Mechanical intelligent fault diagnosis method, system and equipment for data imbalance, medium and product

Abstract

The invention discloses a mechanical intelligent fault diagnosis method, system and device for data imbalance, a medium and a product, and relates to the field of mechanical fault diagnosis. Firstly, a generalization classification regularization generative adversarial network composed of a random sampling module, a generation module and a generalization module is constructed; secondly, using original vibration data as generator input to accelerate a generator training process; in addition, a regularization loss item is introduced into a discriminator loss function so as to stabilize the training process of the generative adversarial network under the condition of little data input; and finally, developing a classifier module to distinguish the original samples from the generated samples so as to facilitate fault classification. According to the method, the strong generalization sample can be efficiently generated only by using a small amount of data, and the method is suitable for an intelligent fault diagnosis scene with insufficient training data or unbalanced training data, so that the performance, efficiency and adaptability of the intelligent fault diagnosis method are improved.

Description

Technical Field

[0001] The present application relates to the field of mechanical fault diagnosis, and in particular to a method, system, device, medium and product for mechanical intelligent fault diagnosis for data imbalance. Background Art

[0002] Fault diagnosis is considered to be one of the key technologies to ensure the safe and reliable operation of rotating machinery. In recent years, with the continuous development of artificial intelligence, intelligent fault diagnosis technology has been studied in considerable depth. Deep neural networks represented by convolutional neural networks can effectively extract fault information from massive data and have been widely used in the field of intelligent fault diagnosis. Although many intelligent diagnosis methods have been developed, their application in actual engineering scenarios is still challenging due to limited labeled training data. Deep neural networks require abundant data to ensure excellent diagnostic performance. However, in actual engineering scenarios, rotating machinery is usually not allowed to operate with faults, and even if it is operated, the duration is very short, which leads to a rich number of healthy samples but a very limited number of fault samples. This causes data imbalance between healthy samples and fault samples, which is easy to cause misdiagnosis when using deep neural networks for fault diagnosis. Therefore, it is of great practical significance to study mechanical fault diagnosis under data imbalance conditions in actual engineering application scenarios.

[0003] To solve the problem of data imbalance, data augmentation is one of the effective solutions to expand the limited fault data. Data augmentation methods can be mainly divided into sampling-based methods, feature diversity-based methods and synthesis-based methods. Sampling-based methods include adding noise, overlapping segmentation strategies, etc. These methods aim to make the most of limited data, however, this method may lead to the loss of some key fault information. Feature diversity-based methods extract different feature representations from limited data, but their performance improvement space is limited, especially when there are very few samples. Synthesis-based methods include synthetic minority over-sampling technique (SMOTE) and generative adversarial network (GAN). However, when data is very limited, the quality of samples generated by SMOTE-based methods is low.

[0004] Due to the advantages of feature self-extraction and deep network architecture, GAN can generate samples by designing adversarial training of generators and discriminators. Many improved GAN variants have emerged to improve the quality of data augmentation. In addition, in order to improve the quality of generated samples, data preprocessing techniques are usually applied before feeding the original vibration signal to GAN. Typical data preprocessing methods include utilizing two-dimensional (2D) representation and time-frequency analysis methods.

[0005] Based on the above description, the existing fault diagnosis method has the following disadvantages:

[0006] 1) The input of the generator in GAN is mostly random noise or random noise with some labels, which cannot fully capture the diverse feature representation of the data and reduces the convergence speed of the model.

[0007] 2) The generalization ability of GAN needs to be improved because the quality of generated samples is low when the training data is extremely limited.

[0008] 3) The 2D image calculation time is long, which will reduce the efficiency of fault diagnosis.

[0009] 4) Time-frequency analysis methods require prior knowledge and expertise in signal processing, which may reduce the adaptability of intelligent diagnosis methods. Summary of the invention

[0010] The purpose of this application is to provide a mechanical intelligent fault diagnosis method, system, equipment, medium and product for data imbalance, which can efficiently generate high-quality samples using only a small amount of original vibration signals in the case of data imbalance and extremely limited training data, thereby improving the performance, efficiency and adaptability of the intelligent fault diagnosis method.

[0011] To achieve the above objectives, this application provides the following solutions:

[0012] In a first aspect, the present application provides a machine intelligent fault diagnosis method for data imbalance, comprising:

[0013] Acquire vibration test data under different mechanical states, and use the vibration test data as sample data to generate a data set; the mechanical state includes a healthy state and multiple fault states;

[0014] The data set is divided into a training set and a test set; the number of vibration experimental data in a healthy state in the training set is greater than the number of vibration experimental data in a fault state; the number of vibration experimental data in a healthy state and the number of vibration experimental data in a fault state in the test set are balanced;

[0015] Construct a generalized classification regularized generative adversarial network;

[0016] Using the training set to train the generalized classification regularized generative adversarial network until a first set condition is met, thereby obtaining an original sample data set, a generalized sample, and a trained generalized classification regularized generative adversarial network;

[0017] Using the original sample data set and the generalized sample to verify whether the trained generalized classification regularized generative adversarial network meets the second set condition;

[0018] When the second set condition is not met, the parameters of the generalized classification regularized generative adversarial network are adjusted, and the generalized classification regularized generative adversarial network is trained by returning to the training set until the first set condition is met, thereby obtaining the original sample data set, the generalized sample, and the trained generalized classification regularized generative adversarial network;

[0019] When the second setting condition is met, the trained generalization classification regularized generative adversarial network is used as the sample generation model;

[0020] Inputting the vibration experimental data in the fault state of the training set into the sample generation model to obtain generated samples, and constructing a balanced training set with the vibration experimental data in the healthy state of the training set;

[0021] Build a fault diagnosis classifier;

[0022] Using the balanced training set to train the fault diagnosis classifier until a third set condition is met, thereby obtaining a trained fault diagnosis classifier;

[0023] The test set data or the real-time vibration data of the mechanical equipment to be diagnosed is used as the data to be diagnosed, and the data to be diagnosed is input into the fault diagnosis classifier to obtain the mechanical fault diagnosis result.

[0024] Optionally, construct a generalized classification regularized generative adversarial network, specifically including:

[0025] Constructing a generator module; the generator module is configured to sample and operate the vibration experimental data of the fault state in the training set to obtain an original sample data set, a candidate data set and a generalized sample; the original sample data set includes a signal space and a label space corresponding to the signal space; the generalized sample is used to constitute the optimization target and loss function of the generator module;

[0026] Constructing a classifier module; the classifier module is configured to perform fault classification based on the generalized sample and the original sample data set, obtain output loss, and apply the output loss to the generator module;

[0027] A discriminator module is constructed; the discriminator module is configured to obtain a true sample score and a false sample score based on the generalized sample and the original sample data set; and the samples in the original sample data set are taken as true samples, and the generalized samples are taken as false samples.

[0028] Optionally, build a generator module, specifically including:

[0029] Constructing a random sampling module; the random sampling module is configured to perform sequential sampling and random sampling on the vibration experimental data of the fault state in the training set to obtain an original sample data set and a candidate data set;

[0030] Constructing an initial sample generation module; the initial sample generation module is configured to obtain an initial sample based on the original sample data set;

[0031] Constructing a generalization module; the generalization module is configured to calculate the consistency correlation coefficient between the candidate data set and the initial sample, and generate a generalized sample.

[0032] Optionally, construct an initial sample generation module, specifically including:

[0033] Constructing a first signal / label feature learning module; the first signal / label feature learning module is configured to obtain original sample features based on the original sample data set;

[0034] Construct a generation module; the generation module is configured to obtain the initial sample based on the original sample characteristics.

[0035] Optionally, build a discriminator module, specifically including:

[0036] Constructing a second signal / label feature learning module; the second signal / label feature learning module is configured to obtain sample features based on the generalized sample and the original sample data set;

[0037] Constructing an identification module; the identification module is configured to obtain a true sample score and a false sample score based on the sample features.

[0038] In a second aspect, the present application provides a data imbalance-oriented machine intelligent fault diagnosis system, wherein the data imbalance-oriented machine intelligent fault diagnosis system is used to implement the data imbalance-oriented machine intelligent fault diagnosis method provided above; the system comprises:

[0039] A data acquisition module is used to acquire vibration test data under different mechanical states, and use the vibration test data as sample data to generate a data set; the mechanical state includes a healthy state and multiple fault states;

[0040] A data partitioning module is used to partition the data set to obtain a training set and a test set; the number of vibration experimental data in a healthy state in the training set is greater than the number of vibration experimental data in a fault state; and the number of vibration experimental data in a healthy state and the number of vibration experimental data in a fault state in the test set are balanced;

[0041] A network construction training module is used to construct a generalized classification regularized generative adversarial network and use the training set to train the generalized classification regularized generative adversarial network until the first set condition is met, thereby obtaining an original sample data set, a generalized sample, and a trained generalized classification regularized generative adversarial network;

[0042] A network verification determination module, used to use the original sample data set and the generalized sample to verify whether the trained generalized classification regularized generative adversarial network meets the second set condition; when the second set condition is not met, adjust the parameters of the generalized classification regularized generative adversarial network, and return to execute the steps in the network training module; when the second set condition is met, use the trained generalized classification regularized generative adversarial network as a sample generation model;

[0043] A training set balancing module, used to input the vibration experimental data of the fault state in the training set into the sample generation model to obtain generated samples, and to construct a balanced training set with the vibration experimental data of the healthy state in the training set;

[0044] A classifier construction training module is used to construct a fault diagnosis classifier; the fault diagnosis classifier is trained using the balanced training set until a third set condition is met, thereby obtaining a trained fault diagnosis classifier;

[0045] The fault diagnosis module is used to obtain the data to be diagnosed, including the test set data and the real-time vibration data of the mechanical equipment to be diagnosed, and input the data to be diagnosed into the trained fault diagnosis classifier to obtain the fault diagnosis result.

[0046] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-mentioned data imbalance-oriented mechanical intelligent fault diagnosis method.

[0047] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for mechanical intelligent fault diagnosis for data imbalance.

[0048] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned data imbalance-oriented mechanical intelligent fault diagnosis method.

[0049] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0050] The present application provides a method, system, device, medium and product for mechanical intelligent fault diagnosis in the face of data imbalance. By constructing a sample generation model, high-quality samples can be efficiently generated using only a small amount of original vibration signals of the diagnostic mechanical equipment when data is unbalanced and training data is extremely limited, thereby improving the performance, efficiency and adaptability of the intelligent fault diagnosis method. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 This is an application environment diagram of a machine intelligent fault diagnosis method for data imbalance in one embodiment of the present application;

[0053] Figure 2 A flowchart of a method for mechanical intelligent fault diagnosis for data imbalance provided in one embodiment of the present application;

[0054] Figure 3 A schematic diagram of the structure of a generalized classification regularized generative adversarial network provided in one embodiment of the present application;

[0055] Figure 4 A flowchart of an implementation method of a machine intelligent fault diagnosis method for data imbalance provided in another embodiment of the present application;

[0056] Figure 5 A schematic diagram of the results of using different data enhancement methods to generate sample training and test set data for fault diagnosis in the case of data imbalance and extremely limited data provided by an embodiment of the present application;

[0057] Figure 6 A schematic diagram of ablation experiment results provided in an embodiment of the present application;

[0058] Figure 7 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0060] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0061] The machine intelligent fault diagnosis method for data imbalance provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, or it can be integrated on the server 104, or it can be placed on the cloud or other servers. The terminal 102 can send the vibration data of the mechanical equipment to be diagnosed to the server 104. After the server 104 receives the vibration data of the mechanical equipment to be diagnosed, for the vibration data of the mechanical equipment to be diagnosed, the server 104 inputs the real-time acquired vibration data into the fault diagnosis classifier to obtain the mechanical fault diagnosis result. The server 104 can feedback the obtained mechanical fault diagnosis result to the terminal 102. In addition, in some embodiments, the mechanical intelligent fault diagnosis method for data imbalance can also be implemented separately by the server 104 or the terminal 102, such as the terminal 102 can directly perform mechanical intelligent fault diagnosis on the vibration data of the mechanical equipment to be diagnosed, or the server 104 can obtain the vibration data of the mechanical equipment to be diagnosed from the data storage system, and perform mechanical intelligent fault diagnosis on the vibration data of the mechanical equipment to be diagnosed.

[0062] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0063] In an exemplary embodiment, Figure 2 As shown, a method for mechanical intelligent fault diagnosis facing data imbalance is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate, including the following steps 200 to 210. Among them:

[0064] Step 200: Acquire vibration test data under different mechanical states, and use the vibration test data as sample data to generate a data set. The mechanical state includes a healthy state and multiple fault states.

[0065] In actual application, the training set (i.e., training vibration data set) and test set (i.e., test vibration data set) under different fault conditions can be obtained through sensors installed on rotating machinery. Among them, the test set is to facilitate the subsequent verification of the performance of the trained generalized classification regularized generative adversarial network. The data in the data set are mainly the vibration signal data of the gears, and these data can reflect the different health states of the gears. The different health states of the gears include normal gears, broken gears, missing gear teeth, cracks at the root of the gears, and wear on the surface of the gears.

[0066] Step 201: Divide the data set to obtain a training set and a test set. The number of vibration test data in the healthy state in the training set is much larger than the number of vibration test data in the fault state, that is, there are a large number of healthy state samples in the training set, and very few samples of various fault states. The number of vibration test data in the healthy state and the number of vibration test data in the fault state in the test set are balanced.

[0067] Step 202: Construct a generalized classification regularized generative adversarial network. The generalized classification regularized generative adversarial network is essentially a conditional generative adversarial network that uses the original signal feature representation as the network input, adds a generalization module and a classifier module to assist in training, and adds an L2 norm regularization term to the discriminator objective function. The trained generalized classification regularized generative adversarial network takes vibration data as input and generates samples as output. The constructed generalized classification regularized generative adversarial network is as follows: Figure 3 shown.

[0068] Step 203: Use the training set to train the generalized classification regularized generative adversarial network until the first set condition is met, thereby obtaining the original sample data set, the generalized sample, and the trained generalized classification regularized generative adversarial network.

[0069] Step 204: Use the original sample data set and the generalized sample to verify whether the trained generalized classification regularized generative adversarial network meets the second set condition.

[0070] Step 205: When the second setting condition is not met, adjust the parameters of the generalized classification regularized generative adversarial network and return to step 203.

[0071] Step 206: When the second set condition is met, the trained generalized classification regularized generative adversarial network is used as a sample generation model.

[0072] Step 207: Input the vibration experimental data of the fault state in the training set into the sample generation model to obtain generated samples, and construct a balanced training set with the vibration experimental data of the healthy state in the training set.

[0073] Step 208: Construct a fault diagnosis classifier. The fault diagnosis model uses the original sample data set and the generalized sample as network input and uses the fault classification result of the sample as output.

[0074] Step 209: Use the balanced training set to train the fault diagnosis classifier until the third set condition is met, thereby obtaining a trained fault diagnosis classifier.

[0075] Step 210: Obtain vibration data of the mechanical equipment to be diagnosed, and input the vibration data into a fault diagnosis classifier to obtain a mechanical fault diagnosis result.

[0076] By implementing the above steps 200 to 210 and constructing a sample generation model, a small amount of original vibration signals in the vibration data of the mechanical equipment to be diagnosed can be efficiently used to generate high-quality samples in the case of data imbalance and extremely limited training data, so as to improve the efficiency of generating high-quality samples. When using a deep neural network-based method for fault diagnosis, the use of these high-quality samples can improve the performance and efficiency of the fault diagnosis classifier, which makes the neural network-based fault diagnosis method more suitable for scenarios with data imbalance and extremely limited training data.

[0077] In another exemplary embodiment of the present application, in order to improve the quality of generated samples, the generalization ability of generated samples can be effectively enhanced under extremely limited training data. In the actual application process, step 202 is essentially the process of initializing the generalized classification regularized generative adversarial network to fully learn the data distribution and then output a balanced training set, such as Figure 4 As shown in Figure 1, the process of constructing a generalized classification regularized generative adversarial network includes:

[0078] (1) Constructing a generator module. The generator module is configured to sample and operate the vibration experimental data of the fault state in the training set to obtain the original sample data set, the candidate data set, and the generalized samples. The original sample data set includes the signal space and the label space corresponding to the signal space. The generalized samples are used to form the optimization objective and loss function of the generator module.

[0079] The process of building a generator module includes:

[0080] (1-1) Construct a random sampling module. The random sampling module is configured to sequentially sample and randomly sample the vibration experimental data of the fault state in the training set to obtain the original sample data set and the candidate data set.

[0081] In the actual application process, the original sample data set and the candidate data set are obtained under very few original vibration signals. Among them, the extremely limited vibration data set under c fault states can be expressed as L D =[L1,L2,…,L c ], where Li Represents the vibration data set in the i-th state, and the data length in each state is L. Sequential sampling is performed in each state to generate m samples, and the length of each sample is l. represents the i-th sample after sampling in c states, [·] represents the L according to the index D Do sampling, x i Represents the vibration signal data points in each state, and finally generates the signal space, which is expressed as The i-th sample has a label Corresponding to the generated label space The two constitute the original sample data set.

[0082] At the same time, for L D Perform random sampling operations to construct two different candidate data sets with corresponding labels, namely and This process can be described as follows:

[0083]

[0084]

[0085] In the formula, and The generation principle of the dataset Similar, but the sample index R i and R j is a randomly generated natural number. and Represent two candidate datasets.

[0086] (1-2) Construct a generalization module. The generalization module is configured to calculate the consistency correlation coefficient between the candidate data set and the initial sample and generate a generalized sample. The output of the random sampling module is the input of the generalization module. The consistency correlation coefficient between the initial generated sample and the original sample is calculated by combining the two candidate data sets and generating a generalized sample. i and β i As the i-th initial generated sample and and The consistency correlation coefficient of the original sample in is as follows:

[0087]

[0088] In the formula, ρ i and ρ i ′ is the Pearson correlation coefficient between the i-th initial generated sample and the samples in the two candidate data sets, and The data sets are and The standard deviation of and Represent the signal space X r and The mean of .

[0089] For the final generated generalization sample It can be expressed as:

[0090]

[0091] The optimization objective and loss function of the proposed generator module composed of generalized samples are expressed as:

[0092]

[0093] Where, L G is the generator module loss, It means that the loss function is minimized by adjusting the parameters of the generator module G. Indicates that the input data follows X r The expectation of the sample distribution of the data set, g G (·) is the output layer that converts the result calculated by the discriminator D into the sample classification score, c C (·) is the output layer that converts the result calculated by the classifier C into the sample classification score, Y g is the label of the fake sample, θ c are the training parameters of the classifier.

[0094] (2) Constructing an initial sample generation module: The initial sample generation module is configured to obtain an initial sample based on the original sample data set.

[0095] In actual application, the construction process of the initial sample generation module includes:

[0096] (2-1) Constructing a first signal / label feature learning module The first signal / label feature learning module is configured to obtain original sample features based on an original sample data set.

[0097] The original samples are input into the generator to fully capture different feature representations and generate initial samples. In order to avoid the label features being covered by the signal features, two independent modules are used to implement the feature learning of the signal and the label. The signal and the label are converted into high-level transferable features through two fully connected (FC) layers. In order to avoid gradient explosion and disappearance, the activation functions of normalization and linear rectification function (ReLu) are used in the network. Considering the model performance and computational efficiency, batch normalization (BN) and instance normalization (IN) are used alternately. After the initial feature learning, the signal feature f x and label feature f y It will be concatenated by the Concatenated(·) function and then input into the generation module. The calculation method of this process is:

[0098] f x =F x (X r ,θ g ), f y =F y (Y r ,θ g ).

[0099] f X =Concatenated(f x ,f y ).

[0100] In the formula, F x (·) and F y (·) denotes the feature learners of signals and labels respectively, θ g represents the training parameters of the generator, f x and f y represents the signal features and label features, f X Represents the original sample characteristics.

[0101] (2-2) Constructing a generation module. The generation module is configured to obtain an initial sample based on the original sample features.

[0102] In the subsequent processing, the output of the random sampling module is the input of the generation module, that is, the original sample feature f X Will be input into the generation module to generate the initial sample The generation module consists of a convolutional layer (Conv) and two fully connected layers (FC). Assuming g(·) represents the generation module, the initial generated sample It can be described as:

[0103]

[0104] (3) Construct a classifier module. The classifier module is configured to perform fault classification based on the generalized sample and the original sample dataset, obtain the output loss, and apply the output loss to the generator module.

[0105] The purpose of building the classifier module is to enable the generator module to pay more attention to class differences, thereby improving the stability of the generated samples. Finally, the classifier module outputs the loss and and will lose Used in the generator module. The optimization objective and loss function formulas are as follows:

[0106]

[0107] In the formula, It means that the loss function value is minimized by adjusting the parameters of the classifier module C. Indicates that the data follows X r The expectation when the sample distribution is .

[0108] (4) Construct a discriminator module. The discriminator module is configured to obtain true sample scores and false sample scores based on the generalized sample and the original sample dataset. The samples in the original sample dataset are taken as true samples, and the generalized samples are taken as false samples. That is, the original sample (true sample) or the generalized sample (false sample) is input to identify the category of the sample.

[0109] In order to improve the generalization ability of samples and the stability of the model, a discriminator module containing a new regularized loss function is developed. Similarly, in order to increase the weight of the label feature in the entire sample, a second signal / label feature learning module is proposed. Then, the signal and label features will be cascaded in the cascade layer and input into the discriminator module. Finally, the discriminator module will output the score for distinguishing true samples from fake samples. It is described as follows:

[0110]

[0111] In the formula, S F and S R They represent the false sample fraction and the true sample fraction respectively. Indicates that the data follows Dataset or The expectation of the sample distribution of f. D and f G θ are the output layers that convert the results calculated by the discriminator module D into classification scores for real samples and fake samples, respectively. d represents the training parameters of the discriminator.

[0112] In order to improve the stability of the training model under the condition of extremely limited training data, the loss L in the discriminator module is D A new regularization term R is introduced in LC (D), the optimization objective and loss function formula are expressed as:

[0113]

[0114] Where, L D is the discriminator module loss, represents the minimization of the function value by adjusting the parameters of the discriminator module D, λ represents the coefficient of the regularization term, α F and α R are two variables called anchors that track the discriminator’s predictions for real and fake samples, ||·|| 2 represents the quadratic norm. R LC (D) is an intermediate quantity that serves to simplify the formula.

[0115] Based on this, the process of building the discriminator module can be summarized as:

[0116] (4-1) Constructing a second signal / label feature learning module. The second signal / label feature learning module is configured to obtain sample features based on the generalized sample and the original sample data set.

[0117] (4-2) Constructing a discrimination module: The discrimination module is configured to obtain a true sample score and a false sample score based on sample features.

[0118] In another exemplary embodiment of the present application, based on the specific construction process of the generalized classification regularized generative adversarial network given above, the training method can be:

[0119] The samples in the original sample data set are sequentially input into the constructed generalized classification regularized generative adversarial network. Each time a sample is input, feature extraction, initial sample generation, generalized sample generation, discriminator parameter update, classifier parameter update and generator parameter update are performed in sequence. The cycle is repeated until the generalized classification regularized generative adversarial network reaches the Nash equilibrium (i.e., the first setting condition is met), and a trained generalized classification regularized generative adversarial network is obtained.

[0120] In another exemplary embodiment of the present application, the process of constructing the balanced training set in step 207 can be described as follows:

[0121] (1) Input the vibration experimental data of the fault state in the training set into the sample generation model to obtain the generated samples under different fault states. In this embodiment, the essence is to input extremely limited fault data into the trained generalized classification regularized generative adversarial network to obtain the generated samples under different fault states.

[0122] (2) The generated samples are combined with the vibration experimental data of the healthy state in the training set to construct a balanced training set.

[0123] For example, in actual application, the original sample data set X r After the samples in are input into the trained generalized classification regularized generative adversarial network process, the generalized sample generated by a single sample is Xg, until a specified number of samples of all fault types are generated, and they are combined with healthy samples to form a balanced training set.

[0124] Furthermore, in the actual application process, in order to further ensure the integrity of the generated balanced training set and improve the accuracy of fault diagnosis, the Fourier transform visualization technology can also be used to evaluate the quality of the generated samples. If necessary, the generalized classification regularized generative adversarial network can be retrained. The generated samples and the original samples are Fourier transformed together, and the quality of the generated samples is analyzed based on the differences between the two samples in the time domain and the frequency domain. If the quality of the generated samples does not meet the similarity requirements with the original samples, the generalized classification regularized generative adversarial network is retrained until the generated samples are highly similar to the original samples in the frequency domain (that is, the second set condition is met). Finally, the trained generalized classification regularized generative adversarial network is saved, and the trained generalized classification regularized generative adversarial network is used to expand the unbalanced training set to generate a balanced training set.

[0125] (3) The fault diagnosis classifier is trained using the balanced training set until the classification accuracy of the fault diagnosis classifier converges (i.e., the third setting condition is met), and the trained fault diagnosis classifier is used as the fault diagnosis model. In the application process, the balanced training set can be divided into a training set and a validation set to complete the training of the fault diagnosis classifier.

[0126] Based on the above description, the present application first introduces the original sample data set and the candidate data set, and performs sequence sampling and random sampling on the extremely limited fault samples. Secondly, the generation module in the generalized classification regularized generative adversarial network is used to perform feature learning and initial sample generation on the original data, and the generalized sample is generated in conjunction with the candidate data set. Then, with the assistance of the discriminator module and the classifier module, the generation performance is improved, and the L2 norm regularization is integrated into the loss function of the discriminator module. After training, the loss function of the generalized classification regularized generative adversarial network reaches the Nash equilibrium, and the generalized classification regularized generative adversarial network will output samples. Subsequently, the trained generator module is extracted, and the vibration experimental data of the fault state in the training set is used to generate samples, which are combined with the vibration experimental data of the healthy state in the training set to form a balanced training set. Finally, the fault diagnosis classifier is trained and used to classify the vibration data acquired in real time. This application fully considers the impact of data imbalance and extremely limited training data on diagnosis in industrial scenarios, provides a new method for state monitoring and fault diagnosis of rotating machinery, and provides important technical support for the safe and reliable operation of rotating machinery equipment.

[0127] In another exemplary embodiment of the present application, Figure 5 As shown, the generalization classification regularization generative adversarial network (GCRGAN) is the method proposed in this application. CGAN represents a conditional generative adversarial network, WGAN-GP represents a generative adversarial network with Wasserstein distance and gradient penalty, ACGAN represents a generative adversarial network with auxiliary classifiers, ACWGAN-GP represents an ACGAN with Wasserstein distance and gradient penalty, Overlapping segmentation represents sample expansion using oversampling of original data, and Direct diagnosis represents a fault diagnosis method using an unbalanced data set and a one-dimensional convolutional neural network. In this embodiment, the proposed GCRGAN method and the above method are used to identify the health status of planetary gearbox test data. The results show that the proposed GCRGAN method can accurately and effectively identify the fault type of 2500 test samples (fault type unknown). In five repeated runs, the highest result of the method's diagnostic accuracy is higher than that of existing advanced methods, and the average diagnostic accuracy reaches 99.28%, with a standard deviation of 0.19%, proving the high accuracy and strong stability of the GCRGAN method proposed in this application. This demonstrates the superiority of the provided method for data augmentation in the face of data imbalance and extremely limited data.

[0128] In another exemplary embodiment of the present application, Figure 6 As shown, an ablation experiment was conducted to verify the role of each component in the method proposed in this application. It should be noted that the GCRGAN (random noise) method refers to a method in which the input of the generator in GCRGAN is changed to random noise, the GCGAN method refers to the GCRGAN method in which the L2 regularization loss function term is missing in the discriminator loss, the GRGAN method refers to the GCRGAN method in which the generator module lacks the classifier module, and the CRGAN method refers to the GCRGAN method in which the generator module lacks the generalization module. Figure 6 The convergence curves and running times of the five methods are shown in Figure 6 Part (a) is a schematic diagram of the ablation experiment results of the method provided in this application. Figure 6 Part (b) is a schematic diagram of the ablation experiment results of the GCRGAN (random noise) method. Figure 6 Part (c) is a schematic diagram of the ablation experiment results of the GCGAN method. Figure 6 Part (d) is a schematic diagram of the ablation experiment results of the GRGAN method. Figure 6 Part (e) is a schematic diagram of the ablation experiment results of the CRGAN method. Figure 6 Part (f) provides a comparison chart of the experimental results of the method provided in this application and GCRGAN (random noise). Compared with the GCRGAN (random noise) method, the GCRGAN method proposed in this application has a lower generator loss and a faster convergence speed. This shows that inputting the original vibration signal can improve the performance of GCRGAN. Compared with the GCGAN method, the generator loss of the GCRGAN method proposed in this application is more stable, which indicates that it can generate more high-quality samples. Compared with the GRGAN method, the discriminator loss of the GCRGAN method proposed in this application is more stable and can better improve the generator. It is worth noting that the CRGAN method is overfitting, which makes it unable to generate high-quality samples.

[0129] In summary, compared with the prior art, the solution provided by this application has the following improvements:

[0130] 1. In order to improve the quality of generated samples, this application designs a new generator consisting of a random sampling module, a generation module and a generalization module, which can effectively enhance the generalization ability of generated samples under extremely limited training data.

[0131] 2. In order to fully capture different feature representations and accelerate model convergence, this application uses the original vibration signal as the input of the new generator. This method can improve the efficiency of high-quality sample generation under extremely limited training data.

[0132] 3. In order to further improve the generalization ability of the GAN model trained based on extremely limited training data, this application introduces a new regularization loss term in the loss function of the discriminator module to effectively stabilize the learning process of GAN under extremely limited training data.

[0133] 4. In order for the generator to pay more attention to class differences and generate more high-quality samples, this application develops a classifier module in the proposed GAN model to distinguish generated samples from real samples.

[0134] Based on the same inventive concept, the embodiment of the present application also provides a data imbalance-oriented machine intelligent fault diagnosis system for implementing the data imbalance-oriented machine intelligent fault diagnosis method involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the data imbalance-oriented machine intelligent fault diagnosis system provided below can refer to the limitations of the data imbalance-oriented machine intelligent fault diagnosis method above, and will not be repeated here.

[0135] In an exemplary embodiment, the present application provides a data imbalance-oriented mechanical intelligent fault diagnosis system comprising:

[0136] The data acquisition module is used to acquire vibration test data under different mechanical states and use the vibration test data as sample data to generate a data set. The mechanical state includes a healthy state and multiple fault states.

[0137] The data partitioning module is used to partition the data set to obtain a training set and a test set. The number of vibration test data in the healthy state in the training set is greater than the number of vibration test data in the fault state. The number of vibration test data in the healthy state and the number of vibration test data in the fault state in the test set are balanced.

[0138] The network construction training module is used to construct a generalized classification regularized generative adversarial network and use a training set to train the generalized classification regularized generative adversarial network until the first set condition is met, thereby obtaining an original sample data set, a generalized sample, and a trained generalized classification regularized generative adversarial network.

[0139] The network verification determination module is used to use the original sample data set and the generalized sample to verify whether the trained generalized classification regularized generative adversarial network meets the second set condition. When the second set condition is not met, the parameters of the generalized classification regularized generative adversarial network are adjusted, and the steps in the network training module are returned to be executed. When the second set condition is met, the trained generalized classification regularized generative adversarial network is used as the sample generation model.

[0140] The training set balancing module is used to input the vibration experimental data of the fault state in the training set into the sample generation model to obtain the generated samples, and construct a balanced training set with the vibration experimental data of the healthy state in the training set.

[0141] The classifier construction training module is used to construct a fault diagnosis classifier; the fault diagnosis classifier is trained using a balanced training set until the third set condition is met, thereby obtaining a trained fault diagnosis classifier.

[0142] The fault diagnosis module is used to obtain the data to be diagnosed, including the test set data and the real-time vibration data of the mechanical equipment to be diagnosed, and input the data to be diagnosed into the trained fault diagnosis classifier to obtain the fault diagnosis result.

[0143] Based on the above description, compared with the prior art, the present application has the following advantages:

[0144] 1. This application designs a feature learning module for original samples, and uses the feature expressions of samples and their categories as input to the generator module, which speeds up the convergence of the model.

[0145] 2. This application designs a conditional generative adversarial network with a loss function with L2 norm regularization, and introduces the regularization term into the loss of the discriminator module, making the learning process of GAN more stable under extremely limited training data.

[0146] 3. This application designs a classifier module and its loss function, and introduces them into the conditional generative adversarial network, so that the generator module can generate higher quality samples.

[0147] 4. This application proposes a sample generation method based on consistency correlation coefficient. By calculating the consistency correlation coefficient between the initial generated sample and two sets of candidate data sets, the two samples are weighted to obtain a generalized sample, thereby improving the quality of the generated sample.

[0148] 5. The method proposed in this application does not use preprocessing methods such as 2D image generation. The fault diagnosis classifier model is simple, the real-time calculation is fast, and the fault diagnosis efficiency is high.

[0149] 6. This application can perform condition monitoring and fault diagnosis on rotating machinery when some categories of experimental data are extremely limited or the data set is unbalanced. It can learn the distribution of extremely limited samples and complete the data without time-frequency analysis process, which broadens the adaptability of intelligent diagnosis methods.

[0150] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 7As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store mechanical intelligent fault diagnosis data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a mechanical intelligent fault diagnosis method for data imbalance is implemented.

[0151] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0152] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0153] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0154] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0155] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0156] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0157] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0158] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A machine intelligent fault diagnosis method for data imbalance, characterized in that: The mechanical intelligent fault diagnosis method for data imbalance includes: Acquire vibration test data under different mechanical states, and use the vibration test data as sample data to generate a data set; the mechanical state includes a healthy state and multiple fault states; The data set is divided into a training set and a test set; the number of vibration experimental data in a healthy state in the training set is greater than the number of vibration experimental data in a fault state; the number of vibration experimental data in a healthy state and the number of vibration experimental data in a fault state in the test set are balanced; Construct a generalized classification regularized generative adversarial network; Using the training set to train the generalized classification regularized generative adversarial network until a first set condition is met, thereby obtaining an original sample data set, a generalized sample, and a trained generalized classification regularized generative adversarial network; Using the original sample data set and the generalized sample to verify whether the trained generalized classification regularized generative adversarial network meets the second set condition; When the second set condition is not met, the parameters of the generalized classification regularized generative adversarial network are adjusted, and the generalized classification regularized generative adversarial network is trained by returning to the training set until the first set condition is met, thereby obtaining the original sample data set, the generalized sample, and the trained generalized classification regularized generative adversarial network; When the second setting condition is met, the trained generalization classification regularized generative adversarial network is used as the sample generation model; Inputting the vibration experimental data in the fault state of the training set into the sample generation model to obtain generated samples, and constructing a balanced training set with the vibration experimental data in the healthy state of the training set; Build a fault diagnosis classifier; Using the balanced training set to train the fault diagnosis classifier until a third set condition is met, thereby obtaining a trained fault diagnosis classifier; The test set data or the real-time vibration data of the mechanical equipment to be diagnosed is used as the data to be diagnosed, and the data to be diagnosed is input into the fault diagnosis classifier to obtain the mechanical fault diagnosis result.

2. The machine intelligent fault diagnosis method for data imbalance according to claim 1 is characterized in that: Construct a generalized classification regularized generative adversarial network, including: Constructing a generator module; the generator module is configured to sample and operate the vibration experimental data of the fault state in the training set to obtain an original sample data set, a candidate data set and a generalized sample; the original sample data set includes a signal space and a label space corresponding to the signal space; the generalized sample is used to constitute the optimization target and loss function of the generator module; Constructing a classifier module; the classifier module is configured to perform fault classification based on the generalized sample and the original sample data set, obtain output loss, and apply the output loss to the generator module; A discriminator module is constructed; the discriminator module is configured to obtain a true sample score and a false sample score based on the generalized sample and the original sample data set; and the samples in the original sample data set are taken as true samples, and the generalized samples are taken as false samples.

3. The machine intelligent fault diagnosis method for data imbalance according to claim 2 is characterized in that: Build the generator module, including: Constructing a random sampling module; the random sampling module is configured to perform sequential sampling and random sampling on the vibration experimental data of the fault state in the training set to obtain an original sample data set and a candidate data set; Constructing an initial sample generation module; the initial sample generation module is configured to obtain an initial sample based on the original sample data set; Constructing a generalization module; the generalization module is configured to calculate the consistency correlation coefficient between the candidate data set and the initial sample, and generate a generalized sample.

4. The machine intelligent fault diagnosis method for data imbalance according to claim 2 is characterized in that: Construct the initial sample generation module, including: Constructing a first signal / label feature learning module; the first signal / label feature learning module is configured to obtain original sample features based on the original sample data set; Construct a generation module; the generation module is configured to obtain the initial sample based on the original sample characteristics.

5. The machine intelligent fault diagnosis method for data imbalance according to claim 2 is characterized in that: Build the discriminator module, including: Constructing a second signal / label feature learning module; the second signal / label feature learning module is configured to obtain sample features based on the generalized sample and the original sample data set; Constructing an identification module; the identification module is configured to obtain a true sample score and a false sample score based on the sample features.

6. A machine intelligent fault diagnosis system for data imbalance, characterized in that: The data imbalance-oriented machine intelligent fault diagnosis system is used to implement the data imbalance-oriented machine intelligent fault diagnosis method according to any one of claims 1 to 5; the data imbalance-oriented machine intelligent fault diagnosis system includes: A data acquisition module is used to acquire vibration test data under different mechanical states, and use the vibration test data as sample data to generate a data set; the mechanical state includes a healthy state and multiple fault states; A data partitioning module is used to partition the data set to obtain a training set and a test set; the number of vibration experimental data in a healthy state in the training set is greater than the number of vibration experimental data in a fault state; and the number of vibration experimental data in a healthy state and the number of vibration experimental data in a fault state in the test set are balanced; A network construction training module is used to construct a generalized classification regularized generative adversarial network and use the training set to train the generalized classification regularized generative adversarial network until the first set condition is met, thereby obtaining an original sample data set, a generalized sample, and a trained generalized classification regularized generative adversarial network; A network verification determination module, used to use the original sample data set and the generalized sample to verify whether the trained generalized classification regularized generative adversarial network meets the second set condition; when the second set condition is not met, adjust the parameters of the generalized classification regularized generative adversarial network, and return to execute the steps in the network training module; when the second set condition is met, use the trained generalized classification regularized generative adversarial network as a sample generation model; A training set balancing module, used to input the vibration experimental data of the fault state in the training set into the sample generation model to obtain generated samples, and to construct a balanced training set with the vibration experimental data of the healthy state in the training set; A classifier construction training module is used to construct a fault diagnosis classifier; the fault diagnosis classifier is trained using the balanced training set until a third set condition is met, thereby obtaining a trained fault diagnosis classifier; The fault diagnosis module is used to obtain the data to be diagnosed, including the test set data and the real-time vibration data of the mechanical equipment to be diagnosed, and input the data to be diagnosed into the trained fault diagnosis classifier to obtain the fault diagnosis result.

7. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the mechanical intelligent fault diagnosis method for data imbalance described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for mechanical intelligent fault diagnosis for data imbalance described in any one of claims 1 to 5 is implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for mechanical intelligent fault diagnosis for data imbalance described in any one of claims 1 to 5 is implemented.

Images