Belt conveyor vibration number fault identification method and system, storage medium and equipment
By generating pseudo-data through short-time Fourier transform and classification generative adversarial network, the problem of belt conveyor vibration fault identification relying on human experience and high cost is solved, and low-cost and fast fault identification is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG CONTWELL COMM TECH CO LTD
- Filing Date
- 2023-06-07
- Publication Date
- 2026-07-31
AI Technical Summary
In existing technologies, vibration fault identification of underground conveyor belts mainly relies on human experience and requires a large number of abnormal vibration signals as training data, resulting in high fault identification costs and difficulty in efficient identification.
A method based on short-time Fourier transform and classification generative adversarial network (CatGAN) is adopted to generate random interference signals as pseudo data, mix them with normal vibration signals, train the classification generative adversarial network, and identify belt conveyor faults.
By generating pseudo-data to train the model, the reliance on real fault vibration signals is reduced, enabling fast and low-cost belt conveyor fault identification and reducing the dependence on field experience.
Smart Images
Figure CN116878871B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of mining equipment technology, specifically to a method, system, storage medium, and device for identifying vibration faults in belt conveyors. Background Technology
[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.
[0003] Belt conveyors are equipment used for transporting and hoisting ore in underground mining operations. Due to the harsh underground environment, belt conveyors are prone to various malfunctions during operation, which manifest as vibrations. Currently, the identification of belt conveyor vibration faults still relies mainly on human experience. Some identification algorithms that use vibration signals for fault diagnosis require a large amount of abnormal vibration signals of different types as training data. These abnormal vibration signals originate from the vibration of the belt conveyor under fault conditions, making it difficult to obtain such data and resulting in high costs for fault identification algorithms. Summary of the Invention
[0004] To address the technical problems mentioned above, this invention provides a method, system, storage medium, and device for identifying vibration data faults in belt conveyors. Based on short-time Fourier transform generative adversarial network (CatGAN) technology, the vibration signal of the belt conveyor during normal operation is preprocessed, and a random interference signal is generated as pseudo-data using a model constructed by a classification generative adversarial network (CatGAN). This pseudo-data is mixed with the preprocessed normal vibration signal and used as input to the CatGAN classifier. After training, the CatGAN is able to identify fiber optic vibration data faults in the belt conveyor.
[0005] To achieve the above objectives, the present invention adopts the following technical solution:
[0006] The first aspect of the present invention provides a method for identifying vibration faults in belt conveyors, comprising the following steps:
[0007] Vibration signals of the belt conveyor acquired by fiber optic sensors are used to determine whether the belt conveyor has malfunctioned using a trained recognition model.
[0008] The training process of the recognition model is as follows:
[0009] Vibration signals of the belt conveyor under normal operating conditions are acquired using fiber optic sensors and preprocessed.
[0010] Pseudo-data samples are obtained by generating random interference signals based on classification generative adversarial networks and then preprocessed.
[0011] The preprocessed pseudo-data samples and vibration signals from the conveyor belt under normal operating conditions are mixed and used as training inputs for the recognition model. The classifier in the generative adversarial network is trained, and the training is completed after the evaluation criteria are met.
[0012] The pseudo data samples and the vibration signals under normal operating conditions of the belt conveyor undergo the same preprocessing process. Specifically, based on the short-time Fourier transform, the vibration signals are converted into a two-dimensional data matrix containing time and frequency domain information, and then normalized.
[0013] The normalized two-dimensional data matrix serves as the input data for the classification generative adversarial network.
[0014] The activation function for classification-based generative adversarial networks is the softmax function.
[0015] The classification generative adversarial network randomly generates interference signals to obtain pseudo-data samples, which are then mixed with the vibration signals of the belt conveyor under normal operation after preprocessing, and used as training input data for the classification generative adversarial network.
[0016] After training, the classification generative adversarial network is evaluated based on the accuracy index and test data.
[0017] The accuracy evaluation model is shown in the following formula:
[0018]
[0019] The true category to which each test sample belongs is y1, y2, ..., y3. k The discriminator clustering yields categories c1, c2, ..., c k c i Assigned to the true category y j Through a one-to-one correspondence, each pseudo-class corresponds to only one true class. Assume that the number of samples belonging to the true class in each cluster category are n1, n2, ..., n k n ij This indicates clustering to belong to y j c i The number of samples in c i Corresponding to the real category y j That is, n ij This indicates the number of samples that were correctly classified.
[0020] A second aspect of the present invention provides a system for implementing the above-described method, comprising:
[0021] The identification module is configured to: determine whether the belt conveyor has malfunctioned based on the vibration signal of the belt conveyor acquired by the fiber optic sensor and the trained identification model.
[0022] The training process of the recognition model loaded in the recognition module is as follows:
[0023] Vibration signals of the belt conveyor under normal operating conditions are acquired using fiber optic sensors and preprocessed.
[0024] Pseudo-data samples are obtained by generating random interference signals based on classification generative adversarial networks and then preprocessed.
[0025] The preprocessed pseudo-data samples and vibration signals from the conveyor belt under normal operating conditions are mixed and used as training inputs for the recognition model. The classifier in the generative adversarial network is trained, and the training is completed after the evaluation criteria are met.
[0026] A third aspect of the present invention provides a computer-readable storage medium.
[0027] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the belt conveyor vibration fault identification method described above.
[0028] A fourth aspect of the present invention provides a computer device.
[0029] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the belt conveyor vibration fault identification method described above.
[0030] Compared with existing technologies, one or more of the above technical solutions have the following beneficial effects:
[0031] 1. Random interference signals generated by a classification generative adversarial network (CatGAN) are used as pseudo-data to simulate abnormal vibration signals under fault conditions. After being mixed with preprocessed normal vibration signals, a dataset is formed and used as input to the CatGAN classifier. The mixed dataset contains vibration signals during normal operation of the belt conveyor and abnormal vibration signals simulating fault conditions. After the classification generative adversarial network is trained, it can identify abnormal vibration data contained in the vibration signals of the belt conveyor obtained by using fiber optic sensors, thereby quickly identifying the fault conditions of the belt conveyor.
[0032] 2. The abnormal vibration signals used in training the classification generative adversarial network (CatGAN) are generated by random interference signals, without using the actual fault vibration signals of the belt conveyor. This eliminates the need for a large number of real fault vibration signals as input, resulting in relatively lower computational costs. The trained model can alert staff that the belt conveyor has experienced abnormal vibrations, while the process of analyzing the specific fault type is still carried out by on-site maintenance personnel based on experience, thereby quickly identifying the fault status of the belt conveyor. Attached Figure Description
[0033] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0034] Figure 1 This is a schematic diagram of the classification generative adversarial network (CatGAN) structure in the belt conveyor vibration fault identification process provided by one or more embodiments of the present invention;
[0035] Figure 2 This is a schematic diagram of the training process of a classification generative adversarial network (Cat GAN) provided in one or more embodiments of the present invention. Detailed Implementation
[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0037] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0038] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0039] As described in the background section, the failure of belt conveyors in underground mining operations manifests as vibration. Currently, the identification of belt conveyor vibration failures still relies mainly on human experience. Some identification algorithms that use vibration signals as fault diagnosis require a large amount of abnormal vibration signals of different types as training data. These abnormal vibration signals originate from the vibration of the belt conveyor under fault conditions, and obtaining this type of data is difficult, resulting in high cost of fault identification algorithms.
[0040] Therefore, the following embodiments provide a method, system, storage medium, and device for identifying vibration data faults in belt conveyors. Based on short-time Fourier transform generative adversarial network (CatGAN) technology, the vibration signal of the belt conveyor during normal operation is preprocessed, and a random interference signal is generated as pseudo-data using a model constructed by a classification generative adversarial network (CatGAN). This pseudo-data is mixed with the preprocessed normal vibration signal and used as the input to the CatGAN classifier. After training, the CatGAN can identify fiber optic vibration data faults in the belt conveyor.
[0041] Example 1:
[0042] like Figure 1-2 As shown, the method for identifying vibration faults in belt conveyors includes the following steps:
[0043] Vibration signals of the belt conveyor acquired by fiber optic sensors are used to determine whether the belt conveyor has malfunctioned using a trained recognition model.
[0044] The training process of the recognition model is as follows:
[0045] Vibration signals of the belt conveyor under normal operating conditions are acquired using fiber optic sensors and preprocessed.
[0046] Pseudo-data samples are obtained by generating random interference signals based on classification generative adversarial networks and then preprocessed.
[0047] The preprocessed pseudo-data samples and vibration signals from the conveyor belt under normal operating conditions are mixed and used as training inputs for the recognition model. The classifier in the generative adversarial network is trained, and the training is completed after the evaluation criteria are met.
[0048] Specifically:
[0049] 1. Vibration signals of the belt conveyor are collected using a fiber optic acoustic vibration system. The collected one-dimensional vibration signal is preprocessed using a Short-Time Fourier Transform (STFT) to convert the vibration signal into a two-dimensional data matrix containing time and frequency domain information. This matrix is then normalized. The STFT calculation formula is as follows:
[0050]
[0051] Where x(t) is a time-domain signal, t and s represent time, f represents frequency, and g(st) is a window function centered at time t.
[0052] The length of the window function determines the time and frequency resolution of the spectrum. For different specific signals, the appropriate window length needs to be selected. The formulas for calculating the time resolution T and frequency resolution F are as follows:
[0053]
[0054]
[0055] Where N x The length of the signal to be processed is given by N0, where N is the overlap length of the window function during the translation process. w The length of the window function.
[0056] A one-dimensional vibration signal is transformed into a two-dimensional data matrix through a short-time Fourier transform. The two-dimensional matrix is normalized, and each normalized matrix is considered as an input to CatGAN.
[0057] 2. Classification Generative Adversarial Networks (CatGAN) replace the sigmoid function with the softmax function on top of the GAN structure, transforming the discriminator from a two-classifier to a multi-classifier. The structure of CatGAN is as follows: Figure 1 As shown. After constructing the CatGAN model:
[0058] 1) Randomly generate one-dimensional signal data with the same sampling length as the actual sample length, and learn upsampling through multiple deconvolution layers of the generator model to obtain generated pseudo data samples;
[0059] 2) The pseudo data samples and real data samples are preprocessed and calculated in step 1, and then mixed together as the input to the CatGAN classifier;
[0060] 3) Train the classifier and update the classifier model parameters;
[0061] 4) Determine if the number of iterations has been reached. If not, update the generator model parameters and repeat steps 1)-3); if the number of iterations has been reached, training is complete.
[0062] 3. After training, the model's performance is evaluated using an accuracy metric. The diagnostic performance is tested using test data with known correct labels. Each test sample belongs to a true category y1, y2, ..., yk, and the discriminator clusters the samples into categories c1, c2, ..., ck. ci is assigned to the true category yj. Through a one-to-one correspondence, each pseudo-category corresponds to only one true category. Assuming the number of samples belonging to the true category in each cluster is n1, n2, ..., nk, nij represents the number of samples clustered into ci belonging to yj. Since ci corresponds to the true category yj, nij represents the number of correctly classified samples. The accuracy metric is calculated as follows:
[0063]
[0064] Once the accuracy requirements are met, the model can be used to monitor abnormal vibrations of the belt conveyor.
[0065] During coal conveyor transport, abnormal vibration signals occur when belt conveyor rollers derail, stop rotating, or coal falls onto the belt and rollers. Based on Short-Time Fourier Transform Generative Adversarial Network (SFT) technology, a model constructed using a classification-based Generative Adversarial Network (CatGAN) generates random interference signals as pseudo-data to simulate abnormal vibration signals under fault conditions. This pseudo-data is then mixed with preprocessed normal vibration signals to form a dataset that serves as input to the CatGAN classifier. This mixed dataset contains vibration signals from normal belt conveyor operation and abnormal vibration signals generated by the classification-based Generative Adversarial Network to simulate fault conditions. After training, CatGAN can identify abnormal vibration data in the vibration data acquired by fiber optic sensors, thus quickly identifying abnormal vibrations in the belt conveyor.
[0066] The abnormal vibration signals used to train the classification generative adversarial network (CatGAN) are generated by random interference signals instead of actual fault vibration signals of the conveyor belt. This eliminates the need for a large number of real fault vibration signals as input, resulting in lower computational costs. The trained model can alert staff that the conveyor belt has experienced abnormal vibrations, while the process of analyzing the specific fault type is still carried out by on-site maintenance personnel based on experience, thus quickly identifying the fault status of the conveyor belt.
[0067] Example 2:
[0068] A system for implementing the above method includes:
[0069] The identification module is configured to: determine whether the belt conveyor has malfunctioned based on the vibration signal of the belt conveyor acquired by the fiber optic sensor and the trained identification model.
[0070] The training process of the recognition model loaded in the recognition module is as follows:
[0071] Vibration signals of the belt conveyor under normal operating conditions are acquired using fiber optic sensors and preprocessed.
[0072] Pseudo-data samples are obtained by generating random interference signals based on classification generative adversarial networks and then preprocessed.
[0073] The preprocessed pseudo-data samples and vibration signals from the conveyor belt under normal operating conditions are mixed and used as training inputs for the recognition model. The classifier in the generative adversarial network is trained, and the training is completed after the evaluation criteria are met.
[0074] The system utilizes a model constructed using a classification-based generative adversarial network (CatGAN) to generate random interference signals as pseudo-data, simulating abnormal vibration signals under fault conditions. This pseudo-data is then mixed with preprocessed normal vibration signals to form a dataset that serves as input to the CatGAN classifier. This mixed dataset includes vibration signals from the conveyor belt during normal operation and abnormal vibration signals generated by the classification-based generative adversarial network, simulating fault conditions. After training, CatGAN can identify abnormal vibration data within the vibration data acquired by fiber optic sensors, thus quickly identifying abnormal vibrations in the conveyor belt.
[0075] Example 3:
[0076] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the belt conveyor vibration fault identification method described in Embodiment 1 above.
[0077] The belt conveyor vibration fault identification method utilizes random interference signals generated by a model constructed using a classification generative adversarial network (CatGAN) as pseudo-data to simulate abnormal vibration signals under fault conditions. This pseudo-data is then mixed with preprocessed normal vibration signals to form a dataset that serves as input to the CatGAN classifier. This mixed dataset contains vibration signals from normal belt conveyor operation and abnormal vibration signals generated by the classification generative adversarial network, simulating fault conditions. After training, the CatGAN can identify abnormal vibration data contained in the vibration data acquired by fiber optic sensors, thus quickly identifying abnormal vibrations in the belt conveyor.
[0078] Example 4:
[0079] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the belt conveyor vibration fault identification method described in Embodiment 1 above.
[0080] The belt conveyor vibration fault identification method utilizes random interference signals generated by a model constructed using a classification generative adversarial network (CatGAN) as pseudo-data to simulate abnormal vibration signals under fault conditions. This pseudo-data is then mixed with preprocessed normal vibration signals to form a dataset that serves as input to the CatGAN classifier. This mixed dataset contains vibration signals from normal belt conveyor operation and abnormal vibration signals generated by the classification generative adversarial network, simulating fault conditions. After training, the CatGAN can identify abnormal vibration data contained in the vibration data acquired by fiber optic sensors, thus quickly identifying abnormal vibrations in the belt conveyor.
[0081] The steps or modules involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.
[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for identifying a belt conveyor vibration number fault, characterized in that: Includes the following steps: Vibration signals of the belt conveyor acquired by fiber optic sensors are used to determine whether the belt conveyor has malfunctioned using a trained recognition model. The training process of the recognition model is as follows: Vibration signals of the belt conveyor under normal operating conditions are acquired using fiber optic sensors and preprocessed. Pseudo-data samples are obtained and preprocessed based on random interference signals generated by a classification-based generative adversarial network. The activation function of the classification-based generative adversarial network is the softmax function. The preprocessing process specifically involves: converting the vibration signal into a two-dimensional data matrix containing time and frequency domain information based on the short-time Fourier transform, and then normalizing it. The preprocessed pseudo-data samples and the vibration signals of the belt conveyor under normal operating conditions are mixed as the training input of the recognition model. The actual fault vibration signals of the belt conveyor are not used to train the classifier in the generative adversarial network. The training is completed after the evaluation of the indicators meets the termination conditions.
2. The belt conveyor vibration number fault identification method of claim 1, wherein: The pseudo-data samples and the vibration signals from the belt conveyor under normal operating conditions undergo the same preprocessing procedure.
3. The method of claim 1, wherein: The normalized two-dimensional data matrix serves as the input data for the classification generative adversarial network.
4. The method of claim 1, wherein: The classification generative adversarial network randomly generates interference signals to obtain pseudo-data samples, which are then mixed with the vibration signals of the belt conveyor under normal operating conditions after preprocessing, and used as training input data for the classification generative adversarial network.
5. The method of claim 1, wherein: After training, the classification generative adversarial network is evaluated based on the accuracy index and test data.
6. A belt conveyor vibration number fault identification system characterized by: include: The identification module is configured to: determine whether the belt conveyor has malfunctioned based on the vibration signal of the belt conveyor acquired by the fiber optic sensor and the trained identification model. The training process of the recognition model loaded in the recognition module is as follows: Vibration signals of the belt conveyor under normal operating conditions are acquired using fiber optic sensors and preprocessed. Pseudo-data samples are obtained and preprocessed based on random interference signals generated by a classification-based generative adversarial network. The activation function of the classification-based generative adversarial network is the softmax function. The preprocessing process specifically involves: converting the vibration signal into a two-dimensional data matrix containing time and frequency domain information based on the short-time Fourier transform, and then normalizing it. The preprocessed pseudo-data samples and the vibration signals of the belt conveyor under normal operating conditions are mixed as the training input of the recognition model. The actual fault vibration signals of the belt conveyor are not used to train the classifier in the generative adversarial network. The training is completed after the evaluation of the indicators meets the termination conditions.
7. A computer readable storage medium characterized in that, It stores a computer program that, when executed by a processor, implements the steps in the belt conveyor vibration fault identification method as described in any one of claims 1-5.
8. A computer device, comprising: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the belt conveyor vibration number fault identification method as described in any one of claims 1-5.