor expansibility or to the easy generation of over-fitting results. In this case, restrictions on the privacy and confidentiality of data significantly affect the effect of machine learning, resulting in a decrease in failure diagnosis accuracy.
Disclosure of Invention
The invention aims to provide a federal learning-based unbalanced bearing fault diagnosis method and a federal learning-based unbalanced bearing fault diagnosis system, so that an accurate fault diagnosis result is obtained.
The technical solution for realizing the purpose of the invention is as follows: a federal learning-based unbalanced bearing fault diagnosis method, the method operating at a plurality of local user nodes and a server aggregation node, comprising the steps of:
step 1, preprocessing an unbalanced bearing fault data set by each user, wherein the preprocessing comprises sliding window sampling and normalization processing;
step 2, each user pretrains the local classifier by the preprocessed data, and sends the data to the server to aggregate the data into a global classifier, and then sends the global classifier to the user to continuously pretrain the local classifier, and continuously pretrains N 1 The wheel then goes to step 3;
step 3, each user receives the global classifier pre-trained in the step 2 as a local classifier, trains the local classifier and the local data enhancer in a combined way, and sends the local classifier and the local data enhancer to a server aggregation node;
step 4, the server side aggregation node receives the local classifier and the local data enhancer sent by all local user nodes, and weights and aggregates all the local classifier and the local data enhancer according to a weighting strategy to obtain a global classifier and a global data enhancer;
step 5, broadcasting the global classifier and the global data enhancer to the participating local user nodes, and the local userThe node receives the global classifier and the global data enhancer sent by the service end aggregation node as the local classifier and the local data enhancer, returns to the step 4 to continuously carry out weighted sum aggregation on all the local classifiers and the local data enhancers, and lasts for N 2 The wheel then goes to step 6;
and 6, after training, each user uses the local classifier and the local data enhancer obtained in the step 5 to locally detect bearing faults of the test data, and obtains the accuracy of the local classifier of each user and an confusion matrix diagram of the test data.
A federal learning-based unbalanced bearing fault diagnosis system which, when in operation, performs the steps of the unbalanced bearing fault diagnosis method.
Compared with the prior art, the invention has the remarkable advantages that: (1) The centralized model training part in the Internet of things is transferred to the edge equipment, data among all users are combined, and meanwhile, the calculation load of a cloud end or a server end is reduced; (2) The data is always kept in the local of the edge equipment, so that the safety of the data can be improved; (3) For the problem of unbalanced data among edge devices in an industrial scene, the aggregation mode is improved through the joint optimization of a data enhancer and a classifier, so that the quality of a final model is improved; (4) After training, a data enhancer for generating high-quality fault samples and a high-precision fault classifier can be obtained at the same time, and the method is an end-to-end data enhancement mode.
Drawings
FIG. 1 is a block diagram of an unbalanced bearing fault diagnosis system based on federal learning according to the present invention.
Detailed Description
According to the unbalanced bearing fault diagnosis method based on federal learning, each user side is provided with a fault classifier network CNN and a data enhancement network GAN, false fault data generated by the GAN optimize the classifier CNN, and meanwhile, diagnosis results of the CNN optimize the generation of the GAN. Finally, the data enhancement network can generate high-quality fault samples, and the fault classifier can also obtain high-precision fault diagnosis results. Meanwhile, the problem of unbalanced data distribution among training users is considered, and the influence of gradient noise is reduced by adopting a geometric mean aggregation method. The invention solves the unbalanced problem of bearing fault diagnosis under the condition of federal learning, including category unbalance and distribution unbalance, and can obtain accurate fault diagnosis results.
The invention relates to a federal learning-based unbalanced bearing fault diagnosis method, which is operated on a plurality of local user nodes and a service end aggregation node and comprises the following steps:
step 1, preprocessing an unbalanced bearing fault data set by each user, wherein the preprocessing comprises sliding window sampling and normalization processing;
step 2, each user pretrains the local classifier by the preprocessed data, and sends the data to the server to aggregate the data into a global classifier, and then sends the global classifier to the user to continuously pretrain the local classifier, and continuously pretrains N 1 The wheel then goes to step 3;
step 3, each user receives the global classifier pre-trained in the step 2 as a local classifier, trains the local classifier and the local data enhancer in a combined way, and sends the local classifier and the local data enhancer to a server aggregation node;
step 4, the server side aggregation node receives the local classifier and the local data enhancer sent by all local user nodes, and weights and aggregates all the local classifier and the local data enhancer according to a weighting strategy to obtain a global classifier and a global data enhancer;
step 5, broadcasting the global classifier and the global data enhancer to the participating local user nodes, wherein the local user nodes receive the global classifier and the global data enhancer sent by the service end aggregation node and serve as the local classifier and the local data enhancer, returning to the step 4 to continuously weight and aggregate all the local classifiers and the local data enhancers for N duration 2 The wheel then goes to step 6;
and 6, after training, each user uses the local classifier and the local data enhancer obtained in the step 5 to locally detect bearing faults of the test data, and obtains the accuracy of the local classifier of each user and an confusion matrix diagram of the test data.
As a specific example, each local user node is an edge server, and is provided with sensors of various components and environments in an industrial equipment workshop and a collector of control equipment;
an edge server in an industrial equipment workshop gathers sensor network data of equipment, so that processing of bottom data by an edge side is realized;
the local user node is used as an initiator and a participant of the federal training task, is responsible for training a local model and storing data, and is transmitted to a server cloud, namely a server through a wireless network;
the server aggregation node is arranged at the cloud end of the server and serves as a trusted third party to aggregate and issue the model through the wireless network.
As a specific example, in step 1, the sliding window sampling formula is as follows:
Figure BDA0004106673320000031
wherein, N is the length of the bearing data signal, step is the length of the sampling signal, and W is the data overlap ratio.
As a specific example, the normalization process in step 1 is specifically:
mapping data to a range of 0-1 for processing, wherein x is the current sample data value, x max For the maximum value of the sample data, xmin is the minimum value of the sample data, and the normalization processing formula is as follows:
Figure BDA0004106673320000041
as a specific example, in step 2, the classifier is a CNN network, where the CNN network includes an input layer, a convolution layer 3×3×32, a pooling layer 2×2, a convolution layer 3×3×64, a pooling layer 2×2, a convolution layer 3×3×128, a pooling layer 2×2, a full connection layer 2048×1024, and an output layer sequentially arranged, and the output layer activation function is sigmoid.
As a specific example, in step 3, the local data enhancer uses a generation of an antagonism network GAN model, the model comprising an embedded network, a recovery network, a generator, a discriminator, wherein:
the embedded network is used for learning potential characteristics of the sample;
the recovery network is used for restoring the potential characteristics into an original sample;
the generator is used for generating potential characteristics with distribution close to the original samples, and restoring the potential characteristics into generated samples with distribution close to the original samples by using the recovery network, and sending the generated samples into the discriminator and the fault classifier;
the discriminator is used for judging whether the sample is from an original sample or a generated sample;
the embedded network, the recovery network, the generator and the discriminator are all realized based on the three-layer cyclic neural network GRU.
As a specific example, in step 3, the local classifier and the local data enhancer are jointly trained, and the specific process is as follows:
(1) training an embedded network and a recovery network in the data enhancer, and extracting potential spatial features of the data;
(2) training a discriminator of the data enhancer, and sending the sample generated by the generator into the discriminator;
(3) generating samples by using a generator of a data enhancer, and expanding an unbalanced fault sample set by marking the generated samples with matched category labels and mixing the generated samples with unbalanced real samples;
(4) training a classifier CNN, and sending the mixed sample set into a pre-trained classifier;
(5) a generator of training data enhancer.
As a specific example, the loss function of the generator is designed to:
Figure BDA0004106673320000042
wherein L is g Representing loss from discriminator, L feature-error Representing losses, L, from embedded and recovery networks fault-error Representing the loss from the classifier; k is the number of samples generated per round, d fake Xfeature for discriminating correct probability of each round of discriminator real Xfeature, a potential feature of a real sample fake To generate potential features of the sample.
As a specific example, in step 4, all local classifiers and local data enhancers are weighted and aggregated according to a weighting policy, the user data distribution is unbalanced, the gradient aggregation is noisy, and the server side adopts a geometric mean GM when calculating the model gradient:
Figure BDA0004106673320000051
wherein ω is model gradient, R d For the Euclidean space subset, i represents the user number, m is the total number of users, alpha i Weights, ω, for the ith user model i Is the ith user model gradient.
The invention also provides a unbalanced bearing fault diagnosis system based on federal learning, and the system executes the steps of the unbalanced bearing fault diagnosis method when in operation.
The technical solutions in the embodiments of the present invention will be clearly and specifically described below with reference to the drawings in the embodiments of the present invention. The described embodiments are only a few embodiments of the present invention.
Examples
The target bearing of the unbalanced federal bearing fault diagnosis system is a motor bearing with the sampling frequency of 12K, and the motor bearing comprises a normal state, an outer ring fault, an inner ring fault and a ball fault, wherein the number of samples with one type of faults is obviously less than that of other types, and unbalanced data are required to be subjected to fault diagnosis.
Fig. 1 shows the overall framework of the system, comprising two types of nodes: the server side aggregates the nodes and the user local nodes. Each local node is an edge server, and a controller for controlling the sensors of each component and the environment and the controller are arranged in the local node, and the edge server in the workshop gathers the data of the sensor network to realize the processing of the data of the bottom layer by the edge side. It is assumed herein that there is a "data islanding" problem between local nodes, i.e., the local nodes cannot transfer data to each other. The aggregation node is used as a trusted third party and is responsible for encryption, decryption and aggregation of the model. As sponsors and participants of federal training tasks, the local nodes are responsible for training the local model. The data transmitted by the sensor is raw data, exists in the form of vibration signals, and needs to be preprocessed before model training.
First, the user normalizes the unbalanced bearing fault data set, including fault categories with large sample size (normal, inner ring fault, ball fault) and fault categories with small sample size (outer ring fault). Carrying out sliding window enhancement processing on the data, wherein the window length is 400, the step length is 20, time sequence slice data are obtained, and the time sequence slice data are sent into a classifier for pre-training, the classifier is a CNN network, and the CNN network comprises the following components: input layer, convolution layer 3 x 32, pooling layer 2 x 2, convolution layer 3 x 64, pooling layer 2 x 2, convolution layer 3 x 128, pooling layer 2 x 2, full link layer 2048 x 1024, output layer, wherein the output layer activation function is sigmoid. Thereafter, the joint training was performed locally, and the training procedure is shown in table 1, with I set to 10.
When the local training is completed, the local model parameters are aggregated according to the improved aggregation strategy shown in table 2, and a new round of federal learning is started using the aggregated model until the iteration number e=100.
Table 1 local model training schematic
Figure BDA0004106673320000061
TABLE 2 Server-side model aggregation illustration
Figure BDA0004106673320000062
In summary, the federal fault diagnosis algorithm designed by the invention can realize the migration of the traditional data modeling technology from the cloud to the edge computing equipment, and can increase the quality of the training model and the training speed. The combination optimization and aggregation strategy of the geometric median of the data enhancer and the classifier improves the recognition effect of the federal model on the class with small data quantity in the data unbalanced scene. The safe and effective sharing of data among the edge devices is realized, so that the operation efficiency of the whole fault diagnosis system is improved.

Claims (10)

1. The unbalanced bearing fault diagnosis method based on federal learning is characterized by being operated on a plurality of local user nodes and a service end aggregation node and comprising the following steps of:
step 1, preprocessing an unbalanced bearing fault data set by each user, wherein the preprocessing comprises sliding window sampling and normalization processing;
step 2, each user pretrains the local classifier by the preprocessed data, and sends the data to the server to aggregate the data into a global classifier, and then sends the global classifier to the user to continuously pretrain the local classifier, and continuously pretrains N 1 The wheel then goes to step 3;
step 3, each user receives the global classifier pre-trained in the step 2 as a local classifier, trains the local classifier and the local data enhancer in a combined way, and sends the local classifier and the local data enhancer to a server aggregation node;
step 4, the server side aggregation node receives the local classifier and the local data enhancer sent by all local user nodes, and weights and aggregates all the local classifier and the local data enhancer according to a weighting strategy to obtain a global classifier and a global data enhancer;
step 5, broadcasting the global classifier and the global data enhancer to the participating local user nodes, wherein the local user nodes receive the global classifier and the global data enhancer sent by the service end aggregation node and serve as the local classifier and the local data enhancer, returning to the step 4 to continuously weight and aggregate all the local classifiers and the local data enhancers for N duration 2 The wheel then goes to step 6;
and 6, after training, each user uses the local classifier and the local data enhancer obtained in the step 5 to locally detect bearing faults of the test data, and obtains the accuracy of the local classifier of each user and an confusion matrix diagram of the test data.
2. The federal learning-based unbalanced bearing fault diagnosis method of claim 1, wherein each local user node is an edge server provided with sensors of various components and environments in an industrial equipment workshop and a collector of control equipment;
an edge server in an industrial equipment workshop gathers sensor network data of equipment, so that processing of bottom data by an edge side is realized;
the local user node is used as an initiator and a participant of the federal training task, is responsible for training a local model and storing data, and is transmitted to a server cloud, namely a server through a wireless network;
the server aggregation node is arranged at the cloud end of the server and serves as a trusted third party to aggregate and issue the model through the wireless network.
3. The federal learning-based unbalanced bearing fault diagnosis method according to claim 1, wherein in step 1, the sliding window sampling formula is as follows:
Figure FDA0004106673310000021
wherein, N is the length of the bearing data signal, step is the length of the sampling signal, and W is the data overlap ratio.
4. The unbalanced bearing fault diagnosis method based on federal learning according to claim 1, wherein the normalization processing in step 1 is specifically:
mapping data to a range of 0-1 for processing, wherein x is the current sample data value, x max For the maximum value of the sample data, xmin is the minimum value of the sample data, and the normalization processing formula is as follows:
Figure FDA0004106673310000022
5. the method of claim 1, wherein in step 2, the classifier is a CNN network, the CNN network includes an input layer, a convolution layer 3×3×32, a pooling layer 2×2, a convolution layer 3×3×64, a pooling layer 2×2, a convolution layer 3×3×128, a pooling layer 2×2, a full connection layer 2048×1024, and an output layer sequentially arranged, and the output layer activation function is sigmoid.
6. The federally learned unbalanced bearing fault diagnosis method of claim 1, wherein in step 3, the local data enhancer uses a generated countermeasure network GAN model, the model comprising an embedded network, a recovery network, a generator, a discriminator, wherein:
the embedded network is used for learning potential characteristics of the sample;
the recovery network is used for restoring the potential characteristics into an original sample;
the generator is used for generating potential characteristics with distribution close to the original samples, and restoring the potential characteristics into generated samples with distribution close to the original samples by using the recovery network, and sending the generated samples into the discriminator and the fault classifier;
the discriminator is used for judging whether the sample is from an original sample or a generated sample;
the embedded network, the recovery network, the generator and the discriminator are all realized based on the three-layer cyclic neural network GRU.
7. The unbalanced bearing fault diagnosis method based on federal learning of claim 6, wherein in step 3, the local classifier and the local data enhancer are jointly trained, and the specific process is as follows:
(1) training an embedded network and a recovery network in the data enhancer, and extracting potential spatial features of the data;
(2) training a discriminator of the data enhancer, and sending the sample generated by the generator into the discriminator;
(3) generating samples by using a generator of a data enhancer, and expanding an unbalanced fault sample set by marking the generated samples with matched category labels and mixing the generated samples with unbalanced real samples;
(4) training a classifier CNN, and sending the mixed sample set into a pre-trained classifier;
(5) a generator of training data enhancer.
8. The federally learned unbalanced bearing fault diagnosis method according to claim 7, wherein the generator loss function is designed to:
Figure FDA0004106673310000031
wherein L is g Representing loss from discriminator, L feature-error Representing losses, L, from embedded and recovery networks fault-error Representing the loss from the classifier; k is the number of samples generated per round, d fake Xfeature for discriminating correct probability of each round of discriminator real Xfeature, a potential feature of a real sample fake To generate potential features of the sample.
9. The unbalanced bearing fault diagnosis method based on federal learning of claim 8, wherein in step 4, all local classifiers and local data enhancers are weighted and aggregated according to a weighting strategy, user data distribution is unbalanced, noise exists in gradient aggregation, and a geometric mean GM is adopted by a server when calculating a model gradient:
Figure FDA0004106673310000032
wherein ω is a modulusGradient, R d For the Euclidean space subset, i represents the user number, m is the total number of users, alpha i Weights, ω, for the ith user model i Is the ith user model gradient.
10. A federal learning-based unbalanced bearing fault diagnosis system, wherein the system is operable to perform the steps of the unbalanced bearing fault diagnosis method of any one of claims 1 to 9.
CN202310194574.6A 2023-03-03 2023-03-03 Unbalanced bearing fault diagnosis method and system based on federal learning Pending CN116244640A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310194574.6A CN116244640A (en) 2023-03-03 2023-03-03 Unbalanced bearing fault diagnosis method and system based on federal learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310194574.6A CN116244640A (en) 2023-03-03 2023-03-03 Unbalanced bearing fault diagnosis method and system based on federal learning

Publications (1)

Publication Number Publication Date
CN116244640A true CN116244640A (en) 2023-06-09

Family

ID=86632857

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310194574.6A Pending CN116244640A (en) 2023-03-03 2023-03-03 Unbalanced bearing fault diagnosis method and system based on federal learning

Country Status (1)

Country Link
CN (1) CN116244640A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117272211A (en) * 2023-11-20 2023-12-22 北京邮电大学 Lightweight spacecraft fault detection classification method based on long-short-term memory model

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117272211A (en) * 2023-11-20 2023-12-22 北京邮电大学 Lightweight spacecraft fault detection classification method based on long-short-term memory model
CN117272211B (en) * 2023-11-20 2024-02-13 北京邮电大学 Lightweight spacecraft fault detection classification method based on long-short-term memory model

Similar Documents

Publication Publication Date Title
Shao et al. A novel approach of multisensory fusion to collaborative fault diagnosis in maintenance
Zhao et al. Deep multi-scale convolutional transfer learning network: A novel method for intelligent fault diagnosis of rolling bearings under variable working conditions and domains
Xia et al. Fault diagnosis for rotating machinery using multiple sensors and convolutional neural networks
Liu et al. An intrusion detection model with hierarchical attention mechanism
CN109902399A (en) Rolling bearing fault recognition methods under a kind of variable working condition based on ATT-CNN
Geng et al. Bearing fault diagnosis based on improved federated learning algorithm
CN108614548A (en) A kind of intelligent failure diagnosis method based on multi-modal fusion deep learning
Ding et al. HYBRID‐CNN: An Efficient Scheme for Abnormal Flow Detection in the SDN‐Based Smart Grid
Chen et al. Industrial edge intelligence: Federated-meta learning framework for few-shot fault diagnosis
CN116415506A (en) Industrial robot fault diagnosis method and system based on federal learning
CN112763215B (en) Multi-working-condition online fault diagnosis method based on modular federal deep learning
CN116244640A (en) Unbalanced bearing fault diagnosis method and system based on federal learning
CN115314287B (en) Counter anomaly detection system based on deep clustering
CN113541985A (en) Internet of things fault diagnosis method, training method of model and related device
Wu et al. Fault diagnosis of rotating machinery using Gaussian process and EEMD‐treelet
CN114564987A (en) Rotary machine fault diagnosis method and system based on graph data
Yang et al. A balanced deep transfer network for bearing fault diagnosis
Wen et al. Bearing fault diagnosis via fusing small samples and training multi-state siamese neural networks
Hou et al. Bearing fault diagnosis under small data set condition: A Bayesian network method with transfer learning for parameter estimation
Liu et al. Data privacy protection: A novel federated transfer learning scheme for bearing fault diagnosis
CN116992336B (en) Bearing fault diagnosis method based on federal local migration learning
Sun et al. SCG-GFFE: A Self-Constructed graph fault feature extractor based on graph Auto-encoder algorithm for unlabeled single-variable vibration signals of harmonic reducer
Li et al. Class imbalanced fault diagnosis via combining K-means clustering algorithm with generative adversarial networks
Li et al. An auto-regulated universal domain adaptation network for uncertain diagnostic scenarios of rotating machinery
CN117008570A (en) Cross-working condition fault diagnosis method and system based on open set federation reactance domain adaptation

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