Coal mine ventilation motor vibration fault detection method and system based on deep learning

Abstract

The invention discloses a coal mine ventilation motor vibration fault detection method and system based on deep learning, and the method comprises the steps: collecting the multivariate time series data of a coal mine ventilation motor, carrying out the preprocessing of the multivariate time series data, and extracting the feature vector of the preprocessed multivariate time series data; fusing and reconstructing the feature vectors, training a network model based on deep learning through reconstructed features, and obtaining a vibration fault detection model; single-fault independent detection verification and multi-fault synchronous detection verification are carried out on the vibration fault detection model, the detection performance of the vibration fault detection model is determined according to the verification result, the vibration fault detection model is optimized, and deployment and fault detection are carried out on the optimized vibration fault detection model. Compared with the prior art, the method improves the judgment precision, does not need manual intervention, can achieve the fault detection through the detection data of the sensor, improves the detection precision and efficiency, and guarantees the objectivity of a detection result.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and particularly to a vibration fault detection method and system for coal mine ventilation motors based on deep learning. Background Art

[0002] At present, in the coal mine industry, ventilation motors are important equipment to ensure the safe and stable production of coal mines. However, due to the special nature of the coal mine environment, such as high temperature, high humidity, dust pollution, etc., ventilation motors are prone to various faults after long-term use. Among them, vibration faults are the most common type. Such faults may lead to abnormal operation of the motor, performance degradation, and even major safety accidents. Therefore, timely and accurate detection and diagnosis of vibration faults in ventilation motors play a crucial role in ensuring the normal production of coal mines. However, traditional fault detection methods are generally carried out through manual inspections, which not only require a large amount of manpower and time, but also, although the auditory and visual methods are simple, they are limited by human subjective factors and are prone to misjudgment, resulting in low objectivity of the detection results and reducing the detection efficiency and accuracy. Summary of the Invention

[0003] In view of the problems shown above, the present invention provides a vibration fault detection method and system for coal mine ventilation motors based on deep learning to solve the problems mentioned in the background art that vibration fault detection through manual inspection not only requires a large amount of manpower and time, but also, although the auditory and visual methods are simple, they are limited by human subjective factors and are prone to misjudgment, resulting in low objectivity of the detection results and reducing the detection efficiency and accuracy.

[0004] A vibration fault detection method for coal mine ventilation motors based on deep learning includes the following steps:

[0005] Collect multivariate time-series data of coal mine ventilation motors and preprocess it, and extract the feature vectors of the preprocessed multivariate time-series data;

[0006] Fuse and reconstruct the feature vectors, and train a deep learning-based network model through the reconstructed features to obtain a vibration fault detection model;

[0007] Perform single-fault independent detection verification and multi-fault synchronous detection verification on the vibration fault detection model through vibration data samples of ventilation motors with different fault types to obtain verification results;

[0008] Determine the detection performance of the vibration fault detection model according to the verification results, optimize the vibration fault detection model based on the detection performance, and deploy and perform fault detection on the optimized vibration fault detection model.

[0009] Preferably, collecting the multi - temporal data of the coal mine ventilation motor and pre - processing it, and extracting the feature vectors of the pre - processed multi - temporal data, including:

[0010] Collecting the multi - temporal data of the coal mine ventilation motor through vibration sensors, temperature sensors and pressure sensors, and performing data cleaning and smoothing pre - processing on the multi - temporal data;

[0011] Converting the pre - processed multi - temporal data into a unified format, and obtaining the statistical features of the converted multi - temporal data as the basic feature vectors;

[0012] Obtaining the high - order dimension parameters of each type of signal in the multi - temporal data and using them as advanced feature vectors;

[0013] Integrating the basic feature vectors and the advanced feature vectors to generate the final feature vectors of the multi - temporal data.

[0014] Preferably, fusing and reconstructing the feature vectors, and training a deep - learning - based network model through the reconstructed features to obtain a vibration fault detection model, including:

[0015] Using a variational auto - encoder model to fuse and reconstruct the feature vectors, obtaining reconstructed features, and establishing a selective state - space deep - learning model as an identification model;

[0016] Obtaining the identification behavior strategy of the identification model for the coal mine ventilation motor in the fault state through the reconstructed features as model inputs;

[0017] Obtaining vibration fault data with labels, converting the vibration fault data into representation vectors and performing low - dimensional mapping;

[0018] Training the identification model according to the identification behavior strategy and the mapped data to obtain a vibration fault detection model.

[0019] Preferably, performing single - fault independent detection verification and multi - fault synchronous detection verification on the vibration fault detection model through vibration data samples of ventilation motors with different fault types, and obtaining verification results, including:

[0020] Respectively obtaining the first energized fan vibration data samples with different fault types and the second energized fan vibration data samples in the non - fault state;

[0021] Inputting the first energized fan vibration data samples into the vibration fault detection model respectively, and confirming the single - fault independent detection accuracy according to the first model output results;

[0022] Randomly select multiple first-powered fan vibration data samples from the first-powered fan vibration data samples and input them into the vibration fault detection model at the same time, and confirm the multi-fault synchronous detection accuracy according to the output result of the second model;

[0023] Based on the single-fault independent detection accuracy and the multi-fault synchronous detection accuracy, confirm the verification result of the detection logic of the vibration fault detection model.

[0024] Preferably, determine the detection performance of the vibration fault detection model according to the verification result, optimize the vibration fault detection model based on the detection performance, and deploy and detect faults for the optimized vibration fault detection model, including:

[0025] Determine the detection accuracy and detection integrity of the vibration fault detection model according to the verification result, and determine the detection performance according to the detection accuracy and detection integrity;

[0026] Determine the model adjustment parameters based on the detection performance, and optimize the vibration fault detection model based on the model adjustment parameters;

[0027] Determine the vibration detection link of the coal mine ventilation motor, determine the model deployment requirements according to the detection link, and configure the deployment environment based on the model deployment requirements;

[0028] Deploy the optimized vibration fault detection model to the deployment environment for vibration fault detection of the ventilation motor.

[0029] A coal mine ventilation motor vibration fault detection system based on deep learning, the system includes:

[0030] An extraction module for collecting multivariate time-series data of a coal mine ventilation motor, preprocessing it, and extracting feature vectors of the preprocessed multivariate time-series data;

[0031] A training module for fusing and reconstructing the feature vectors, training a deep learning-based network model through the reconstructed features, and obtaining a vibration fault detection model;

[0032] A verification module for performing single-fault independent detection verification and multi-fault synchronous detection verification on the vibration fault detection model through ventilation motor vibration data samples of different fault types, and obtaining a verification result;

[0033] An optimization module for determining the detection performance of the vibration fault detection model according to the verification result, optimizing the vibration fault detection model based on the detection performance, and deploying and detecting faults for the optimized vibration fault detection model.

[0034] Preferably, the extraction module includes:

[0035] A preprocessing sub-module for collecting multivariate time-series data of a coal mine ventilation motor through vibration sensors, temperature sensors, and pressure sensors, and performing data cleaning and smoothing preprocessing on the multivariate time-series data;

[0036] A first acquisition sub-module for converting the preprocessed multivariate time-series data into a unified format and obtaining the statistical features of the converted multivariate time-series data as basic feature vectors;

[0037] A second acquisition sub-module for obtaining the high-order dimensional parameters of each type of signal in the multivariate time-series data and using them as advanced feature vectors;

[0038] An integration sub-module for integrating the basic feature vectors and the advanced feature vectors to generate the final feature vector of the multivariate time-series data.

[0039] Preferably, the training module includes:

[0040] A feature fusion and reconstruction sub-module for using a variational autoencoder model to fuse and reconstruct the feature vectors, obtaining the reconstructed features, and establishing a selective state-space deep learning model as an identification model;

[0041] A third acquisition sub-module for obtaining the identification behavior strategy of the identification model for the coal mine ventilation motor in a fault state by using the reconstructed features as model inputs;

[0042] A conversion sub-module for obtaining vibration fault data with labels, converting the vibration fault data into representation vectors and performing low-dimensional mapping;

[0043] A training sub-module for training the identification model according to the identification behavior strategy and the mapped data to obtain a vibration fault detection model.

[0044] Preferably, the verification module includes:

[0045] A fourth acquisition sub-module for respectively obtaining a first energized fan vibration data sample of different fault types and a second energized fan vibration data sample in a non-fault state;

[0046] A first confirmation sub-module for respectively inputting the first energized fan vibration data samples into the vibration fault detection model and confirming the single-fault independent detection accuracy according to the first model output result;

[0047] A second confirmation sub-module for randomly selecting multiple first energized fan vibration data samples from the first energized fan vibration data samples and inputting them into the vibration fault detection model at the same time, and confirming the multi-fault synchronous detection accuracy according to the second model output result;

[0048] The third confirmation sub-module is used to confirm the verification result of the detection logic of the vibration fault detection model based on the single-fault independent detection accuracy and the multi-fault synchronous detection accuracy.

[0049] Preferably, the optimization module includes:

[0050] A determination sub-module is used to determine the detection accuracy and detection integrity of the vibration fault detection model according to the verification result, and determine the detection performance according to the detection accuracy and detection integrity;

[0051] An optimization sub-module is used to determine the model adjustment parameters based on the detection performance, and optimize the vibration fault detection model based on the model adjustment parameters;

[0052] A configuration sub-module is used to determine the vibration detection link of the coal mine ventilation motor, determine the model deployment requirements according to the detection link, and configure the deployment environment based on the model deployment requirements;

[0053] A deployment and fault detection sub-module is used to deploy the optimized vibration fault detection model into the deployment environment to detect the vibration fault of the ventilation motor.

[0054] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structures specifically pointed out in the written specification and the drawings.

[0055] The technical solutions of the present invention will be further described in detail below through the drawings and embodiments. Description of the Drawings

[0056] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention.

[0057] Figure 1 It is a working flowchart of a method for detecting vibration faults of a coal mine ventilation motor based on deep learning provided by the present invention;

[0058] Figure 2 It is another working flowchart of a method for detecting vibration faults of a coal mine ventilation motor based on deep learning provided by the present invention;

[0059] Figure 3 It is a structural schematic diagram of a system for detecting vibration faults of a coal mine ventilation motor based on deep learning provided by the present invention;

[0060] Figure 4 It is a structural schematic diagram of an extraction module in a system for detecting vibration faults of a coal mine ventilation motor based on deep learning provided by the present invention. Detailed implementation manners

[0061] Here, exemplary embodiments will be described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0062] Currently, in the coal mining industry, ventilation motors are important equipment to ensure the safe and stable production of mines. However, due to the particularity of the coal mine environment, such as high temperature, high humidity, dust pollution, etc., ventilation motors are prone to various faults after long-term use. Among them, vibration faults are the most common. Such faults may lead to abnormal operation of the motor, performance degradation, and even major safety accidents. Therefore, timely and accurately detecting and diagnosing the vibration faults of ventilation motors plays a crucial role in ensuring the normal production of mines. However, traditional fault detection methods are generally carried out through manual inspections. Not only does this require a large amount of manpower and time, but although the auditory and visual methods are simple, they are limited by human subjective factors and are prone to misjudgment, resulting in low objectivity of the detection results and reducing the detection efficiency and accuracy. To solve the above problems, this embodiment discloses a method for detecting vibration faults of coal mine ventilation motors based on deep learning. To solve the above problems, this embodiment discloses a method for intelligent fault detection based on a deep learning model.

[0063] A method for detecting vibration faults of coal mine ventilation motors based on deep learning, as Figure 1 shown, includes the following steps:

[0064] Step S101: Collect multivariate time-series data of a coal mine ventilation motor and preprocess it, and extract the feature vectors of the preprocessed multivariate time-series data;

[0065] Step S102: Fuse and reconstruct the feature vectors, train a deep learning-based network model through the reconstructed features, and obtain a vibration fault detection model;

[0066] Step S103: Perform single-fault independent detection verification and multi-fault synchronous detection verification on the vibration fault detection model through vibration data samples of ventilation motors with different fault types, and obtain verification results;

[0067] Step S104: Determine the detection performance of the vibration fault detection model according to the verification results, optimize the vibration fault detection model based on the detection performance, and deploy and perform fault detection on the optimized vibration fault detection model.

[0068] The working principle of the above technical solution is as follows: collect the multi-source time-series data of the coal mine ventilation motor and preprocess it, and extract the feature vectors of the preprocessed multi-source time-series data; fuse and reconstruct the feature vectors, train the network model based on deep learning through the reconstructed features, and obtain the vibration fault detection model; perform single-fault independent detection verification and multi-fault synchronous detection verification on the vibration fault detection model through the vibration data samples of different fault types to obtain the verification results; determine the detection performance of the vibration fault detection model according to the verification results, optimize the vibration fault detection model based on the detection performance, and deploy and detect faults with the optimized vibration fault detection model.

[0069] The beneficial effects of the above technical solution are as follows: by constructing a vibration fault detection model, accurate model training can be carried out according to the data samples of vibration faults, and then high-precision and effective fault detection and determination can be carried out through the model, improving the determination accuracy. At the same time, no human intervention is required, and fault detection can be realized through the detection data of sensors, improving the detection accuracy and efficiency while ensuring the objectivity of the detection results. Further, aiming at the problem that it is difficult for the deep learning model to learn few-sample faults due to unbalanced fault data, VAE is used for data reconstruction to increase the fault samples, effectively improving the recognition accuracy of the subsequent model. At the same time, the feature fusion of the data enables the subsequent model to perform multi-fault recognition. It solves the problems in the prior art that vibration fault detection through manual inspection requires a large amount of manpower and time. Although the auditory and visual methods are simple, they are limited by human subjective factors and are prone to misjudgment, resulting in low objectivity of the detection results and reducing the detection efficiency and accuracy.

[0070] In one embodiment, as Figure 2 shown, the collection of the multi-source time-series data of the coal mine ventilation motor and its preprocessing, and the extraction of the feature vectors of the preprocessed multi-source time-series data include:

[0071] Step S201: Collect the multi-source time-series data of the coal mine ventilation motor through vibration sensors, temperature sensors and pressure sensors, and perform data cleaning and smoothing preprocessing on the multi-source time-series data;

[0072] Step S202: Convert the preprocessed multi-source time-series data into a unified format, and obtain the statistical features of the converted multi-source time-series data as the basic feature vectors;

[0073] Step S203: Obtain the high-order dimensional parameters of each type of signal in the multi-source time-series data and use them as advanced feature vectors;

[0074] Step S204: Integrate the basic feature vectors and the advanced feature vectors to generate the final feature vectors of the multi-source time-series data.

[0075] The beneficial effects of the above technical solution are as follows: By obtaining the multi-source time-series data of the coal mine ventilation motor and performing preprocessing, the quality and effectiveness of the data can be ensured. Further, by obtaining the basic feature vectors and advanced feature vectors of the multi-source time-series data and integrating them to generate the final feature vectors of the multi-source time-series data, richer and more comprehensive feature information can be provided. At the same time, it helps to better understand the operating state and behavior pattern of the equipment, discover the state changes and evolution trends of the coal mine ventilation motor at different time points, and thus provide more accurate fault prediction and maintenance suggestions.

[0076] In one embodiment, the fusion and reconstruction of the feature vectors, and the training of the deep learning-based network model through the reconstructed features to obtain the vibration fault detection model include:

[0077] Using a variational autoencoder model to fuse and reconstruct the feature vectors to obtain reconstructed features, and establishing a selective state space deep learning model as the recognition model;

[0078] Obtaining the recognition behavior strategy of the recognition model for the coal mine ventilation motor in the fault state by using the reconstructed features as the model input;

[0079] Obtaining vibration fault data with labels, converting the vibration fault data into a representation vector and performing low-dimensional mapping;

[0080] Training the recognition model according to the recognition behavior strategy and the mapped data to obtain the vibration fault detection model.

[0081] The beneficial effects of the above technical solution are as follows: By fusing and reconstructing the feature vectors, obtaining the recognition behavior strategy of the recognition model for the coal mine ventilation motor in the fault state, and training the model in combination with the mapped data of the vibration fault data, the generalization ability of the model can be improved, and the detection efficiency and accuracy of vibration faults can be improved.

[0082] In this embodiment, obtaining the recognition behavior strategy of the recognition model for the coal mine ventilation motor in the fault state by using the reconstructed features as the model input includes:

[0083] Obtaining the vibration signal corresponding to the reconstructed features, performing frequency band analysis on the vibration signal, and determining the time domain feature vectors and frequency domain feature vectors of each frequency band according to the analysis results;

[0084] Determining the vibration skewness factor, vibration kurtosis factor, and vibration valley factor corresponding to the fault state of the reconstructed features according to the time domain feature vectors and frequency domain feature vectors of each frequency band and the signal curve of the vibration signal;

[0085] Determining the recognition deviation response corresponding to the fault state of the reconstructed features based on the vibration skewness factor, vibration kurtosis factor, and vibration valley factor;

[0086] Determine the potential misjudged fault states corresponding to the reconstructed features according to the recognition deviation, and obtain the motor vibration mechanism of the potential misjudged fault states.

[0087] Determine the motor vibration frequency and vibration energy characteristics under the potential misjudged fault states according to the motor vibration mechanism, and determine the key monitoring parameters for the coal mine ventilation motor according to the motor vibration frequency and vibration energy characteristics.

[0088] Determine the synchronous monitoring time window and asynchronous monitoring time window of the key monitoring parameters, and respectively obtain the integrated operation characteristics under the synchronous monitoring time window and the cross-operation characteristics of the key monitoring parameters under the asynchronous monitoring time window.

[0089] Construct an operation characteristic matrix of the coal mine ventilation motor under potential misjudged faults for the integrated operation characteristics and cross-operation characteristics.

[0090] Determine the process change law of the key monitoring parameters of the coal mine ventilation motor under potential misjudged faults according to the operation characteristic matrix, and generate an interference recognition strategy for potential misjudged faults according to the process change law.

[0091] Obtain the recognition behavior strategy of the recognition model for the coal mine ventilation motor in the fault state corresponding to the reconstructed feature according to the interference recognition strategy of the potential misjudged fault and the shallow and deep recognition features of the fault state corresponding to the reconstructed feature.

[0092] The beneficial effects of the above technical solutions are as follows: By determining the interference recognition strategy of the potential misjudged faults corresponding to the reconstructed features, the occurrence of misrecognition of faults in the reconstructed features by the recognition model can be effectively avoided, ensuring the fault recognition accuracy and efficiency. Further, by obtaining the recognition behavior strategy of the recognition model for the coal mine ventilation motor in the fault state corresponding to the reconstructed feature according to the interference recognition strategy of the potential misjudged fault and the shallow and deep recognition features of the fault state corresponding to the reconstructed feature, a specific recognition strategy can be accurately formulated based on the deep recognition features of the fault state corresponding to the reconstructed feature, ensuring the accuracy and precision of the recognition, and improving the model performance and working stability.

[0093] In one embodiment, the vibration fault detection model is independently detected and verified for single faults and synchronously detected and verified for multiple faults by using the vibration data samples of ventilation motors of different fault types, and the verification results are obtained, including:

[0094] Respectively obtain the first energized fan vibration data samples of different fault types and the second energized fan vibration data samples in the non-fault state.

[0095] Input the first energized fan vibration data samples into the vibration fault detection model respectively, and confirm the single-fault independent detection accuracy according to the first model output results.

[0096] Randomly select multiple first-powered fan vibration data samples from the first-powered fan vibration data samples and input them into the vibration fault detection model at the same time, and confirm the multi-fault synchronous detection accuracy according to the output result of the second model;

[0097] Based on the single-fault independent detection accuracy and the multi-fault synchronous detection accuracy, confirm the verification result of the detection logic of the vibration fault detection model.

[0098] The beneficial effects of the above technical solution are as follows: confirm the single-fault independent detection accuracy according to the vibration data samples of the powered fan with different fault types, select multiple first-powered fan vibration data samples to confirm the multi-fault synchronous detection accuracy, and confirm the verification result of the detection logic of the vibration fault detection model, which can improve the detection accuracy of vibration faults, reduce the occurrence of false alarms and missed alarms. At the same time, the operation efficiency of the ventilation motor is improved, and the normal operation of the equipment can be better guaranteed.

[0099] In one embodiment, determine the detection performance of the vibration fault detection model according to the verification result, optimize the vibration fault detection model based on the detection performance, and deploy and detect faults for the optimized vibration fault detection model, including:

[0100] Determine the detection accuracy and detection integrity of the vibration fault detection model according to the verification result, and determine the detection performance according to the detection accuracy and detection integrity;

[0101] Determine the model adjustment parameters based on the detection performance, and optimize the vibration fault detection model based on the model adjustment parameters;

[0102] Determine the vibration detection link of the coal mine ventilation motor, determine the model deployment requirements according to the detection link, and configure the deployment environment based on the model deployment requirements;

[0103] Deploy the optimized vibration fault detection model to the deployment environment for vibration fault detection of the ventilation motor.

[0104] The beneficial effects of the above technical solution are as follows: determine the model adjustment parameters according to the detection performance and optimize the vibration fault detection model, which can improve the detection accuracy and efficiency of the model. Further, drive the model deployment requirements according to the detection link and configure the deployment environment, which can ensure the detection result of the model and quickly determine the vibration fault of the ventilation motor.

[0105] In one embodiment, this embodiment also discloses a coal mine ventilation motor vibration fault detection system based on deep learning, as Figure 3 shown, the system includes:

[0106] An extraction module 301, configured to collect multivariate time-series data of a coal mine ventilation motor, preprocess the data, and extract feature vectors of the preprocessed multivariate time-series data;

[0107] A training module 302, configured to fuse and reconstruct the feature vectors, train a deep learning-based network model through the reconstructed features, and obtain a vibration fault detection model;

[0108] A verification module 303, configured to perform single-fault independent detection verification and multi-fault synchronous detection verification on the vibration fault detection model through vibration data samples of ventilation motors with different fault types, and obtain verification results;

[0109] An optimization module 304, configured to determine the detection performance of the vibration fault detection model according to the verification results, optimize the vibration fault detection model based on the detection performance, and deploy and perform fault detection on the optimized vibration fault detection model.

[0110] The working principle and beneficial effects of the above technical solution have been described in the method embodiment, and will not be elaborated here.

[0111] In one embodiment, as Figure 4 shown, the extraction module 301 includes:

[0112] A preprocessing sub-module 3011, configured to collect multivariate time-series data of a coal mine ventilation motor through vibration sensors, temperature sensors, and pressure sensors, and perform data cleaning and smoothing preprocessing on the multivariate time-series data;

[0113] A first acquisition sub-module 3012, configured to convert the preprocessed multivariate time-series data into a unified format, and obtain statistical features of the converted multivariate time-series data as basic feature vectors;

[0114] A second acquisition sub-module 3013, configured to obtain high-order dimension parameters of each type of signal in the multivariate time-series data and use them as advanced feature vectors;

[0115] An integration sub-module 3014, configured to integrate the basic feature vectors and the advanced feature vectors to generate final feature vectors of the multivariate time-series data.

[0116] In one embodiment, the training module includes:

[0117] A feature fusion and reconstruction sub-module, configured to fuse and reconstruct the feature vectors using a variational autoencoder model, obtain reconstructed features, and establish a selective state space deep learning model as an identification model;

[0118] A third acquisition sub-module, configured to obtain the identification behavior strategy of the identification model for the coal mine ventilation motor in a fault state through the reconstructed features as model inputs;

[0119] A conversion sub-module, configured to obtain vibration fault data with tags, convert the vibration fault data into a representation vector, and perform low-dimensional mapping;

[0120] A training sub-module, configured to train an identification model according to an identification behavior strategy and mapping data, and obtain a vibration fault detection model.

[0121] In one embodiment, the verification module includes:

[0122] A fourth acquisition sub-module, configured to respectively obtain a first energized fan vibration data sample of different fault types and a second energized fan vibration data sample in a non-fault state;

[0123] A first confirmation sub-module, configured to respectively input the first energized fan vibration data sample into the vibration fault detection model, and confirm the single-fault independent detection accuracy according to the first model output result;

[0124] A second confirmation sub-module, configured to randomly select multiple first energized fan vibration data samples from the first energized fan vibration data samples and input them into the vibration fault detection model simultaneously, and confirm the multi-fault synchronous detection accuracy according to the second model output result;

[0125] A third confirmation sub-module, configured to confirm the verification result of the detection logic of the vibration fault detection model based on the single-fault independent detection accuracy and the multi-fault synchronous detection accuracy.

[0126] In one embodiment, the optimization module includes:

[0127] A determination sub-module, configured to determine the detection accuracy and detection integrity of the vibration fault detection model according to the verification result, and determine the detection performance according to the detection accuracy and detection integrity;

[0128] An optimization sub-module, configured to determine model adjustment parameters based on the detection performance, and optimize the vibration fault detection model based on the model adjustment parameters;

[0129] A configuration sub-module, configured to determine the vibration detection link of the coal mine ventilation motor, determine the model deployment requirements according to the detection link, and configure the deployment environment based on the model deployment requirements;

[0130] A deployment and fault detection sub-module, configured to deploy the optimized vibration fault detection model to the deployment environment to perform vibration fault detection on the ventilation motor.

[0131] Those skilled in the art should understand that the first and second in the present invention refer to different application stages.

[0132] Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include known common general knowledge or conventional technical means in the technical field not disclosed herein. The specification and examples are only to be considered as exemplary, and the true scope and spirit of the present disclosure are pointed out by the following claims.

[0133] It should be understood that the present disclosure is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is only limited by the appended claims.

Claims

1. A method for detecting vibration faults of coal mine ventilation motors based on deep learning, characterized in that: The following steps are involved: Collect and preprocess the multivariate time series data of coal mine ventilation motors, and extract the feature vector of the preprocessed multivariate time series data; The feature vectors are fused and reconstructed, and the deep learning-based network model is trained through the reconstructed features to obtain a vibration fault detection model; The vibration fault detection model is verified by single fault independent detection and multi-fault synchronous detection through ventilation motor vibration data samples of different fault types to obtain verification results; The detection performance of the vibration fault detection model is determined according to the verification results, the vibration fault detection model is optimized based on the detection performance, and the optimized vibration fault detection model is deployed and faults are detected.

2. The method for detecting vibration faults of coal mine ventilation motors based on deep learning according to claim 1 is characterized in that: The method of collecting multivariate time series data of the coal mine ventilation motor and preprocessing the data, and extracting the feature vector of the preprocessed multivariate time series data, includes: Vibration sensors, temperature sensors and pressure sensors are used to collect multivariate time series data of coal mine ventilation motors, and data cleaning and smoothing preprocessing are performed on the multivariate time series data. Convert the preprocessed multivariate time series data into a unified format, and obtain the statistical features of the converted multivariate time series data as basic feature vectors; Obtain high-order dimensional parameters of each type of signal in multivariate time series data and use them as advanced feature vectors; The basic eigenvectors and advanced eigenvectors are integrated to generate the final eigenvector of multivariate time series data.

3. The method for detecting vibration faults of coal mine ventilation motors based on deep learning according to claim 1 is characterized in that: The feature vectors are fused and reconstructed, and a network model based on deep learning is trained by reconstructing the features to obtain a vibration fault detection model, including: Use the variational autoencoder model to fuse and reconstruct the feature vectors, obtain the reconstructed features, and establish a selection state space deep learning model as the recognition model; The identification behavior strategy of the identification model for the coal mine ventilation motor under fault condition is obtained by reconstructing the features as the model input; Obtain vibration fault data with labels, convert the vibration fault data into representation vectors and perform low-dimensional mapping; The recognition model is trained according to the recognition behavior strategy and the mapping data to obtain a vibration fault detection model.

4. The method for detecting vibration faults of coal mine ventilation motors based on deep learning according to claim 1, characterized in that: The vibration fault detection model is verified by performing single fault independent detection and multi-fault synchronous detection on the vibration fault detection model through the ventilation motor vibration data samples of different fault types, and the verification results are obtained, including: Respectively obtain vibration data samples of the first powered-on fan in different fault types and the second powered-on fan in a non-fault state; Inputting the vibration data samples of the first powered-on fan into the vibration fault detection model respectively, and confirming the single fault independent detection accuracy according to the output result of the first model; Randomly select multiple first energized fan vibration data samples from the first energized fan vibration data samples and simultaneously input them into the vibration fault detection model, and confirm the multi-fault synchronous detection accuracy according to the second model output result; The verification results of the detection logic of the vibration fault detection model are confirmed based on the single fault independent detection accuracy and the multi-fault synchronous detection accuracy.

5. The method for detecting vibration faults of coal mine ventilation motors based on deep learning according to claim 1, characterized in that: The method of determining the detection performance of the vibration fault detection model according to the verification results, optimizing the vibration fault detection model based on the detection performance, and deploying and detecting faults on the optimized vibration fault detection model includes: Determine the detection accuracy and detection integrity of the vibration fault detection model according to the verification results, and determine the detection performance according to the detection accuracy and detection integrity; Determine model adjustment parameters based on detection performance, and optimize the vibration fault detection model based on the model adjustment parameters; Determine the vibration detection link of the coal mine ventilation motor, determine the model deployment requirements based on the detection link, and configure the deployment environment based on the model deployment requirements; The optimized vibration fault detection model is deployed in the deployment environment to perform vibration fault detection on ventilation motors.

6. A coal mine ventilation motor vibration fault detection system based on deep learning, characterized in that: The system includes: An extraction module is used to collect and preprocess the multivariate time series data of the coal mine ventilation motor, and extract the feature vector of the preprocessed multivariate time series data; A training module is used to fuse and reconstruct feature vectors, train a deep learning-based network model through reconstructed features, and obtain a vibration fault detection model; A verification module is used to perform single fault independent detection verification and multi-fault synchronous detection verification on the vibration fault detection model through ventilation motor vibration data samples of different fault types to obtain verification results; The optimization module is used to determine the detection performance of the vibration fault detection model according to the verification result, optimize the vibration fault detection model based on the detection performance, and deploy and detect faults with the optimized vibration fault detection model.

7. The deep learning-based coal mine ventilation motor vibration fault detection system according to claim 6 is characterized in that: The extraction module comprises: The preprocessing submodule is used to collect multivariate time series data of coal mine ventilation motors through vibration sensors, temperature sensors and pressure sensors, and perform data cleaning and smoothing preprocessing on the multivariate time series data; A first acquisition submodule is used to convert the preprocessed multivariate time series data into a unified format, and obtain statistical features of the converted multivariate time series data as basic feature vectors; The second acquisition submodule is used to obtain high-order dimensional parameters of various types of signals in the multivariate time series data and use them as advanced feature vectors; The integration submodule is used to integrate the basic feature vectors and the advanced feature vectors to generate the final feature vector of the multivariate time series data.

8. The method for detecting vibration faults of coal mine ventilation motors based on deep learning according to claim 6 is characterized in that: The training module comprises: The feature fusion and reconstruction submodule is used to fuse and reconstruct feature vectors using a variational autoencoder model, obtain reconstructed features, and establish a selection state space deep learning model as a recognition model; The third acquisition submodule is used to obtain the identification behavior strategy of the recognition model for the coal mine ventilation motor in a fault state by using the reconstructed features as the model input; A conversion submodule is used to obtain vibration fault data with labels, convert the vibration fault data into a representation vector and perform low-dimensional mapping; The training submodule is used to train the recognition model according to the recognition behavior strategy and mapping data to obtain the vibration fault detection model.

9. The deep learning-based coal mine ventilation motor vibration fault detection system according to claim 6 is characterized in that: The verification module comprises: A fourth acquisition submodule, used to respectively acquire vibration data samples of the first powered-on fan of different fault types and vibration data samples of the second powered-on fan in a non-fault state; A first confirmation submodule, used to input the vibration data samples of the first powered-on fan into the vibration fault detection model respectively, and confirm the single fault independent detection accuracy according to the output result of the first model; A second confirmation submodule is used to randomly select multiple first energized fan vibration data samples from the first energized fan vibration data samples and simultaneously input them into the vibration fault detection model, and confirm the multi-fault synchronous detection accuracy according to the second model output result; The third confirmation submodule is used to confirm the verification result of the detection logic of the vibration fault detection model based on the single fault independent detection accuracy and the multi-fault synchronous detection accuracy.

10. The deep learning-based coal mine ventilation motor vibration fault detection system according to claim 6, characterized in that: The optimization module comprises: A determination submodule is used to determine the detection accuracy and detection integrity of the vibration fault detection model according to the verification result, and determine the detection performance according to the detection accuracy and detection integrity; An optimization submodule, used for determining model adjustment parameters based on detection performance, and optimizing the vibration fault detection model based on the model adjustment parameters; The configuration submodule is used to determine the vibration detection link of the coal mine ventilation motor, determine the model deployment requirements according to the detection link, and configure the deployment environment based on the model deployment requirements; The deployment and fault detection submodule is used to deploy the optimized vibration fault detection model into the deployment environment to perform vibration fault detection of the ventilation motor.

Images