Fault prediction method, system and program product for UAV rotor bearings

Abstract

A method, system, and program product for predicting faults in unmanned aerial vehicle (UAV) rotor bearings are disclosed, belonging to the field of artificial intelligence technology. The method comprises the following steps: obtaining samples to be diagnosed; extracting latent feature vectors using a feature extractor; inputting the latent feature vectors into a fault diagnosis learner to obtain fault data vectors for the bearing to be diagnosed; and a training process comprising obtaining a small batch of unlabeled training samples; ranking the unlabeled training samples using a ranker, and labeling the top K unlabeled training samples by experts, who then convert them into labeled training samples. This method fully utilizes unlabeled samples, using some of these samples as representations of diagnostic uncertainty and then embedding them into a feature latent code. Through adversarial training, the overall latent subspace of the data is reshaped, thereby introducing diagnostic-related information and enhancing diagnostic perception capabilities.

Description

Technical Field

[0001] The present invention relates to a fault prediction method, system and program product for a UAV rotor bearing, and belongs to the field of artificial intelligence technology. Background Art

[0002] The prior art provides a variety of bearing fault prediction methods. For example, the Chinese invention patent application with publication number CN119337204A discloses a VMD-PIML bearing fault prediction method based on physical information constraints. This method improves the FNGO algorithm by introducing a chaotic map to initialize the eagle group distribution, and introduces a sine function and a refraction reverse learning strategy to enhance the algorithm's computational accuracy and search capabilities. The improved FNGO algorithm is used to iterate the important VMD parameter components [K, ɑ] to extract signal features. Finally, the improved FNGO algorithm is used to optimize the hyperparameters in the PIML prediction model. The PIML prediction model can combine physical constraints with machine learning models, integrate bearing signal features, and improve the accuracy and robustness of model predictions. This invention constrains the PIML prediction model with physical knowledge and uses the improved NGO algorithm to optimize the VMD denoising model and the PIML prediction model, thereby improving the generalization and robustness of bearing fault prediction.

[0003] Chinese invention patent application publication number CN119246068A discloses a method for predicting early-stage rolling bearing failures based on a Bayesian-optimized DCAE-RNN. The method comprises the following steps: constructing a DCAE-RNN model, comprising a DCAE model and an RNN model. The DCAE model is composed of a SADE and CAE, and optimizing the hyperparameters of the SDAE and CAE using a Bayesian optimization function; training the DCAE-RNN model using a training set to obtain a trained DCAE-RNN model; inputting the real-time vibration signal of the rolling bearing into the trained DCAE-RNN model, and calculating the residual between the predicted value and the true value output by the model to detect early-stage rolling bearing failures. This invention can effectively describe the operating status of rolling bearings, predicting early-stage rolling bearing failures earlier and more stably with high accuracy and robustness. Furthermore, the model is easy to deploy, which is very important for monitoring the degradation of rolling bearing operating conditions.

[0004] However, these invention patent applications do not disclose how to utilize the uncertain information of unlabeled samples. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a fault prediction method, system and program product for drone rotor bearings, which make full use of unlabeled samples, use some unlabeled samples as representations of diagnostic uncertainty, and then embed them into feature latent codes. Through adversarial training, the overall latent subspace of the data is reshaped to introduce diagnosis-related information, thereby enhancing the ability of diagnostic perception.

[0006] To achieve the above-mentioned object of the invention, the present invention provides a method for predicting the fault of a UAV rotor bearing, which comprises the following steps:

[0007] Step 1: Obtain the sample to be diagnosed ;

[0008] Step 2: Use the feature extractor to identify the sample to be diagnosed Extracting latent feature vectors ;

[0009] Step 3: Substitute the latent feature vector Input into the fault diagnosis learner to obtain the fault data vector of the diagnosed bearing , To find the optimal parameters of the fault diagnosis learner, the training process of the fault diagnosis learner includes:

[0010] S01: Get small batches of unlabeled training samples and labeled training samples , N and M are positive integers greater than or equal to 2;

[0011] S02: Using Ranking Machine For unlabeled training samples Ranking to get ranking value , sort the ranking values from large to small, and give the top K unlabeled training samples to experts for labeling, and convert them into labeled training samples, is the parameter of the ranker; calculate the ranking loss ;

[0012] S03: From labeled training samples through feature extractor Extract the latent feature vector ;Fault diagnosis learner According to the latent feature vector Get the fault data vector of the diagnosed bearing , , are the parameters of the fault diagnosis learner.

[0013] The present invention also provides a fault prediction system for a UAV rotor bearing, which includes: a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor executes the computer program to implement the above-mentioned UAV rotor bearing fault prediction method.

[0014] To achieve the aforementioned object of the invention, the present invention further provides a computer program product, which includes a computer program code, and the computer program code can be called by a processor to execute the aforementioned method for predicting the fault of a UAV rotor bearing.

[0015] The present invention has the following beneficial effects:

[0016] The present invention provides a fault prediction method, system and program product for drone rotor bearings. By fully utilizing unlabeled samples, some unlabeled samples are used as representations of diagnostic uncertainty, which are then embedded into feature latent codes. The overall latent subspace of the data is reshaped through adversarial training, thereby introducing diagnosis-related information and enhancing the ability of diagnostic perception. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of a method for predicting a fault of a UAV rotor bearing provided by the first embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] First embodiment

[0020] Figure 1 FIG. 1 is a flow chart of a method for predicting a fault of a UAV rotor bearing provided by the first embodiment of the present invention. Figure 1 As shown, the first embodiment of the present invention provides a method for predicting a fault of a UAV rotor bearing, which includes the following steps:

[0021] Step 1: Obtain the sample to be diagnosed ;

[0022] Step 2: Use the feature extractor to identify the sample to be diagnosed Extracting latent feature vectors ;

[0023] Step 3: Substitute the latent feature vector Input into the fault diagnosis learner to obtain the fault data vector of the diagnosed bearing , is the optimal parameter of the fault diagnosis learner.

[0024] In the first embodiment, the training process of the fault diagnosis learner includes:

[0025] S01: Get small batches of unlabeled training samples and labeled training samples , N and M are positive integers greater than or equal to 2.

[0026] In the first embodiment, the training process of the fault diagnosis learner further includes:

[0027] S02: Using Ranking Machine For unlabeled training samples Ranking to get ranking value , sort the ranking values from large to small, and give the top K unlabeled training samples to experts for labeling, and convert them into labeled training samples, is the parameter of the ranker; calculate the ranking loss ;

[0028] In the present invention, the ranking loss is calculated by the following formula: :

[0029] ,

[0030] Where, For the ranker Ranking value obtained by ranking; Test data for ranking; are the parameters of the ranker, is the activation function.

[0031] In the first embodiment, the training process of the fault diagnosis learner further includes:

[0032] S03: From labeled training samples through feature extractor Extract the latent feature vector ;Fault diagnosis learner According to the latent feature vector Get the fault data vector of the diagnosed bearing , , are the parameters of the fault diagnosis learner.

[0033] The training process of the fault diagnosis learner provided by the present invention also includes:

[0034] S04: Computational Fault Diagnosis Learner loss :

[0035] ,

[0036] Where, is the measured fault data vector of the diagnosed bearing, It is the fault data vector of the diagnosed bearing output by the fault diagnosis learner.

[0037] The training process of the fault diagnosis learner provided by the present invention also includes:

[0038] S05: Update the parameters of the fault diagnosis learner and the parameters of the ranker :

[0039] ; ,

[0040] is the ranking loss The scale adjustment hyperparameter of is the first learning coefficient; is the second learning coefficient; For the parameters The gradient, For the parameters gradient.

[0041] The training process of the fault diagnosis learner provided by the present invention also includes:

[0042] S06: Calculate the reconstruction loss of the VAE network (Variational Autoencoder Network) and combat losses :

[0043] ,

[0044] Where, For expectations; is the encoder of the VAE network, are the parameters of the encoder, is a hyperparameter; for 's predicted ranking; is the divergence; is the decoder of the VAE network, are the decoder parameters, For The extracted latent feature vector, for The predicted ranking of is the distribution of the latent feature vector z, i=1,…,M+K, j=1,…NK;

[0045] ,

[0046] Where, For expectations; represents the judge, is the parameter of the judge;

[0047] Update the parameters of the VAE network encoder by the following formula and decoder parameters :

[0048] ;

[0049] ,

[0050] Where, To combat losses The scale adjustment hyperparameter of is the third learning coefficient; For the parameters gradient; is the fourth learning coefficient; For the parameters gradient.

[0051] The training process of the fault diagnosis learner provided by the present invention also includes:

[0052] S07: Calculate the loss of the discriminator :

[0053] ,

[0054] Where, For expectations; Represents the judge, is the parameter of the judge; is the decoder of the VAE network, are the parameters of the decoder of the VAE network; i=1,…,M+K, j=1,…,NK.

[0055] Update the parameters of the discriminator by :

[0056] ,

[0057] Where, is the fifth learning coefficient; For the parameters gradient.

[0058] The training process of the fault diagnosis learner provided by the present invention also includes:

[0059] S08: Repeat steps S02 to S07 until the stopping condition is reached and the best 、 、 、 、 .

[0060] The first embodiment of the present invention makes full use of unlabeled samples, takes some unlabeled samples as representations of diagnostic uncertainty, and then embeds them into latent features. It reshapes the overall latent subspace of the data through adversarial training, thereby introducing diagnosis-related information and enhancing the ability of diagnostic perception.

[0061] Second embodiment

[0062] The second embodiment of the present invention only describes the contents that are different from the first embodiment, and the same contents will not be described again.

[0063] A second embodiment of the present invention provides a fault prediction system for a drone rotor bearing, which includes a storage device and a processor. The storage device stores a computer program that can be run on the processor, and the processor executes the computer program to implement the fault prediction method for the drone rotor bearing described in the first embodiment.

[0064] The beneficial effects of the second embodiment of the present invention are the same as those of the first embodiment and will not be described again here.

[0065] Third embodiment

[0066] The second embodiment of the present invention only describes the contents that are different from the first embodiment, and the same contents will not be described again.

[0067] A third embodiment of the present invention provides a program product, which includes computer program code, and the computer program code can be called by a processor to execute the method described in the first embodiment.

[0068] The beneficial effects of the third embodiment of the present invention are the same as those of the first embodiment and will not be described again here.

[0069] It should be noted that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0070] The basic principles, main features and advantages of the present invention are shown and described above. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and modifications fall within the scope of the invention as claimed.

Claims

1. A method for predicting the fault of a UAV rotor bearing, characterized in that: The steps include: Step 1: Obtain the sample to be diagnosed ; Step 2: Use the feature extractor to identify the sample to be diagnosed Extracting latent feature vectors ; Step 3: Substitute the latent feature vector Input into the fault diagnosis learner to obtain the fault data vector of the diagnosed bearing , To find the optimal parameters of the fault diagnosis learner, the training process of the fault diagnosis learner includes: S01: Get small batches of unlabeled training samples and labeled training samples , N and M are positive integers greater than or equal to 2; S02: Using Ranking Machine For unlabeled training samples Ranking to get ranking value , sort the ranking values from large to small, and give the top K unlabeled training samples to experts for labeling, and convert them into labeled training samples, is the parameter of the ranker; calculate the ranking loss ; S03: From labeled training samples through feature extractor Extract the latent feature vector ;Fault diagnosis learner According to the latent feature vector Get the fault data vector of the diagnosed bearing , , are the parameters of the fault diagnosis learner; S04: Computational Fault Diagnosis Learner loss ; S05: Update the parameters of the fault diagnosis learner and the parameters of the ranker : ; , is the ranking loss The scale adjustment hyperparameter of is the first learning coefficient; is the second learning coefficient; For the parameters The gradient, For the parameters The gradient; S06: Calculate the reconstruction loss of the VAE network and combat losses , update the parameters of the VAE network encoder by the following formula and decoder parameters : ； ， Where, To combat losses The scale adjustment hyperparameter of is the third learning coefficient; For the parameters gradient; is the fourth learning coefficient; For the parameters The gradient; S07: Calculate the loss of the discriminator , update the discriminator parameters by : ， Where, is the fifth learning coefficient; For the parameters The gradient; S08: Repeat steps S02 to S07 until the stopping condition is reached and the best 、 、 、 、 .

2. The fault prediction method for a UAV rotor bearing according to claim 1, characterized in that: ， Where, is the measured fault data vector of the diagnosed bearing, It is the fault data vector of the diagnosed bearing output by the fault diagnosis learner.

3. The fault prediction method for a UAV rotor bearing according to claim 2, characterized in that: ， Where, For the ranker Ranking value obtained by ranking; Test data for ranking; are the parameters of the ranker, is the activation function.

4. The method for predicting the failure of a UAV rotor bearing according to claim 3, wherein: ， For expectations; is the encoder of the VAE network, are the parameters of the encoder, is a hyperparameter; for 's predicted ranking; is the divergence; is the decoder of the VAE network, are the decoder parameters, For The extracted latent feature vector, for The predicted ranking of is the distribution of the latent feature vector z, i=1,…,M+K, j=1,…NK.

5. The method for predicting the failure of a UAV rotor bearing according to claim 4, wherein: ， Where, For expectations; Represents the judge, Parameters of the decision maker.

6. The method for predicting the failure of a UAV rotor bearing according to claim 5, characterized in that: ， Where, For expectations; Represents the judge, are the parameters of the decision maker; i=1,…,M+K, j=1,…,NK.

7. A fault prediction system for a UAV rotor bearing, comprising: A memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor executes the computer program to implement the fault prediction method for the drone rotor bearing according to any one of claims 1 to 6.

8. A computer program product comprising computer program code, wherein the computer program code can be called by a processor to execute the fault prediction method for a UAV rotor bearing according to any one of claims 1 to 6.

Images