CN115618267A - Unsupervised domain self-adaption and entropy optimization equipment sensing diagnosis method and system - Google Patents
Unsupervised domain self-adaption and entropy optimization equipment sensing diagnosis method and system Download PDFInfo
- Publication number
- CN115618267A CN115618267A CN202211424445.3A CN202211424445A CN115618267A CN 115618267 A CN115618267 A CN 115618267A CN 202211424445 A CN202211424445 A CN 202211424445A CN 115618267 A CN115618267 A CN 115618267A
- Authority
- CN
- China
- Prior art keywords
- data
- domain
- training
- sample
- label
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000005457 optimization Methods 0.000 title claims abstract description 7
- 238000012549 training Methods 0.000 claims abstract description 48
- 238000009826 distribution Methods 0.000 claims abstract description 26
- 238000005516 engineering process Methods 0.000 claims abstract description 16
- 230000036541 health Effects 0.000 claims abstract description 13
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 12
- 238000000605 extraction Methods 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000001172 regenerating effect Effects 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 238000013500 data storage Methods 0.000 claims description 2
- 230000003044 adaptive effect Effects 0.000 claims 5
- 238000002405 diagnostic procedure Methods 0.000 claims 4
- 230000001953 sensory effect Effects 0.000 claims 3
- 238000005192 partition Methods 0.000 claims 1
- 238000012546 transfer Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012631 diagnostic technique Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000008646 thermal stress Effects 0.000 description 1
Images
Classifications
-
- 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
- G01M13/045—Acoustic or vibration analysis
-
- 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
- G06N3/088—Non-supervised learning, e.g. competitive learning
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Acoustics & Sound (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
An unsupervised domain self-adaption and entropy optimization device sensing diagnosis method and system. The method comprises the following steps: 1) Collecting vibration sensing signals of equipment under different working conditions, and preprocessing data by utilizing a gram angular field technology; 2) Constructing a fault diagnosis model based on a convolutional neural network, wherein the fault diagnosis model comprises a feature extraction module, a feature consistency judgment module and a feature identification module; 3) Training a source domain labeled sample on a feature extraction module and a feature recognition module of a fault diagnosis model; 4) Carrying out consistency training on the target domain label-free training sample on a fault diagnosis model by using a data enhancement strategy; 5) And inputting the sample of which the target domain is not labeled and participates in training into the trained diagnostic model and obtaining the health state of the equipment. By combining the RandAugment technology and the domain self-adaptive technology, the identification precision of the model to the health state of the equipment can be improved under the conditions of data imbalance and data distribution transfer.
Description
Technical Field
The invention belongs to the technical field of mechanical equipment health management and computer artificial intelligence, and particularly relates to an equipment sensing diagnosis method based on unsupervised domain self-adaptation and convolutional neural network entropy optimization under multiple working conditions and data imbalance.
Background
Rolling bearings are one of the most widely used components in mechanical equipment (e.g., pharmaceutical equipment) but are susceptible to failure due to minor failure. The fault diagnosis based on the equipment sensing information can support the maintenance of reliable operation of the device, prolong the service life of the equipment and prevent major accidents. Rolling bearings generally operate continuously under random noise, impact loads and thermal stresses. The signal processing-based approach may enable a degree of faulty bearing mechanical diagnosis-by revealing bearing anomaly characteristics. However, this method has high requirements on the professional knowledge of the practitioner. Due to the complex and diverse nature of the work environment, more effective diagnostic techniques are needed to solve the problem. The development of artificial intelligence has promoted the progress of fault diagnosis technology, and a plurality of methods with excellent performance emerge. With the increase of the complexity of the industrial sensor network, a deep learning method which is good at processing complex, nonlinear and variable data has gradually become a mainstream research method.
However, there have been studies with a premise hypothesis: the training data and the test data are distributed in a consistent manner. Under the constraints of working conditions, sensor performance and the like, the data distribution of the target sample and the source domain sample is different, and a data distribution transfer phenomenon can occur, so that the requirement of consistent distribution is too severe in the variable working condition production process of the actual industry, and the applicability is limited. In addition, in practical application, the normal operation time of the equipment is longer than the fault time of the equipment, so that the collected data of the healthy sample and the data of the fault sample are often unbalanced, and the phenomenon of unbalance is presented. However, the application research of the fault diagnosis method for mechanical equipment (such as a pharmaceutical device) under multiple working conditions and data imbalance is quite deficient at present.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide an equipment sensing diagnosis method based on unsupervised domain self-adaption and entropy optimization, which has better self-adaption and higher accuracy, under multiple working conditions and data imbalance. The method utilizes a data enhancement strategy combining a RandAugment technology and a domain self-adaptive technology, improves the fault diagnosis precision under the conditions of data imbalance and multiple working conditions, and relieves the influence of the data imbalance and the data distribution inconsistency of a source domain and a target domain on performance reduction to a certain extent.
The aim of the invention can be realized by the following technical scheme:
step 1: collecting vibration sensing signals at the device under a plurality of operating conditions, each operating condition comprising a plurality of device health states; the data of one working condition is a data field, the data distribution in different data fields is inconsistent, and long-tail distribution is carried out on the data field to divide samples;
step 2: preprocessing the data, converting the collected one-dimensional data into two-dimensional data through a gram angular field algorithm, and storing the two-dimensional data into a magnetic disk; constructing a fault diagnosis model based on a convolutional neural network, wherein the fault diagnosis model comprises a feature extraction module, a feature consistency judgment module and a feature identification module;
and step 3: randomly selecting a domain from a data storage bank as a source domain, and randomly sampling the domain to divide a data set; training the labeled sample on a feature extraction module and a feature recognition module of the fault diagnosis model;
and 4, step 4: randomly selecting one of the rest databases as a target domain, and randomly sampling and dividing a data set; on the basis of the source domain training model in the step 3, training a target domain label-free training sample on a fault diagnosis model by using a data enhancement strategy;
and 5: inputting the sample of the target domain without label and without participation in training into the trained diagnosis model and obtaining the health state of the equipment.
Further, the step 2 specifically comprises the following steps:
acquiring a one-dimensional time sequence vibration signal on equipment, preprocessing the one-dimensional data through a gram angular field algorithm, and converting to obtain two-dimensional data, wherein a gram angular field conversion formula is as follows:
wherein, X = [ X = 1 ,x 2 ,…,x n ]Is a set of n time-series data,
is a regularized data set, α ∈ [0, π]Is the angle value, t, after the regularization data encoding i Is the timestamp, r is the polar radius, and N is a constant factor.
Further, the step 3 specifically includes the following steps:
and modifying the structure of the ReNet50 network, and replacing the last linear layer with a full-connection layer which has an Xavier initialization weight and is free of bias. Extracting features by using the modified convolutional neural network, wherein the process is as follows:
z(i)＝g(w T x(i)+b)#(2)
wherein x (i) = { x 1 ,x 2 ,…,x n Is n input data, g (×) is the linear unit ReLU activation function, w is the weight matrix, b is the bias term, z (×) is the output.
Further, the step 4 specifically includes the following steps:
step 4.1: randomly sampling the label-free image data of the target domain to form a training set and a testing set;
step 4.2: the data enhancement strategy in the invention consists of a RandAugment technology and a self-training algorithm;
step 4.3: the training set image is transformed by RandAugment technology, generating a dataset P = { r = { (r) } 1 (x τ ),r 2 (x τ ),…r n (x τ ) In which x τ Is from the target domain unlabeled training data, r i Is the transform i-th operation function;
step 4.4: inputting the data set P into a data enhancement strategy, selecting the most reliable pseudo sample from the output result, and deciding the consistency of the label and the truth label;
step 4.5: if the decision result is consistency, minimizing the pseudo sample entropy loss value consistent with the true value label, otherwise, maximizing the inconsistent pseudo sample entropy loss value;
step 4.6: combining conditional entropy loss, conditional distribution distance and classification loss of a target domain label-free training set to form a target function, and training model parameters;
further, the consistency in step 4.4 is described as:
wherein the prediction + The generated pseudo label is consistent with the true label, prediction - It is the generated pseudo label that does not coincide with the true label.
Further, the entropy loss in step 4.5 is described by the formula:
wherein L is EO Is the complete function of the entropy minimization,
is a function of the loss of consistency entropy,
is a function of the loss of non-uniformity entropy,the latter two functions are defined as follows:
wherein,
is the moving average exponent of the sample label y and p is the probability distribution function.
Further, the step 4.6 specifically includes the following steps:
step 4.6.1: calculating the condition distribution distance of the target domain, wherein the process description formula is as follows:
wherein, S is the source domain,
is the target domain, X * Is the source domain or the target domain, k is the number of Gaussian kernels, MMD Hk The representative kernel space is embedded with k gaussian kernels, defined as follows:
where φ is a nonlinear mapping function of the regenerating kernel Hilbert space H,
is a sample from either the source domain or the target domain.
Step 4.6.2: combining the classified cross entropy loss and the conditional distribution distance loss with the conditional entropy loss to form an objective function to be optimized, wherein the formula of the objective function is as follows:
L final ＝minL c +αL M +βL EO #(9)
where α and β are penalty factors.
The invention also provides a system of the unsupervised domain self-adaptive entropy-optimized equipment sensing diagnosis method, which is characterized by comprising the following steps:
the feature extraction module is used for extracting high-dimensional feature representation from a source domain sample of the preprocessed image;
the characteristic consistency module is used for judging the consistency of the pseudo samples generated by the self-training algorithm and the real samples and selecting the most reliable pseudo samples out of the training model;
and the characteristic identification module is used for identifying the sample which is not labeled in the target domain and does not participate in training so as to judge the corresponding health category of the test sample.
Furthermore, the network frameworks of the feature extraction module and the feature recognition module are convolutional neural networks, and the basis of the feature consistency module is a RandAugment technology and a self-training algorithm.
The invention has the beneficial effects that:
1. the deep features of the data can be automatically mined through the deep learning convolutional neural network without manually extracting the features, so that the manpower and the time are saved, and the efficiency is improved.
2. The characteristic consistency module in the invention can improve the quality of pseudo samples generated by semi-supervised learning and select the most reliable pseudo samples for model training.
3. Under the framework of a convolutional neural network, the RandAugment technology and the domain self-adaptive technology are combined, the conditional entropy loss and the conditional distribution distance loss are calculated and optimized as a part of a target function, the limitation of a diagnosis model under multiple conditions and unbalanced data can be well broken through, the diagnosis model can be well adapted to diagnosis tasks under different working conditions and unbalanced sample sizes, the self-adaptability and the generalization capability of the model are improved, and the method can be widely applied to equipment health monitoring tasks of complex systems such as machinery, aviation, railways and the like.
Drawings
FIG. 1 is a graph of outer race, inner race, and cage failure preprocessing images for a tag in an embodiment of the present invention.
Fig. 2 is a data diagram of transformation using the RandAugemnt technique in an embodiment of the present invention.
FIG. 3 is a schematic diagram of a model structure of the deep learning framework in the present invention.
Fig. 4 is a flow chart of a method corresponding to the present invention.
FIG. 5 is a diagram illustrating the classification result of the test data according to the embodiment of the present invention.
Detailed Description
The invention is described in further detail below with reference to the figures and examples. The following examples are some examples of the present invention and are intended to illustrate the present invention, but not to limit the scope of the present invention.
Example one
Experimental data an LDK UER204 ac motor was collected at a sampling frequency of 64khz using the bearing data set of the seian university of transportation. The failure type occurs mainly in the inner race, the outer race and the cage.
As shown in fig. 4, the present invention comprises the steps of:
step 1: vibration signals are collected at the device under a plurality of operating conditions, each operating condition comprising a plurality of device health states. The data of one working condition is a data field, the data distribution in different data fields is inconsistent, and long-tail distribution is carried out on the data field to divide samples;
step 2: preprocessing the data, converting the collected one-dimensional data into two-dimensional data through a gram angular field algorithm, and storing the two-dimensional data into a magnetic disk; constructing a fault diagnosis model based on a convolutional neural network, wherein the fault diagnosis model comprises a feature extraction module, a feature consistency judgment module and a feature identification module;
and step 3: randomly selecting a domain from the data repository as a source domain, and randomly sampling the domain to divide the data set; training the labeled sample on a feature extraction module and a feature recognition module of the fault diagnosis model;
and 4, step 4: randomly selecting one of the rest databases as a target domain, and randomly sampling and dividing a data set; on the basis of the model in the step 3, training a target domain label-free training sample on a fault diagnosis model by using a data enhancement strategy;
and 5: inputting the sample of the target domain without label and without participation in training into the trained diagnosis model and obtaining the health state of the equipment.
Further, the step 1 specifically includes the following steps:
step 1.1: the present embodiment includes 3 operating conditions, which are 1hp, 2hp, and 3hp, respectively. 3 working conditions form 3 data sets, each data set is named by a working condition name and comprises two fault states, and the two fault states are shown in a table I;
step 1.2: because some fault categories are easier to appear than other categories, power law distribution is adopted to subdivide the data set sample distribution, including two data unbalanced distribution settings;
table one: data set detail description
The diagnostic task in the embodiment is represented by the symbol X AB_C It is shown that the source domain is the Ahp dataset, the target domain is the Bhp dataset, and C is the data imbalance distribution 1 or 2. Under two different data unbalanced distributions, a source domain labeled sample and a target domain unlabeled training sample participate in a training stage, and a target domain unlabeled test sample is only used in a testing stage.
Further, the step 2 specifically comprises the following steps:
collecting one-dimensional time sequence vibration signals on equipment, preprocessing the one-dimensional data through a gram angular field algorithm, and converting to obtain two-dimensional data, wherein a gram angular field conversion formula is as follows:
wherein, X = [ X ] 1 ,x 2 ,…,x n ]Is a set of n time-series data,
is a regularized data set, α ∈ [0, π ∈ ]]Is the angle value, t, after the regularization data encoding i Is the timestamp, r is the polar radius, and N is a constant factor.
Further, the step 3 specifically includes the following steps:
the fault diagnosis model established in the present embodiment is shown in fig. 2, and the fault diagnosis process is shown in fig. 3.
And modifying the structure of the ReNet50 network, and replacing the last linear layer with a full-connection layer which has an Xavier initialization weight and is free of bias. Extracting features by using the modified convolutional neural network, wherein the process is as follows:
z(i)＝g(w T x(i)+b)#(2)
wherein x (i) = { x 1 ,x 2 ,…,x n Is n input data, g (×) is the linear unit ReLU activation function, w is the weight matrix, b is the bias term, z (×) is the output.
Further, the step 4 specifically includes the following steps:
step 4.1: randomly sampling the label-free image data of the target domain to form a training set and a testing set;
step 4.2: the data enhancement strategy in the invention consists of a RandAugment technology and a self-training algorithm;
step 4.3: transforming the training set image by a RandAugment technology to generate a data set P = { r = { (r) } 1 (x τ ),r 2 (x τ ),…r n (x τ ) In which x is τ Is from the target domain unlabeled training data, r i Is the operation function of transforming i times, the transformation result of each sample is shown in fig. 4;
step 4.4: inputting the data set P into a self-training algorithm, selecting the most reliable pseudo sample from the output result, and deciding the consistency of the label and the truth value label;
step 4.5: if the decision result is consistency, minimizing the pseudo sample entropy loss value consistent with the true value label, otherwise, maximizing the inconsistent pseudo sample entropy loss value;
step 4.6: combining conditional entropy loss, conditional distribution distance and classification loss of a target domain label-free training set to form a target function, and training model parameters;
further, the consistency in step 4.4 is described as:
wherein the prediction + The generated pseudo label is consistent with the true label, prediction - It is the generated pseudo label that does not coincide with the true label.
Further, the entropy loss in step 4.5 is described by the formula:
wherein L is EO Is the complete entropy-minimizing function of the entropy,
is a function of the loss of consistency entropy,
is an inconsistent entropy loss function, the latter two functions are defined as follows:
wherein,
is the moving average exponent of the sample label y and p is the probability distribution function.
Further, the step 4.6 specifically includes the following steps:
step 4.6.1: calculating the condition distribution distance of the target domain, wherein the process description formula is as follows:
wherein, S is the source domain,
is the target domain, X * Is the source domain or the target domain, k is the number of Gaussian kernels, MMD Hk The representative kernel space is embedded with k gaussian kernels, defined as follows:
where φ is a nonlinear mapping function of the regenerative kernel Hilbert space H,
is a sample from either the source domain or the target domain.
Step 4.6.2: combining the classified cross entropy loss and the conditional distribution distance loss with the conditional entropy loss to form an objective function to be optimized, wherein the formula of the objective function is as follows:
L final ＝minL c +αL M +βL EO #(9)
where α and β are penalty factors. The test results for the 8 operating conditions are shown in fig. 5. The classification result in fig. 5 shows that the average diagnostic accuracy of the model under 8 tasks reaches 99.32%, which indicates that the model under the working conditions can maintain the generalization performance and the mobility under the conditions of multiple working conditions and data imbalance, and is beneficial to monitoring the health state of the equipment under the working conditions.
The above-described embodiments are some of the embodiments of the present invention, and are used for illustrating the present invention, but are not intended to limit the scope of the present invention. Various modifications and alterations to this invention will become apparent to those skilled in the art; any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.
Claims (6)
1. An unsupervised domain self-adaption and entropy optimization device sensing diagnosis method and system are characterized by comprising the following steps:
step 1: collecting vibration sensing signals at the device under a plurality of operating conditions, each operating condition comprising a plurality of device health states; the data of one working condition is a data domain, the data distribution in different data domains is inconsistent, and long tail distribution is carried out on the data domain to divide samples;
and 2, step: preprocessing the data, converting the collected one-dimensional data into two-dimensional data through a gram angular field algorithm, and storing the two-dimensional data into a magnetic disk; constructing a fault diagnosis model based on a convolutional neural network, wherein the fault diagnosis model at least comprises a feature extraction module, a feature consistency judgment module and a feature identification module;
and step 3: randomly selecting a domain from a data storage bank as a source domain, and randomly sampling the domain to divide a data set; training the labeled sample on a feature extraction module and a feature recognition module of the fault diagnosis model;
and 4, step 4: one of the remaining domains in the data repository is randomly selected as a target domain and randomly sampled to partition the data set. On the basis of the fault diagnosis model trained in the source domain in the step 3, training a target domain label-free training sample on the fault diagnosis model by using a data enhancement strategy;
and 5: inputting the sample of the target domain without label and without participation in training into the trained diagnosis model and obtaining the health state of the equipment.
2. An unsupervised domain adaptive and entropy optimized device sensory diagnostic method according to claim 1, wherein the step 2 comprises:
acquiring a one-dimensional time sequence vibration sensing signal on equipment, preprocessing the one-dimensional data through a gram angular field algorithm, and converting to obtain two-dimensional data, wherein a gram angular field conversion formula is as follows:
wherein, X = [ X ] 1 ,x 2 ,…,x n ]Is a set of n time-series data,
is a regularized data set, α ∈ [0, π ∈ ]]Is the angle value, t, after the regularization data encoding i Is the timestamp, r is the polar radius, and N is a constant factor.
3. An unsupervised domain adaptive and entropy optimized device sensory diagnostic method according to claim 1, wherein the step 3 comprises:
and modifying the structure of the ReNet50 network, and replacing the last linear layer with a full-connection layer which has an Xavier initialization weight and is free of bias. Extracting features by using the modified convolutional neural network, wherein the process is as follows:
z(i)＝g(w T x(i)+b)
wherein, x (i) = { x 1 ，x 2 ，...，x n Is n input data, g (, is a linear unit ReLU activation function, w is a weight matrix, b is a bias term, and z (, is an output).
4. An unsupervised domain adaptive and entropy optimized device sensory diagnostic method according to claim 1, wherein the step 4 comprises:
step 4.1: randomly sampling the label-free image data of the target domain to form a training set and a testing set;
step 4.2: the data enhancement strategy in the invention consists of a RandAugment technology and a self-training algorithm;
step 4.3: transforming the training set image by a RandAugment technology to generate a data set and generate a data set P = { r = { (r) } 1 (x τ )，r 2 (x τ )，...r n (x τ ) In which x τ Is from the target domain unlabeled training data, r i Is the transform i-th operation function;
step 4.4: inputting the data set P into a data enhancement strategy, selecting the most reliable pseudo sample from the output result, and deciding the consistency of the label and the truth label;
step 4.5: if the decision result is consistency, minimizing the pseudo sample entropy loss value consistent with the true value label, otherwise, maximizing the inconsistent pseudo sample entropy loss value;
step 4.6: combining conditional entropy loss, conditional distribution distance and classification loss of a target domain label-free training set to form a target function, and training model parameters;
the consistency in step 4.4 is described as:
wherein the prediction + The generated pseudo label is consistent with the true label, prediction _ It is the generated pseudo label that does not coincide with the true label.
The entropy loss description formula in step 4.5 is:
wherein L is EO Is the complete entropy minimization function, L prediction+ Is a consistent entropy loss function, L prediction_ The latter two functions are defined as follows:
wherein,
is the moving average exponent of the sample label y and p is the probability distribution function.
The step 4.6 specifically comprises the following steps:
step 4.6.1: calculating the condition distribution distance of the target domain, wherein the process description formula is as follows:
wherein, S is the source domain,
is the target domain, X * Is the source domain or the target domain, k is the number of Gaussian kernels, MMD Hk The representative kernel space embeds k gaussian kernels, defined as follows:
where φ is a nonlinear mapping function of the regenerative kernel Hilbert space H,
is a sample from either the source domain or the target domain.
Step 4.6.2: combining the classified cross entropy loss and the conditional distribution distance loss with the conditional entropy loss to form an objective function to be optimized, wherein the formula of the objective function is as follows:
L final ＝minL c +αL M +βL EO
where α and β are penalty factors.
5. A system employing the unsupervised domain adaptive and entropy optimized device sensing diagnostic method of any of claims 1-5, the system comprising:
the feature extraction module is used for extracting high-dimensional feature representation from a source domain sample of the preprocessed image;
the characteristic consistency module is used for judging the consistency of the pseudo samples generated by the self-training algorithm and the real samples and selecting the most reliable pseudo samples out of the training model;
and the characteristic identification module is used for identifying the sample which is not labeled in the target domain and does not participate in training so as to judge the corresponding health category of the test sample.
6. The system for the sensoring diagnosis of the unsupervised domain adaptive and entropy optimization device according to claim 5, wherein the network frameworks of the feature extraction module and the feature recognition module are convolutional neural networks, and the basis of the feature consistency module is a RandAugment technology and a self-training algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211424445.3A CN115618267A (en) | 2022-11-15 | 2022-11-15 | Unsupervised domain self-adaption and entropy optimization equipment sensing diagnosis method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211424445.3A CN115618267A (en) | 2022-11-15 | 2022-11-15 | Unsupervised domain self-adaption and entropy optimization equipment sensing diagnosis method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115618267A true CN115618267A (en) | 2023-01-17 |
Family
ID=84879486
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211424445.3A Pending CN115618267A (en) | 2022-11-15 | 2022-11-15 | Unsupervised domain self-adaption and entropy optimization equipment sensing diagnosis method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115618267A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116385813A (en) * | 2023-06-07 | 2023-07-04 | 南京隼眼电子科技有限公司 | ISAR image classification method, ISAR image classification device and storage medium |
| CN117054872A (en) * | 2023-09-15 | 2023-11-14 | 合肥融讯电子科技有限公司 | Motor fault prediction detection system based on data model |
-
2022
- 2022-11-15 CN CN202211424445.3A patent/CN115618267A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116385813A (en) * | 2023-06-07 | 2023-07-04 | 南京隼眼电子科技有限公司 | ISAR image classification method, ISAR image classification device and storage medium |
| CN116385813B (en) * | 2023-06-07 | 2023-08-29 | 南京隼眼电子科技有限公司 | ISAR image space target classification method, device and storage medium based on unsupervised contrast learning |
| CN117054872A (en) * | 2023-09-15 | 2023-11-14 | 合肥融讯电子科技有限公司 | Motor fault prediction detection system based on data model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Han et al. | Deep transfer learning with limited data for machinery fault diagnosis | |
| Wang et al. | A deep learning method for bearing fault diagnosis based on time-frequency image | |
| Wang et al. | Self-supervised signal representation learning for machinery fault diagnosis under limited annotation data | |
| Zhang et al. | A fault diagnosis method based on improved convolutional neural network for bearings under variable working conditions | |
| CN115618267A (en) | Unsupervised domain self-adaption and entropy optimization equipment sensing diagnosis method and system | |
| CN108304941A (en) | A kind of failure prediction method based on machine learning | |
| Li et al. | Joint attention feature transfer network for gearbox fault diagnosis with imbalanced data | |
| Liu et al. | A fault diagnosis method based on transfer convolutional neural networks | |
| Yang et al. | Learn generalization feature via convolutional neural network: A fault diagnosis scheme toward unseen operating conditions | |
| Zhang et al. | Manifold sparse auto-encoder for machine fault diagnosis | |
| Wang et al. | A light weight multisensory fusion model for induction motor fault diagnosis | |
| Zhang et al. | Intelligent machine fault diagnosis using convolutional neural networks and transfer learning | |
| Li et al. | Cross-attribute adaptation networks: Distilling transferable features from multiple sampling-frequency source domains for fault diagnosis of wind turbine gearboxes | |
| Amruthnath et al. | Factor analysis in fault diagnostics using random forest | |
| CN116975718A (en) | Rolling bearing cross-domain fault diagnosis method based on self-supervision learning | |
| Huang et al. | Intelligent fault identification for industrial internet of things via prototype-guided partial domain adaptation with momentum weight | |
| Han et al. | The multilabel fault diagnosis model of bearing based on integrated convolutional neural network and gated recurrent unit | |
| Saufi et al. | Machinery fault diagnosis based on a modified hybrid deep sparse autoencoder using a raw vibration time-series signal | |
| Zheng et al. | An unsupervised transfer learning method based on SOCNN and FBNN and its application on bearing fault diagnosis | |
| Xu et al. | Bearing Fault Diagnosis in the Mixed Domain Based on Crossover‐Mutation Chaotic Particle Swarm | |
| Zhao et al. | A capsnet-based fault diagnosis method for a digital twin of a wind turbine gearbox | |
| Chen et al. | Transfer learning with unsupervised domain adaptation method for bearing fault diagnosis | |
| CN116383757B (en) | Bearing fault diagnosis method based on multi-scale feature fusion and migration learning | |
| CN117390411A (en) | Variable working condition bearing fault diagnosis method based on meta-shift learning | |
| Qin et al. | Rolling bearings fault diagnosis via 1d convolution networks |
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 |