Mine mining electromechanical equipment fault intelligent diagnosis system and diagnosis method

Abstract

The invention discloses an intelligent fault diagnosis system and method for mine mining electromechanical equipment, and relates to the technical field of equipment fault diagnosis. Attention visualization processing is performed on an image by using a gradient activation mapping algorithm, so that a thermodynamic diagram of a hydraulic support is generated; feature extraction is conducted on the thermodynamic diagram, external risk factors of the hydraulic support are output, after operation state data of the hydraulic support are monitored through sensing equipment, internal risk factors of the hydraulic support are output through a linear regression analysis model, and operation risk coefficients of the hydraulic support are output in combination with the external risk factors and the internal risk factors; and after comprehensively analyzing the operation risk coefficients of all the hydraulic supports, predicting the mining risk of the monitoring area, and generating a corresponding management strategy. The diagnosis system combines the operation risk coefficients of all the hydraulic supports to generate mining risk prediction of the whole monitoring area, potential risks can be found in advance, preventive measures are taken, and the safety in the mine mining process is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment fault diagnosis, and in particular to an intelligent fault diagnosis system and method for mining electromechanical equipment. Background Art

[0002] In mining production, mining electromechanical equipment undertakes key tasks and directly affects the production efficiency and safety of the mine. These equipments operate in a complex and changeable underground environment for a long time and are easily affected by factors such as dust, high temperature, high humidity, impact and vibration, and have a high probability of failure. Hydraulic support is the core equipment for roof support of coal mining working face. It is mainly used to support and control the roof of coal mining working face to prevent roof collapse and ensure safe and efficient mining of coal mine. Since hydraulic support works in harsh environment of high pressure, high humidity, dust and strong impact for a long time, its hydraulic system, electronic control system, mechanical structure and other parts are prone to failure. Therefore, it is necessary to analyze the operating status of hydraulic support online through diagnostic system.

[0003] Traditional fault diagnosis methods usually only evaluate the operating status of a single hydraulic support, but fail to fully consider the coupling between supports and their impact on the overall mining system. In actual mining operations, hydraulic supports work together in groups, and their stability depends not only on individual operating conditions, but also on factors such as mechanical transmission between adjacent supports, changes in surrounding rock pressure, and fluctuations in mining conditions. Therefore, the abnormality of a single support may form a chain effect in a local area, thereby causing instability of the local support group and even affecting the safety of the entire working face. However, existing diagnostic methods are mainly based on isolated equipment status monitoring and lack the ability to comprehensively analyze grouped support systems. It is difficult to accurately predict regional-level mining risks, and thus it is impossible to provide forward-looking safety warnings and optimization and control strategies.

[0004] Based on this, the present application proposes an intelligent diagnosis system and method for faults of electromechanical equipment in mining, which combines the operating risk coefficients of all hydraulic supports to generate a mining risk forecast for the overall monitoring area, and can detect potential risks in advance and take preventive measures to effectively improve safety during mining. Summary of the invention

[0005] The purpose of the present invention is to provide an intelligent fault diagnosis system and method for mining electromechanical equipment to solve the shortcomings of the background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: an intelligent fault diagnosis method for mining electromechanical equipment, the diagnostic method comprising the following steps: Based on the mine management platform, the hydraulic support information in the monitoring area is obtained, and all hydraulic supports in the monitoring area are numbered; In the real-time mining process, after the image of the hydraulic support is acquired through the industrial camera, the gradient class activation mapping algorithm is used to perform attention visualization processing on the image to generate the heat map of the hydraulic support. The features of the heat map are extracted and the external risk factors of the hydraulic support are output. After the operating status data of the hydraulic support is monitored by the sensor equipment, the linear regression analysis model is used to output the internal risk factors of the hydraulic support. The operation risk coefficient of the hydraulic support is output by combining the external risk factors with the internal risk factors. After comprehensive analysis of the operation risk coefficients of all hydraulic supports, the mining risk in the monitored area is predicted and the corresponding management strategy is generated.

[0007] In a preferred embodiment, the external risk factor and the internal risk factor are combined to output the operation risk coefficient of the hydraulic support, and the expression is: , where is the operational risk factor, For external risk factors, As an intrinsic risk factor, , is the adjustment coefficient, and the adjustment coefficient is greater than 0.

[0008] In a preferred embodiment, after comprehensively analyzing the operating risk coefficients of all hydraulic supports, the mining risk of the monitored area is predicted, including the following steps: After obtaining the weight coefficient of the hydraulic support based on the relative importance of the hydraulic support, the risk index is obtained by weighting the operation risk coefficients of all the hydraulic supports in the monitoring area; The obtained risk index is compared with the preset index threshold, and the index threshold is used to predict the mining risk of the monitoring area; If the risk index is less than or equal to the index threshold, it is predicted that the mining risk in the monitoring area is small. If the risk index is greater than the index threshold, it is predicted that the mining risk in the monitoring area is high.

[0009] In a preferred embodiment, the risk index is obtained by weighted calculation of the operation risk coefficients of all hydraulic supports in the monitoring area, and the expression is: , where is the risk index, is the number of hydraulic supports in the monitoring area, For the The weight coefficient of a hydraulic support is For the The operation risk factor of a hydraulic support.

[0010] In a preferred embodiment, obtaining the weight coefficient of the hydraulic support based on the relative importance of the hydraulic support includes the following steps: The distance index and range excavation volume of the hydraulic support are obtained, and the distance index and range excavation volume are normalized so that the value range of the distance index and the range excavation volume is mapped to [0,1], and the normalized value of the distance index and the normalized value of the range excavation volume are obtained. The relative importance value of the hydraulic support is obtained by subtracting the normalized value of the distance index from the normalized value of the range excavation volume. The weight coefficient of each hydraulic support is calculated based on the relative importance value. The expression is: , where For the The weight coefficient of a hydraulic support is is the number of hydraulic supports in the monitoring area, For the The relative importance value of each hydraulic support.

[0011] In a preferred embodiment, feature extraction is performed on the thermal map and the external risk factors of the hydraulic support are output; In the thermal map, a square area of ​​the same size is generated with each oil leakage point as the center, and the oil leakage point is the center point of the square area. The square area is subdivided into several grids, and the number of grids occupied by oil stains in the square area is obtained. The number of grids occupied by oil stains is divided by the total number of grids in the square area to obtain the oil leakage ratio value of the square area, and the oil leakage ratio values ​​of all square areas are summed to obtain the oil leakage index of the hydraulic support; In the heat map, identify the number of peripheral pipelines in each warning area, sum the number of peripheral pipelines in all warning areas to obtain the total number, and divide the total number by the number of warning areas to obtain the pipeline index; The oil leakage index and pipeline index are normalized so that their value ranges are mapped to [0,1]. The normalized oil leakage index is added to the pipeline index to obtain the external risk factor of the hydraulic support.

[0012] In a preferred embodiment, after monitoring the operating status data of the hydraulic support by a sensor device, the inherent risk factor of the hydraulic support is outputted using a linear regression analysis model; The sensor equipment installed on the hydraulic support collects the operating status data, including the hydraulic system pressure, support displacement, temperature, flow rate, and vibration amplitude; Calculate and obtain the hydraulic system pressure deviation, temperature deviation, and flow deviation based on the operating status data; The hydraulic system pressure deviation, temperature deviation, flow deviation, bracket displacement and vibration amplitude are substituted as variables into the linear regression analysis model, and the linear regression analysis model outputs the intrinsic risk factors.

[0013] In a preferred embodiment, in the real-time mining process, after acquiring an image of the hydraulic support through an industrial camera, the image is subjected to attention visualization processing using a gradient class activation mapping algorithm to generate a heat map of the hydraulic support, including the following steps: The hydraulic support image is acquired through an industrial camera, and the monitoring structure of the hydraulic support is located using a target detection algorithm; Use deep convolutional neural network to extract features from hydraulic support images and extract the morphology, texture, edge, and damage feature information of the monitored structure; For abnormal targets of hydraulic supports, the gradient of the feature map of the last convolutional layer output by the deep convolutional neural network model is calculated. After mapping the attention of different structures through gradient weighting, a Grad-CAM map is generated and color-coded. The Grad-CAM map is fused with the original hydraulic support image to obtain the final thermal map.

[0014] In a preferred embodiment, the distance index acquisition logic is: obtain the hydraulic support position coordinates , and obtain the lane center coordinates of the monitoring area , Lane boundary coordinates And the coordinates of the turning area , calculate the distance between the hydraulic support position coordinates and the lane center coordinates, lane boundary coordinates and turning area coordinates, marked as , then the distance index .

[0015] The present invention also provides an intelligent fault diagnosis system for mining electromechanical equipment, comprising an information acquisition module, an external risk analysis module, an internal risk analysis module, and a prediction module; Information acquisition module: obtains the hydraulic support information in the monitoring area based on the mine management platform, and numbers all the hydraulic supports in the monitoring area; External risk analysis module: During the real-time mining process, after obtaining the image of the hydraulic support through the industrial camera, the gradient class activation mapping algorithm is used to perform attention visualization processing on the image to generate a heat map of the hydraulic support, extract features from the heat map and output the external risk factors of the hydraulic support; Intrinsic risk analysis module: After monitoring the operating status data of the hydraulic support through the sensor equipment, the linear regression analysis model is used to output the intrinsic risk factors of the hydraulic support; Prediction module: Combine external risk factors and internal risk factors to output the operation risk coefficient of the hydraulic support. After comprehensive analysis of the operation risk coefficients of all hydraulic supports, predict the mining risk in the monitored area and generate corresponding management strategies.

[0016] In the above technical solution, the technical effects and advantages provided by the present invention are: The present invention uses the gradient class activation mapping algorithm to perform attention visualization processing on the image, thereby generating a heat map of the hydraulic support, extracting features from the heat map and outputting the external risk factors of the hydraulic support, and after monitoring the operating status data of the hydraulic support through the sensing equipment, using the linear regression analysis model to output the internal risk factors of the hydraulic support, combining the external risk factors with the internal risk factors to output the operating risk coefficient of the hydraulic support, and after comprehensively analyzing the operating risk coefficients of all hydraulic supports, predicting the mining risk of the monitoring area, and generating corresponding management strategies. The diagnostic system combines the operating risk coefficients of all hydraulic supports to generate a mining risk prediction for the overall monitoring area, which can detect potential risks in advance, take preventive measures, and effectively improve the safety of the mining process. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0018] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Example 1: Please refer to Figure 1 As shown, the present embodiment provides an intelligent fault diagnosis method for mining electromechanical equipment, the diagnosis method comprising the following steps: The diagnostic system obtains the hydraulic support information in the monitoring area based on the mine management platform. The hydraulic support information includes the number of hydraulic supports and the location of hydraulic supports, and numbers all hydraulic supports in the monitoring area. The hydraulic support monitoring system is connected through the mine management platform to automatically collect basic information of all hydraulic supports in the monitoring area, including support model, quantity, spatial distribution location, etc. According to the arrangement of hydraulic supports in the mining working face, a support numbering rule is established to ensure clear numbering logic and facilitate subsequent data association and analysis.

[0021] Based on the geological structure and mining plan of the mine, a three-dimensional spatial coordinate system is established, and the actual position information of the hydraulic support is mapped to the coordinate system. Combined with geological exploration data, key parameters such as formation pressure and surrounding rock characteristics of different structures are calibrated to provide a basis for the stability assessment of the hydraulic support.

[0022] In the real-time mining process, after the image of the hydraulic support is acquired through the industrial camera, the gradient class activation mapping algorithm (Grad-CAM) is used to perform attention visualization on the image to generate a heat map of the hydraulic support. In the heat map, red indicates the high-attention area of ​​the hydraulic support, yellow indicates the secondary-attention area of ​​the hydraulic support, and blue indicates the low-attention area of ​​the hydraulic support. Industrial cameras are deployed in the mining area to ensure real-time image acquisition of all hydraulic supports, and a fixed acquisition frequency is set to obtain clear and complete equipment status images. The original images are pre-processed by denoising, contrast enhancement, white balance adjustment, etc. to reduce the impact of ambient light and improve the recognizability of key parts of the hydraulic supports. Target detection algorithms (such as YOLO) are used to locate the monitoring structures of the hydraulic supports (cylinders, connectors, peripheral pipelines, etc.), and irrelevant background information is filtered out to improve computing efficiency.

[0023] Use deep convolutional neural network (CNN) to extract features from hydraulic support images, extracting structural morphology, texture, edge, damage and other feature information; For abnormal targets of hydraulic supports (such as oil leakage and stress risks of peripheral pipelines), the model outputs the gradient of the feature map of the last convolutional layer, maps the attention of different structures through gradient weighting, generates a Grad-CAM map, and performs color coding: 1. Red (high concern area): 1) Oil leakage point: Check whether there is any sign of oil leakage in the hydraulic cylinder, pipeline joints and other parts.

[0024] 2) External pipeline warning area: Identify whether the external pipeline has moved to a position where it may be clamped during the operation of the hydraulic support.

[0025] 2. Yellow (minor concern area): Indicates the load-bearing parts of the hydraulic support, such as the main beam and connection parts, indicating potential risk of wear or deformation.

[0026] 3. Blue (low concern area): represents the parts that are currently operating stably and have no obvious abnormalities.

[0027] The thermal map generated by Grad-CAM is fused with the original hydraulic support image to obtain the final thermal map, making the monitored structure intuitive and visible, which is conducive to improving the efficiency of manual inspection and automatic diagnosis. When the high-concern area (red area) exceeds the set threshold, or the external pipeline enters the warning area, the system will automatically trigger an alarm and push it to the mine management platform for timely processing.

[0028] The hydraulic support image is acquired through an industrial camera, and the monitoring structure (cylinder, connector, peripheral pipeline, etc.) of the hydraulic support is located using the YOLO algorithm, including the following steps: Input images captured by industrial cameras, use the trained YOLO model to perform target detection, and output results, including: Bounding Box of each target structure; Category labels (cylinders, connections, peripheral pipelines, etc.).

[0029] The pre-training of the YOLO model consists of the following steps: Use annotation tools (such as LabelImg, Roboflow) to select the key components of the hydraulic support and generate a YOLO format annotation file (.txt) in the following format: <object_class><x_center><y_center> <width> <height>; Where: object_class is the class number (such as hydraulic support = 0, cylinder = 1, connector = 2). (x_center, y_center) is the normalized center coordinate of the object. (width, height) is the normalized width and height of the object.

[0030] Use YOLO weights pre-trained on common datasets such as COCO and ImageNet as initial model weights (such as yolov5s.pt and yolov8n.pt). Through transfer learning, reduce the dependence on a large amount of mining data and speed up the training convergence.

[0031] Configure training parameters: Learning rate: such as 0.001, and optimized using the cosine annealing decay strategy.

[0032] Batch size: Choose an appropriate size (such as 16 or 32) based on the GPU memory.

[0033] Optimizer: Use SGD or Adam for gradient optimization.

[0034] Loss functions: include target classification loss, bounding box regression loss, and confidence loss.

[0035] Number of training epochs: usually 50-300, depending on the data size and training effect.

[0036] The model training is as follows: python_train.py --img 640 --batch 16 --epochs 100 --data custom.yaml --weights yolov5s.pt --device 0 Among them: --img 640: input image size 640×640 pixels. --batch 16: 16 images per training batch. --epochs 100: training 100 epochs. --data custom.yaml: custom dataset configuration file. --weights yolov5s.pt: use pre-trained weights for transfer learning. --device 0: use GPU (CUDA 0) for acceleration.

[0037] Monitor classification loss, bounding box loss, and confidence loss, adjust learning rate and training strategy, use mAP (meanAverage Precision) to evaluate model performance, if the model is overfitting, increase data enhancement or use regularization (such as Dropout), if the model converges slowly, adjust the learning rate or use a deeper YOLO variant, test on data that has not participated in training, and calculate mAP, IoU and other indicators: python_val.py --weights best.pt --data custom.yaml --img 640; Deploy the model, take images of supports at the mine site for real-time detection, and observe the YOLO model's ability to identify oil leaks, pipeline anomalies, loose connections, and other situations.

[0038] The deep convolutional neural network (CNN) is used to extract features from hydraulic support images to extract structural morphology, texture, edge, damage and other feature information, including the following steps: Use labeling tools (such as LabelImg, VIA) to label the key areas in the image, including the various components of the hydraulic support (cylinders, pipelines, connectors, etc.) and possible faults (oil leaks, damaged areas, etc.), and adjust the image to a size suitable for ResNet, usually 224×224 or 256×256 pixels.

[0039] The preprocessed hydraulic support image is input into the ResNet model. ResNet extracts features of different levels in the image through multiple convolutional layers (Conv_Layers): Structural morphology: The basic shape and contour features of the image are extracted through low-layer convolution, such as the outline of the oil cylinder and the connection method of the pipeline.

[0040] Texture features: The convolution kernel in the middle layer extracts the detailed texture in the image, such as scratches and oil stains on the surface of the hydraulic support.

[0041] Edge information: ResNet obtains boundary information of hydraulic supports and their components through edge detection in multiple convolutional layers to help identify damage and component contact areas.

[0042] Damage features: High-level convolutional layers can capture complex damage features of hydraulic supports, such as cracks on the cylinder surface, deformation of connectors, or pressure points on peripheral pipelines.

[0043] For abnormal targets of hydraulic supports (such as oil leakage and stress risk of peripheral pipelines), the gradient of the deep convolutional neural network (CNN) model output for the feature map of the last convolution layer is calculated. After mapping the attention of different structures through gradient weighting, a Grad-CAM map is generated and color-coded, including the following steps: The prediction results of the target category are obtained through the output of the last layer of the CNN model, such as the oil leakage point of the hydraulic support, the stress of the peripheral pipeline, and other abnormalities. Calculate the gradient of the feature map of the last convolutional layer for a specific target class (e.g. "oil leak" or "pipeline stress risk") ,Right now: ,in, is the predicted score of the target category, It is the feature map of the kth channel of the last convolutional layer. For each channel k of the convolutional feature map, the gradient is pooled by global average pooling to obtain the average gradient of each channel. : , where is the convolutional feature map The value of position (i, j) in the convolutional feature map. H and W are the height and width of the convolutional feature map. is the average gradient value, i.e., the weight; For each channel k of the convolutional feature map, its feature map is weighted, and the weight is , this weighting process is performed by multiplying the average gradient value with the activation value in the feature map: , where is the weighted feature map value; By weighting the gradient value and the feature map, a two-dimensional weighted feature map, namely the Grad-CAM map, is generated, which reflects the contribution of each region to the target category. The generated Grad-CAM map is upsampled to make its size consistent with the original input image, which is convenient for comparison with the original image.

[0044] Color coding: Color-code the Grad-CAM graph, usually visualized using a heat map: Red indicates the areas that the model is most concerned about (such as oil leakage points of hydraulic supports and high-risk areas for pipeline stress).

[0045] Yellow indicates areas of minor concern.

[0046] Blue or green indicates areas of low concern.

[0047] Generate a heat map: The Grad-CAM map is fused with the original image to generate a color-coded heat map.

[0048] Superimpose the heat map on the original image to highlight the key areas of the hydraulic support, such as oil leaks and pipeline stress areas.

[0049] Visual output: Output images with thermal maps to help monitoring personnel intuitively see the risk areas of hydraulic supports.

[0050] Extract features from the thermal map and output the external risk factors of the hydraulic support; In the thermal map, a square area with the same size (length and width) is generated with each oil leakage point as the center. The oil leakage point is the center point of the square area. The square area is subdivided into several grids, and the number of grids occupied by oil stains in the square area is obtained (as long as there is oil stain in the grid, it is judged as occupied). The number of grids occupied by oil stains is divided by the total number of grids in the square area to obtain the oil leakage ratio value of the square area. The oil leakage ratio values ​​of all square areas are summed to obtain the oil leakage index of the hydraulic support; In the heat map, identify the number of external pipelines in each warning area. The warning area is the structural activity area of ​​the hydraulic support. When the external pipeline enters the warning area (it may be affected by equipment vibration or environmental vibration), there is a risk of being clamped. Sum the number of external pipelines in all warning areas to obtain the total number, and divide the total number by the number of warning areas to obtain the pipeline index; The oil leakage index and pipeline index are normalized so that their value ranges are mapped to [0,1]. The normalized oil leakage index is added to the pipeline index to obtain the external risk factor of the hydraulic support.

[0051] After monitoring the operating status data of the hydraulic support through the sensor equipment, the linear regression analysis model is used to output the inherent risk factors of the hydraulic support; The key operating status data is collected through various sensing devices installed on the hydraulic support (such as pressure sensors, displacement sensors, temperature sensors, flow sensors, vibration sensors, etc.), including but not limited to: Hydraulic system pressure: actual pressure changes in hydraulic cylinders and pipelines.

[0052] Bracket displacement: The change in position of the bracket after being subjected to force, to determine whether there is abnormal settlement.

[0053] Temperature: Changes in oil temperature and ambient temperature to avoid equipment failure caused by overheating or overcooling.

[0054] Flow rate: Hydraulic oil flow rate, to determine whether there is leakage or abnormal oil supply in the system.

[0055] Vibration Amplitude: Abnormal vibrations may indicate a loose connection or excessive impact forces.

[0056] The hydraulic system pressure deviation, temperature deviation, and flow deviation are calculated based on the operating status data. The parameter deviation is calculated using the general formula: Calculate and obtain, where is the parameter deviation, is the real-time parameter value, is the standard parameter value; Substituting the hydraulic system pressure deviation, temperature deviation, flow deviation, bracket displacement and vibration amplitude as variables into the linear regression analysis model, the model expression is: , where As an intrinsic risk factor, For each variable, is the intercept term, which indicates the basic risk when there is no influence of monitoring data. In this application, , is the regression coefficient of each variable, and the regression coefficient is greater than 0. Through linear regression analysis, the intrinsic risk factor is calculated using the monitoring data of the hydraulic support such as pressure, displacement, temperature, flow, and vibration. The larger the intrinsic risk factor, the more unstable the operation of the hydraulic support.

[0057] Combine external risk factors and internal risk factors to output the operation risk coefficient of hydraulic support; The larger the external risk factor, the greater the external operation risk of the hydraulic support predicted by analyzing the thermal map, and the larger the internal risk factor, the greater the overall operation failure risk of the hydraulic support; Combining external risk factors with internal risk factors, the operation risk coefficient of the hydraulic support is output, and the expression is: , where is the operational risk factor, For external risk factors, As an intrinsic risk factor, , is the adjustment coefficient, and the adjustment coefficient is greater than 0.

[0058] The greater the operation risk coefficient, the greater the risk of failure of the hydraulic support when it continues to operate, that is, the greater the impact on the safe construction in the monitored area.

[0059] After comprehensive analysis of the operational risk factors of all hydraulic supports, the mining risk in the monitored area is predicted and corresponding management strategies are generated; After obtaining the weight coefficient of the hydraulic support based on the relative importance of the hydraulic support, the risk index is obtained by weighted calculation of the operation risk coefficients of all hydraulic supports in the monitoring area. The expression is: , where is the risk index, is the number of hydraulic supports in the monitoring area, For the The weight coefficient of a hydraulic support is For the The operation risk factor of a hydraulic support.

[0060] The larger the risk index of the monitoring area, the greater the overall support failure risk of the monitoring area, that is, the greater the construction safety risk. The obtained risk index is compared with the preset index threshold. The index threshold is used to predict the mining risk of the monitoring area. If the risk index is less than or equal to the index threshold, the mining risk of the monitoring area is predicted to be small. If the risk index is greater than the index threshold, the mining risk of the monitoring area is predicted to be large.

[0061] The weight coefficient of the hydraulic support is obtained based on the relative importance of the hydraulic support, including the following steps: The distance index and range excavation volume of the hydraulic support are obtained, and the distance index and range excavation volume are normalized so that the value range of the distance index and the range excavation volume is mapped to [0,1], and the normalized value of the distance index and the normalized value of the range excavation volume are obtained. The relative importance value of the hydraulic support is obtained by subtracting the normalized value of the distance index from the normalized value of the range excavation volume. The weight coefficient of each hydraulic support is calculated based on the relative importance value. The expression is: , where For the The weight coefficient of a hydraulic support is is the number of hydraulic supports in the monitoring area, For the The relative importance value of each hydraulic support.

[0062] The logic of obtaining the distance index is: Get the position coordinates of the hydraulic support , and obtain the lane center coordinates of the monitoring area , Lane boundary coordinates And the coordinates of the turning area , calculate the distance between the hydraulic support position coordinates and the lane center coordinates, lane boundary coordinates and turning area coordinates, marked as , then the distance index The smaller the distance index of the hydraulic support, the closer the hydraulic support is to the core position of the monitoring area, and the greater the weight. The distance between the hydraulic support and each core position is obtained by the general Euclidean distance calculation formula: For example, we calculate the distance between the hydraulic support and the center of the lane, and the expression is: , where is the distance between the hydraulic support and the center of the roadway, is the position coordinate of the hydraulic support, is the center coordinate of the laneway. The calculation methods for the distance between the hydraulic support and the laneway boundary and the distance between the hydraulic support and the turning area are the same as above and will not be introduced one by one in this application.

[0063] The logic for obtaining the range mining volume is: with the hydraulic support as the center and R as the radius, a circular monitoring range is generated for the hydraulic support, and the mining volume in the circular monitoring range is counted as the range mining volume.

[0064] When the mining risk in the predicted monitoring area is small and the mining risk in the predicted monitoring area is large, the generated management strategies are as follows: 1) When the mining risk in the monitoring area is predicted to be low (low risk): When the calculated risk index is low, it indicates that the hydraulic supports in the monitoring area are operating stably as a whole and there are no obvious abnormalities in the mining area. The management strategies are as follows: Maintain normal production plans without additional adjustments to the mining process. Perform routine maintenance on hydraulic supports according to the established maintenance cycle to ensure long-term stable operation of the equipment. Combine historical data trends to continuously track the status of hydraulic supports to prevent the accumulation of potential faults. Adjust the pace of mining work to improve work efficiency while ensuring safety. If the risk factor of individual supports is slightly higher but does not reach the danger threshold, local enhanced inspections can be arranged, focusing on abnormal supports. According to the health status of the hydraulic supports, optimize equipment and personnel scheduling to ensure efficient operation.

[0065] 2) When the mining risk in the monitoring area is predicted to be high (high risk): When the calculated risk index is high, it indicates that there may be safety hazards in the hydraulic supports in the monitoring area, and active measures need to be taken to reduce the risk. The management strategies are as follows: Reduce the mining intensity or suspend operations in high-risk areas to avoid further expansion of the fault. Re-evaluate the status of the support and analyze whether there are problems such as overload operation of the support and uneven force between supports. If the risk area is small, optimize the construction sequence and adjust the force of the support to avoid local load concentration. Immediately dispatch an inspection team to conduct on-site inspections of high-risk supports to confirm the specific fault points. Conduct in-depth fault detection, including: Review the Grad-CAM heat map to identify whether the abnormal area continues to expand.

[0066] Read sensor data to determine whether pressure, displacement, vibration and other indicators exceed the limit.

[0067] Combined with linear regression prediction, determine whether the equipment has a trend of further deterioration.

[0068] If a serious failure risk is confirmed, worn or faulty parts such as hydraulic cylinders, seals, and pipes should be replaced in advance.

[0069] Issue safety alerts to notify on-site workers to be alert and make emergency preparations. Dispatch emergency plans, such as: Establish support reinforcement measures to avoid landslide accidents caused by support instability.

[0070] If the risk is extremely high, shut down the machine immediately for inspection to ensure the safety of the mining personnel.

[0071] Readjust the hydraulic system to prevent large-scale support instability due to abnormal pressure.

[0072] If high-risk areas appear repeatedly, it is necessary to consider improving the support design or optimizing the mining plan. Through big data analysis, we can find out the long-standing weak links in mining operations and make targeted improvements. We will provide special training for mine management personnel and maintenance teams to improve fault handling and prediction capabilities.

[0073] In summary: When the risk is low, the main measures are to optimize operation, local maintenance and intelligent prediction to ensure that the equipment remains stable and improve production efficiency. When the risk is high, emergency intervention, adjustment of mining plans, enhanced inspections, and even shutdowns are required to prevent the expansion of faults and ensure mining safety.

[0074] Through the above strategies, the mine diagnostic system can effectively reduce the equipment failure rate, reduce the occurrence of safety accidents, and ensure the efficient, safe and sustainable operation of mining operations.

[0075] This application uses the gradient class activation mapping algorithm to perform attention visualization processing on the image, thereby generating a heat map of the hydraulic support, extracting features from the heat map and outputting the external risk factors of the hydraulic support, and using the linear regression analysis model to output the internal risk factors of the hydraulic support after monitoring the operating status data of the hydraulic support through the sensor equipment. The operating risk coefficient of the hydraulic support is output by combining the external risk factor and the internal risk factor. After comprehensively analyzing the operating risk coefficients of all hydraulic supports, the mining risk of the monitored area is predicted and the corresponding management strategy is generated. The diagnostic system combines the operating risk coefficients of all hydraulic supports to generate a mining risk prediction for the overall monitoring area, which can detect potential risks in advance and take preventive measures to effectively improve the safety of the mining process.

[0076] Embodiment 2: The intelligent fault diagnosis system for mining electromechanical equipment described in this embodiment includes an information acquisition module, an external risk analysis module, an internal risk analysis module, and a prediction module; Information acquisition module: Based on the mine management platform, the hydraulic support information in the monitoring area is obtained. The hydraulic support information includes the number of hydraulic supports and the location of hydraulic supports. All hydraulic supports in the monitoring area are numbered and sent to the external risk analysis module, the internal risk analysis module and the prediction module. External risk analysis module: In the real-time mining process, after the image of the hydraulic support is obtained through the industrial camera, the gradient class activation mapping algorithm (Grad-CAM) is used to perform attention visualization processing on the image to generate a heat map of the hydraulic support. In the heat map, red indicates the high-attention area of ​​the hydraulic support, yellow indicates the secondary-attention area of ​​the hydraulic support, and blue indicates the low-attention area of ​​the hydraulic support. The heat map is feature extracted and the external risk factors of the hydraulic support are output. The external risk factors are sent to the prediction module; Intrinsic risk analysis module: After monitoring the operating status data of the hydraulic support through the sensor equipment, the linear regression analysis model is used to output the intrinsic risk factors of the hydraulic support, and the intrinsic risk factors are sent to the prediction module; Prediction module: Combine external risk factors and internal risk factors to output the operation risk coefficient of the hydraulic support. After comprehensive analysis of the operation risk coefficients of all hydraulic supports, predict the mining risk in the monitored area and generate corresponding management strategies.

[0077] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0078] It should be understood that the term "and / or" in this article is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the related objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0079] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0080] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0081] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.< / height> < / width>

Claims

1. An intelligent fault diagnosis method for mining electromechanical equipment, characterized in that: The diagnostic method comprises the following steps: Based on the mine management platform, the hydraulic support information in the monitoring area is obtained, and all hydraulic supports in the monitoring area are numbered; In the real-time mining process, after the image of the hydraulic support is acquired through the industrial camera, the gradient class activation mapping algorithm is used to perform attention visualization processing on the image to generate the heat map of the hydraulic support. The features of the heat map are extracted and the external risk factors of the hydraulic support are output. After the operating status data of the hydraulic support is monitored by the sensor equipment, the linear regression analysis model is used to output the internal risk factors of the hydraulic support. The operation risk coefficient of the hydraulic support is output by combining the external risk factors with the internal risk factors. After comprehensive analysis of the operation risk coefficients of all hydraulic supports, the mining risk in the monitored area is predicted and the corresponding management strategy is generated.

2. The intelligent fault diagnosis method for mining electromechanical equipment according to claim 1 is characterized by: Combining external risk factors with internal risk factors, the operation risk coefficient of the hydraulic support is output, and the expression is: , where is the operational risk factor, For external risk factors, As an intrinsic risk factor, , is the adjustment coefficient, and the adjustment coefficient is greater than 0.

3. The intelligent fault diagnosis method for mining electromechanical equipment according to claim 2 is characterized by: After comprehensive analysis of the operating risk factors of all hydraulic supports, the mining risk of the monitored area is predicted, including the following steps: After obtaining the weight coefficient of the hydraulic support based on the relative importance of the hydraulic support, the risk index is obtained by weighting the operation risk coefficients of all the hydraulic supports in the monitoring area; The obtained risk index is compared with a preset index threshold, and the index threshold is used to predict the mining risk in the monitoring area; If the risk index is less than or equal to the index threshold, it is predicted that the mining risk in the monitoring area is small. If the risk index is greater than the index threshold, it is predicted that the mining risk in the monitoring area is high.

4. The intelligent fault diagnosis method for mining electromechanical equipment according to claim 3 is characterized by: The risk index is obtained by weighting the operating risk coefficients of all hydraulic supports in the monitoring area. The expression is: , where is the risk index, is the number of hydraulic supports in the monitoring area, For the The weight coefficient of a hydraulic support is For the The operation risk factor of a hydraulic support.

5. The intelligent fault diagnosis method for mining electromechanical equipment according to claim 4 is characterized in that: The weight coefficient of the hydraulic support is obtained based on the relative importance of the hydraulic support, including the following steps: The distance index and range excavation volume of the hydraulic support are obtained, and the distance index and range excavation volume are normalized so that the value range of the distance index and the range excavation volume is mapped to [0,1], and the normalized value of the distance index and the normalized value of the range excavation volume are obtained. The relative importance value of the hydraulic support is obtained by subtracting the normalized value of the distance index from the normalized value of the range excavation volume. The weight coefficient of each hydraulic support is calculated based on the relative importance value. The expression is: , where For the The weight coefficient of a hydraulic support is is the number of hydraulic supports in the monitoring area, For the The relative importance value of each hydraulic support.

6. The intelligent fault diagnosis method for mining electromechanical equipment according to claim 2 is characterized by: Extract features from the thermal map and output the external risk factors of the hydraulic support; In the thermal map, a square area of ​​the same size is generated with each oil leakage point as the center, and the oil leakage point is the center point of the square area. The square area is subdivided into several grids, and the number of grids occupied by oil stains in the square area is obtained. The number of grids occupied by oil stains is divided by the total number of grids in the square area to obtain the oil leakage ratio value of the square area, and the oil leakage ratio values ​​of all square areas are summed to obtain the oil leakage index of the hydraulic support; In the heat map, identify the number of peripheral pipelines in each warning area, sum the number of peripheral pipelines in all warning areas to obtain the total number, and divide the total number by the number of warning areas to obtain the pipeline index; The oil leakage index and pipeline index are normalized so that their value ranges are mapped to [0,1]. The normalized oil leakage index is added to the pipeline index to obtain the external risk factor of the hydraulic support.

7. The intelligent fault diagnosis method for mining electromechanical equipment according to claim 6 is characterized by: After monitoring the operating status data of the hydraulic support through the sensor equipment, the linear regression analysis model is used to output the inherent risk factors of the hydraulic support; The sensor equipment installed on the hydraulic support collects the operating status data, including the hydraulic system pressure, support displacement, temperature, flow rate, and vibration amplitude; Calculate and obtain the hydraulic system pressure deviation, temperature deviation, and flow deviation based on the operating status data; The hydraulic system pressure deviation, temperature deviation, flow deviation, bracket displacement and vibration amplitude are substituted as variables into the linear regression analysis model, and the linear regression analysis model outputs the intrinsic risk factors.

8. The intelligent fault diagnosis method for mining electromechanical equipment according to claim 7 is characterized by: In the real-time mining process, after the image of the hydraulic support is acquired through the industrial camera, the gradient class activation mapping algorithm is used to perform attention visualization on the image to generate the heat map of the hydraulic support, including the following steps: The hydraulic support image is acquired through an industrial camera, and the monitoring structure of the hydraulic support is located using a target detection algorithm; Use deep convolutional neural network to extract features from hydraulic support images and extract the morphology, texture, edge, and damage feature information of the monitored structure; For abnormal targets of hydraulic supports, the gradient of the feature map of the last convolutional layer output by the deep convolutional neural network model is calculated. After mapping the attention of different structures through gradient weighting, a Grad-CAM map is generated and color-coded. The Grad-CAM map is fused with the original hydraulic support image to obtain the final thermal map.

9. The intelligent fault diagnosis method for mining electromechanical equipment according to claim 5, characterized in that: The distance index acquisition logic is: Get the hydraulic support position coordinates , and obtain the lane center coordinates of the monitoring area , Lane boundary coordinates And the turning area coordinates , calculate the distance between the hydraulic support position coordinates and the lane center coordinates, lane boundary coordinates and turning area coordinates, marked as , then the distance index .

10. An intelligent fault diagnosis system for mining electromechanical equipment, used to implement the diagnosis method according to any one of claims 1 to 9, characterized in that: It includes information acquisition module, external risk analysis module, internal risk analysis module and prediction module; Information acquisition module: obtains the hydraulic support information in the monitoring area based on the mine management platform, and numbers all the hydraulic supports in the monitoring area; External risk analysis module: During the real-time mining process, after obtaining the image of the hydraulic support through the industrial camera, the gradient class activation mapping algorithm is used to perform attention visualization processing on the image to generate a heat map of the hydraulic support, extract features from the heat map and output the external risk factors of the hydraulic support; Intrinsic risk analysis module: After monitoring the operating status data of the hydraulic support through the sensor equipment, the linear regression analysis model is used to output the intrinsic risk factors of the hydraulic support; Prediction module: Combine external risk factors and internal risk factors to output the operation risk coefficient of the hydraulic support. After comprehensive analysis of the operation risk coefficients of all hydraulic supports, predict the mining risk in the monitored area and generate corresponding management strategies.

Images