CN114757286A - Multi-class fault data generation method based on conditional countermeasure generation network - Google Patents
Multi-class fault data generation method based on conditional countermeasure generation network Download PDFInfo
- Publication number
- CN114757286A CN114757286A CN202210409795.6A CN202210409795A CN114757286A CN 114757286 A CN114757286 A CN 114757286A CN 202210409795 A CN202210409795 A CN 202210409795A CN 114757286 A CN114757286 A CN 114757286A
- Authority
- CN
- China
- Prior art keywords
- fault
- fault data
- data
- generating
- generator
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000012549 training Methods 0.000 claims abstract description 62
- 239000008358 core component Substances 0.000 claims abstract description 43
- 230000006870 function Effects 0.000 claims description 33
- 238000009826 distribution Methods 0.000 claims description 23
- 238000005457 optimization Methods 0.000 claims description 18
- 238000005070 sampling Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 8
- 238000005315 distribution function Methods 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 230000003042 antagnostic effect Effects 0.000 claims 1
- 238000003745 diagnosis Methods 0.000 abstract description 9
- 230000008485 antagonism Effects 0.000 abstract 3
- 238000004590 computer program Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000013526 transfer learning Methods 0.000 description 5
- 238000004519 manufacturing process Methods 0.000 description 4
- 239000000306 component Substances 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 229910052744 lithium Inorganic materials 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002243 precursor Substances 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 239000013598 vector Substances 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
- G06F18/21—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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Numerical Control (AREA)
Abstract
The invention discloses a multi-class fault data generation method based on a conditional countermeasure generation network, which comprises the following steps: acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool, extracting characteristics, and creating a fault data set; classifying a fault data set, defining a fault label for each fault category, and defining the fault label as condition information; constructing an antagonism generation network, and updating and training generators and discriminators in the antagonism generation network based on random noise data and condition information to obtain a condition antagonism generation network; and (3) generating a network based on the condition confrontation, inputting random noise data and condition information, and generating fault data of the multi-axis industrial numerical control machine tool core components of different categories. According to the invention, the fault data of the multi-axis industrial numerical control machine tool core components of different categories can be generated through the training condition confrontation generation network, the problem of the sparsity of the fault data of the numerical control machine tool is effectively solved, and the training of a downstream fault diagnosis model is supported.
Description
Technical Field
The invention relates to the technical field of mechanical control engineering, in particular to a multi-class fault data generation method based on a conditional countermeasure generation network.
Background
The numerical control machine tool is used as an industrial master machine and is core basic equipment in the manufacturing industry. High-end multiaxis digit control machine tool is used for precision manufacturing and instrument processing mostly, and its operational failure not only can cause the product quality problem, increases the cost of enterprise, still can lead to the technical bottleneck. With the development of multi-axis, high-speed, high-precision and multifunctional numerical control machine tools, the control difficulty of three-axis, four-axis and five-axis machine tools is gradually increased, the complexity and uncertainty of faults are increased, and the diagnosis is difficult to be carried out by using the traditional manual or simple sensing method. Therefore, artificial intelligence models based on a neural network, a sequence model, a convolutional network and the like are used for developing the research and development of the fault diagnosis system of the industrial multi-axis numerical control machine tool, the cost of enterprise-related fault detection and disposal manpower and material resources is reduced, and the method has important significance for the economic and technical development of the manufacturing industry. However, most of the operation data of the industrial multi-axis numerical control machine tool belongs to normal working condition data, and the lack of fault precursor data leads to insufficient effective data which can be used for training an intelligent fault diagnosis model, the model is difficult to learn the association rule between fault variables and results, the accuracy, the generalization and the robustness performance of the model are seriously influenced, and the model cannot be used in an actual industrial scene.
The existing fault data generation technology mainly comprises a transfer learning and naive confrontation generation network. The drawback of transfer learning is that it is generally only suitable for small data set processing, and its strong heuristic assumption that other domain labels are still suitable for the target domain may also cause severe labeling noise. The disadvantage of the naive countermeasure generation network is that data simulation is performed only based on original data, knowledge supervision in the external field is lacked, the quality of generated fault data is generally low, and the corresponding distribution is also greatly different from actual fault data.
Thus, there is a need for improvements and enhancements in the art.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a method for generating multi-class fault data based on a conditional countermeasure generation network, aiming at solving the problems in the prior art that the quality of the fault data generated by the conventional fault data generation method is generally low, the fault data is sparse and lacking, and the corresponding distribution is also different from the actual fault data.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
in a first aspect, the present invention provides a multi-class fault data generation method for generating a network based on conditional countermeasure, wherein the method includes:
Acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool, extracting characteristics of the real fault data, and creating a fault data set;
classifying the fault data set to obtain a plurality of fault categories, defining a fault label for each fault category, and defining the fault label as condition information;
constructing a countermeasure generation network, and performing update training on a generator and a discriminator in the countermeasure generation network based on random noise data and the condition information to obtain a condition countermeasure generation network;
and generating a network based on condition confrontation, inputting the random noise data and the condition information, generating fault data of core components of the multi-axis industrial numerical control machine tool of different categories, and supporting downstream fault recognition model training.
In one implementation, the acquiring real fault data of a core component of a multi-axis industrial numerical control machine tool, performing feature extraction on the real fault data, and creating a fault data set includes:
acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool;
and sampling time-frequency domains and extracting features of the real fault data to obtain fault features, and constructing the fault data set according to the fault features.
In an implementation manner, the classifying the fault data set to obtain a plurality of fault categories, defining a fault label for each fault category, and defining the fault label as condition information includes:
classifying the fault data set by adopting a clustering algorithm based on the fault characteristics to obtain fault categories corresponding to the fault characteristics;
defining a fault label corresponding to each fault category;
and taking the fault label as the condition information, wherein the condition information is a logic condition for describing the distribution of the real fault data.
In one implementation, the performing update training on generators and discriminators in the countermeasure generation network based on random noise data and the condition information to obtain a conditional countermeasure generation network includes:
defining a loss function of a generator and a discriminator of the countermeasure generation network, wherein the loss function of the generator is used for measuring and generating errors between fault data of the candidate industrial machine tool core component and real fault data distribution, and the loss function of the discriminator is used for judging the authenticity of the fault data;
Defining a parameter iterative optimization algorithm of the generator and the discriminator;
inputting random noise data and the condition information into the generator, and outputting artificial fault data;
inputting the artificial fault data and the real fault data into the discriminator to obtain an authenticity result of the artificial fault data;
updating and training the parameters of the generator and the discriminator based on the authenticity result, the parameter iterative optimization algorithm and the loss functions of the generator and the discriminator;
and repeating the updating training of the generator and the arbiter so that the training of the generator is carried out until the threshold is minimum, and the training of the arbiter is carried out until the accuracy of the binary judgment of 0-1 exceeds a preset threshold, thereby obtaining the conditional countermeasure generating network.
In one implementation, the update training of the parameters of the generator and the arbiter based on the authenticity result, the parameter iterative optimization algorithm, and the loss function of the generator and the arbiter comprises:
updating the parameters of the generator by superimposing a random gradient based on the authenticity result, the parameter iterative optimization algorithm and the loss functions of the generator and the discriminator;
And updating the parameters of the discriminator by decreasing the random gradient.
In one implementation, the repeating update-trains the generator and the arbiter, including:
firstly, updating the parameters of the generator k times, and then updating the parameters of the discriminator 1 time.
In one implementation, the objective loss function of the conditional adversary generation network is:
In a second aspect, an embodiment of the present invention further provides a multi-class fault data generation apparatus for generating a network based on conditional countermeasure, where the apparatus includes:
the fault data set construction module is used for acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool, extracting the characteristics of the real fault data and creating a fault data set;
the classification and label definition module is used for classifying the fault data set to obtain a plurality of fault categories, defining a fault label for each fault category and defining the fault label as condition information;
The conditional countermeasure generation network training module is used for constructing a countermeasure generation network and carrying out update training on a generator and a discriminator in the countermeasure generation network based on random noise data and the condition information to obtain a conditional countermeasure generation network;
and the fault data generation module is used for generating a network based on conditional countermeasure, inputting the random noise data and the condition information, generating fault data of core components of the multi-axis industrial numerical control machine tool of different categories and supporting downstream fault recognition model training.
In a third aspect, an embodiment of the present invention further provides a terminal device, where the terminal device includes a memory, a processor, and a multi-class fault data generation program stored in the memory and executable on the processor and generating a network based on a conditional countermeasure, and when the processor executes the multi-class fault data generation program generating a network based on a conditional countermeasure, the method for generating multi-class fault data based on a conditional countermeasure generation network according to any of the foregoing solutions is implemented.
In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, where a multi-class fault data generation program for generating a network based on conditional countermeasure is stored, and when the multi-class fault data generation program for generating a network based on conditional countermeasure is executed by a processor, the steps of the multi-class fault data generation method for generating a network based on conditional countermeasure according to any one of the foregoing solutions are implemented.
Has the beneficial effects that: compared with the prior art, the invention provides a multi-class fault data generation method based on a conditional countermeasure generation network, which comprises the steps of firstly obtaining real fault data of a core component of a multi-axis industrial numerical control machine tool, carrying out feature extraction on the real fault data, and creating a fault data set; classifying the fault data set to obtain a plurality of fault categories, defining a fault label for each fault category, and defining the fault label as condition information; constructing a countermeasure generation network, and performing update training on a generator and a discriminator in the countermeasure generation network based on random noise data and the condition information to obtain a condition countermeasure generation network; and generating a network based on the conditional countermeasure, inputting the real fault data and the condition information, and generating fault data of core components of the multi-axis industrial numerical control machine tool of different categories. According to the invention, the fault data of the core components of the multi-axis industrial numerical control machine tool in different categories can be generated through the training condition confrontation generation network, the problem of sparsity of the fault data of the numerical control machine tool is effectively solved, and the training of a downstream fault diagnosis model is effectively supported.
Drawings
Fig. 1 is a flowchart of a multi-class fault data generation method for generating a network based on conditional countermeasure according to an embodiment of the present invention.
Fig. 2 is an overall flowchart of a multi-class fault data generation method for generating a network based on conditional countermeasure according to an embodiment of the present invention.
Fig. 3 is a schematic block diagram of a multi-class fault data generation apparatus for generating a network based on conditional countermeasure according to an embodiment of the present invention.
Fig. 4 is a schematic block diagram of a terminal device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and effects of the present invention clearer and clearer, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
At present, no fault data generation method for multi-axis industrial numerical control machine tools is available temporarily in the industry. In other related fields such as lithium battery manufacturing, nuclear power equipment detection and the like, the main solution is to generate simulation fault data by a technical means. The prior art generates countermeasure networks based mostly on transfer learning or naive. The transfer learning refers to selecting the associated field with fully labeled data; and uniformly mapping the data characteristics of the field and the target field to a new characteristic space, constructing similar data distribution, and training a model by using the annotation data in the associated field. The naive generation countermeasure network is composed of a generator and a discriminator. Under a fault diagnosis scene, the generator is responsible for constructing fault data, and the discriminator is responsible for judging whether the data accords with the characteristics of real fault data; the two methods resist iteration, and the data distribution rule can be learned from a small number of real fault data labels to automatically generate fault data. The existing fault data generation technology mainly comprises a migration learning and naive confrontation generation network. The drawback of transfer learning is that it is generally only suitable for small data set processing, and its strong heuristic assumption that other domain labels are still suitable for the target domain may also cause severe labeling noise. The disadvantage of the naive countermeasure generation network is that data simulation is performed only based on a small amount of original data, external field knowledge supervision is lacked, the quality of generated fault data is generally low, and the difference between corresponding distribution and actual fault data is large.
In order to solve the above technical problem, this embodiment provides a multi-class fault data generation method for generating a network based on conditional countermeasure, and in specific implementation, this embodiment first obtains real fault data of a multi-axis industrial numerical control machine core component, and performs feature extraction on the real fault data, and creates a fault data set. And classifying the fault data set to obtain a plurality of fault categories, defining a fault label for each fault category, and defining the fault label as condition information. And then, constructing a countermeasure generation network, and updating and training generators and discriminators in the countermeasure generation network based on the fault data set and the condition information to obtain the conditional countermeasure generation network. And finally, generating a network based on conditional countermeasure, inputting the real fault data and the condition information, and generating fault data of core components of the multi-axis industrial numerical control machine tool of different categories. The invention can generate the fault data of the multi-axis industrial numerical control machine tool core components of different categories through the training condition confrontation generation network, and effectively supports the training of a downstream fault diagnosis model.
Exemplary method
The multi-class fault data generation method based on the conditional countermeasure generation network can be applied to terminal equipment, and the terminal equipment can be intelligent equipment such as a computer. Specifically, as shown in fig. 1, the multi-class fault data generation method for a conditional countermeasure generation network in the embodiment includes the following steps:
s100, acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool, performing feature extraction on the real fault data, and creating a fault data set.
The method comprises the steps of firstly extracting real fault data of a core component of the multi-axis industrial numerical control machine tool, wherein the real fault data can be used for collecting an operation state in real time and recording fault data when the multi-axis industrial numerical control machine tool operates, and the fault data is a fault which really occurs when the multi-axis industrial numerical control machine tool operates. After the real fault data is acquired, the present embodiment performs feature extraction on the real fault data, and creates a fault data set based on the extracted fault features. In the present embodiment, the original fault data set is obtained based on real fault data of the core component of the multi-axis industrial numerical control machine tool, and therefore, the present embodiment is intended to analyze fault data of the core component of the multi-axis industrial numerical control machine tool.
In one implementation, the embodiment includes the following steps when creating the failure data set:
s101, acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool;
and S102, performing time-frequency domain sampling and feature extraction on the real fault data to obtain fault features, and constructing the fault data set according to the fault features.
After the real fault data is extracted, the time-frequency domain sampling and the feature extraction are performed on the data signal of the real fault data. In this embodiment, the time-frequency domain samples include time-domain samples and frequency-domain samples. When time-domain sampling is performed on the data signal of the real fault data, the embodiment may perform sampling at a time interval T, that is, perform sampling every T times from the data signal of the real fault data to obtain a first data sample. In the frequency domain sampling of the data signal of the real fault data, in this embodiment, sampling points are set at equal intervals on a frequency spectrum of the data signal, and then the data signal of the real fault data is subjected to frequency domain sampling based on the set sampling points to obtain a second data sample. Next, the present embodiment combines the first data sample and the second data sample to form a sample data set.
After the sample data set is obtained, the embodiment extracts the fault characteristics of each fault data in the sample data set, namely extracts the fault characteristics of each fault data, and then generates the fault data set according to the extracted fault characteristics. In this embodiment, the fault data set is composed of a plurality of fault features, and the fault features are extracted from the sample data set obtained from the real fault data through time domain sampling and frequency domain sampling, so that the fault data set in this embodiment has rich fault features, fully embodies the diversity of samples, and is beneficial to training on generating a countermeasure network for a condition in a subsequent process.
And S200, classifying the fault data set to obtain a plurality of fault categories, defining a fault label for each fault category, and defining the fault label as condition information.
After the fault data set is constructed, the fault features in the fault data set are classified to obtain a plurality of fault categories. In order to distinguish and mark each fault category, the present embodiment defines a fault label for each fault category to distinguish each fault category. Next, the present embodiment takes the trouble ticket as the condition information.
In an implementation manner, the step S200 specifically includes the following steps:
step S201, based on the fault characteristics, classifying the fault data set by adopting a clustering algorithm to obtain fault categories corresponding to the fault characteristics;
step S202, defining a fault label corresponding to each fault category;
step S203, using the fault label as the condition information, where the condition information is a logic condition for describing the distribution of the real fault data.
Specifically, in this embodiment, all fault features are extracted from the fault data set, and then all fault features are classified, and a clustering algorithm is used for the classification. The clustering algorithm is based on similarity, so that the embodiment can analyze all fault characteristics, analyze the similarity among the fault characteristics, then use the fault characteristics with the similarity exceeding a threshold as the same fault category, and analyze all fault categories based on the similarity. Then, the present embodiment manually defines and sets a type of fault label for each fault category, which can be used to distinguish each fault category, and then uses these fault labels as condition information, as shown in fig. 2. Since the condition information plays a role of constraint in the countermeasure generation network, the present embodiment uses the fault label as the condition information of the countermeasure generation network, and can refer to data generation in a subsequent step.
Step S300, constructing a countermeasure generation network, and updating and training generators and discriminators in the countermeasure generation network based on the fault data set and the condition information to obtain the conditional countermeasure generation network.
The embodiment constructs a countermeasure generation network, which includes a generator and a discriminator, and updates and trains the generator and the discriminator in the countermeasure generation network based on a fault data set and set condition information to obtain a conditional countermeasure generation network, which can be used for generating fault data.
In one implementation manner, the embodiment includes, in the training process of generating a network for a challenge:
step S301, defining a loss function of a generator and an arbiter of the countermeasure generation network, wherein the loss function of the generator is used for measuring and generating errors between fault data of the candidate industrial machine tool core component and real fault data distribution, and the loss function of the arbiter is used for judging authenticity of the fault data;
step S302, defining a parameter iterative optimization algorithm of the generator and the discriminator;
step S303, inputting the random noise data and the condition information into the generator, and outputting artificial fault data;
Step S304, inputting the artificial fault data and the real fault data to the discriminator to obtain an authenticity result of the artificial fault data;
step S305, updating and training the parameters of the generator and the discriminator based on the authenticity result, the parameter iterative optimization algorithm and the loss functions of the generator and the discriminator;
and S306, repeatedly carrying out updating training on the generator and the arbiter so that the generator is trained to the minimum threshold value, and the accuracy of the binary judgment of the arbiter when the training is 0-1 exceeds a preset threshold value, so as to obtain the conditional countermeasure generation network.
Specifically, after the countermeasure generation network is constructed, the present embodiment defines the loss functions of the generator and the arbiter of the countermeasure generation network, and defines the parameter iterative optimization algorithm (i.e., the optimizer in fig. 2) of the countermeasure generation network. The loss function of the generator is used for measuring and generating errors between the fault data of the candidate industrial machine tool core component and the real fault data distribution, and the loss function of the discriminator is used for judging the authenticity of the fault data. In the embodiment, the authenticity judgment of the fault data is a binary judgment, that is, 0 represents that the fault data is not true by the discriminator, and 1 represents that the fault data is true; the generator, in contrast, generates artificial fault data based on random noise data, and the loss function of the discriminator determines the difference in distance between the artificial fault data and the true fault data, both of which are represented by vectors.
The objective loss function of the conditional countermeasure generation network in this embodiment is:
In training, as shown in fig. 2, the present embodiment inputs the random noise data and the condition information into the generator, through which artificial fault data is output. And then inputting the artificial fault data and the real fault data into the discriminator, and outputting the authenticity result of the artificial fault data through the discriminator. Then, the method and the device perform updating training on the parameters of the generator and the arbiter based on the authenticity result, the parameter iterative optimization algorithm and the loss functions of the generator and the arbiter. Specifically, the present embodiment updates the parameters of the generator by superimposing a random gradient based on the authenticity result, the parameter iterative optimization algorithm, and the loss function of the generator and the discriminator. The embodiment further updates the parameters of the discriminator by decreasing the random gradient based on the authenticity result, the parameter iterative optimization algorithm and the loss function of the generator and the discriminator. After the parameters of the generator and the discriminator are updated, the updating training of the generator and the discriminator can be repeated, and during the updating training, the parameter of the generator is updated k times (introducing a hyper-parameter k) first, and then the parameter of the discriminator is updated 1 time, so that the distribution difference between the artificial fault data output by the generator and the fault data set is smaller than a preset threshold value, and the conditional countermeasure generating network is obtained.
And S400, generating a network based on the conditional countermeasure, inputting the random noise data and the condition information, and generating fault data of the core components of the multi-axis industrial numerical control machine tool of different types.
The present embodiment inputs random noise data and the condition information to the conditional countermeasure generating network after training of the conditional countermeasure generating network is completed, and the conditional countermeasure generating network can generate the fault data, whereas the conditional countermeasure generating network in the present embodiment is trained based on the random noise data and the condition information of the multi-axis industrial nc machine tool core components, and therefore, after the random noise data and the condition information are input to the conditional countermeasure generating network, fault data of different types of multi-axis industrial nc machine tool core components can be output.
Compared with the traditional model, the method has the advantages that condition information is introduced into the confrontation generating structure, so that the feature selection during the generation of different types of fault data is better controlled, the data noise is reduced, and the generation quality of the fault data is improved; meanwhile, a hyper-parameter k is introduced in the training iteration process, so that the convergence speed of the discriminator is increased, and the model training efficiency is improved. The method can automatically generate and expand the fault data of the core components of the multi-axis industrial numerical control machine, effectively support the training and testing of downstream related fault diagnosis models and enhance the operation control stability of the industrial numerical control machine.
Exemplary devices
Based on the foregoing embodiment, the present invention further provides a multi-class fault data generation apparatus for generating a network based on conditional countermeasure, as shown in fig. 3, the apparatus in this embodiment includes: a fault data set building module 10, a classification and label definition module 20, a conditional countermeasure generation network training module 30, and a fault data generation module 40. Specifically, the fault data set building module 10 is configured to obtain real fault data of a core component of the multi-axis industrial numerical control machine tool, perform feature extraction on the real fault data, and create a fault data set. The classification and label definition module 20 is configured to classify the fault data set to obtain a plurality of fault categories, define a fault label for each fault category, and define the fault label as condition information. The conditional countermeasure generation network training module 30 is configured to construct a countermeasure generation network, and update and train a generator and a discriminator in the countermeasure generation network based on the random noise data and the condition information to obtain a conditional countermeasure generation network. The fault data generation module 40 is configured to generate a network based on conditional countermeasure, input the random noise data and the condition information, and generate fault data of core components of the multi-axis industrial numerical control machine tool of different categories.
In one implementation, the fault data set building module 10 includes:
the data acquisition unit is used for acquiring real fault data of the core component of the multi-axis industrial numerical control machine tool;
and the data set construction unit is used for carrying out time-frequency domain sampling and feature extraction on the real fault data to obtain fault features and constructing the fault data set according to the fault features.
In one implementation, the classification and label definition module 20 includes:
the fault classification unit is used for classifying the fault data set by adopting a clustering algorithm based on the fault characteristics to obtain fault categories corresponding to the fault characteristics;
a label definition unit, configured to define a fault label corresponding to each fault category;
and the condition defining unit is used for taking the fault label as the condition information, wherein the condition information is a logic condition for describing the distribution of the real fault data.
In one implementation, the conditional countermeasure generation network training module 30 includes:
the loss function definition unit is used for defining a loss function of a generator and an arbiter of the countermeasure generation network, wherein the loss function of the generator is used for measuring errors between fault data of the candidate industrial machine tool core component and real fault data distribution, and the loss function of the arbiter is used for judging the authenticity of the fault data;
The optimization algorithm defining unit is used for defining a parameter iterative optimization algorithm of the generator and the discriminator;
an artificial data generation unit for inputting the random noise data and the condition information into the generator and outputting artificial fault data;
the authenticity judging unit is used for inputting the artificial fault data and the real fault data to the discriminator to obtain an authenticity result of the artificial fault data;
the updating and training unit is used for updating and training the parameters of the generator and the discriminator based on the authenticity result, the parameter iterative optimization algorithm and the loss functions of the generator and the discriminator;
the repeated training unit is used for repeatedly carrying out updating training on the generator and the discriminator so that the generator is trained to be minimum in threshold value, and the discriminator is trained until the accuracy of 0-1 binary judgment exceeds a preset threshold value, so that the conditional countermeasure generating network is obtained;
wherein the objective loss function of the conditional countermeasure generation network is:
In one implementation, the update training unit includes:
a generator parameter updating subunit, configured to update a parameter of the generator by superimposing a random gradient based on the authenticity result, the parameter iterative optimization algorithm, and a loss function of the generator and the discriminator;
and the discriminator parameter updating subunit is used for updating the parameters of the discriminator by descending the random gradient.
In one implementation, the repetitive training unit includes:
and the training subunit is used for firstly updating the parameters of the generator k times and then updating the parameters of the discriminator 1 time.
The working principle of each module in the multi-class fault data generation device based on the conditional countermeasure generation network in this embodiment is the same as that of each step in the above method embodiment, and is not described here again.
Based on the above embodiments, the present invention further provides a terminal device, and a schematic block diagram of the terminal device may be as shown in fig. 4. The terminal equipment comprises a processor and a memory which are connected through a system bus, and the processor and the memory are arranged in a host. Wherein the processor of the terminal device is configured to provide computing and control capabilities. The memory of the terminal equipment comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the terminal equipment is used for communicating with an external terminal through network communication connection. The computer program is executed by a processor to implement a multi-class failure data generation method for generating a network based on conditional countermeasure.
It will be understood by those skilled in the art that the block diagram of fig. 4 is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation to the terminal equipment to which the solution of the present invention is applied, and a specific terminal equipment may include more or less components than those shown in the figure, or may combine some components, or have different arrangements of components.
In one embodiment, a terminal device is provided, where the terminal device includes a memory, a processor, and a multi-class fault data generation method program stored in the memory and executable on the processor for generating a network based on conditional countermeasure, and when the processor executes the multi-class fault data generation method program for generating a network based on conditional countermeasure, the following operation instructions are implemented:
acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool, extracting characteristics of the real fault data, and creating a fault data set;
classifying the fault data set to obtain a plurality of fault categories, defining a fault label for each fault category, and defining the fault label as condition information;
constructing a countermeasure generation network, and performing update training on a generator and a discriminator in the countermeasure generation network based on random noise data and the condition information to obtain a condition countermeasure generation network;
And inputting the random noise data and the condition information based on a condition countermeasure generation network to generate fault data of core components of the multi-axis industrial numerical control machine tool in different categories.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by a computer program, which may be stored in a non-volatile computer-readable storage medium, and the computer program may include the processes of the embodiments of the methods described above when executed. Any reference to memory, storage, operations databases, or other media used in the embodiments provided herein may include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), dual-rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), Rambus (Rambus) direct RAM (RDRAM), direct bused dynamic RAM (DRDRAM), and bused dynamic RAM (RDRAM).
In summary, the invention discloses a multi-class fault data generation method for generating a network based on conditional countermeasure, which comprises the following steps: acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool, extracting characteristics, and creating a fault data set; classifying fault data sets, defining fault labels for each fault category, and defining the fault labels as condition information; constructing a countermeasure generation network, and updating and training a generator and a discriminator in the countermeasure generation network based on random noise data and condition information to obtain a condition countermeasure generation network; and (3) inputting random noise data and condition information based on the condition countermeasure generation network, and generating fault data of the core components of the multi-axis industrial numerical control machine tool of different categories. According to the invention, the fault data of the core components of the multi-axis industrial numerical control machine tool in different categories can be generated through the training condition confrontation generation network, the problem of sparsity of the fault data of the numerical control machine tool is effectively solved, and the training of a downstream fault diagnosis model is effectively supported.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, and not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical 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 of the embodiments of the present invention.
Claims (10)
1. A method for generating multi-class fault data for a conditional countermeasure generation network, the method comprising:
acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool, extracting characteristics of the real fault data, and creating a fault data set;
classifying the fault data set to obtain a plurality of fault categories, defining a fault label for each fault category, and defining the fault label as condition information;
constructing a countermeasure generation network, and updating and training a generator and a discriminator in the countermeasure generation network based on random noise data and the condition information to obtain a condition countermeasure generation network;
and generating a network based on condition confrontation, inputting the random noise data and the condition information, generating fault data of core components of the multi-axis industrial numerical control machine tool of different categories, and supporting downstream fault recognition model training.
2. The method for generating multi-class fault data based on conditional countermeasure generation network according to claim 1, wherein the obtaining real fault data of a multi-axis industrial numerical control machine core component, performing feature extraction on the real fault data, and creating a fault data set comprises:
Acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool;
and performing time-frequency domain sampling and feature extraction on the real fault data to obtain fault features, and constructing the fault data set according to the fault features.
3. The method according to claim 1, wherein the classifying the real failure data set to obtain a plurality of failure categories, defining a failure label for each failure category, and defining the failure label as the condition information comprises:
based on the fault characteristics, classifying the fault data set by adopting a clustering algorithm to obtain fault categories corresponding to the fault characteristics;
defining a fault label corresponding to each fault category;
and taking the fault label as the condition information, wherein the condition information is a logic condition for describing the distribution of the real fault data.
4. The method as claimed in claim 1, wherein the step of performing update training on generators and discriminators in the countermeasure generation network based on random noise data and the condition information to obtain the conditional countermeasure generation network comprises:
Defining a loss function of a generator and a discriminator of the antagonistic generation network, wherein the loss function of the generator is used for measuring and generating errors between fault data of the candidate industrial machine tool core component and real fault data distribution, and the loss function of the discriminator is used for judging the authenticity of the fault data;
defining a parameter iterative optimization algorithm of the generator and the discriminator;
inputting the random noise data and the condition information into the generator, and outputting artificial fault data;
inputting the artificial fault data and the real fault data into the discriminator to obtain an authenticity result of the artificial fault data;
updating and training the parameters of the generator and the discriminator based on the authenticity result, the parameter iterative optimization algorithm and the loss functions of the generator and the discriminator;
and repeating the updating training of the generator and the arbiter so that the training of the generator is carried out until the threshold is minimum, and the training of the arbiter is carried out until the accuracy of the binary judgment of 0-1 exceeds a preset threshold, thereby obtaining the conditional countermeasure generating network.
5. The method according to claim 4, wherein the training of updating the parameters of the generator and the arbiter based on the authenticity result, the iterative optimization algorithm of the parameters, and the loss function of the generator and the arbiter comprises:
Updating the parameters of the generator by superimposing a random gradient based on the authenticity result, the parameter iterative optimization algorithm and the loss functions of the generator and the discriminator;
and updating the parameters of the discriminator by decreasing the random gradient.
6. The method of claim 4, wherein the repeating of the training of the generator and the arbiter comprises:
firstly, updating the parameters of the generator k times, and then updating the parameters of the discriminator 1 time.
7. The method of generating multi-class fault data for a conditional countermeasure generation network according to claim 4, wherein the objective loss function of the conditional countermeasure generation network is:
8. An apparatus for generating multi-class failure data of a generation network based on conditional countermeasure, the apparatus comprising:
the fault data set construction module is used for acquiring real fault data of a core component of the multi-axis industrial numerical control machine tool, extracting the characteristics of the real fault data and creating a fault data set;
The classification and label definition module is used for classifying the fault data set to obtain a plurality of fault categories, defining a fault label for each fault category and defining the fault label as condition information;
the conditional countermeasure generation network training module is used for constructing a countermeasure generation network and carrying out update training on a generator and a discriminator in the countermeasure generation network based on random noise data and the condition information to obtain a conditional countermeasure generation network;
and the fault data generation module is used for generating a network based on conditional countermeasure, inputting the random noise data and the condition information, generating fault data of core components of the multi-axis industrial numerical control machine tool of different categories and supporting downstream fault recognition model training.
9. A terminal device characterized in that the terminal device comprises a memory, a processor, and a multi-class failure data generating program stored in the memory and executable on the processor for generating a network based on conditional countermeasure, and the processor implements the steps of the multi-class failure data generating method for generating a network based on conditional countermeasure according to any one of claims 1 to 7 when executing the multi-class failure data generating program for generating a network based on conditional countermeasure.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a multi-class failure data generation program for generating a network based on conditional countermeasure, the multi-class failure data generation program for generating a network based on conditional countermeasure being executed by a processor for implementing the steps of the multi-class failure data generation method for generating a network based on conditional countermeasure according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210409795.6A CN114757286A (en) | 2022-04-19 | 2022-04-19 | Multi-class fault data generation method based on conditional countermeasure generation network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210409795.6A CN114757286A (en) | 2022-04-19 | 2022-04-19 | Multi-class fault data generation method based on conditional countermeasure generation network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114757286A true CN114757286A (en) | 2022-07-15 |
Family
ID=82330798
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210409795.6A Pending CN114757286A (en) | 2022-04-19 | 2022-04-19 | Multi-class fault data generation method based on conditional countermeasure generation network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114757286A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115356599A (en) * | 2022-10-21 | 2022-11-18 | 国网天津市电力公司城西供电分公司 | Multi-mode urban power grid fault diagnosis method and system |
| CN116049719A (en) * | 2023-01-09 | 2023-05-02 | 长沙理工大学 | Fault detection and estimation method, system and medium based on migration type generation model |
| CN116226676A (en) * | 2023-05-08 | 2023-06-06 | 中科航迈数控软件(深圳)有限公司 | Machine tool fault prediction model generation method suitable for extreme environment and related equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109186964A (en) * | 2018-07-28 | 2019-01-11 | 华中科技大学 | Rotary machinery fault diagnosis method based on angle resampling and ROC-SVM |
| CN112649198A (en) * | 2021-01-05 | 2021-04-13 | 西交思创智能科技研究院(西安)有限公司 | Intelligent fault diagnosis method, system and equipment for quasi-unbalanced rolling bearing and application |
| CN113569685A (en) * | 2021-07-20 | 2021-10-29 | 华中科技大学 | Method and system for establishing fault diagnosis model and fault diagnosis of machine tool spindle bearing |
| CN114326655A (en) * | 2021-11-30 | 2022-04-12 | 深圳先进技术研究院 | Industrial robot fault data generation method, system, terminal and storage medium |
-
2022
- 2022-04-19 CN CN202210409795.6A patent/CN114757286A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109186964A (en) * | 2018-07-28 | 2019-01-11 | 华中科技大学 | Rotary machinery fault diagnosis method based on angle resampling and ROC-SVM |
| CN112649198A (en) * | 2021-01-05 | 2021-04-13 | 西交思创智能科技研究院(西安)有限公司 | Intelligent fault diagnosis method, system and equipment for quasi-unbalanced rolling bearing and application |
| CN113569685A (en) * | 2021-07-20 | 2021-10-29 | 华中科技大学 | Method and system for establishing fault diagnosis model and fault diagnosis of machine tool spindle bearing |
| CN114326655A (en) * | 2021-11-30 | 2022-04-12 | 深圳先进技术研究院 | Industrial robot fault data generation method, system, terminal and storage medium |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115356599A (en) * | 2022-10-21 | 2022-11-18 | 国网天津市电力公司城西供电分公司 | Multi-mode urban power grid fault diagnosis method and system |
| CN116049719A (en) * | 2023-01-09 | 2023-05-02 | 长沙理工大学 | Fault detection and estimation method, system and medium based on migration type generation model |
| CN116226676A (en) * | 2023-05-08 | 2023-06-06 | 中科航迈数控软件(深圳)有限公司 | Machine tool fault prediction model generation method suitable for extreme environment and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114757286A (en) | Multi-class fault data generation method based on conditional countermeasure generation network | |
| CN110232280B (en) | Software security vulnerability detection method based on tree structure convolutional neural network | |
| CN108694502B (en) | Self-adaptive scheduling method for robot manufacturing unit based on XGboost algorithm | |
| CN114004155B (en) | Transient stability evaluation method and device considering topological structure characteristics of power system | |
| CN108879732B (en) | Transient stability evaluation method and device for power system | |
| CN110162972B (en) | UAF vulnerability detection method based on statement joint coding deep neural network | |
| CN111275108A (en) | Method for performing sample expansion on partial discharge data based on generation countermeasure network | |
| CN111478314A (en) | Transient stability assessment method for power system | |
| CN114169231A (en) | Method for obtaining distribution line fault classification, positioning and line selection deep learning model based on transfer learning | |
| CN116087692B (en) | Distribution network tree line discharge fault identification method, system, terminal and medium | |
| Li et al. | Structure optimization for echo state network based on contribution | |
| CN116150757A (en) | Intelligent contract unknown vulnerability detection method based on CNN-LSTM multi-classification model | |
| CN112862211A (en) | Method and device for assigning orders of dynamic ring defects of communication management system | |
| CN117370568A (en) | Power grid main equipment knowledge graph completion method based on pre-training language model | |
| CN115860056B (en) | Sensor array neural network method for mixed gas concentration prediction | |
| CN115358473A (en) | Power load prediction method and prediction system based on deep learning | |
| Muškardin et al. | Testing-based black-box extraction of simple models from RNNs and transformers | |
| CN115470854A (en) | Information system fault classification method and classification system | |
| Zhang et al. | An improved genetic algorithm for the re-entrant and flexible job-shop scheduling problem | |
| CN113139332A (en) | Automatic model construction method, device and equipment | |
| Yang et al. | Data-Driven Based System Identification for Linear Time-Invariant Systems | |
| CN118296387B (en) | Model bi-directional iteration-based training sample optimization method and model bi-directional iteration-based training sample optimization system | |
| CN113591484B (en) | Expert system rule base establishment method based on semantic analysis | |
| CN116681102A (en) | Neural network maintenance and test method and device based on semantic modularization | |
| CN117746090A (en) | Power system self-organizing criticality identification method based on graph convolution neural network |
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: 20220715 |
|
| RJ01 | Rejection of invention patent application after publication |