CN109947086B - Mechanical fault migration diagnosis method and system based on counterstudy - Google Patents
Mechanical fault migration diagnosis method and system based on counterstudy Download PDFInfo
- Publication number
- CN109947086B CN109947086B CN201910289486.8A CN201910289486A CN109947086B CN 109947086 B CN109947086 B CN 109947086B CN 201910289486 A CN201910289486 A CN 201910289486A CN 109947086 B CN109947086 B CN 109947086B
- Authority
- CN
- China
- Prior art keywords
- training
- domain
- data set
- model
- 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.)
- Active
Links
- 238000013508 migration Methods 0.000 title claims abstract description 123
- 230000005012 migration Effects 0.000 title claims abstract description 117
- 238000003745 diagnosis Methods 0.000 title claims abstract description 108
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000012549 training Methods 0.000 claims abstract description 208
- 238000012360 testing method Methods 0.000 claims abstract description 62
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 33
- 238000009826 distribution Methods 0.000 claims description 37
- 238000013507 mapping Methods 0.000 claims description 29
- 238000011156 evaluation Methods 0.000 claims description 22
- 238000003062 neural network model Methods 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 description 14
- 238000013528 artificial neural network Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 238000012800 visualization Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 239000004576 sand Substances 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 230000003042 antagnostic effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The invention discloses a mechanical fault migration diagnosis method and system based on countermeasure learning, wherein the method comprises the following steps: acquiring original signals of mechanical faults under different working conditions, and analyzing to generate a source domain training data set with a label, a source domain training data set without a label and a target domain testing data set under different working conditions; training a deep convolutional neural network model according to a source domain training data set with a label and a back propagation algorithm to generate a fault diagnosis model; training a fault diagnosis model according to the source domain training data set without the label and the target domain testing data set; fine-tuning the trained fault diagnosis model according to the source domain training data set with the label and a back propagation algorithm; inputting the target domain test data set without the label into the fault diagnosis model after fine adjustment, and outputting the fault category of the sample to be tested. The method obtains the domain invariant feature through the counterstudy method, realizes the migration before different domains, and realizes the intelligent diagnosis of the variable working condition mechanical fault.
Description
Technical Field
The invention relates to the technical field of fault diagnosis of mechanical equipment, in particular to a mechanical fault migration diagnosis method and system based on countermeasure learning.
Background
With the gradual development of industrial technology, the demand for industrial equipment is continuously increased, the integrated scale of the current industrial system is larger and larger, the structure of a single equipment is more and more complex, the coupling degree between different equipment in the same system is higher and higher, the factors lay a solid foundation for realizing complex behaviors of mechanical equipment on one hand, and meanwhile, the probability of the fault of the whole system is greatly increased.
The existing industrial system usually operates continuously and stably for a long time, the failure frequency is low, once the failure occurs, the failure deterioration speed is high, the failure causes great threat, and if the failure is not controlled in time, the serious accident of machine damage and human death can be caused. Therefore, the method has important social significance for ensuring the long-period stable operation of the industrial system and avoiding serious accidents, and can bring great economic and social benefits.
In order to guarantee the safety of an industrial system and promote the development of intelligent manufacturing, more and more industrial complex systems utilize an industrial Internet of things platform to establish an equipment operation state monitoring system, so that the industrial system collects and stores massive industrial equipment operation data, and a sufficient data source is provided for a data-driven fault diagnosis method. However, the opportunity is often accompanied by a challenge, and the industrial system has the characteristics of high environmental complexity, incomplete information and the like, so that the fault diagnosis technology based on data driving has a huge challenge in real application. For the fault diagnosis technology based on data driving, the main problems are derived from data, due to the characteristics of industrial equipment, the difficulty of collecting fault samples of the equipment is high, the existing actual fault data are insufficient in types and few in samples, the data are incomplete, and the migration learning method is a key technology for solving the problem.
The deep learning model based on data driving has strong dependence on samples, and a large amount of sample data is needed to train to obtain a model meeting actual needs. The good effect of the depth model requires that two conditions are met simultaneously: 1) there are sufficient training samples; 2) the distribution of the training data and the target application data should be as similar as possible. However, in an actual fault diagnosis task, the two conditions are often difficult to be simultaneously satisfied, for example, an industrial system usually needs to adjust power load according to actual requirements, and the change of the load changes spatial distribution of data obtained by an acquisition system, and when a model obtained by training data collected under a low-load condition is used for diagnosing data under a high-load condition, the performance of the model is greatly affected. However, collecting data again under target application test conditions can be costly and, in many cases, much less likely to be practical. Thus, migrating an already trained diagnostic model based on source domain training data and target application test data is an important means to solve this problem.
At present, the problem of migration fault diagnosis still belongs to the preliminary stage, and a great number of technical difficulties need to be deeply researched and broken through. The method for diagnosing the mechanical fault migration based on the counterstudy under the variable working conditions is an effective technical method for solving the technical problem.
Disclosure of Invention
The present invention is directed to solving, at least to some extent, one of the technical problems in the related art.
Therefore, one objective of the present invention is to provide a mechanical fault migration diagnosis method based on countermeasure learning, which obtains a domain invariant feature through the countermeasure learning method, thereby implementing migration before different domains, and implementing intelligent diagnosis of variable-condition mechanical faults.
Another object of the present invention is to provide a mechanical fault migration diagnostic system based on countermeasure learning.
In order to achieve the above object, an embodiment of an aspect of the present invention provides a method for diagnosing mechanical fault migration based on countermeasure learning, including: s1, acquiring original signals of mechanical faults of a sample to be tested under different working conditions, and analyzing the original signals of the mechanical faults under different working conditions to generate a source domain training data set with a label, a source domain training data set without the label and a target domain testing data set without the label under different working conditions;
s2, training a deep convolutional neural network model according to the labeled source domain training data set and the back propagation algorithm, and generating a fault diagnosis model;
s3, performing migration training on the fault diagnosis model according to the unlabeled source domain training data set, the unlabeled target domain testing data set, a migration method of countermeasure learning and a Wasserstein distance-guided countermeasure network;
s4, fine-tuning the fault diagnosis model after migration training according to the source domain training data set with the label and a back propagation algorithm;
and S5, inputting the target domain test data set without the label into the fault diagnosis model after fine adjustment, and outputting the fault category of the sample to be tested.
According to the mechanical fault migration diagnosis method based on the countermeasure learning, the original signals of the mechanical faults under different working conditions are obtained, and source domain training data with labels and target domain test data without the labels are generated; optimizing a convolutional neural network serving as a basic diagnosis model by using a back propagation algorithm by using source domain training data with labels; obtaining a domain invariant feature by using source domain training data without a label and target domain test data through a Wasserstein distance-guided counterstudy method, and realizing cross-domain migration; the source domain training data with the labels are utilized, and a back propagation algorithm is used for fine tuning of the migration model, so that the problem of over-migration is avoided; and inputting the data of the label-free target domain to be tested into the migrated model to obtain a fault classification result.
In addition, the mechanical failure migration diagnosis method based on the countermeasure learning according to the above embodiment of the present invention may further have the following additional technical features:
further, in an embodiment of the present invention, the S2 further includes:
s21, the labeled source domain training data set X is processedsRandomly distributed to generate 70% of training data
And 30% of the test data
S22, from the training data
Extract data of size m
Optimizing the convolution characteristic mapping network unit and the full-connection classification network unit in the convolution neural network model through the back propagation algorithm;
s23, testing data through the optimized convolutional neural network model
Predicting and calculating the accuracy, if the accuracy is less than the preset accuracy, executing the step S22And otherwise, stopping training and storing all the obtained parameters to generate the fault diagnosis model.
Further, in an embodiment of the present invention, the S3 further includes:
migrating the fault diagnosis model generated in step S2 based on the migration method of countermeasure learning by using the source domain data without a tag and the target domain data without a tag, exploring a complex feature space by using the Wasserstein distance-guided countermeasure network, narrowing distribution differences between different domain data features by a countermeasure training strategy to obtain domain invariant features, and finally implementing cross-domain migration.
Further, in an embodiment of the present invention, the step S3 and the step S4 specifically include:
step 1, training a data set X from the labeled source domainsWith the unlabeled target domain data set X to be testedtMiddle-extracting migration training data with size of m
And
step 2, optimizing the domain evaluation network D by using the back propagation algorithmjAnd a full junction layer Fcj：
Step 3, training data set X from the labeled source domainsExtracting m-sized micro-call training data
Step 4, optimizing a convolution feature mapping network unit M and a full-connection classification network unit C in the convolution neural network model by using the back propagation algorithm;
step 5, evaluating the network D in the current domainjThe output Wessenstein distance is less than a preset threshold ThrewStopping training, keeping the obtained parameters, and completing the fault diagnosis modulePerforming migration training and fine adjustment on the model; otherwise, returning to the step 1 and re-executing the steps 1-5.
Further, in an embodiment of the present invention, the step S3 further includes:
and the fault diagnosis model after the migration training realizes the mapping of the features through the deep convolutional neural network, the domain evaluation network guides the domain-invariant feature learning process, and when the distribution between the source domain training data and the target domain training data is not biased, the domain-invariant feature mapping is obtained.
In order to achieve the above object, another embodiment of the present invention provides a system for diagnosing mechanical failure migration based on countermeasure learning, including: the data generation module is used for acquiring original signals of mechanical faults of a sample to be tested under different working conditions, and analyzing the original signals of the mechanical faults under the different working conditions to generate a source domain training data set with a label, a source domain training data set without the label and a target domain testing data set without the label under the different working conditions;
the model generation module is used for training a deep convolutional neural network model according to the source domain training data set with the label and a back propagation algorithm to generate a fault diagnosis model;
a model training module for performing migration training on the fault diagnosis model according to the unlabeled source domain training dataset, the unlabeled target domain testing dataset, a migration method of countermeasure learning, and a Wasserstein distance-guided countermeasure network;
the model fine-tuning module is used for fine-tuning the fault diagnosis model after the migration training according to the source domain training data set with the label and a back propagation algorithm;
and the diagnosis module is used for inputting the target domain test data set without the label into the fault diagnosis model after fine adjustment and outputting the fault category of the sample to be tested.
According to the mechanical fault migration diagnosis system based on the countermeasure learning, the original signals of the mechanical faults under different working conditions are obtained, and source domain training data with labels and target domain test data without the labels are generated; optimizing a convolutional neural network serving as a basic diagnosis model by using a back propagation algorithm by using source domain training data with labels; obtaining a domain invariant feature by using source domain training data without a label and target domain test data through a Wasserstein distance-guided counterstudy method, and realizing cross-domain migration; the source domain training data with the labels are utilized, and a back propagation algorithm is used for fine tuning of the migration model, so that the problem of over-migration is avoided; and inputting the data of the label-free target domain to be tested into the migrated model to obtain a fault classification result.
In addition, the mechanical failure migration diagnosis system based on the countermeasure learning according to the above embodiment of the present invention may further have the following additional technical features:
further, in an embodiment of the present invention, the model generation module includes: the device comprises a data distribution unit, an optimization unit and a determination unit;
the data distribution unit is used for distributing the labeled source domain training data set XsRandomly distributed to generate 70% of training data
And 30% of the test data
The optimization unit is used for obtaining training data
Extract data of size m
Optimizing the convolution characteristic mapping network unit and the full-connection classification network unit in the convolution neural network model through the back propagation algorithm;
the determining unit is used for testing data through the optimized convolutional neural network model
And predicting and calculating the accuracy, and stopping training and storing all the obtained parameters to generate the fault diagnosis model when the accuracy is greater than the preset accuracy.
Further, in one embodiment of the present invention, the model training module is further configured to,
migrating the generated fault diagnosis model based on the migration method of the countermeasure learning by using the source domain data without the label and the target domain data without the label, exploring a complex feature space by adopting the Wasserstein distance-guided countermeasure network, reducing distribution difference among different domain data features by a countermeasure training strategy to obtain domain invariant features, and finally realizing cross-domain migration.
Further, in one embodiment of the present invention, the model training module and the model fine tuning module are further configured to,
training a dataset X from the tagged source domainsWith the unlabeled target domain data set X to be testedtMiddle-extracting migration training data with size of m
And
and optimizing the domain evaluation network D using the back propagation algorithmjAnd a full junction layer Fcj：
Training a dataset X from the tagged source domainsExtracting m-sized micro-call training data
Optimizing a convolution feature mapping network unit M and a full-connection classification network unit C in the convolution neural network model by using the back propagation algorithm;
local domain evaluation network DjThe output Wessenstein distance is less than a preset threshold ThrewStopping training, keeping the obtained parameters, and completing the trainingAnd (5) migration training and fine adjustment of the fault diagnosis model.
Further, in an embodiment of the present invention, the model training module is further configured to,
and the fault diagnosis model after the migration training realizes the mapping of the features through the deep convolutional neural network, the domain evaluation network guides the domain-invariant feature learning process, and when the distribution between the source domain training data and the target domain training data is not biased, the domain-invariant feature mapping is obtained.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The foregoing and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a flow chart of a method for mechanical fault migration diagnosis based on countermeasure learning according to an embodiment of the present invention;
FIG. 2 is a flow chart of a method for mechanical fault migration diagnosis based on countermeasure learning according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an overview of a mechanical failure migration diagnostic model based on countermeasure learning, according to an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating visualization of base model features for source domain training data in a D → B migration task according to an embodiment of the present invention;
FIG. 5 is a schematic diagram illustrating visualization of fundamental model features of target domain test data in a D → B migration task according to an embodiment of the present invention;
FIG. 6 is a schematic diagram illustrating visualization of migration model features for source domain training data in a D → B migration task according to an embodiment of the present invention;
FIG. 7 is a schematic diagram illustrating visualization of migration model features for target domain test data in a D → B migration task according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of a mechanical failure migration diagnosis system based on countermeasure learning according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
The following describes a mechanical failure migration diagnosis method and system based on countermeasure learning according to an embodiment of the present invention with reference to the drawings.
First, a proposed countermeasure learning-based mechanical failure migration diagnosis method according to an embodiment of the present invention will be described with reference to the drawings.
Fig. 1 is a flowchart of a method for diagnosing mechanical failure migration based on countermeasure learning according to an embodiment of the present invention.
As shown in fig. 1, the method for diagnosing mechanical fault migration based on countermeasure learning includes the following steps:
in step S1, original signals of mechanical faults of the sample to be tested under different conditions are obtained, and the original signals of mechanical faults under different conditions are analyzed to generate a source domain training data set with a label, a source domain training data set without a label, and a target domain testing data set without a label under different conditions.
Specifically, original signals of mechanical faults under different working conditions are obtained, and a labeled source domain training data set under the 1 st working condition is sorted and generated
And 2, target domain data set to be tested without label in working condition
Wherein
And
are respectively a data set XsAnd XtAnd the ith and jth samples of (1), each sample having n data points, msAnd mtRespectively the total number of samples of the two data sets,
is a data set XsThe fault category label of the ith sample of (1).
In step S2, a deep convolutional neural network model is trained according to the labeled source domain training data set and the back propagation algorithm, and a fault diagnosis model is generated.
In particular, utilizing tagged source domain training data
And training a deep convolutional neural network model constructed by the convolutional layer and the fully-connected layer by using a back propagation algorithm to obtain a fault diagnosis model capable of accurately classifying training data.
It will be appreciated that a convolutional neural network is a type of neural network that uses a convolutional calculation method instead of a general matrix calculation. A typical convolution calculation is for image classification, and a two-dimensional convolution is defined as:
for the problem of migration fault diagnosis of mechanical faults, 1-dimensional vibration signals are processed, and therefore the embodiment of the invention uses a one-dimensional convolution neural network. As is readily apparent from equation (1), when m is equal to 1, the two-dimensional convolution is converted to a one-dimensional convolution. The one-dimensional convolution calculation formula is as follows:
wherein,
the first convolutional layer having a dimension n × 1 × j
The jth convolution kernel of (a);
is the signal segment of the ith input; bijIs a deviation; phi is the activation function.
For the fault classification problem of the embodiment of the invention, a fully-connected layer is added after the continuous convolutional layers to output the classification result, and the formula of the fully-connected layer is defined as follows:
yl＝φ(Wlyl-1+bl) (3)
wherein, WlIs a direct weight parameter of two fully-connected layers, yl-1Is a feature map of an upper layer, blIs the deviation of the current layer.
Training a fault-classified convolutional neural network using a back-propagation method, the classification network including convolutional feature mapping network elements M (with a parameter θ) for extracting multi-layer fault features of a data setM) And a fully connected classifying network unit C (with a parameter theta) for establishing a mapping between the fault signature and the sample signaturec) The loss of the model is defined as the cross entropy between the SoftMax function prediction probability distribution and the One-hot encoding of the source domain data tag:
wherein,
is a function of the indication of the function,
is the predicted distribution value of the K-th dimension, and K is the number of classes.
Further, in an embodiment of the present invention, step S2 specifically includes:
s21, the source domain training data set X with labelssRandomly distributed to generate 70% of training data
And 30% of the test data
S22, from the training data
Extract data of size m
Optimizing a convolution feature mapping network unit M and a full-connection classification network unit C in the convolution neural network by using a back propagation algorithm, and updating the convolution neural network according to the following formula:
s23, testing data through the optimized convolutional neural network model
Predicting and calculating the accuracy, if the accuracy is less than the preset accuracy ThrecThen step S22 is executed, otherwise, the training is stopped and all the obtained parameters are saved, and the fault diagnosis model is generated.
In step S3, the fault diagnosis model is migration-trained according to the unlabeled source domain training dataset, the unlabeled target domain testing dataset, the migration method of the countermeasure learning, and the Wasserstein distance-guided countermeasure network.
Further, migration is performed on the basic diagnosis model obtained in the previous step by using source domain data and target domain data without labels based on a migration method of countermeasure learning, a complicated feature space is explored by adopting a Wasserstein distance-guided countermeasure network, domain invariant features are obtained by reducing distribution differences among different domain data features through a countermeasure training strategy, and cross-domain migration is finally achieved.
In step S4, the fault diagnosis model after the migration training is trimmed according to the labeled source domain training data set and the back propagation algorithm.
In particular, utilizing tagged source domain training data
And (3) fine-tuning the diagnosis model obtained after migration by using a back propagation algorithm, ensuring the applicability of the diagnosis model to a source domain data set and avoiding the problem of over-migration.
It is understood that the Wasserstein distance is a distributed measurement method. The Wasserstein metric is a metric for measuring the distance between data distributions in the Polish metric space (M, p). For the measurement of two different probability distributions P and Q, the Wasserstein distance formula is defined as follows:
in an embodiment of the present invention, a 1 st order Wasserstein distance function is used to guide the antagonistic network training process, the 1 st order Wasserstein distance function formula, satisfying the Kantorovich-Rubinstein duality, is as follows:
where the upper bound covers all 1-L ipschitz functions f: x → R.
The Wasserstein distance-guided countermeasure training strategy, the objective function of the countermeasure network used in the embodiment of the present invention is defined as follows:
wherein D is a 1-L ipschitz function, PgIs the distribution of generative models
z-p (z) to satisfy the L ipschitz constraint, a penalty term is added to the objective function to solve the optimization difficulty caused by the constraint.
λ is a penalty factor.
In the proposed migration fault diagnosis model, a deep convolutional neural network is used for realizing the mapping of the features, and a domain evaluation network is used for guiding the learning process of the domain invariant features. When the distribution between the source domain and the target domain is not biased, a domain-invariant feature map will be obtained. Meanwhile, the domain evaluator is used for estimating the Wasserstein distance of the feature distribution of the source domain and the target domain. Source domain feature distribution
And target domain feature distribution
Wasserstein distance of (a) by maximizing a loss function L of domain evaluation network DwdIs estimated, wherein fs＝F(xs) And ft＝F(xt). The loss function is defined as follows:
wherein x issAnd xtIs from the source domain XsAnd a target domain XtTo satisfy the L ipsitz constraint on Wasserstein distanceConditional, putting forward a penalty term for adding constraint condition to original target function
The formula is defined as follows:
wherein,
is from distribution
Is sampled randomly.
And obtaining the domain-invariant feature through the training strategy learning. Firstly, training a domain evaluator network to maximize the Wasserstein distance of two domains, then, fixing the parameters of the domain evaluator network, and reducing the Wasserstein distance by adjusting the feature mapping in the depth model. The learning strategy formula of the domain invariant feature is expressed as follows:
where λ is a penalty coefficient.
Acquiring domain invariant features, Fc, by migrating fully-connected classification network units C before a source domain and a target domain for a convolutional neural network of a basic diagnostic modelj(parameters thereof)
) Is the jth fully-connected layer in a fully-connected network, and evaluates the network D using multiple domainsj(parameters thereof)
) To estimate the Wasserstein distance between the source domain and the target domain. Thus, the penalty function for the resistance training strategy is defined as:
wherein,
is a gradient penalty term and λ is a penalty coefficient.
According to the proposed fine tuning of the migration model, in the process of domain invariant feature confrontation iterative training, after each confrontation network optimization adjustment, a source domain training data set is used, and a convolution feature mapping network unit M and a full-connection classification network unit C in a convolution neural network are adjusted through a back propagation algorithm of a formula (5) and a formula (6), so that the migration model is fine tuned.
Further, in an embodiment of the present invention, the steps S3 and S4 specifically include:
step 1, training data set X from source domain with labelsWith the unlabeled target domain data set X to be testedtMiddle-extracting migration training data with size of m
And
step 2, optimizing the domain evaluation network D by using a back propagation algorithmjAnd a full connection layer
And updating the network according to the following formula:
step 3, training data set X from the source domain with labelssExtracting m-sized micro-call training data
And 4, optimizing a convolution feature mapping network unit M and a fully-connected classification network unit in the convolution neural network model by using a back propagation algorithm to update the convolution neural network according to the following formula:
step 5, evaluating the network D in the current domainjThe output Wessenstein distance is less than a preset threshold ThrewStopping training, keeping the obtained parameters, and completing the migration training and fine adjustment of the fault diagnosis model; otherwise, returning to the step 1 and re-executing the steps 1-5.
In step S5, the unlabeled target domain test data set is input into the trimmed fault diagnosis model, and the fault category of the sample to be tested is output.
Specifically, target domain data to be tested
And inputting the migrated diagnosis model to obtain the fault category of the test sample.
The present invention will be described in detail with reference to the accompanying drawings and specific embodiments.
An overall scheme of a mechanical fault migration diagnosis model based on countermeasure learning under variable working conditions is shown in FIG. 2. Using a deep convolutional neural network as a basic diagnostic model, including multiple convolutional layers, a fully-connected layer, and a final Softmax output layer, a source domain training dataset X is utilizedsComputing the cross entropy of the model prediction output and the data labels, optimizing the underlying diagnostics by back-propagation algorithmAnd the model is subjected to iterative training to achieve the target that the source domain training data can be accurately classified. On the basis of a basic model, a countermeasure training strategy is used for learning domain invariant features in each layer of features in a depth model respectively, a domain evaluation network is used for evaluating Wasserstein distance of feature distribution of a source domain and a target domain, and the distribution distance across the domains is reduced step by step through iterative training. The method aims to solve the problem of migration caused by incomplete data in the practical application of fault diagnosis by adjusting and learning the feature distribution of target domain test data to be close to the feature distribution of source domain data through an antagonistic learning strategy on the basis of accurately classifying the source domain training data. Therefore, after the Wasserstein distance of different domains is reduced by the countertraining of each generation, the migration model is finely adjusted by using the labeled source domain data, the applicability of the migration model to the source domain data is ensured, and the problem of over-migration is avoided.
Taking a rolling bearing fault diagnosis task under 4 different working conditions as an example, the mechanical fault migration diagnosis method based on the countermeasure network is verified. Under the load conditions of 4 different driving motors, vibration signals of the bearings are collected through a vibration sensor, and the sampling frequency is set to be 12 kHz. The fault database comprises 3 faults of inner ring defects, outer ring defects and rolling body defects, wherein each fault has 3 fault sizes, 10 types of faults are added in a normal state, 500 sample data are provided for each type, and specifically shown in table 1, and table 1 is a bearing fault data set description under different working conditions.
TABLE 1
According to the framework provided by the embodiment of the invention, for the bearing fault diagnosis problem, as shown in fig. 3, a convolutional neural network is used for constructing a basic diagnosis model, the framework and the detailed information of the convolutional neural network are shown in table 2, and table 2 is the detailed information of the convolutional neural network model. Using a domain evaluation network to estimate the Wasserstein distance between the cross-domain distributions of each layer, the domain detail information of the domain evaluation network is shown in table 3, and table 3 is the domain evaluation network detail information.
TABLE 2
TABLE 3
Training a convolutional neural network of basic diagnosis:
step 1: training data set X of source domainsRandomly distributed to obtain 70% of training data
And 30% of the test data
Step 2: from training data
Extract data of size m
And step 3: optimizing a convolution feature mapping network unit M and a full-connection classification network unit C in the convolution neural network by using a back propagation algorithm, and updating the convolution neural network according to the following formula:
and 4, step 4: using convolutionNeural network pair test data
Predicting and calculating the accuracy, and when the accuracy is less than a set threshold ThrecReturning to the step 2, and re-executing the steps 2-4; otherwise, stopping training and storing all the obtained parameters, and completing the pre-training of the basic diagnosis model.
The method comprises the following steps of confrontation training and fine tuning of a migration model:
step 1: training dataset X from Source DomainsTesting data set X with target domaintMiddle-extracting migration training data with size of m
And
step 2: optimizing domain evaluation network D using back-propagation algorithmjAnd a full junction layer FcjThe network is updated according to the following formula:
and step 3: training dataset X from Source DomainsExtracting m-sized micro-call training data
And 4, step 4: updating the convolutional neural network by using the convolutional characteristic mapping network unit M and the fully-connected classification network unit C optimized by the back propagation algorithm according to the following formula:
and 5: when the Wessestein distance output by the domain evaluation network is smaller than a set threshold ThrewStopping training and keeping the obtained parameters to finish the training and fine adjustment of the migration model; otherwise, returning to the step 1 and re-executing the steps 1-5.
The hyper-parameters in the training process are set as learning rate α being 0.0001, mini-batch size being m being 64, penalty coefficient lambda being 10, basic diagnosis model test accuracy threshold Threc99%, Wasserstein distance threshold, Threw＝0.0001。
4 different types of working condition data sets are designed to be migrated mutually to verify the method of the embodiment. E.g., A → B, with data set A as the source domain training data set and data set B as the target domain testing data set. The comparative calculation results are shown in table 4, and table 4 shows the diagnosis accuracy of the basic model and the migration model in different migration tasks. The result shows that the migration method of the embodiment of the invention can greatly improve the diagnosis accuracy of the basic model to the target domain test data.
| A→B | A→C | A→D | B→A | B→C | B→D | |
| Basic model | 87.93％ | 89.00％ | 80.73％ | 97.47％ | 99.40％ | 89.00％ |
| Migration model | 99.73％ | 99.67％ | 100.00％ | 99.13％ | 100.00％ | 99.93％ |
| C→A | C→B | C→D | D→A | D→B | D→C | |
| Basic model | 97.00％ | 97.20％ | 89.53％ | 90.20％ | 75.53％ | 79.26％ |
| Transfer moldModel (III) | 98.53％ | 99.80％ | 100.00％ | 98.07％ | 98.27％ | 99.53％ |
FIGS. 4-7 are graphs showing the results of visualization of the diagnostic model domain invariant feature in the D → B migration task using s-TNE techniques. Fig. 4 and 5 are respectively the visualization results of the characteristics of the basic model on the source domain training data and the target domain test data, and fig. 6 and 7 are respectively the visualization results of the migration model on the source domain training data and the target domain test data. The result shows that the difference of the characteristic distribution of the basic diagnosis model of the target test data and the source domain training data is obvious, particularly the categories 7, 8 and 9 are most prominent, the spatial distance of the characteristic distribution is very large, and the basic model is obviously difficult to accurately classify and identify the target domain test data. After the migration is carried out by the method of the embodiment of the invention, the distribution distances of all the categories in the feature space are basically consistent, so that the model after the migration can accurately classify and diagnose the source domain training data and the target domain test data.
According to the mechanical fault migration diagnosis method based on the countermeasure learning, provided by the embodiment of the invention, the original signals of the mechanical fault under different working conditions are obtained, and the source domain training data with the label and the target domain test data without the label are generated; optimizing a convolutional neural network serving as a basic diagnosis model by using a back propagation algorithm by using source domain training data with labels; obtaining a domain invariant feature by using source domain training data without a label and target domain test data through a Wasserstein distance-guided counterstudy method, and realizing cross-domain migration; the source domain training data with the labels are utilized, and a back propagation algorithm is used for fine tuning of the migration model, so that the problem of over-migration is avoided; and inputting the data of the label-free target domain to be tested into the migrated model to obtain a fault classification result.
Next, a mechanical failure migration diagnosis system based on countermeasure learning proposed according to an embodiment of the present invention is described with reference to the drawings.
Fig. 8 is a schematic structural diagram of a mechanical failure migration diagnosis system based on countermeasure learning according to an embodiment of the present invention.
As shown in fig. 8, the countermeasure learning-based mechanical failure migration diagnosis system includes: data generation module 100, model generation module 200, model training module 300, model fine tuning module 400, and diagnostic module 500.
The data generating module 100 is configured to obtain original signals of mechanical faults of a sample to be tested under different working conditions, and analyze the original signals of the mechanical faults under the different working conditions to generate a source domain training data set with a label, a source domain training data set without the label, and a target domain testing data set without the label under the different working conditions.
The model generation module 200 is configured to train a deep convolutional neural network model according to the labeled source domain training data set and the back propagation algorithm, and generate a fault diagnosis model.
The model training module 300 is used for performing migration training on the fault diagnosis model according to the unlabeled source domain training data set, the unlabeled target domain testing data set, the migration method of the countermeasure learning and the Wasserstein distance-guided countermeasure network.
The model fine-tuning module 400 is configured to fine-tune the fault diagnosis model after the migration training according to the labeled source domain training data set and the back propagation algorithm.
The diagnosis module 500 is configured to input the target domain test data set without the label into the trimmed fault diagnosis model, and output the fault category of the sample to be tested.
The mechanical fault migration diagnosis system 10 based on the countermeasure learning obtains the domain invariant feature through the countermeasure learning method, thereby realizing the migration before different domains and realizing the intelligent diagnosis of the variable working condition mechanical fault.
Further, in one embodiment of the present invention, the model generation module includes: the device comprises a data distribution unit, an optimization unit and a determination unit;
a data distribution unit for distributing the labeled source domain training data set XsRandomly distributed to generate 70% of training data
And 30% of the test data
An optimization unit for deriving training data
Extract data of size m
Optimizing a convolution characteristic mapping network unit and a full-connection classification network unit in the convolution neural network model through a back propagation algorithm;
a determination unit for determining the test data by the optimized convolutional neural network model
And predicting and calculating the accuracy, and stopping training and storing all the obtained parameters to generate a fault diagnosis model when the accuracy is greater than the preset accuracy.
Further, in one embodiment of the present invention, the model training module is further configured to,
the method comprises the steps of migrating a generated fault diagnosis model by using source domain data without a label and target domain data without a label based on a migration method of countermeasure learning, exploring a complex feature space by using a Wasserstein distance-guided countermeasure network, reducing distribution difference among different domain data features by using a countermeasure training strategy to obtain domain invariant features, and finally realizing cross-domain migration.
Further, in one embodiment of the present invention, the model training module and the model fine tuning module are further configured to,
slave beltSource domain training dataset X of labelssWith the unlabeled target domain data set X to be testedtMiddle-extracting migration training data with size of m
And
and optimizing the domain evaluation network D using a back propagation algorithmjAnd a full junction layer Fcj：
Training dataset X from tagged source domainsExtracting m-sized micro-call training data
Optimizing a convolution characteristic mapping network unit M and a full-connection classification network unit C in the convolution neural network model by using a back propagation algorithm;
local domain evaluation network DjThe output Wessenstein distance is less than a preset threshold ThrewAnd stopping training, keeping the obtained parameters, and completing the migration training and fine adjustment of the fault diagnosis model.
Further, in one embodiment of the present invention, the model training module is further configured to,
and the fault diagnosis model after the migration training realizes the mapping of the features through the deep convolutional neural network, the domain evaluation network guides the domain-invariant feature learning process, and when the distribution between the source domain training data and the target domain training data is not biased, the domain-invariant feature mapping is obtained.
It should be noted that the foregoing explanation of the embodiment of the mechanical fault migration diagnosis method based on countermeasure learning also applies to the system of this embodiment, and details are not described here.
According to the mechanical fault migration diagnosis system based on the countermeasure learning, provided by the embodiment of the invention, the original signals of the mechanical faults under different working conditions are obtained, and the source domain training data with the labels and the target domain test data without the labels are generated; optimizing a convolutional neural network serving as a basic diagnosis model by using a back propagation algorithm by using source domain training data with labels; obtaining a domain invariant feature by using source domain training data without a label and target domain test data through a Wasserstein distance-guided counterstudy method, and realizing cross-domain migration; the source domain training data with the labels are utilized, and a back propagation algorithm is used for fine tuning of the migration model, so that the problem of over-migration is avoided; and inputting the data of the label-free target domain to be tested into the migrated model to obtain a fault classification result.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (10)
1. A mechanical fault migration diagnosis method based on countermeasure learning is characterized by comprising the following steps:
s1, acquiring original signals of mechanical faults of a sample to be tested under different working conditions, and analyzing the original signals of the mechanical faults under different working conditions to generate a source domain training data set with a label, a source domain training data set without the label and a target domain testing data set without the label under different working conditions;
s2, training a deep convolutional neural network model according to the labeled source domain training data set and the back propagation algorithm, and generating a fault diagnosis model;
s3, performing migration training on the fault diagnosis model according to the unlabeled source domain training data set, the unlabeled target domain testing data set, a migration method of countermeasure learning and a Wasserstein distance-guided countermeasure network;
s4, fine-tuning the fault diagnosis model after migration training according to the source domain training data set with the label and a back propagation algorithm;
and S5, inputting the target domain test data set without the label into the fault diagnosis model after fine adjustment, and outputting the fault category of the sample to be tested.
2. The method according to claim 1, wherein the S2 further comprises:
s21, the labeled source domain training data set X is processedsRandomly distributed to generate 70% of training data
And 30% of the test data
S22, from the training data
Extract data of size m
Optimizing the convolution characteristic mapping network unit and the full-connection classification network unit in the convolution neural network model through the back propagation algorithm;
s23, testing data through the optimized convolutional neural network model
And predicting and calculating the accuracy, executing the step S22 if the accuracy is less than the preset accuracy, otherwise, stopping training and storing all the obtained parameters, and generating the fault diagnosis model.
3. The method according to claim 1, wherein the S3 further comprises:
migrating the fault diagnosis model generated in the step S2 based on the migration method of the countermeasure learning by using the source domain training dataset without a label and the target domain test dataset without a label, exploring a complex feature space by using the Wasserstein distance-guided countermeasure network, reducing distribution differences among different domain data features by using a countermeasure training strategy to obtain domain invariant features, and finally implementing cross-domain migration.
4. The method according to claim 1, wherein the steps S3 and S4 specifically include:
step 1, training a data set X from the labeled source domainsTesting the data set X with the unlabeled target domaintMiddle-extracting migration training data with size of m
And
step 2, using the reverse transmissionBroadcast algorithm optimization domain evaluation network DjAnd a full junction layer Fcj：
Step 3, training data set X from the labeled source domainsExtracting m-sized micro-call training data
Step 4, optimizing a convolution feature mapping network unit M and a full-connection classification network unit C in the convolution neural network model by using the back propagation algorithm;
step 5, evaluating the network D in the current domainjThe output Wessenstein distance is less than a preset threshold ThrewStopping training, keeping the obtained parameters, and completing the migration training and fine adjustment of the fault diagnosis model; otherwise, returning to the step 1 and re-executing the steps 1-5.
5. The method according to claim 1, wherein the step S3 further comprises:
and the fault diagnosis model after the migration training realizes the mapping of the features through the deep convolutional neural network, the domain evaluation network guides the domain-invariant feature learning process, and when the distribution between the source domain training data and the target domain training data is not biased, the domain-invariant feature mapping is obtained.
6. A system for mechanical fault migration diagnosis based on countermeasure learning, comprising:
the data generation module is used for acquiring original signals of mechanical faults of a sample to be tested under different working conditions, and analyzing the original signals of the mechanical faults under the different working conditions to generate a source domain training data set with a label, a source domain training data set without the label and a target domain testing data set without the label under the different working conditions;
the model generation module is used for training a deep convolutional neural network model according to the source domain training data set with the label and a back propagation algorithm to generate a fault diagnosis model;
a model training module for performing migration training on the fault diagnosis model according to the unlabeled source domain training dataset, the unlabeled target domain testing dataset, a migration method of countermeasure learning, and a Wasserstein distance-guided countermeasure network;
the model fine-tuning module is used for fine-tuning the fault diagnosis model after the migration training according to the source domain training data set with the label and a back propagation algorithm;
and the diagnosis module is used for inputting the target domain test data set without the label into the fault diagnosis model after fine adjustment and outputting the fault category of the sample to be tested.
7. The system of claim 6, wherein the model generation module comprises: the device comprises a data distribution unit, an optimization unit and a determination unit;
the data distribution unit is used for distributing the labeled source domain training data set XsRandomly distributed to generate 70% of training data
And 30% of the test data
The optimization unit is used for obtaining training data
Extract data of size m
Optimizing the convolution characteristic mapping network unit and the full-connection classification network unit in the convolution neural network model through the back propagation algorithm;
the determining unit is used for testing data through the optimized convolutional neural network model
And predicting and calculating the accuracy, and stopping training and storing all the obtained parameters to generate the fault diagnosis model when the accuracy is greater than the preset accuracy.
8. The system of claim 6, wherein the model training module is further configured to,
migrating the generated fault diagnosis model by using the source domain training data set without the label and the target domain testing data set without the label based on the migration method of the countermeasure learning, exploring a complex feature space by using the Wasserstein distance-guided countermeasure network, reducing distribution difference among different domain data features by using a countermeasure training strategy to obtain domain invariant features, and finally realizing cross-domain migration.
9. The system of claim 6, wherein the model training module and the model fine tuning module are further configured to,
training a dataset X from the tagged source domainsTesting the data set X with the unlabeled target domaintMiddle-extracting migration training data with size of m
And
and optimizing the domain evaluation network D using the back propagation algorithmjAnd a full junction layer Fcj：
Training a dataset X from the tagged source domainsExtracting m-sized micro-call training data
Optimizing a convolution feature mapping network unit M and a full-connection classification network unit C in the convolution neural network model by using the back propagation algorithm;
local domain evaluation network DjThe output Wessenstein distance is less than a preset threshold ThrewAnd stopping training, keeping the obtained parameters, and completing the migration training and fine adjustment of the fault diagnosis model.
10. The system of claim 6, wherein the model training module is further configured to,
and the fault diagnosis model after the migration training realizes the mapping of the features through the deep convolutional neural network, the domain evaluation network guides the domain-invariant feature learning process, and when the distribution between the source domain training data and the target domain training data is not biased, the domain-invariant feature mapping is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910289486.8A CN109947086B (en) | 2019-04-11 | 2019-04-11 | Mechanical fault migration diagnosis method and system based on counterstudy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910289486.8A CN109947086B (en) | 2019-04-11 | 2019-04-11 | Mechanical fault migration diagnosis method and system based on counterstudy |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109947086A CN109947086A (en) | 2019-06-28 |
| CN109947086B true CN109947086B (en) | 2020-07-28 |
Family
ID=67014823
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910289486.8A Active CN109947086B (en) | 2019-04-11 | 2019-04-11 | Mechanical fault migration diagnosis method and system based on counterstudy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109947086B (en) |
Families Citing this family (56)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414383A (en) * | 2019-07-11 | 2019-11-05 | 华中科技大学 | Convolutional neural networks based on Wasserstein distance fight transfer learning method and its application |
| CN110567720B (en) * | 2019-08-07 | 2020-12-18 | 东北电力大学 | Method for diagnosing depth confrontation of fault of fan bearing under unbalanced small sample scene |
| CN110555273B (en) * | 2019-09-05 | 2023-03-24 | 苏州大学 | Bearing life prediction method based on hidden Markov model and transfer learning |
| CN110738107A (en) * | 2019-09-06 | 2020-01-31 | 上海衡道医学病理诊断中心有限公司 | microscopic image recognition and segmentation method with model migration function |
| CN110907176B (en) * | 2019-09-30 | 2021-02-02 | 合肥工业大学 | Wasserstein distance-based fault diagnosis method for deep countermeasure migration network |
| CN110751207B (en) * | 2019-10-18 | 2022-08-05 | 四川大学 | Fault diagnosis method for anti-migration learning based on deep convolution domain |
| CN110728377B (en) * | 2019-10-21 | 2020-06-09 | 山东大学 | Intelligent fault diagnosis method and system for electromechanical equipment |
| CN110796199B (en) * | 2019-10-30 | 2021-05-28 | 腾讯科技（深圳）有限公司 | Image processing method and device and electronic medical equipment |
| CN110866365B (en) * | 2019-11-22 | 2021-06-01 | 北京航空航天大学 | Mechanical equipment intelligent fault diagnosis method based on partial migration convolutional network |
| CN111158964B (en) * | 2019-11-26 | 2021-06-08 | 北京邮电大学 | Disk failure prediction method, system, device and storage medium |
| CN111027678B (en) * | 2019-12-04 | 2023-08-04 | 湃方科技(北京)有限责任公司 | Data migration method and device |
| CN111046581B (en) * | 2019-12-27 | 2022-10-04 | 国网江苏省电力有限公司电力科学研究院 | Power transmission line fault type identification method and system |
| CN111060318B (en) * | 2020-01-09 | 2021-12-28 | 山东科技大学 | Bearing fault diagnosis method based on deep countermeasure migration network |
| CN111442926B (en) * | 2020-01-11 | 2021-09-21 | 哈尔滨理工大学 | Fault diagnosis method for rolling bearings of different models under variable load based on deep characteristic migration |
| CN111314113B (en) * | 2020-01-19 | 2023-04-07 | 赣江新区智慧物联研究院有限公司 | Internet of things node fault detection method and device, storage medium and computer equipment |
| CN111444780A (en) * | 2020-03-06 | 2020-07-24 | 同济大学 | Bearing fault diagnosis method based on deep sparse noise reduction self-coding network |
| CN111444952B (en) * | 2020-03-24 | 2024-02-20 | 腾讯科技（深圳）有限公司 | Sample recognition model generation method, device, computer equipment and storage medium |
| WO2021199194A1 (en) * | 2020-03-30 | 2021-10-07 | 三菱電機株式会社 | Diagnosis system, diagnosis method, and program |
| CN111428803A (en) * | 2020-03-31 | 2020-07-17 | 山东大学 | Wasserstein distance-based depth domain adaptive image classification method |
| CN111612035A (en) * | 2020-04-18 | 2020-09-01 | 华为技术有限公司 | Method for training migration model, method and device for detecting fault |
| CN111614215B (en) * | 2020-05-11 | 2021-11-12 | 东南大学 | Method for designing driving motor for electric vehicle based on generation of countermeasure network |
| CN111598161A (en) * | 2020-05-14 | 2020-08-28 | 哈尔滨工业大学（威海） | Engine gas circuit state diagnosis system based on CNN transfer learning |
| CN111626345A (en) * | 2020-05-15 | 2020-09-04 | 北京航空航天大学 | Multi-stage deep convolution transfer learning fault diagnosis method between different bearing devices |
| CN111651937B (en) * | 2020-06-03 | 2023-07-25 | 苏州大学 | Method for diagnosing faults of in-class self-adaptive bearing under variable working conditions |
| CN111898634B (en) * | 2020-06-22 | 2022-08-16 | 西安交通大学 | Intelligent fault diagnosis method based on depth-to-reactance-domain self-adaption |
| CN111814396B (en) * | 2020-07-02 | 2024-02-20 | 重庆大学 | Centrifugal fan fault early warning method based on transfer learning |
| CN111949459B (en) * | 2020-08-10 | 2022-02-01 | 南京航空航天大学 | Hard disk failure prediction method and system based on transfer learning and active learning |
| CN111949796B (en) * | 2020-08-24 | 2023-10-20 | 云知声智能科技股份有限公司 | Method and system for analyzing front-end text of voice synthesis of resource-limited language |
| CN111998936B (en) * | 2020-08-25 | 2022-04-15 | 四川长虹电器股份有限公司 | Equipment abnormal sound detection method and system based on transfer learning |
| CN112161784B (en) * | 2020-09-07 | 2022-01-18 | 华南理工大学 | Mechanical fault diagnosis method based on multi-sensor information fusion migration network |
| CN112149726B (en) * | 2020-09-21 | 2024-02-09 | 浙江工业大学 | Totally-enclosed compressor fault diagnosis method based on knowledge sharing and model migration |
| CN112149350B (en) * | 2020-09-21 | 2024-04-26 | 浙江工业大学 | Model migration method for solving data dependency problem in heat pump system fault diagnosis |
| CN112329329B (en) * | 2020-09-22 | 2024-02-20 | 东北大学 | Simulation data driven rotary machine depth semi-supervised migration diagnosis method |
| CN112257851A (en) * | 2020-10-29 | 2021-01-22 | 重庆紫光华山智安科技有限公司 | Model confrontation training method, medium and terminal |
| CN112330063B (en) * | 2020-11-25 | 2024-03-26 | 新奥新智科技有限公司 | Equipment fault prediction method, equipment fault prediction device and computer readable storage medium |
| CN112668633B (en) * | 2020-12-25 | 2022-10-14 | 浙江大学 | Adaptive graph migration learning method based on fine granularity field |
| CN112629863B (en) | 2020-12-31 | 2022-03-01 | 苏州大学 | Bearing fault diagnosis method for dynamic joint distribution alignment network under variable working conditions |
| CN112733943B (en) * | 2021-01-13 | 2024-03-22 | 浙江工业大学 | Heat pump fault diagnosis model migration method based on data mixed shearing technology |
| CN113569887B (en) * | 2021-01-18 | 2022-10-11 | 腾讯医疗健康(深圳)有限公司 | Picture recognition model training and picture recognition method, device and storage medium |
| CN112836896A (en) * | 2021-03-03 | 2021-05-25 | 西门子工厂自动化工程有限公司 | Method for maintaining equipment and system for maintaining equipment |
| CN113076920B (en) * | 2021-04-20 | 2022-06-03 | 同济大学 | Intelligent fault diagnosis method based on asymmetric domain confrontation self-adaptive model |
| CN113204280B (en) * | 2021-05-08 | 2023-09-26 | 山东英信计算机技术有限公司 | Method, system, equipment and medium for diagnosing power failure |
| CN113392881B (en) * | 2021-05-27 | 2023-04-18 | 重庆大学 | Rotary machine fault diagnosis method based on transfer learning |
| CN113538353B (en) * | 2021-07-05 | 2023-09-01 | 华北电力大学（保定） | Five-phase asynchronous motor rolling bearing fault diagnosis method based on single-channel diagram data enhancement and migration training residual error network |
| CN114136622B (en) * | 2021-08-10 | 2023-04-18 | 南京航空航天大学 | DBN-DTL-based aeroengine gas circuit fault diagnosis method |
| CN114021610B (en) * | 2021-09-10 | 2023-04-07 | 中国大唐集团科学技术研究院有限公司火力发电技术研究院 | Fan fault recognition model training method and system based on transfer learning |
| CN114065945B (en) * | 2021-10-29 | 2024-08-02 | 宜宾电子科技大学研究院 | Intelligent fault diagnosis method based on DJTN transfer learning |
| CN114021285B (en) * | 2021-11-17 | 2024-04-12 | 上海大学 | Rotary machine fault diagnosis method based on mutual local countermeasure migration learning |
| CN114167837B (en) * | 2021-12-02 | 2023-09-15 | 中国路桥工程有限责任公司 | Intelligent fault diagnosis method and system for railway signal system |
| CN114305446B (en) * | 2021-12-25 | 2024-08-16 | 肇庆星网医疗科技有限公司 | Atrial fibrillation detection method and system based on artificial intelligence |
| CN114239859B (en) * | 2022-02-25 | 2022-07-08 | 杭州海康威视数字技术股份有限公司 | Power consumption data prediction method and device based on transfer learning and storage medium |
| CN114969998B (en) * | 2022-03-08 | 2024-08-23 | 西安交通大学 | Engine fault diagnosis method based on fault characteristic migration and progressive fine adjustment |
| CN115562029B (en) * | 2022-10-17 | 2023-06-16 | 杭州天然气有限公司 | Intelligent control method and system for natural gas turbine expansion generator set |
| CN116009480B (en) * | 2023-03-24 | 2023-06-09 | 中科航迈数控软件(深圳)有限公司 | Fault monitoring method, device and equipment of numerical control machine tool and storage medium |
| CN116992953B (en) * | 2023-09-27 | 2024-04-19 | 苏州捷杰传感技术有限公司 | Model training method, fault diagnosis method and device |
| CN118519424B (en) * | 2024-07-23 | 2024-09-20 | 四川环龙技术织物有限公司 | Fault detection method and system for papermaking clothing production control system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009205615A (en) * | 2008-02-29 | 2009-09-10 | Internatl Business Mach Corp <Ibm> | Change analysis system, method, and program |
| CN105628383A (en) * | 2016-02-01 | 2016-06-01 | 东南大学 | Bearing fault diagnosis method and system based on improved LSSVM transfer learning |
| CN108009633A (en) * | 2017-12-15 | 2018-05-08 | 清华大学 | A kind of Multi net voting towards cross-cutting intellectual analysis resists learning method and system |
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| CN108875918A (en) * | 2018-08-14 | 2018-11-23 | 西安交通大学 | It is a kind of that diagnostic method is migrated based on the mechanical breakdown for being adapted to shared depth residual error network |
| CN109165695A (en) * | 2018-09-17 | 2019-01-08 | 重庆交通大学 | Piler method for diagnosing faults based on fault tree and transfer learning |
-
2019
- 2019-04-11 CN CN201910289486.8A patent/CN109947086B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009205615A (en) * | 2008-02-29 | 2009-09-10 | Internatl Business Mach Corp <Ibm> | Change analysis system, method, and program |
| CN105628383A (en) * | 2016-02-01 | 2016-06-01 | 东南大学 | Bearing fault diagnosis method and system based on improved LSSVM transfer learning |
| CN108009633A (en) * | 2017-12-15 | 2018-05-08 | 清华大学 | A kind of Multi net voting towards cross-cutting intellectual analysis resists learning method and system |
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| CN108875918A (en) * | 2018-08-14 | 2018-11-23 | 西安交通大学 | It is a kind of that diagnostic method is migrated based on the mechanical breakdown for being adapted to shared depth residual error network |
| CN109165695A (en) * | 2018-09-17 | 2019-01-08 | 重庆交通大学 | Piler method for diagnosing faults based on fault tree and transfer learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109947086A (en) | 2019-06-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109947086B (en) | Mechanical fault migration diagnosis method and system based on counterstudy | |
| Chen et al. | Combining empirical mode decomposition and deep recurrent neural networks for predictive maintenance of lithium-ion battery | |
| Grezmak et al. | Interpretable convolutional neural network through layer-wise relevance propagation for machine fault diagnosis | |
| Parisi et al. | Automated location of steel truss bridge damage using machine learning and raw strain sensor data | |
| Ayodeji et al. | Causal augmented ConvNet: A temporal memory dilated convolution model for long-sequence time series prediction | |
| CN112633339A (en) | Bearing fault intelligent diagnosis method, bearing fault intelligent diagnosis system, computer equipment and medium | |
| Moradi et al. | Intelligent health indicator construction for prognostics of composite structures utilizing a semi-supervised deep neural network and SHM data | |
| US11954923B2 (en) | Method for rating a state of a three-dimensional test object, and corresponding rating system | |
| CN103630244A (en) | Equipment fault diagnosis method and system of electric power system | |
| Kapteyn et al. | From physics-based models to predictive digital twins via interpretable machine learning | |
| Shi et al. | Intelligent fault diagnosis of rolling mills based on dual attention-guided deep learning method under imbalanced data conditions | |
| CN115618732A (en) | Nuclear reactor digital twin key parameter autonomous optimization data inversion method | |
| Qi et al. | Application of EMD combined with deep learning and knowledge graph in bearing fault | |
| CN114861349B (en) | Rolling bearing RUL prediction method based on model migration and wiener process | |
| CN116702076A (en) | Small sample migration learning fault diagnosis method, system, computer and storage medium based on CNN feature fusion | |
| CN115221973A (en) | Aviation bearing fault diagnosis method based on enhanced weighted heterogeneous ensemble learning | |
| CN116842379A (en) | Mechanical bearing residual service life prediction method based on DRSN-CS and BiGRU+MLP models | |
| Zhao | A method of power supply health state estimation based on grey clustering and fuzzy comprehensive evaluation | |
| Esfahani et al. | Remaining useful life prognostics based on stochastic degradation modeling: turbofan engine as case study | |
| Martins et al. | Prediction maintenance based on vibration analysis and deep learning–A case study of a drying press supported on a hidden Markov model | |
| Ovseenko et al. | The possibility of artificial neural network application in prototyping in instrument making industry | |
| Ren et al. | A Lightweight Group Transformer-Based Time Series Reduction Network for Edge Intelligence and Its Application in Industrial RUL Prediction | |
| CN116714437B (en) | Hydrogen fuel cell automobile safety monitoring system and monitoring method based on big data | |
| Li et al. | A remaining useful life prediction method considering the dimension optimization and the iterative speed | |
| CN117872219A (en) | Open-circuit fault diagnosis method and system for train traction converter |
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 |