Cloud side-end collaborative coal mine transportation system state monitoring method and model

Abstract

The invention relates to the technical field of coal mine transportation system state monitoring, in particular to a cloud side-end collaborative coal mine transportation system state monitoring method and a cloud side-end collaborative coal mine transportation system state monitoring model. In order to solve the problems of difficult multi-modal data fusion, poor dynamic response and insufficient model generalization of a coal mine transportation system, the invention provides a new cloud side-end collaborative coal mine transportation system state monitoring method. Comprising the following steps of multi-modal data acquisition, multi-modal data synchronous transmission, edge computing node data compression, lightweight sensitive feature extraction and transmission, federal learning cross-mine data collaboration, cloud teacher large model construction, knowledge distillation driven student model construction and diagnosis early warning. According to the method, knowledge distillation and federal learning are fused, so that real-time monitoring, accurate diagnosis and efficient maintenance of the coal mine transportation system are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mine transportation system status monitoring, and specifically to a method and model for coal mine transportation system status monitoring with cloud-edge-end collaboration. Background Art

[0002] The coal mine transportation system is a key link in mine production, and its operating status directly affects the production efficiency and safety of coal. However, the core equipment of the coal mine transportation system (such as belt conveyors, scraper conveyors, and hoists) operates in a harsh environment of high humidity, high dust, strong noise, and heavy loads for a long time, and is extremely prone to faults such as conveyor belt tearing, chain breaking, and bearing overheating, resulting in a decrease in production efficiency and even safety accidents. Therefore, the monitoring and maintenance of the coal mine transportation system are crucial, and the complexity of the underground mine environment poses a huge challenge to the monitoring and maintenance of the coal mine transportation system.

[0003] Traditional manual inspection and regular maintenance methods are not only inefficient, but also have large monitoring blind spots, rely on experience for diagnosis, and have the defect of lagging response, making it difficult to accurately capture early fault signals of equipment. With the rapid development of Internet of Things, big data, and artificial intelligence technologies, building an intelligent operation and maintenance model has become an effective way to solve this problem. Existing large general models have shown excellent performance and broad application potential in a wide range of information processing and interaction tasks, such as general situation recognition, natural language question answering, etc. However, in the face of specific professional fields, especially the complex and highly specialized application scenario of the coal mine transportation system, their effectiveness has been significantly limited.

[0004] The coal mine transportation system is a complex system integrating multiple disciplines such as machinery, electricity, geology, and safety. Its operation and maintenance management not only require a high degree of professionalism and accuracy, but also need to have the ability to respond immediately to complex working conditions. Therefore, building an intelligent model for the coal mine transportation system needs to face many technical bottlenecks, such as: First, the limitations of general large models do not match the diverse data processing requirements of coal mine transportation systems. Currently, generative general large models have achieved remarkable results in the scope of information processing and interaction tasks, capable of efficiently extracting key features and quickly and accurately answering various questions. However, when facing the field of safety monitoring and intelligent early warning of coal mine transportation systems, which combines complexity and professionalism, their response effects are difficult to meet the actual needs. The coal mine transportation system involves the coordinated operation of numerous complex devices, and the working conditions are harsh and changeable. At present, most mainstream intelligent operation and maintenance systems build diagnostic models based on single-sensor data, making it difficult to comprehensively capture and analyze the unique fault patterns, operating rules, and complex environmental factors of coal mine transportation systems, resulting in inaccurate diagnoses and untimely early warnings in actual applications, which in turn affect the stability and safety of coal mine transportation systems. In view of this, there is an urgent need to develop a large model that can integrate multiple data types and effectively handle various abnormal phenomena to meet the complex requirements of coal mine transportation systems.

[0005] Second, there are obvious deficiencies in the efficiency and accuracy of large-scale high-dimensional data analysis. The real-time status monitoring data of coal mine transportation systems is huge in scale and high in dimension, including but not limited to various types such as motor current, equipment vibration, sound, temperature, video, environmental parameters, and operating speed. It is often accompanied by noise pollution and redundancy problems, resulting in a large amount of data, occupying a lot of storage space, and incurring huge storage costs, seriously affecting data transmission efficiency. In addition, traditional methods that solely rely on preset thresholds for data judgment often ignore the complex and changeable patterns and correlations behind the data, making it difficult to capture potential and valuable information. Moreover, the setting of thresholds is often based on manual experience or historical data, lacking dynamic adaptability and flexibility, and it is difficult to cope with the changing data environment and business requirements. When facing the massive data of underground coal mine transportation systems, the efficiency of existing data preprocessing and feature mining methods is extremely low, making it difficult to meet the needs of real-time fault diagnosis and early warning, and may lead to the failure to detect and handle faults in a timely manner, thus triggering greater safety risks.

[0006] Thirdly, the isolated monitoring data of the coal mine transportation system restricts the fault diagnosis accuracy and model generalization ability. Due to the complexity and particularity of the coal mine operation environment, it is often difficult to obtain sufficient and high-quality equipment operation data for fault monitoring and analysis. There are differences in the details of the monitoring data of different coal mine transportation systems, but there are also many commonalities among them. However, at the current stage, under the multiple constraints of privacy protection requirements, lack of industry norms, and enterprise competition considerations, the data of each coal mine has not yet achieved the goal of sharing. The intelligent solutions of each mine often build diagnostic models for specific equipment and fault types. When new equipment is added or new fault types occur in the system, the original model will not be able to effectively diagnose and predict. To address this situation, it is usually necessary to collect data again, build a model, and conduct training. This process is not only time-consuming and laborious but also requires the participation of professional technical personnel, resulting in high operation and maintenance costs. The generalization ability of the model has become one of the key factors restricting the application of intelligent solutions. Summary of the Invention

[0007] In order to solve the problems of difficult multi-modal data fusion, poor dynamic response, and insufficient model generalization in the coal mine transportation system, the present invention provides a new method and model for state monitoring of the coal mine transportation system with cloud-edge-end collaboration.

[0008] The present invention is implemented by the following technical solutions: A method for state monitoring of a coal mine transportation system with cloud-edge-end collaboration includes the following steps: 1) Multi-modal data acquisition: Deploy a variety of sensors for real-time acquisition of multi-modal data of the coal mine transportation system at key parts of the core equipment in multiple mines; 2) Multi-modal data synchronous transmission: The multi-modal data collected by a variety of sensors achieves timestamp alignment through the 5G network clock synchronization protocol to ensure the spatio-temporal consistency of the multi-modal data, and then the multi-modal data is transmitted to the edge computing node; 3) Edge computing node data compression: Compress the multi-modal data at the edge computing node; 4) Lightweight sensitive feature extraction and transmission: Extract lightweight sensitive features from various types of data; 5) Federated learning cross-mine data collaboration: Build a federated learning framework for the coal mine transportation system across mines. Through encryption technology and multi-party secure computing protocols, each mine uploads the encrypted multi-modal lightweight sensitive features to the cloud through the 5G network for interaction, realizing the cross-mine secure collaboration of multi-modal lightweight sensitive features under privacy protection; 6) Build a cloud-based teacher large model: Build a multi-modal Transformer architecture teacher large model in the cloud. After receiving the encrypted multi-modal lightweight sensitive features, the cloud determines the weights based on the data volume and data quality factors of each mine to ensure that the teacher large model can fully absorb the effective features in the data of each mine, comprehensively learn the features of multi-modal data, and train and test the multi-modal Transformer architecture teacher large model; 7) Student models driven by knowledge distillation: Use knowledge distillation technology in the cloud to transfer the knowledge of the trained teacher large model to multiple targeted lightweight student models according to the data characteristics and diagnostic tasks of each mine for subsequent anomaly detection; 8) Diagnostic warning: The cloud uses the optimized lightweight student models with high adaptability to perform anomaly detection on the real-time collected data of the coal mine transportation systems of each mine, and gives the diagnostic results of equipment status, anomaly type, and fault degree to the corresponding centralized control system.

[0009] Further, the sensors in step 1) include various sensors configured on the belt conveyor, various sensors installed on the scraper conveyor, and various sensors carried on the hoist.

[0010] Further, in step 1), the various sensors configured on the belt conveyor include laser displacement sensors, infrared thermal imagers, vibration sensors, machine vision cameras, acoustic emission sensors, voltage and current sensors, and gas sensors; the various sensors installed on the scraper conveyor include strain gauges, vibration sensors, machine vision cameras, and gas sensors; the various sensors carried on the hoist include fiber Bragg grating sensors, acoustic array sensors, vibration sensors, and machine vision cameras.

[0011] Further, in step 2), when the multi-modal data realizes timestamp alignment through the 5G network clock synchronization protocol, the time error is controlled within 1 ms.

[0012] Further, in step 3), for image / video data, a non-uniform compression algorithm combining wavelet transform and sparse coding is used for compression, with a compression ratio of 10:1, retaining key texture information (such as cracks on the conveyor belt surface); vibration and acoustic data are compressed using the compressive sensing algorithm to achieve high signal reconstruction accuracy at a lower sampling rate; for environmental gas parameter data, an efficient compression algorithm combining principal component analysis and quantization coding is used for compression, precisely retaining key parameter information such as dangerous gas concentration and environmental temperature; for other discrete data, a hybrid compression strategy combining Huffman coding and run-length coding is used to achieve efficient compression for high-frequency occurrence items and continuous repeated items in the data.

[0013] Further, in step 4), the image data extracts the defect area features through a lightweight convolutional neural network model; the video data uses a lightweight deep learning model for spatio-temporal feature extraction to capture the static features in the video frames and analyze the dynamic information between video frames; the vibration signal captures abnormal fluctuations through joint time-domain and frequency-domain analysis; the acoustic signal uses a combination of Mel-frequency cepstral coefficients and convolutional recurrent networks to identify abnormal sound patterns; the environmental gas parameter data combines time series analysis to extract key features such as gas concentration change trends and periodic fluctuations; other discrete data uses a decision tree-based feature selection method to screen out the most influential feature variables for classification or prediction tasks.

[0014] Further, in step 5), the multi-modal lightweight sensitive features achieve cross-modal feature interaction through the self-attention mechanism, and the number of model parameters reaches the billion level.

[0015] Further, in step 6), the data for training the multi-modal Transformer architecture teacher large model includes the historical fault database, measured data, and cross-mine data of federated learning.

[0016] Further, in step 7), the student model adopts a CNN+GRU hybrid architecture, and the number of parameters is only 1% of that of the teacher large model. It uses a two-stage distillation strategy: 1) Response distillation: minimizing the difference in the output probability distributions of the student and the teacher large model; 2) Feature distillation: aligning the multi-modal feature maps of the intermediate layers of the teacher large model through attention-guided mean squared error loss.

[0017] Further, in step 8), the centralized control system displays the normal, abnormal, or faulty state of the equipment through dashboards and status indicator lights, and shows potential safety hazard areas through heat maps and warning signals.

[0018] Further, in step 8), based on the diagnostic warning results, the student model combines various factors such as the historical operation data of the coal mine transportation system, the law of fault development, and real-time working condition monitoring information to accurately predict the time, location, possible impact degree, and detailed remaining service life information of the fault occurrence of the coal mine transportation system under the current operating conditions, and transmits this information to the corresponding centralized control system for display. At the same time, a comprehensive maintenance plan is formulated based on the fault prediction results and transmitted to the corresponding centralized control system for display, facilitating the operators to immediately take measures to relieve abnormal symptoms, and feedbacking the maintenance plan to all links of equipment selection, installation and commissioning, daily maintenance, and operation procedure formulation to improve the predictive maintenance system and enhance the overall reliability and stability of the coal mine transportation system.

[0019] A cloud-edge-end collaborative state monitoring model for a coal mine transportation system, which is a model established based on the cloud-edge-end collaborative state monitoring method for a coal mine transportation system as described above.

[0020] Beneficial effects of the present invention: 1. Significantly improve the intelligent operation and maintenance level of the coal mine transportation system: The present invention integrates the cloud, edge, and terminal devices to construct a multi-modal data fusion large model architecture covering the coal mine transportation system. By integrating multi-source heterogeneous data such as time-series signals, images, videos, and discrete parameters, and using advanced technologies of large models combined with deep learning, the problems of alignment and complementary analysis of multi-source heterogeneous data in the coal mine transportation system are solved. It can efficiently capture complex fault modes, environmental risks, and personnel violations during the operation of coal mine transportation system equipment (belt conveyors, scraper conveyors, and hoists), avoiding the limitations of single-modal data, enhancing the rapidity, comprehensiveness, and accuracy of anomaly detection, enabling a more comprehensive understanding of the operation state of the coal mine transportation system, and ensuring the safe and stable operation of the coal mine transportation system.

[0021] 2. Reduce the cloud computing pressure and system response speed delay: Through knowledge distillation technology, the present invention migrates the knowledge ability of the originally large and computationally intensive complex teacher large model to a lightweight student model with a small volume, high computational efficiency, and high adaptability in the cloud. It can still accurately capture and diagnose key information in the characteristic data of each mine, ensuring the high accuracy and reliability of the diagnosis results. Moreover, it greatly reduces the computational burden on the cloud, enabling the cloud to handle massive mine characteristic data more easily, significantly shortening the time for data processing and result output, and enabling the system to respond more quickly to the real-time needs of the mine site.

[0022] 3. Achieve data collaboration for cross-mine coal mine transportation systems: The present invention designs a federated learning framework for data privacy protection of cross-mine coal mine transportation systems, and proposes a technical solution integrating knowledge distillation and federated learning. Through advanced encryption technologies and multi-party secure computing protocols, the security of monitoring data characteristics during the interaction process of uploading to the cloud server is ensured, breaking the limitations of data islands in each mine, and achieving cross-mine secure collaboration of monitoring data characteristics under privacy protection. The cloud server determines the weights based on the data volume and data quality factors of each mine, ensuring that the teacher large model can fully absorb and integrate the effective characteristics in the data of each mine, realizing the organic combination and continuous injection of cross-mine global knowledge and personalized characteristics of each mine in the large model database, and significantly improving the generalization and robustness of the teacher large model and the high-adaptability lightweight student model.

[0023] 4. Comprehensive and Fine Abnormality Diagnosis and Precise Maintenance Guidance Strategy: In the cloud, the present invention uses a highly adaptable and lightweight model to detect the data of the coal mine transportation system in each mine in real time and displays it through visualization means. It can combine historical data with real-time monitoring to accurately predict the details of abnormalities, formulate a comprehensive maintenance plan, take measures to relieve abnormal symptoms, and feedback to each link to strengthen the predictive maintenance system. Compared with traditional fault diagnosis methods, such as rule-based logical judgment or simple statistical analysis of a single data source, the present invention can provide more specific fault diagnosis suggestions for the personalized characteristics of the coal mine transportation system in each mine, provide scientific and reasonable maintenance strategies for maintenance personnel, thus significantly improving the accuracy and work efficiency of fault diagnosis, reducing maintenance costs and production losses caused by frequent shutdowns for maintenance, and providing solid technical support for the safe production and efficient operation of coal mine enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings herein are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present invention and, together with the specification, are used to explain the principles of the present invention.

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0026] Figure 1 It is a flowchart of the method for monitoring the state of the coal mine transportation system with cloud-edge-end collaboration according to the present invention; Figure 2 It is a structural diagram of the model for monitoring the state of the coal mine transportation system with cloud-edge-end collaboration according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] In order to be able to more clearly understand the above-mentioned objects, features, and advantages of the present invention, the following will further describe the solution of the present invention. It should be noted that, without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other.

[0028] In the description, it should be noted that the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. It should be noted that unless otherwise clearly specified and limited, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal communication of two components. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific situations.

[0029] Numerous specific details are set forth in the following description to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0030] The following specifically describes the embodiments of the present invention with reference to the accompanying drawings.

[0031] As Figure 1 and 2 shown, a method for monitoring the state of a coal mine transportation system with cloud-edge-end collaboration includes the following steps: 1) Multimodal data acquisition: Deploy a variety of sensors for real-time acquisition of multimodal data of the coal mine transportation system at key parts of multiple mine core devices, including laser displacement sensors, infrared thermal imagers, vibration sensors, machine vision cameras, acoustic emission sensors, voltage and current sensors, and gas sensors configured on belt conveyors; a variety of sensors installed on scraper conveyors include strain gauges, vibration sensors, machine vision cameras, and gas sensors; a variety of sensors carried on hoists include fiber Bragg grating sensors, acoustic array sensors, vibration sensors, and machine vision cameras.

[0032] 2) Multimodal data synchronous transmission: The multimodal data collected by a variety of sensors achieves timestamp alignment through the 5G network clock synchronization protocol to ensure the spatio-temporal consistency of the multimodal data, with the time error controlled within 1 ms, and then the multimodal data is transmitted to the edge computing node; 3) Edge computing node data compression: Compress the multimodal data at the edge computing node; for image / video data, use a non-uniform compression algorithm combining wavelet transform and sparse coding for compression, with a compression ratio of 10:1, retaining key texture information (such as cracks on the conveyor belt surface); use a compressive sensing algorithm for compressing vibration and acoustic data to achieve high signal reconstruction accuracy at a lower sampling rate; for environmental gas parameter data, use an efficient compression algorithm combining principal component analysis and quantization coding to accurately retain key parameter information such as dangerous gas concentration and environmental temperature; for other discrete data, use a hybrid compression strategy combining Huffman coding and run-length coding to achieve efficient compression for high-frequency occurrence items and continuous repeated items in the data; 4) Lightweight Sensitive Feature Extraction and Transmission: Extract lightweight sensitive features from various types of data; for image data, extract defect area features through a lightweight convolutional neural network model; for video data, use a lightweight deep learning model for spatio-temporal feature extraction to capture static features in video frames and analyze dynamic information between video frames; for vibration signals, capture abnormal fluctuations through joint time-domain and frequency-domain analysis; for acoustic signals, identify abnormal sound patterns by combining Mel Frequency Cepstral Coefficients and a convolutional recurrent network; for environmental gas parameter data, extract key features such as gas concentration change trends and periodic fluctuations by combining time series analysis; for other discrete data, use a decision tree-based feature selection method to screen out the most influential feature variables for classification or prediction tasks. 5) Federated Learning Cross-Mine Data Collaboration: Build a federated learning framework for the cross-mine coal transportation system. Through encryption technology and multi-party secure computing protocols, each mine uploads encrypted multi-modal lightweight sensitive features to the cloud via a 5G network for interaction, achieving cross-mine secure collaboration of multi-modal lightweight sensitive features under privacy protection; multi-modal lightweight sensitive features achieve cross-modal feature interaction through the self-attention mechanism, and the number of model parameters reaches the billion level. 6) Build a Cloud Teacher Large Model: Build a multi-modal Transformer architecture teacher large model in the cloud. After receiving the encrypted multi-modal lightweight sensitive features, the cloud determines the weights based on the data volume and data quality factors of each mine to ensure that the teacher large model can fully absorb the effective features in the data of each mine, comprehensively learn the features of multi-modal data, and train and test the multi-modal Transformer architecture teacher large model; the data used to train the multi-modal Transformer architecture teacher large model includes the historical fault database, measured data, and cross-mine data from federated learning. 7) Knowledge Distillation-Driven Student Model: In the cloud, use knowledge distillation technology to transfer the knowledge of the trained teacher large model to multiple targeted lightweight student models according to the data characteristics of each mine and the diagnostic tasks for subsequent anomaly detection. The student model uses a CNN+GRU hybrid architecture, and the number of parameters is only 1% of that of the teacher large model. Use a two-stage distillation strategy: 1) Response distillation: Minimize the difference in output probability distributions between the student and the teacher large model; 2) Feature distillation: Align the multi-modal feature maps of the intermediate layers of the teacher large model through attention-guided mean squared error loss. 8) Diagnostic warning: The cloud uses the optimized lightweight student models with high adaptability to perform anomaly detection on the real-time collected data of each coal mine transportation system in the mine, and gives the diagnostic results of equipment status, anomaly type, and fault degree to the corresponding centralized control system. The centralized control system displays the normal, abnormal, or faulty status of the equipment through dashboards and status indicator lights, and shows potential safety hazard areas through heat maps and warning signals. Based on the diagnostic warning results, the student model combines various factors such as the historical operation data of the coal mine transportation system, the law of fault development, and real-time working condition monitoring information to accurately predict the time, location, possible impact degree, and detailed remaining service life of the coal mine transportation system under the current operating conditions, and transmits this information to the corresponding centralized control system for display. At the same time, a comprehensive maintenance plan is formulated according to the fault prediction results and the maintenance plan is transmitted to the corresponding centralized control system for display, so that the operators can immediately take measures to relieve the abnormal symptoms, and feedback the maintenance plan to all links of equipment selection, installation and commissioning, daily maintenance, and operation procedure formulation to improve the predictive maintenance system, so as to enhance the overall reliability and stability of the coal mine transportation system.

[0033] A state monitoring model for a coal mine transportation system with cloud-edge-end collaboration, which is a model established based on the state monitoring method for a coal mine transportation system with cloud-edge-end collaboration as described above.

[0034] The above are only specific embodiments of the present invention, which enable those skilled in the art to understand or implement the present invention. Although the above embodiments have been described in detail, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the above embodiments, or equivalently replace some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the above embodiments, and they should all be covered by the protection scope of the claims.

Claims

1. A method for monitoring the state of a coal mine transportation system with cloud-edge-terminal collaboration, characterized in that, It includes the following steps: 1) Multi-modal data collection: Deploy a variety of sensors for real-time collection of multi-modal data of the coal mine transportation system at key parts of multiple mine core equipment; 2) Multi-modal data synchronous transmission: The multi-modal data collected by a variety of sensors achieve timestamp alignment through the 5G network clock synchronization protocol to ensure the spatio-temporal consistency of the multi-modal data, and then transmit the multi-modal data to the edge computing node; 3) Edge computing node data compression: Compress the multi-modal data at the edge computing node; 4) Lightweight sensitive feature extraction and transmission: Extract lightweight sensitive features from various types of data; 5) Federated learning cross-mine data collaboration: Build a federated learning framework for the cross-mine coal mine transportation system. Through encryption technology and multi-party secure computing protocol, each mine uploads the encrypted multi-modal lightweight sensitive features to the cloud through the 5G network for interaction, realizing the cross-mine secure collaboration of multi-modal lightweight sensitive features under privacy protection; 6) Build a cloud teacher large model: Build a multi-modal Transformer architecture teacher large model in the cloud. After receiving the encrypted multi-modal lightweight sensitive features, the cloud determines the weights according to the data volume and data quality factors of each mine to ensure that the teacher large model can fully absorb the effective features in the data of each mine, comprehensively learn the features of multi-modal data, and train and test the multi-modal Transformer architecture teacher large model; 7) Knowledge distillation-driven student model: Use knowledge distillation technology in the cloud to transfer the knowledge of the trained teacher large model to multiple targeted lightweight student models according to the data characteristics and diagnostic tasks of each mine for subsequent anomaly detection; 8) Diagnosis and early warning: The cloud uses the optimized lightweight student models with high adaptability to perform anomaly detection on the real-time collected data of the coal mine transportation system of each mine, and gives the diagnosis results of equipment status, anomaly type, and fault degree to the corresponding centralized control system.

2. The method for monitoring the state of the coal mine transportation system with cloud-edge-terminal collaboration according to claim 1, characterized in that, The sensors in step 1) include a variety of sensors configured on the belt conveyor, a variety of sensors installed on the scraper conveyor, and a variety of sensors carried on the hoist.

3. The method for monitoring the state of the coal mine transportation system with cloud-edge-terminal collaboration according to claim 2, wherein, In step 1), the variety of sensors configured on the belt conveyor include laser displacement sensors, infrared thermal imagers, vibration sensors, machine vision cameras, acoustic emission sensors, voltage and current sensors, and gas sensors; the variety of sensors installed on the scraper conveyor include strain gauges, vibration sensors, machine vision cameras, and gas sensors; the variety of sensors carried on the hoist include fiber Bragg grating sensors, acoustic array sensors, vibration sensors, and machine vision cameras.

4. The method for monitoring the state of the coal mine transportation system with cloud-edge-terminal collaboration according to claim 3, wherein, In step 2), the time error when the multi-modal data achieves timestamp alignment through the 5G network clock synchronization protocol is controlled within 1 ms.

5. The method for monitoring the state of the coal mine transportation system with cloud-edge-terminal collaboration according to claim 4, characterized in that, In step 3), for image / video data, a non-uniform compression algorithm combining wavelet transform and sparse coding is used for compression, with a compression ratio of 10:1, retaining key texture information; vibration and acoustic data are compressed using a compressive sensing algorithm to achieve high signal reconstruction accuracy at a low sampling rate; for environmental gas parameter data, an efficient compression algorithm combining principal component analysis and quantization coding is used for compression, precisely retaining key parameter information such as hazardous gas concentration and environmental temperature; for other discrete data, a hybrid compression strategy combining Huffman coding and run-length coding is employed to achieve efficient compression for high-frequency occurrence items and continuous repeated items in the data.

6. The method for monitoring the state of the coal mine transportation system with cloud-edge-terminal collaboration according to claim 5, wherein, In step 4), defect region features are extracted from image data through a lightweight convolutional neural network model; spatio-temporal features are extracted from video data using a lightweight deep learning model to capture static features in video frames and analyze dynamic information between video frames; abnormal fluctuations are captured from vibration signals through joint time-frequency domain analysis; abnormal sound patterns are identified from acoustic signals by combining Mel-frequency cepstral coefficients and a convolutional recurrent network; key features such as gas concentration change trends and periodic fluctuations are extracted from environmental gas parameter data by combining time series analysis; for other discrete data, a decision tree-based feature selection method is used to screen out the most influential feature variables for classification or prediction tasks.

7. The method for monitoring the state of the coal mine transportation system with cloud-edge-terminal collaboration according to claim 6, wherein, In step 5), cross-modal feature interaction of multi-modal lightweight sensitive features is achieved through a self-attention mechanism, and the number of model parameters reaches the billion level.

8. The method for monitoring the state of the coal mine transportation system with cloud-edge-terminal collaboration according to claim 7, wherein In step 7), the student model adopts a CNN+GRU hybrid architecture, with the number of parameters being only 1% of that of the teacher large model, and a two-stage distillation strategy is used: 1) Response distillation: minimizing the difference in output probability distributions between the student and the teacher large model; 2) Feature distillation: aligning the multi-modal feature maps of the intermediate layers of the teacher large model through an attention-guided mean square error loss.

9. The method for monitoring the state of the coal mine transportation system with cloud-edge-terminal collaboration according to claim 8, characterized in that, In step 8), based on the diagnosis and early warning results, the student model combines various factors such as the historical operation data of the coal mine transportation system, the law of fault development, and real-time working condition monitoring information to accurately predict the time, location, possible impact degree, and detailed remaining service life of the coal mine transportation system under the current operating conditions, and transmits this information to the corresponding centralized control system for display. At the same time, a comprehensive maintenance plan is formulated based on the fault prediction results and transmitted to the corresponding centralized control system for display.

10. A state monitoring model for a coal mine transportation system with cloud-edge-end collaboration, characterized in that, This model is established based on the cloud-edge-terminal collaborative state monitoring method for coal mine transportation systems described in any one of claims 1 to 9.

Images