CN114742122A - Equipment fault diagnosis method and device, electronic equipment and storage medium - Google Patents
Equipment fault diagnosis method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN114742122A CN114742122A CN202210178941.9A CN202210178941A CN114742122A CN 114742122 A CN114742122 A CN 114742122A CN 202210178941 A CN202210178941 A CN 202210178941A CN 114742122 A CN114742122 A CN 114742122A
- Authority
- CN
- China
- Prior art keywords
- model
- fault diagnosis
- determining
- data set
- fault
- 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 196
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000003860 storage Methods 0.000 title claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 264
- 238000004590 computer program Methods 0.000 claims description 7
- 208000037408 Device failure Diseases 0.000 claims description 5
- 238000012546 transfer Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 6
- 238000000605 extraction Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000009795 derivation Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 239000000853 adhesive Substances 0.000 description 1
- 230000001070 adhesive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Abstract
The application provides a device fault diagnosis method, a device, an electronic device and a storage medium, wherein the method comprises the following steps: determining a source domain training data set and a target domain training data set according to target fault data; the target fault data is fault data of equipment with the same fault type and different fault modes as the equipment to be diagnosed; determining a first model parameter according to the source domain training data set and the fault diagnosis training model; determining a second model parameter according to the first model parameter, the source domain training data set, the target domain training data set, the preset model parameter and the fault diagnosis training model; determining a fault diagnosis model according to the second model parameters and the fault diagnosis training model; and carrying out fault diagnosis on the equipment to be diagnosed according to the fault diagnosis model. The fault diagnosis model of the equipment to be diagnosed is trained through the fault data of other equipment, the knowledge transfer from the existing fault data to the fault diagnosis model of the equipment to be diagnosed is realized, and the diversity of the trained fault diagnosis model is improved.
Description
Technical Field
The present disclosure relates to the field of device fault diagnosis technologies, and in particular, to a device fault diagnosis method and apparatus, an electronic device, and a storage medium.
Background
With the continuous advance of large-scale, complicated, integrated and intelligent industrial equipment, the degradation process and the failure mode of the industrial equipment become more and more complex, and the demand and the requirement on the failure detection and diagnosis algorithm are increasingly increased. Under the background of industrial big data, the intelligent fault diagnosis driven by data becomes automatic and effective due to the powerful model construction and learning capability of deep learning.
However, the operation process of the industrial equipment is complex, and the acquired data has the characteristics of large scale, high dimensionality, multiple modes, nonlinearity and the like, so that the data-driven intelligent fault diagnosis technology can face the following problems in practical application: because the equipment to be diagnosed can not obtain the fault data in advance in the operation process, the fault diagnosis model can not be trained in a targeted manner, so that the fault condition of the equipment to be diagnosed can be determined.
Disclosure of Invention
The application provides a device fault diagnosis method, a device, an electronic device and a storage medium, which are used for overcoming the defect that in the prior art, a device to be diagnosed cannot obtain fault data in advance in the operation process, so that a fault diagnosis model cannot be trained in a targeted manner, and realizing the training of the fault diagnosis model of the device to be diagnosed through the fault data of other devices, so that the existing fault data is migrated to the knowledge of the fault diagnosis model of the device to be diagnosed, and the diversity of the trained fault diagnosis model is improved.
The application provides an equipment fault diagnosis method, which comprises the following steps:
determining a source domain training data set and a target domain training data set according to target fault data; the target fault data is fault data of equipment with the same fault type and different fault modes as the equipment to be diagnosed;
determining a first model parameter according to the source domain training data set and a fault diagnosis training model;
determining a second model parameter according to the first model parameter, the source domain training data set, the target domain training data set, a preset model parameter and the fault diagnosis training model;
determining a fault diagnosis model according to the second model parameters and the fault diagnosis training model;
and carrying out fault diagnosis on the equipment to be diagnosed according to the fault diagnosis model.
According to the equipment fault diagnosis method provided by the application, determining the first model parameter according to the source domain training data set and the fault diagnosis training model comprises the following steps:
determining a first supervised learning loss value according to the source domain training data set and the fault diagnosis training model;
and determining the first model parameter according to the first supervised learning loss value and the preset model parameter.
Correspondingly, the determining a second model parameter according to the first model parameter, the source domain training data set, the target domain training data set, a preset model parameter, and the fault diagnosis training model includes:
updating the fault diagnosis training model according to the first model parameter to obtain an updated fault diagnosis training model;
determining a second supervised learning loss value according to the target domain training data set and the updated fault diagnosis training model;
determining a distance loss value and a discrimination loss value according to the source domain training data set and the target domain training data set;
and determining the second model parameter according to the preset model parameter, the first supervised learning loss value, the second supervised learning loss value, the discrimination loss value and the distance loss value.
Correspondingly, the determining a distance loss value and a discriminant loss value according to the source domain training data set and the target domain training data set includes:
determining the same category data between the source domain training data set and the target domain training data set;
determining a feature mean of the same category data;
and determining the distance loss value according to the characteristic mean value.
Correspondingly, the determining the feature mean of the same category data includes:
determining data characteristics and sample quantity corresponding to the same type of data;
determining the feature mean according to the data features and the number of samples.
Correspondingly, the determining a distance loss value and a discriminant loss value according to the source domain training data set and the target domain training data set includes:
determining Euclidean distance between the source domain training data set and sample data in the target domain training data set;
and determining the discrimination loss value according to the label of the sample data and the Euclidean distance.
Correspondingly, the determining the discriminant loss value according to the label of the sample data and the euclidean distance includes:
if the labels of the sample data are the same, determining the difference loss value according to the Euclidean distance;
and if the labels of the sample data are different, determining the difference loss value according to a preset value and the Euclidean distance.
The present application further provides an apparatus for diagnosing a device failure, including:
the first determining module is used for determining a source domain training data set and a target domain training data set according to the target fault data; the target fault data is fault data of equipment with the same fault type and different fault modes as the equipment to be diagnosed;
the training module is used for determining a first model parameter according to the source domain training data set and the fault diagnosis training model;
a second determining module, configured to determine a second model parameter according to the first model parameter, the source domain training data set, the target domain training data set, a preset model parameter, and the fault diagnosis training model;
the updating module is used for determining a fault diagnosis model according to the second model parameters and the fault diagnosis training model;
and the diagnosis module is used for carrying out fault diagnosis on the equipment to be diagnosed according to the fault diagnosis model.
The present application also provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the device fault diagnosis method according to any one of the above methods when executing the program.
The present application also provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of diagnosing a fault in a device as described in any of the above.
According to the equipment fault diagnosis method, the device, the electronic equipment and the storage medium, the source domain training data set and the target domain training data set are determined according to the fault data of the equipment with the same fault type and different fault modes as the equipment to be diagnosed, then the second model parameter is determined according to the preset model parameter, the source domain training data set, the target domain training data set and the fault diagnosis training model, and the fault diagnosis model is determined according to the second model parameter and the fault diagnosis training model, so that the fault diagnosis model of the equipment to be diagnosed is trained through the fault data of other equipment, the knowledge transfer of the existing fault data to the fault diagnosis model of the equipment to be diagnosed is realized, the fault diagnosis model can be trained based on the fault data of other equipment when the equipment to be diagnosed cannot acquire the fault data in advance in the operation process, therefore, the diversity of the training fault diagnosis model is improved.
Drawings
In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic flow diagram of a method for diagnosing equipment failure provided herein;
FIG. 2 is a second schematic flow chart of the apparatus fault diagnosis method provided in the present application;
FIG. 3 is a schematic flow chart of a method for determining a fault diagnosis model provided herein;
FIG. 4 is a schematic flow chart of the determination of model parameters provided herein;
FIG. 5 is a schematic flow chart of determining parameters of a second model provided herein;
FIG. 6 is a schematic structural diagram of a device failure diagnosis apparatus provided in the present application;
fig. 7 is a schematic structural diagram of an electronic device provided in the present application.
Detailed Description
To make the purpose, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below with reference to the drawings in the present application, and it is obvious that the described embodiments are some, but not all embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Due to the fact that the operation process of industrial equipment is complex, collected data have the characteristics of large scale, high dimensionality, multiple modes, nonlinearity and the like, the data-driven intelligent fault diagnosis technology has the following problems in practical application:
(1) data acquired by multiple sensors have very high dimensionality, and strong relevance may exist among all the dimensionalities;
(2) the collected data types are seriously unbalanced, the vast majority of the collected data are data in a healthy state, and fault data are sparse;
(3) the target industrial equipment to be diagnosed can not obtain fault data in advance in the operation process and can not train a fault diagnosis model in a targeted manner;
(4) there are differences in the data distribution for the same fault type but different fault modes.
Based on this, the following embodiments are provided, in which a fault diagnosis model parameter optimization process is improved, so that in a process of model learning by using a plurality of different data sources having different fault modes with a device to be diagnosed, a model is prompted to pay attention to fault knowledge common to the different data sources, so that the learned fault knowledge has mobility, and then, a fault diagnosis model of the device to be diagnosed is trained based on the fault knowledge, thereby improving the diversity of the trained fault diagnosis model.
The device failure diagnosis method, apparatus, electronic device, and storage medium of the present application are described below with reference to fig. 1 to 7.
Specifically, the present application provides an apparatus fault diagnosis method, and referring to fig. 1, fig. 1 is one of the flow diagrams of the apparatus fault diagnosis method provided by the present application.
While a logical order is shown in the flow chart, in some cases, the steps shown or described may be performed in a different order than that shown or described.
In the embodiment of the application, an electronic device is taken as an execution subject for example, and the device fault diagnosis method includes:
step S10, determining a source domain training data set and a target domain training data set according to the target fault data; the target fault data is fault data of equipment with the same fault type and different fault modes as the equipment to be diagnosed;
it should be noted that, because the data of the industrial equipment acquired by the multiple sensors has very high dimensionality, and meanwhile, strong correlation may exist between the dimensionalities, on the basis, the fault data of the target industrial equipment can be constructed by using the fault data of other industrial equipment under the condition that the fault data of the equipment to be diagnosed is not acquired.
Specifically, a plurality of different data source data which have the same fault type as the equipment to be diagnosed but have different fault modes are acquired, wherein the data source refers to other industrial equipment, and the data source data refers to equipment data collected from other industrial equipment, such as fault data. The same failure type includes multiple failure modes, for example, assuming that the industrial equipment is a drum, the failure types include bearing outer ring damage, bearing inner ring damage, rolling ball damage, and the like, and the failure modes corresponding to the bearing outer ring damage include adhesive wear, foreign matter peeling, fatigue peeling, single-point (multi-point) pitting, and the like.
Further, usable data sets are constructed based on the data source dataRandomly selecting two data source data from a plurality of different data source data of the available data set, and marking the data source data as a source domain training data set DtrsAnd a target domain training data set DtrtOptionally also labeled as training field data DtrsAnd test domain data Dtrt。
It is understood that different data source data with different failure modes represent that the data in the available data set is acquired differently from the data distribution of the device to be diagnosed, and the different data source data acquisition conditions are different from the device to be diagnosed, such as working conditions, workload, environmental noise, device structure, and the like. Based on the method, the common knowledge in the available data sets is extracted by a knowledge generalization method through the constructed available data sets of the plurality of different data sources, so that the generalization capability of the whole fault diagnosis model is improved.
Step S20, determining a first model parameter according to the source domain training data set and the fault diagnosis training model;
it should be noted that the fault diagnosis training model includes a feature extraction module FψAnd fault discrimination module TθThe model parameter Θ is a preset model parameter, i.e., an initial model parameter (θ, ψ).
Specifically, the first model parameters are determined according to the source domain training data set and the fault diagnosis training model, as described in step S21 to step S22.
Further, the detailed description of steps S21 to S22 is as follows:
step S21, determining a first supervised learning loss value according to the source domain training data set and the fault diagnosis training model;
step S22, determining the first model parameter according to the first supervised learning loss value and the preset model parameter.
It should be noted that the task of supervised learning (supersupervised learning) is to learn a model through training data, so that the model can make a prediction (predicted value is close to true value) on any unknown input and its corresponding output. And the supervised learning loss value is a numerical value corresponding to a loss function in the supervised learning and is used for evaluating the inconsistency between the real value of the sample and the model predicted value.
In particular, a source domain training data set D is employedtrsTraining a fault diagnosis training model, and calculating a first supervised learning loss value L in the model training processtrs(Θ), wherein the first supervised learning loss value is calculated using cross entropy loss, the formula is as follows:
where N is the number of data set samples, xiFor the ith sample, yiIs the label of the ith sample, FψFor the feature extraction module, TθFor the fault discrimination module, log () is a logarithmic function, and the function expression of softmax () is:
wherein h isiAnd (3) an ith dimension scalar quantity output by the C dimension of the model, wherein C is the number of sample classes.
Further, the first supervised learning loss value L is determinedtrs(Θ) after learning the loss value L by first supervisiontrs(Θ) and a preset model parameter Θ to calculate a first model parameter Θ 'of the fault diagnosis training model, wherein a calculation formula of the first model parameter Θ' is as follows:
wherein, theta is a preset model parameter, alpha is a step value of updating the model parameter,a gradient value of the loss value is learned for the first supervision.
In the embodiment, the first supervised learning loss value is determined according to the source domain training data set and the fault diagnosis training model, and the first model parameter is determined according to the first supervised learning loss value and the preset model parameter, so that the inconsistency degree between the real value of the sample and the model predicted value can be evaluated, and the accuracy of training the fault diagnosis model is improved.
Step S30, determining second model parameters according to the first model parameters, the source domain training data set, the target domain training data set, preset model parameters and the fault diagnosis training model;
specifically, a distance loss value and a discrimination loss value are determined according to a source domain training data set and a target domain training data set, and then a second model parameter is determined according to a preset model parameter, a first supervised learning loss value, a second supervised learning loss value, a discrimination loss value and a distance loss value. Specifically, the steps S31 to S34 are described.
Step S40, determining a fault diagnosis model according to the second model parameters and the fault diagnosis training model;
specifically, the fault diagnosis training model is updated by using the first model parameter Θ' to obtain an updated fault diagnosis training model, and then the updated fault diagnosis training model is updated again by using the second model parameter Θ ″ to obtain the fault diagnosis model.
And step S50, performing fault diagnosis on the equipment to be diagnosed according to the fault diagnosis model.
Specifically, after the fault diagnosis model is obtained through training, the fault diagnosis model is applied to a real scene of the industrial equipment, and real-time fault diagnosis of the target industrial equipment is achieved. For example, before the fault data of the device to be diagnosed is not acquired, the device to be diagnosed is diagnosed by using the fault diagnosis model obtained by training the fault data of other devices, so that whether the device to be diagnosed has a fault or not can be determined, if the device to be diagnosed has a fault, a corresponding technician can be notified to perform maintenance, or automatic maintenance is performed based on a maintenance scheme corresponding to the fault problem, and thus the operation safety of the device to be diagnosed is improved.
For example, referring to fig. 3, fig. 3 is a schematic flow chart of determining a fault diagnosis model provided in the present application.
Specifically, the generation of the equipment diagnosis fault model is divided into two stages, namely a training stage and a testing stage. In the training phase, a usable data set is constructed by acquiring a plurality of different data source data which have the same fault type but different fault modes with the equipment to be diagnosed, and then constructing the usable data set based on the data source data. Further, the fault diagnosis training model is trained using data in the available data set, while a first supervised learning loss value is calculated during the training, and then the fault diagnosis training model is updated based on the first supervised learning loss value. Further, the updated fault diagnosis training model is trained by adopting target equipment operation data (namely a target domain training data set), meanwhile, a second supervised learning loss value is calculated in the training process, finally, a second model parameter is determined according to the preset model parameter, the first supervised learning loss value, the second supervised learning loss value and other loss values, and then, the updated fault diagnosis training model is updated again based on the second model parameter to obtain the fault diagnosis model. And finally, applying the fault diagnosis model to a real scene of the industrial equipment to realize real-time judgment of the running state of the target industrial equipment, thereby improving the running safety of the equipment to be diagnosed.
The present embodiment determines the source domain training data set and the target domain training data set by determining from fault data of devices having the same fault type and different fault modes as the device to be diagnosed, then, second model parameters are determined according to the preset model parameters, the source domain training data set, the target domain training data set and the fault diagnosis training model, then the fault diagnosis model is determined according to the second model parameters and the fault diagnosis training model, based on the second model parameters and the fault diagnosis training model, the fault diagnosis model of the equipment to be diagnosed is trained through the fault data of other equipment, the knowledge transfer of the existing fault data to the fault diagnosis model of the equipment to be diagnosed is realized, and when the equipment to be diagnosed cannot acquire the fault data in advance in the operation process, the fault diagnosis model can be trained based on fault data of other equipment, so that the diversity of the trained fault diagnosis model is improved.
Further, referring to fig. 2, fig. 2 is a second schematic flowchart of the device fault diagnosis method provided in the present application, and the step S30 includes:
step S31, updating the fault diagnosis training model according to the first model parameter to obtain an updated fault diagnosis training model;
step S32, determining a second supervised learning loss value according to the target domain training data set and the updated fault diagnosis training model;
step S33, determining a distance loss value and a discrimination loss value according to the source domain training data set and the target domain training data set;
step S34, determining the second model parameter according to the preset model parameter, the first supervised learning loss value, the second supervised learning loss value, the discrimination loss value, and the distance loss value.
Specifically, the fault diagnosis training model is updated by using the first model parameter Θ' to obtain an updated fault diagnosis training model, and then, the target domain training data set D is usedtrtTraining the model with updated parameters, and calculating a second supervised learning loss value L in the model training processtrt(Θ'), wherein the second supervised learning loss value is calculated using cross entropy loss, the formula is as follows:
wherein N is the number of data set samples, xiFor the ith sample, yiIs the label of the ith sample, Tψ'Feature extraction module, T, for updated fault diagnosis training modelsθ'For the fault discrimination module of the updated fault diagnosis training model, log () is a logarithmic function, and the function expression of softmax () is as follows:
wherein h isiAnd (3) an ith dimension scalar quantity output by the C dimension of the model, wherein C is the number of sample classes.
Further, a distance loss value and a discrimination loss value are determined according to the source domain training data set and the target domain training data set, as described in steps S330 to S332, and steps S333 to S334.
Further, the specific description of steps S330 to S332 is as follows:
step S330, determining the same category data between the source domain training data set and the target domain training data set;
step S331, determining a characteristic mean value of the same category data;
and S332, determining the distance loss value according to the characteristic mean value.
Specifically, the feature extraction module F of the fault diagnosis training model after being updatedψ′Extracting the data of the same type from the source domain training data set and the target domain training data set, then determining the data characteristics and the sample number corresponding to the data of the same type, and determining the characteristic mean value according to the data characteristics and the sample number. For example, for class c, the feature mean calculation formula is as follows:andwherein the content of the first and second substances,andrespectively representing the characteristic mean values of the class c data in the source domain training data set and the target domain training data set; n iscNumber of samples of class c;andrespectively representing the ith sample in the source domain training data set and the target domain training data set;andfeature extraction module F for respectively representing updated fault diagnosis training modelψ′And extracting the characteristics of the sample data in the source domain training data set and the target domain training data set.
Further, determining a distance loss value L according to the characteristic mean valuegfr(Θ'), wherein the distance loss value is calculated using JS divergence, the formula is as follows:
wherein D isKL() In order to obtain a KL divergence, the dispersion,where p and q are two different data distributions, C is the number of sample classes,is the feature mean of the class c data in the source domain training dataset,the feature mean of the class c data in the target domain training dataset is used.
In the embodiment, the data characteristics of the data of the same category between the source domain training data set and the target domain training data set can be aligned by extracting the data of the same category from the source domain training data set and the target domain training data set, then determining the characteristic mean value of the data of the same category, and then determining the distance loss value according to the characteristic mean value, so that the accuracy of training the fault diagnosis model is improved.
Further, the detailed description of steps S333 to S334 is as follows:
step S333, determining Euclidean distance between the source domain training data set and the sample data in the target domain training data set;
step S334, determining the discriminant loss value according to the label of the sample data and the euclidean distance.
Mixing the source domain training data set and the target domain training data set, and then calculating a discriminant loss value Llfr(Θ') for constraining features of the same class to be close and features of different classes to be far apart. Specifically, the Euclidean distance between sample data in the source domain training data set and the target domain training data set is determined, and then the discriminant loss value is determined according to the label of the sample data and the Euclidean distance. For example, if the labels of the sample data are the same, determining the respective loss value according to the Euclidean distance; and if the labels of the sample data are different, determining the difference loss value according to the preset value and the Euclidean distance. Wherein, the discrimination loss value is calculated by using the contrast loss, and the formula is as follows:
wherein d(s)n,sm) Representing the Euclidean distance between two sample data features, where yn,ymAnd xi is a label of the sample data n and the sample data m, and xi is a set threshold value.
Further, determining a second model according to the preset model parameters, the first supervised learning loss value, the second supervised learning loss value, the discrimination loss value and the distance loss valueAnd (4) a type parameter. Specifically, the updating step length gamma and the loss weight beta of the fault diagnosis training model are determined1B.b. over parameter2And beta3Then, according to the update step gamma and the loss weight beta1B.b. over parameter2And beta3The method comprises the following steps of presetting a model parameter, a first supervised learning loss value, a second supervised learning loss value, a discrimination loss value and a distance loss value, and determining the second model parameter, wherein the calculation formula of the second model parameter is as follows:
wherein L isfr(Θ′)=β2Lgfr(Θ′)+β3Llfr(Θ′),β2And beta3In order to set the hyper-parameters,the derivation is indicated.
Further, referring to fig. 4, fig. 4 is a schematic flow chart of determining model parameters provided herein;
in this embodiment, the fault diagnosis training model initializes a model parameter, i.e., a preset model parameter Θ, calculates a first supervised learning loss value according to the preset model parameter, determines a first model parameter Θ' according to the first supervised learning loss value, and further determines a second model parameter Θ "according to the preset model parameter, the first supervised learning loss value, the second supervised learning loss value, and other loss values. After the second model parameter Θ ″ is determined, it is necessary to determine whether the model training is terminated, if the model training is not terminated, assign the second model parameter to the preset model parameter, if Θ is equal to Θ ", calculate the first model parameter Θ' and the second model parameter Θ" again according to the newly assigned preset model parameter Θ until the model training is terminated, and at this time, update the fault diagnosis training model by using the second model parameter Θ ″ obtained by the last calculation to obtain the fault diagnosis model. Whether the model training is terminated or not can be judged through the training times or the loss value, for example, if the preset model training times are set to be 5 times, the model training is terminated after the model training is performed for 5 times; alternatively, when the loss value of the model reaches a set threshold, the training is terminated.
In the embodiment, the Euclidean distance between the source domain training data set and the sample data in the target domain training data set is determined, and then the discriminant loss value is determined according to the label of the sample data and the Euclidean distance, so that the approach of the data features of the same category can be restrained, and the distance of the data features of different categories can be restrained, and the accuracy of the training fault diagnosis model can be improved.
For example, referring to fig. 5, fig. 5 is a schematic flow chart for determining the second model parameters provided herein.
In particular, usable data sets are constructed based on the data source dataThe available data set comprises target fault data, two data source data are randomly selected from a plurality of different data source data of the available data set, and the data source data are marked as a source domain training data set DtrsAnd a target domain training data set Dtrt。
Further, the data set D is trained using the source domaintrsTraining a fault diagnosis training model, and calculating a first supervised learning loss value L in the model training processtrs(Θ), then, the loss value L is learned by the first supervisiontrs(theta) and a preset model parameter theta are used for calculating a first model parameter theta 'of the fault diagnosis training model, the first model parameter theta' is used for updating the fault diagnosis training model, and then a target domain training data set D is usedtrtTraining the fault diagnosis training model after the parameters are updated, and calculating a second supervised learning loss value L in the model training processtrt(Θ′)。
Further, a feature extraction module F of the training model for fault diagnosis after updatingψ′Extracting the same type of data from the source domain training data set and the target domain training data set, then determining the data characteristics and the sample number corresponding to the same type of data, and then according to the data characteristics and the sample numberDetermining a characteristic mean value, and determining a distance loss value L according to the characteristic mean valuegfr(Θ′)。
Further, the source domain training data set and the target domain training data set are mixed, and then, a discriminant loss value L is calculatedlfr(Θ') for constraining features of the same class to be close and features of different classes to be far apart. Specifically, the Euclidean distance between sample data in the source domain training data set and the target domain training data set is determined, and then the discriminant loss value is determined according to the label of the sample data and the Euclidean distance.
Further, a preset model parameter theta and a first supervised learning loss value L are utilizedtrs(Θ) and a second supervised learning loss value Ltrt(theta'), distance loss value Lgfr(theta') and the discrimination loss value Llfr(Θ') determining second model parameters. Wherein the formula of the second model parameter is as follows:
wherein gamma is the step length of model updating,denotes derivation, β is the loss weight, and Lfr(Θ′)=β2Lgfr(Θ′)+β3Llfr(Θ′),β2And beta3Is a set hyper-parameter.
And then, updating the second model parameter theta to obtain a fault diagnosis model, and applying the fault diagnosis model to a real scene of the industrial equipment, so that real-time fault diagnosis of the target industrial equipment is realized, and the running safety of the equipment to be diagnosed is improved.
In the embodiment, the second model parameter is determined according to the first model parameter and the fault diagnosis training model, then the distance loss value and the discrimination loss value are determined according to the source domain training data set and the target domain training data set, and the second model parameter is determined according to the preset model parameter and each loss value, so that the updating parameter of the fault diagnosis training model can be determined, and then the fault diagnosis training model is updated, so that the accuracy of the fault diagnosis model is improved, meanwhile, the accuracy of the fault diagnosis model for diagnosing equipment to be diagnosed is improved, and further, the operation safety of the equipment to be diagnosed is improved.
Fig. 6 is a schematic structural diagram of the device fault diagnosis apparatus provided in the present application, and referring to fig. 6, an embodiment of the present application provides a device fault diagnosis apparatus, which includes a first determining module 601, a training module 602, a second determining module 603, an updating module 604, and a diagnosis module 605, wherein,
the first determining module 601 is configured to determine a source domain training data set and a target domain training data set according to target fault data; the target fault data is fault data of equipment with the same fault type and different fault modes as the equipment to be diagnosed;
the training module 602 is configured to determine a first model parameter according to the source domain training data set and the fault diagnosis training model;
the second determining module 603 is configured to determine a second model parameter according to the first model parameter, the source domain training data set, the target domain training data set, a preset model parameter, and the fault diagnosis training model;
the updating module 604 is configured to determine a fault diagnosis model according to the second model parameter and the fault diagnosis training model;
the diagnosis module 605 is configured to perform fault diagnosis on the device to be diagnosed according to the fault diagnosis model.
The device fault diagnosis apparatus provided in this embodiment determines the source domain training data set and the target domain training data set according to the fault data of the device having the same fault type and different fault mode as the device to be diagnosed, then, second model parameters are determined according to the preset model parameters, the source domain training data set, the target domain training data set and the fault diagnosis training model, then the fault diagnosis model is determined according to the second model parameters and the fault diagnosis training model, based on the second model parameters and the fault diagnosis training model, the fault diagnosis model of the equipment to be diagnosed is trained through the fault data of other equipment, the knowledge transfer of the existing fault data to the fault diagnosis model of the equipment to be diagnosed is realized, and when the equipment to be diagnosed cannot acquire the fault data in advance in the operation process, the fault diagnosis model can be trained based on fault data of other equipment, so that the diversity of the trained fault diagnosis model is improved.
In one embodiment, the training module 602 is specifically configured to:
determining a first supervised learning loss value according to the source domain training data set and the fault diagnosis training model;
and determining the first model parameter according to the first supervised learning loss value and the preset model parameter.
In an embodiment, the second determining module 603 is specifically configured to:
updating the fault diagnosis training model according to the first model parameter to obtain an updated fault diagnosis training model;
determining a second supervised learning loss value according to the target domain training data set and the updated fault diagnosis training model;
determining a distance loss value and a discrimination loss value according to the source domain training data set and the target domain training data set;
and determining the second model parameter according to the preset model parameter, the first supervised learning loss value, the second supervised learning loss value, the discrimination loss value and the distance loss value.
In an embodiment, the second determining module 603 is specifically configured to:
determining the same category data between the source domain training data set and the target domain training data set;
determining a feature mean of the same category data;
and determining the distance loss value according to the characteristic mean value.
In an embodiment, the second determining module 603 is specifically configured to:
determining data characteristics and sample quantity corresponding to the same type of data;
determining the feature mean according to the data features and the number of samples.
In an embodiment, the second determining module 603 is specifically configured to:
determining Euclidean distance between the source domain training data set and sample data in the target domain training data set;
and determining the discrimination loss value according to the label of the sample data and the Euclidean distance.
In an embodiment, the second determining module 603 is specifically configured to:
if the labels of the sample data are the same, determining the difference loss value according to the Euclidean distance;
and if the labels of the sample data are different, determining the difference loss value according to a preset value and the Euclidean distance.
Fig. 7 illustrates a physical structure diagram of an electronic device, and as shown in fig. 7, the electronic device may include: a processor (processor)710, a communication Interface (Communications Interface)720, a memory (memory)730, and a communication bus 740, wherein the processor 710, the communication Interface 720, and the memory 730 communicate with each other via the communication bus 740. Processor 710 may call logic instructions in memory 730 to perform a device fault diagnosis method comprising:
determining a source domain training data set and a target domain training data set according to target fault data; the target fault data is fault data of equipment with the same fault type and different fault modes as the equipment to be diagnosed;
determining a first model parameter according to the source domain training data set and a fault diagnosis training model;
determining second model parameters according to the first model parameters, the source domain training data set, the target domain training data set, preset model parameters and the fault diagnosis training model;
determining a fault diagnosis model according to the second model parameters and the fault diagnosis training model;
and carrying out fault diagnosis on the equipment to be diagnosed according to the fault diagnosis model.
In addition, the logic instructions in the memory 730 can be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In another aspect, the present application also provides a non-transitory computer-readable storage medium, on which a computer program is stored, the computer program being implemented by a processor to execute the method for diagnosing equipment failure provided by the above methods, the method including:
determining a source domain training data set and a target domain training data set according to target fault data; the target fault data is fault data of equipment with the same fault type and different fault modes as the equipment to be diagnosed;
determining a first model parameter according to the source domain training data set and a fault diagnosis training model;
determining a second model parameter according to the first model parameter, the source domain training data set, the target domain training data set, a preset model parameter and the fault diagnosis training model;
determining a fault diagnosis model according to the second model parameters and the fault diagnosis training model;
and carrying out fault diagnosis on the equipment to be diagnosed according to the fault diagnosis model.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. An apparatus fault diagnosis method, comprising:
determining a source domain training data set and a target domain training data set according to target fault data; the target fault data is fault data of equipment with the same fault type and different fault modes as the equipment to be diagnosed;
determining a first model parameter according to the source domain training data set and a fault diagnosis training model;
determining second model parameters according to the first model parameters, the source domain training data set, the target domain training data set, preset model parameters and the fault diagnosis training model;
determining a fault diagnosis model according to the second model parameters and the fault diagnosis training model;
and carrying out fault diagnosis on the equipment to be diagnosed according to the fault diagnosis model.
2. The device fault diagnosis method according to claim 1, wherein the determining of the first model parameters from the source domain training data set and the fault diagnosis training model comprises:
determining a first supervised learning loss value according to the source domain training data set and the fault diagnosis training model;
and determining the first model parameter according to the first supervised learning loss value and the preset model parameter.
3. The device fault diagnosis method according to claim 2, wherein the determining second model parameters according to the first model parameters, the source domain training data set, the target domain training data set, preset model parameters and the fault diagnosis training model comprises:
updating the fault diagnosis training model according to the first model parameter to obtain an updated fault diagnosis training model;
determining a second supervised learning loss value according to the target domain training data set and the updated fault diagnosis training model;
determining a distance loss value and a discrimination loss value according to the source domain training data set and the target domain training data set;
and determining the second model parameter according to the preset model parameter, the first supervised learning loss value, the second supervised learning loss value, the discrimination loss value and the distance loss value.
4. The device fault diagnosis method according to claim 3, wherein the determining distance loss values and discrimination loss values from the source domain training data set and the target domain training data set comprises:
determining the same category data between the source domain training data set and the target domain training data set;
determining a feature mean of the same category data;
and determining the distance loss value according to the characteristic mean value.
5. The device fault diagnosis method according to claim 4, wherein the determining the feature mean value of the same category data comprises:
determining data characteristics and sample quantity corresponding to the same type of data;
and determining the feature mean value according to the data features and the sample number.
6. The device fault diagnosis method according to claim 3, wherein the determining distance loss values and discrimination loss values from the source domain training data set and the target domain training data set comprises:
determining Euclidean distance between the source domain training data set and sample data in the target domain training data set;
and determining the discrimination loss value according to the label of the sample data and the Euclidean distance.
7. The apparatus fault diagnosis method according to claim 6, wherein the determining the discrimination loss value according to the label of the sample data and the euclidean distance includes:
if the labels of the sample data are the same, determining the difference loss value according to the Euclidean distance;
and if the labels of the sample data are different, determining the difference loss value according to a preset value and the Euclidean distance.
8. An apparatus for diagnosing a failure of a device, comprising:
the first determining module is used for determining a source domain training data set and a target domain training data set according to the target fault data; the target fault data is fault data of equipment with the same fault type and different fault modes as the equipment to be diagnosed;
the training module is used for determining a first model parameter according to the source domain training data set and the fault diagnosis training model;
a second determining module, configured to determine a second model parameter according to the first model parameter, the source domain training data set, the target domain training data set, a preset model parameter, and the fault diagnosis training model;
the updating module is used for determining a fault diagnosis model according to the second model parameters and the fault diagnosis training model;
and the diagnosis module is used for carrying out fault diagnosis on the equipment to be diagnosed according to the fault diagnosis model.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the device failure diagnosis method according to any one of claims 1 to 7 when executing the program.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the device failure diagnosis method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210178941.9A CN114742122A (en) | 2022-02-25 | 2022-02-25 | Equipment fault diagnosis method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210178941.9A CN114742122A (en) | 2022-02-25 | 2022-02-25 | Equipment fault diagnosis method and device, electronic equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114742122A true CN114742122A (en) | 2022-07-12 |
Family
ID=82275170
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210178941.9A Pending CN114742122A (en) | 2022-02-25 | 2022-02-25 | Equipment fault diagnosis method and device, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114742122A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116910573A (en) * | 2023-09-13 | 2023-10-20 | 中移(苏州)软件技术有限公司 | Training method and device for abnormality diagnosis model, electronic equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111353373A (en) * | 2019-11-29 | 2020-06-30 | 盐城工学院 | Correlation alignment domain adaptive fault diagnosis method |
CN111898095A (en) * | 2020-07-10 | 2020-11-06 | 佛山科学技术学院 | Deep migration learning intelligent fault diagnosis method and device, storage medium and equipment |
CN112183581A (en) * | 2020-09-07 | 2021-01-05 | 华南理工大学 | Semi-supervised mechanical fault diagnosis method based on self-adaptive migration neural network |
CN112200114A (en) * | 2020-10-20 | 2021-01-08 | 上海明略人工智能(集团)有限公司 | Fault diagnosis model training method and device, electronic equipment and storage medium |
WO2021243838A1 (en) * | 2020-06-03 | 2021-12-09 | 苏州大学 | Fault diagnosis method for intra-class self-adaptive bearing under variable working conditions |
-
2022
- 2022-02-25 CN CN202210178941.9A patent/CN114742122A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111353373A (en) * | 2019-11-29 | 2020-06-30 | 盐城工学院 | Correlation alignment domain adaptive fault diagnosis method |
WO2021243838A1 (en) * | 2020-06-03 | 2021-12-09 | 苏州大学 | Fault diagnosis method for intra-class self-adaptive bearing under variable working conditions |
CN111898095A (en) * | 2020-07-10 | 2020-11-06 | 佛山科学技术学院 | Deep migration learning intelligent fault diagnosis method and device, storage medium and equipment |
CN112183581A (en) * | 2020-09-07 | 2021-01-05 | 华南理工大学 | Semi-supervised mechanical fault diagnosis method based on self-adaptive migration neural network |
CN112200114A (en) * | 2020-10-20 | 2021-01-08 | 上海明略人工智能(集团)有限公司 | Fault diagnosis model training method and device, electronic equipment and storage medium |
Non-Patent Citations (2)
Title |
---|
HUANJIE WANG 等: "Generalization on Unseen Domains via Model-Agnostic Learning for Intelligent Fault Diagnosis", 《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》 * |
张根保等: "一种用于轴承故障诊断的迁移学习模型", 《吉林大学学报(工学版)》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116910573A (en) * | 2023-09-13 | 2023-10-20 | 中移(苏州)软件技术有限公司 | Training method and device for abnormality diagnosis model, electronic equipment and storage medium |
CN116910573B (en) * | 2023-09-13 | 2023-12-05 | 中移(苏州)软件技术有限公司 | Training method and device for abnormality diagnosis model, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP7138753B2 (en) | Image defect detection method, device, electronic device, storage medium and product | |
US11650968B2 (en) | Systems and methods for predictive early stopping in neural network training | |
CN112785091A (en) | Method for performing fault prediction and health management on oil field electric submersible pump | |
CN113111820B (en) | Rotary part fault diagnosis method and device based on improved CNN and relation module | |
CN109753408B (en) | Flow abnormity prediction method based on machine learning | |
CN109815855B (en) | Electronic equipment automatic test method and system based on machine learning | |
CN109145373B (en) | Residual life prediction method and device based on improved ESGP and prediction interval | |
CN108961460B (en) | Fault prediction method and device based on sparse ESGP (Enterprise service gateway) and multi-objective optimization | |
CN116089870A (en) | Industrial equipment fault prediction method and device based on meta-learning under small sample condition | |
CN111611294B (en) | Star sensor data anomaly detection method | |
CN114742122A (en) | Equipment fault diagnosis method and device, electronic equipment and storage medium | |
CN112926642A (en) | Multi-source domain self-adaptive intelligent mechanical fault diagnosis method and system | |
CN116306806A (en) | Fault diagnosis model determining method and device and nonvolatile storage medium | |
CN108549899B (en) | Image identification method and device | |
CN111079348B (en) | Method and device for detecting slowly-varying signal | |
CN111400964B (en) | Fault occurrence time prediction method and device | |
CN111415326A (en) | Method and system for detecting abnormal state of railway contact net bolt | |
CN110413482B (en) | Detection method and device | |
CN114720129B (en) | Rolling bearing residual life prediction method and system based on bidirectional GRU | |
CN116227586A (en) | Meta learning fault diagnosis method and device based on depth residual error shrinkage prototype network | |
CN113033845B (en) | Construction method and device for power transmission resource co-construction and sharing | |
JP2020086786A (en) | Detection device and machine learning method | |
CN115292820A (en) | Method for predicting residual service life of urban rail train bearing | |
CN113919540A (en) | Method for monitoring running state of production process and related equipment | |
CN111210409B (en) | Condition-based generation confrontation network structure damage identification method |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220712 |
|
RJ01 | Rejection of invention patent application after publication |