Scraper conveyor, large coal automatic control system and method thereof

Abstract

The invention belongs to the technical field of automatic control of scraper conveyors, and particularly relates to a scraper conveyor, a large coal automatic control system and a method thereof.The method comprises the steps that a large coal range is pre-stored in a YoV5 model and used for judging feature information, captured in real time, of each piece of coal conveyed on the scraper conveyor; if it is judged that the coal size is lower than the large coal range, the scraper conveyor operates normally, if it is judged that the coal size is within the large coal range or exceeds the large coal range, the YoloV5 model sends a first action signal to the controller to make the controller respond to the first action signal to control the scraper conveyor to slow down, and if it is judged that the coal size exceeds the large coal range, the YoloV5 model sends a second action signal to the controller; and the YoV5 model sends a second action signal to the data analysis module, so that the YoV5 model responds to the second action signal to control the scraper conveyor to stop, and the scraper conveyor returns to a normal operation state after large coal blocks are processed without highly depending on manual operation, so that the safety, the reliability and the production efficiency of the whole production process are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of automatic control of scraper conveyors, and in particular relates to a scraper conveyor, a large-lump coal automatic control system and a method thereof. Background Art

[0002] In the coal mining industry, scraper conveyors are a crucial means of transport, widely used to transport coal mined underground to the surface or for processing. However, the varying sizes of coal, particularly large lumps, can easily cause conveyor overload, jamming, and even damage, severely impacting production efficiency and the safe operation of the equipment.

[0003] Traditional solutions mainly rely on manual monitoring of scraper conveyors. If there are large lumps of coal on the scraper conveyor, manual adjustments are required. Workers detect large lumps of coal through visual inspection or with the help of simple sensors, and then manually adjust the conveyor speed or start the crushing device to deal with it. Although these methods can alleviate the problems caused by large lumps of coal to a certain extent, these methods are highly dependent on manual operation, which not only increases labor intensity and safety risks, but also limits the efficiency improvement of the production process. Summary of the Invention

[0004] In view of this, the present invention provides a scraper conveyor, a large-lump coal automatic control system and a method thereof to solve the technical problems in the prior art of increased labor intensity and safety risks and low production efficiency when manually monitoring the scraper conveyor and manually adjusting the large-lump coal.

[0005] To achieve the above objectives, this application adopts the following scheme:

[0006] A scraper conveyor is provided with a controller for controlling the scraper conveyor, and also includes a large-lump coal automatic control system provided on the scraper conveyor, the large-lump coal automatic control system includes a data acquisition module, a data processing module running a YoloV5 model and a data analysis module for optimizing the YoloV5 model, the data acquisition module is communicatively connected to the data processing module, the data processing module is communicatively connected to the data analysis module and the controller, the data acquisition module is used to collect characteristic information of the coal transported on the scraper conveyor, and transmit the characteristic information to the data processing module, after the data processing module receives the characteristic information, According to the YoloV5 model, the feature information is judged and an action signal is generated. The data processing module sends the action signal to the controller. The controller responds to the action signal and adjusts the working state of the scraper conveyor. The data processing module sends the feature information and the corresponding action signal to the data analysis module. After receiving the feature information and the corresponding action signal, the data analysis module generates optimization information according to the feature information and the corresponding action signal. The data analysis module sends the optimization information to the data processing module. The data processing module corrects the YoloV5 model according to the optimization information.

[0007] Preferably, the data acquisition module includes at least three groups of high-definition cameras, infrared sensors and weight sensors, and the characteristic information includes image information, temperature information and weight information. The three groups of high-definition cameras are respectively arranged at the entrance, exit and middle section of the scraper conveyor, and are used to capture the image information of each piece of coal transported on the scraper conveyor in real time. The infrared sensor is used to detect the temperature information of each piece of coal transported on the scraper conveyor in real time, and the weight sensor is used to detect the weight information of each piece of coal transported on the scraper conveyor in real time.

[0008] Preferably, the data analysis module includes a database, a server and a user terminal communicatively connected to the controller, the database is used to receive and store the feature information and the corresponding action signal, the server is used to call the feature information and the corresponding action signal in the database, and generate the optimization information and analysis information based thereon, and the user terminal is used to query the analysis information.

[0009] A large lump coal automation control system is applied to a scraper conveyor and also includes a data acquisition module, a data processing module running a YoloV5 model, and a data analysis module for optimizing the YoloV5 model. The data acquisition module is communicatively connected to the data processing module, and the data processing module is communicatively connected to the data analysis module and the controller. The data acquisition module is used to collect characteristic information of the coal transported on the scraper conveyor and transmit the characteristic information to the data processing module. After receiving the characteristic information, the data processing module judges the characteristic information according to the YoloV5 model and generates an action signal. The data processing module sends the action signal to the controller. The controller responds to the action signal and adjusts the working state of the scraper conveyor. The data processing module sends the characteristic information and the corresponding action signal to the data analysis module. After receiving the characteristic information and the corresponding action signal, the data analysis module generates optimization information according to the characteristic information and the corresponding action signal. The data analysis module sends the optimization information to the data processing module. The data processing module corrects the YoloV5 model according to the optimization information.

[0010] Preferably, the data acquisition module includes at least three groups of high-definition cameras, infrared sensors and weight sensors, and the characteristic information includes image information, temperature information and weight information. The three groups of high-definition cameras are respectively arranged at the entrance, exit and middle section of the scraper conveyor, and are used to capture the image information of each piece of coal transported on the scraper conveyor in real time. The infrared sensor is used to detect the temperature information of each piece of coal transported on the scraper conveyor in real time, and the weight sensor is used to detect the weight information of each piece of coal transported on the scraper conveyor in real time.

[0011] Preferably, the data analysis module includes a database, a server and a user terminal communicatively connected to the controller, the database is used to receive and store the feature information and the corresponding action signal, the server is used to call the feature information and the corresponding action signal in the database, and generate the optimization information and analysis information based thereon, and the user terminal is used to query the analysis information.

[0012] A method for automatically controlling large coal lumps comprises the following steps:

[0013] S10. Pre-store the range of large coal blocks in the YoloV5 model;

[0014] S20. Real-time capture of characteristic information of each piece of coal transported on the scraper conveyor;

[0015] S30. The characteristic information is transmitted to the YoloV5 model. The YoloV5 model determines the coal size based on the characteristic information. If the coal size is determined to be lower than the large coal range, the scraper conveyor operates normally. If the coal size is determined to be within the large coal range or exceeds the large coal range, S40 is executed.

[0016] S40. If the coal size is determined to be within the range of large coal, the YoloV5 model sends a first action signal to the controller. The controller controls the scraper conveyor to decelerate in response to the first action signal. If the coal size is determined to be outside the range of large coal, the YoloV5 model sends a second action signal to the data analysis module. The controller controls the scraper conveyor to stop in response to the second action signal and processes the coal that exceeds the range of large coal.

[0017] S50. The scraper conveyor returns to normal operation.

[0018] Preferably, the method further comprises the following steps:

[0019] S41. Generate first optimization information based on the characteristic information corresponding to the size of the coal located in the large coal range and the first action signal;

[0020] S42. Optimize the YoloV5 model according to the first optimization information and optimize the judgment strategy.

[0021] Preferably, the method further comprises the following steps:

[0022] S43. Generate second optimization information based on the characteristic information corresponding to the coal size exceeding the range of the large coal and the second action signal;

[0023] S44. Optimize the YoloV5 model according to the second optimization information to optimize the judgment strategy.

[0024] Preferably, the step S40 further comprises: after the scraper conveyor stops, controlling the controller to perform a crushing operation on the large pieces of coal on the scraper conveyor that exceed the range of the large pieces of coal.

[0025] Preferably, the total loss function of the YoloV5 model is the weighted sum of the position loss function, the confidence loss function and the classification loss function, and the total loss function is expressed as:

[0026] L total =L loc +L conf +L cls

[0027] Where, L total is the total loss function, L loc is the position loss function, L conf is the confidence loss function, L cls is the classification loss function.

[0028] Preferably, the mathematical formula of the position loss function is expressed as:

[0029]

[0030] Where S is the number of grids into which the image is divided; B is the number of bounding boxes predicted for each grid; is an indicator variable, which is 1 when the jth bounding box in the i-th grid is responsible for predicting an object, and 0 otherwise; coord is the weight factor of position loss, which is used to adjust the impact of position loss; pred x , pred y The center coordinates of the predicted bounding box; true x , true y is the true bounding box center coordinate; pred w , pred h is the predicted bounding box width and height; true w , true h is the actual bounding box width and height;

[0031] The mathematical formula of the confidence loss function is expressed as:

[0032]

[0033] Where λ noobj ,λ cbj are the weights of confidence loss in the case of no object and with object respectively; BCE is the binary cross entropy loss function; IOU(pred ij ,true ij ) is the intersection-over-union ratio between the predicted bounding box and the true bounding box;

[0034] The mathematical formula of the classification loss function is expressed as:

[0035]

[0036] Where, is the predicted class probability vector; is the true category label vector.

[0037] In the above-mentioned large-lump coal automatic control system in the scraper conveyor, the data acquisition module is used to collect the characteristic information of the coal transported on the scraper conveyor and transmit the characteristic information to the data processing module. Since the data processing module is set at the source close to the data acquisition module (i.e., on the scraper conveyor), the data processing module relies on the YoloV5 model to quickly judge the characteristic information and generate an action signal, and the data processing module directly sends the action signal to the controller, thereby improving the speed at which the controller responds to the action signal, so as to adjust the working state of the scraper conveyor more quickly and handle abnormal conditions in time, that is, when large pieces of coal appear, stop or slow down in time to avoid damage or blockage of the scraper conveyor caused by large pieces of coal. There is no need to rely heavily on manual operation, thereby improving the safety, reliability and production efficiency of the entire production process. Subsequently, the data analysis module generates optimization information based on the characteristic information and its corresponding action signal, and the data processing module corrects the YoloV5 model based on the optimization information, thereby further improving the judgment efficiency of the YoloV5 model.

[0038] In the above-mentioned large-lump coal automatic control method, the large-lump coal range is first pre-stored in the YoloV5 model as a basis for judging the coal size, and the characteristic information of each piece of coal transported on the scraper conveyor is captured in real time to judge. If the coal size is judged to be lower than the large-lump coal range, the scraper conveyor operates normally. If the coal size is judged to be in the large-lump coal range or exceeds the large-lump coal range, the YoloV5 model sends a first action signal to the controller, so that it responds to the first action signal to control the scraper conveyor to slow down, so as to avoid damage to the scraper conveyor caused by large-lump coal. If the coal size is identified to exceed the large-lump coal range, the YoloV5 model sends a second action signal to the data analysis module, so that it responds to the second action signal to control the scraper conveyor to stop, so as to avoid blockage of the scraper conveyor caused by large-lump coal. After stopping or slowing down, after processing the large-lump coal, the scraper conveyor resumes normal operation without relying heavily on manual operation, thereby improving the safety, reliability and production efficiency of the entire production process and effectively preventing the occurrence of safety accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a schematic diagram of the module of the large lump coal automatic control system in the present invention.

[0040] Figure 2 The flowchart of the automatic control method of large coal in the present invention. DETAILED DESCRIPTION

[0041] To facilitate understanding of the present application, a more comprehensive description of the present application will be provided below with reference to the accompanying drawings. Preferred embodiments of the present application will also be described. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the disclosure of the present application.

[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein in the specification of this application are intended only to describe specific embodiments and are not intended to limit this application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0043] Please see Figure 1 This embodiment provides a scraper conveyor, which is provided with a controller for controlling the scraper conveyor, and also includes a large-lump coal automatic control system provided on the scraper conveyor, the large-lump coal automatic control system includes a data acquisition module, a data processing module running a YoloV5 model, and a data analysis module for optimizing the YoloV5 model, the data acquisition module is communicatively connected to the data processing module, the data processing module is communicatively connected to the data analysis module and the controller, the data acquisition module is used to collect characteristic information of the coal transported on the scraper conveyor, and transmit the characteristic information to the data processing module, and the data processing module receives the characteristic information Then, the feature information is judged according to the YoloV5 model and an action signal is generated. The data processing module sends the action signal to the controller. The controller responds to the action signal and adjusts the working state of the scraper conveyor. The data processing module sends the feature information and the corresponding action signal to the data analysis module. After receiving the feature information and the corresponding action signal, the data analysis module generates optimization information according to the feature information and the corresponding action signal. The data analysis module sends the optimization information to the data processing module. The data processing module corrects the YoloV5 model according to the optimization information.

[0044] In this embodiment, the data acquisition module is used to collect the characteristic information of the coal transported on the scraper conveyor in real time. The data acquisition module is arranged at the entrance, exit and middle section of the scraper conveyor to ensure that all areas on the conveyor belt can be fully covered; preferably, the data processing module is close to the source position where the data acquisition module collects the characteristic information. Since it does not need to be reported to the control center for centralized processing, it can effectively solve the high latency problem of traditional cloud computing when processing massive data generated by IoT devices, so that the data collected by the data acquisition module can be quickly judged and an action signal can be generated, and the scraper conveyor can be directly controlled to perform the corresponding working state according to the action signal. For example: when the coal transported on the scraper conveyor is large pieces of coal, the data acquisition module captures the characteristic information of the large pieces of coal, and the YoloV5 model running in the data processing module judges the coal size based on the characteristic information of the large pieces of coal. If it is judged that If the coal size is lower than the range of large coal, the scraper conveyor operates normally and does not generate the action signal. If it is determined that the coal size is within the range of large coal, the YoloV5 model sends a deceleration signal to the controller, and the controller controls the scraper conveyor to decelerate in response to the deceleration signal. If it is identified that the coal size exceeds the range of large coal, the YoloV5 model sends a stop signal to the controller, and the controller controls the scraper conveyor to stop in response to the stop signal. When the scraper conveyor decelerates or stops, the data processing module sends an alarm signal to the data analysis module. After receiving the alarm signal, the data analysis module notifies the operator to perform manual intervention and handle the abnormal situation in time (different feature information and its corresponding action signal). The action signal ensures rapid response and effective processing of abnormal conditions through the above-mentioned feedback mechanism, and can complete the entire process from detection to control within a few milliseconds. At the same time, the large-lump coal automatic control system reports abnormal situations to the data analysis module. The data analysis module stores the above-mentioned abnormalities and provides further analysis and processing for the operator, assisting the data processing module in generating optimization information and optimizing the YoloV5 model running in the data processing module. Specifically, the method for optimizing the YoloV5 model running in the data processing module is as follows: the data analysis module sends the optimization information to the data processing module, and the data processing module corrects the YoloV5 model according to the optimization information. Through the above-mentioned implementation process, damage or blockage of the conveyor caused by large-lump coal can be prevented, and there is no need to rely heavily on manual operation, thereby improving the safety, reliability and production efficiency of the entire production process. In addition, when necessary, the crushing device on the scraper conveyor can be started to crush the large-lump coal to improve the degree of automated control.

[0045] In the above embodiment, the YoloV5 model is run on the data processing module, including: first, using a high-definition camera to collect and label a large number of image samples containing large lumps of coal to construct a high-quality data set; then, training the YoloV5 model based on this data set, and applying data enhancement technology during the training process; finally, deploying the optimized model on the data processing module, processing the feature information collected by the data acquisition module in real time, quickly completing the judgment of large lumps of coal, and forming an action signal.

[0046] In the above implementation process, an edge computing node (i.e., the above-mentioned data processing module) is deployed near the source of feature information acquisition of the data acquisition module. The edge computing node relies on a high-performance processor (such as NVIDIA Jetson Xavier NX) to realize data processing and model inference, and is equipped with storage resources such as SSD to save data and models, and maintains an efficient connection with the data analysis module with the help of a 4G / 5G communication module. Its working mechanism covers key steps such as data collection, preliminary processing, intelligent decision-making, result feedback, and security and privacy protection. It can complete data processing near the place where the feature information is generated, thereby improving efficiency.

[0047] Furthermore, in order to form more comprehensive feature information, the data acquisition module includes at least three groups of high-definition cameras, infrared sensors and weight sensors. The feature information includes image information, temperature information and weight information. The three groups of high-definition cameras are respectively arranged at the entrance, exit and middle section of the scraper conveyor, for capturing the image information of each piece of coal transported on the scraper conveyor in real time, and transmitting the image information to the data processing module. The infrared sensor is used to detect the temperature information of each piece of coal transported on the scraper conveyor in real time, and transmit the temperature information to the data processing module. The weight sensor is used to detect the weight information of each piece of coal transported on the scraper conveyor in real time, and transmit the weight information to the data processing module. The temperature information, the weight information and the image information can be used as historical data for training the YoloV5 model.

[0048] Specifically, the resolution, frame rate, and placement of the HD camera are carefully selected to ensure that clear, high-quality images of the coal are captured. Maintaining a sufficient frame rate can enhance the capture of dynamic changes in the coal. For example, the resolution of the HD camera is usually set to 1080p to ensure the clarity of image details; the frame rate is set to 30fps to capture the dynamic changes of the coal on the high-speed conveyor belt. The HD camera captures the image of the coal on the conveyor belt to form image information and transmits the image information to the data processing module. The data processing module is located on-site and is relatively close to the HD camera. Such an arrangement can greatly shorten the delay time of the coal image transmission, ensure the real-time performance of the system, and improve efficiency. Similarly, the temperature information and weight information collected by the infrared sensor and the weight sensor can also be transmitted faster and respond more efficiently in this way.

[0049] Specifically, the data analysis module includes a database, a server and a user terminal communicated with the controller. The database is used to receive and store the feature information and the corresponding action signal, so as to organize it into historical data. The server is used to call the feature information and the corresponding action signal (i.e., historical data) in the database, and generate the optimization information and analysis information based on it, so as to formulate a global scheduling strategy. The user terminal is used to query the analysis information in real time, so as to provide users with real-time monitoring and manual intervention functions. The historical data stored in the database in the data analysis module is used to train the YoloV5 model. The YoloV5 model can regularly analyze historical data, summarize the time, location and frequency of large coal, etc., help operators understand the weak links in the production process, and optimize the strategy of the YoloV5 model based on this information to take preventive measures in advance.

[0050] A large lump coal automatic control system is applied to a scraper conveyor, characterized in that it also includes a data acquisition module, a data processing module running a YoloV5 model and a data analysis module for optimizing the YoloV5 model. The data acquisition module is communicatively connected to the data processing module, and the data processing module is communicatively connected to the data analysis module and the controller. The data acquisition module is used to collect characteristic information of the coal transported on the scraper conveyor and transmit the characteristic information to the data processing module. After receiving the characteristic information, the data processing module judges the characteristic information according to the YoloV5 model and generates an action signal. The data processing module sends the action signal to the The controller responds to the action signal and adjusts the working state of the scraper conveyor, the data processing module sends the feature information and the corresponding action signal to the data analysis module, and after receiving the feature information and the corresponding action signal, the data analysis module generates optimization information according to the feature information and the corresponding action signal. The data analysis module sends the optimization information to the data processing module, and the data processing module corrects the YoloV5 model according to the optimization information. In addition, the operator can also summarize the time distribution of the feature information and the corresponding action signal, formulate a global scheduling strategy, and manage the scraper conveyor according to the global scheduling strategy.

[0051] The above-mentioned large-lump coal automatic control system can be applied to scraper conveyors. By setting up a data acquisition module to collect characteristic information of the coal transported on the scraper conveyor, and transmitting the characteristic information to the data processing module, the YoloV5 model in the data processing module preprocesses and identifies the characteristic information to form an action signal. The controller decelerates and stops the scraper conveyor according to the action signal, which can not only prevent damage or blockage of the conveyor caused by large-lump coal, but also can combine the operator's statistics on time distribution and abnormal state distribution frequency through the data analysis module, so as to timely discover and solve potential safety hazards in the future, thereby improving the safety and reliability of the entire production process in the future, and improving production efficiency in disguise by reducing the failure rate.

[0052] The data analysis module includes a server, a database and a user interface. The server is used to receive the feature information and the corresponding action signal sent by the data processing module, and formulate a global scheduling strategy based on the feature information and the corresponding action signal to optimize and improve the efficiency of the scraper conveyor. The database is used to record historical data and YoloV5 model parameters, and use the historical data and YoloV5 model parameters to train the YoloV5 model. The user end provides real-time monitoring and manual intervention functions.

[0053] Please see Figure 2 This embodiment also provides a method for automatically controlling large coal lumps, comprising the following steps:

[0054] S10. The operator pre-stores the large coal range (ie, a set of feature information corresponding to the large coal size range) in the YoloV5 model in the above data processing module;

[0055] It should be noted that the range of large coal blocks pre-stored in the YoloV5 model is determined by the operator's experience, or the YoloV5 model is trained in advance using the characteristic information of normal coal and the characteristic information of large coal blocks to obtain the characteristic information of coal whose size meets the range of large coal blocks, which serves as the basis for the YoloV5 model to judge the range of large coal blocks. For example: a camera is used to take pictures of normal coal to obtain a normal coal sample set, and a camera is used to take pictures of large coal blocks to obtain a large coal sample set. Then, the normal coal sample set and the large coal sample set are used to train the YoloV5 model, and the trained YoloV5 model has the large coal threshold.

[0056] S20. The data acquisition module captures the characteristic information of each piece of coal transported on the scraper conveyor in real time;

[0057] S30. The data acquisition module transmits the characteristic information to the YoloV5 model. The YoloV5 model determines the coal size based on the characteristic information. If the coal size is determined to be lower than the large coal size range (i.e., the collected characteristic information is a set of characteristic information lower than the minimum value of the large coal size range), the scraper conveyor operates normally. If the coal size is determined to be within the large coal size range or exceeds the large coal size range (i.e., the collected characteristic information is a set of characteristic information within or exceeding the large coal size range), S40 is executed.

[0058] S40. If it is determined that the coal size is within the range of large coal (i.e., the collected characteristic information is within the set of characteristic information corresponding to the range of large coal size), the YoloV5 model sends a first action signal to the controller. In response to the first action signal, the controller controls the scraper conveyor to decelerate and reduce the speed of the scraper conveyor to reduce the impact of large coal on the conveyor and prevent damage to the equipment. If it is determined that the coal size exceeds the range of large coal (i.e., the collected characteristic information is within the set of characteristic information corresponding to the maximum value of the range of large coal size), the YoloV5 model sends a second action signal to the data analysis module. In response to the second action signal, the controller controls the scraper conveyor to stop and processes the coal whose size exceeds the range of large coal (manually processes or starts the crushing device equipped with the scraper conveyor to process) to prevent blockage or damage caused by large coal and ensure the safe operation of the equipment.

[0059] S50. The scraper conveyor returns to normal operation, which can be achieved by notifying the operator to perform a reset operation or automatically resetting by associating corresponding signals.

[0060] In the automated control method for large-lump coal provided in this embodiment, characteristic information of the coal transported on the scraper conveyor is captured in real time, and the YoloV5 model is used to judge the characteristic information. If it is judged to be a large lump of coal, the scraper conveyor is controlled by the controller to stop or slow down. This can not only prevent damage or blockage of the conveyor caused by large-lump coal, but also promptly discover and resolve possible safety hazards, thereby improving the safety and reliability of the entire production process and improving production efficiency.

[0061] In a further implementation process, the following steps are also included:

[0062] S41. Generate first optimization information based on the characteristic information corresponding to the size of the coal located in the large coal range and the first action signal;

[0063] S42. Optimize the YoloV5 model according to the first optimization information and optimize the judgment strategy.

[0064] Specifically, the characteristic information corresponding to the coal size within the range of the large coal and the corresponding first action signal are transmitted to the data analysis module, and the data analysis module optimizes the YoloV5 model accordingly, so that the response speed of the YoloV5 model in judging and processing when encountering large coal is improved. Furthermore, in special circumstances (situations where the YoloV5 model is difficult to handle and the operator needs to troubleshoot abnormalities), the operator's adjustment of the YoloV5 model can also improve the accuracy of the YoloV5 model in judging large coal.

[0065] In a further implementation process, the following steps are also included:

[0066] S43. Generate second optimization information based on the characteristic information corresponding to the coal size exceeding the range of the large coal and the second action signal;

[0067] S44. Optimize the YoloV5 model according to the second optimization information to optimize the judgment strategy.

[0068] Specifically, the characteristic information corresponding to the coal size that exceeds the range of large coal and the corresponding second action signal are transmitted to the data analysis module. The data analysis module optimizes the YoloV5 model accordingly, so that the response speed of the YoloV5 model in judging and processing coal that exceeds the size of large coal is improved. Furthermore, with the assistance of special circumstances (situations where the YoloV5 model is difficult to handle and the operator needs to troubleshoot abnormalities), the operator's adjustment of the YoloV5 model can also improve the accuracy of the YoloV5 model's judgment on coal that exceeds the size of large coal.

[0069] Furthermore, the step S40 further includes: after the scraper conveyor stops, controlling the controller to perform a crushing operation on the large pieces of coal on the scraper conveyor that exceed the range of the large pieces of coal.

[0070] In this embodiment, after the scraper conveyor stops operating, it indicates that the coal exceeds the range of large-lump coal, which may cause the scraper conveyor to be blocked or stuck, and the coal needs to be crushed. Specifically, a crushing device can be set on the scraper conveyor, and the crushing device is started to crush the large-lump coal.

[0071] Further explanation of the YoloV5 model described above:

[0072] The total loss function of the YoloV5 model is the weighted sum of the position loss function, the confidence loss function and the classification loss function, where in the position prediction, i.e., the bounding box prediction, each bounding box consists of 5 values: t x , t y , t w , t h and t O , where the first four values ​​are used to determine the position and size of the bounding box, and the last value represents the confidence level of the object. The coordinates of the center point of the bounding box (b x ,b y ) and width and height (b w ,b h ) can be calculated by the following formula:

[0073] b x =σ(t x )+c x

[0074] b y =σ(t y )+c y

[0075]

[0076] Where σ is the sigmoid function, c x , c y is the coordinate of the upper left corner of the grid cell, p w , p h are the width and height of the prior box (also called the anchor box).

[0077] In confidence prediction, the confidence of objectness t O It represents the probability that the bounding box contains an object. It is converted to a value between 0 and 1 through the sigmoid function, which represents the confidence of the prediction.

[0078] P obj =σ(t o )

[0079] In category prediction, the category probability is calculated using the softmax function to give the probability distribution of the bounding box belonging to each category. i is the unnormalized predicted value output by the network, P(c i |object) is the probability that an object belongs to category i given that there is an object inside the bounding box.

[0080]

[0081] YoloV5's loss function typically consists of several parts: location loss, confidence loss, and classification loss. Location loss is used to measure the difference between the position and size of the predicted bounding box and the true bounding box. Mean squared error (MSE) is used to calculate location loss. The mathematical formula is:

[0082]

[0083] Among them, S is the number of grids into which the image is divided; B is the number of bounding boxes predicted for each grid; is an indicator variable, which is 1 when the jth bounding box in the i-th grid is responsible for predicting an object, and 0 otherwise; coord is the weight factor of position loss, which is used to adjust the impact of position loss; pred x, pred y The center coordinates of the predicted bounding box; true x , true y is the true bounding box center coordinate; pred w , pred h is the predicted bounding box width and height; true w , true h is the actual bounding box width and height.

[0084] Confidence loss is used to evaluate the accuracy of the predicted bounding box's probability of containing an object. It is calculated using Binary Cross Entropy (BCE), and the mathematical formula is:

[0085]

[0086] Among them, λ noobj ,λ cbj are the weights of confidence loss in the case of no object and with object respectively; BCE is the binary cross entropy loss function; IOU(pred ij ,true ij ) is the Intersection over Union (IoU) between the predicted bounding box and the true bounding box.

[0087] Classification loss is used to evaluate whether the predicted category probability is correct. It is also calculated using cross entropy loss (CE), and the mathematical formula is:

[0088]

[0089] in, is the predicted class probability vector; is the true category label vector.

[0090] Therefore, the total loss is the weighted sum of the above three losses, and the mathematical formula is:

[0091] L total =L loc +L conf +L cls

[0092] The YoloV5 model provided in this application can directly predict the bounding box and category probability of the target from the original image (image information), significantly improving the detection speed. In addition, the model effectively enhances the detection ability of targets of different scales by introducing the feature pyramid network (FPN), especially for large pieces of coal of different sizes. The YoloV5 model also uses an anchor mechanism to pre-set a series of rectangular boxes with specific sizes and proportions, and then adjust the model to accurately locate the target boundary, thereby improving the detection accuracy. In order to enhance the generalization and robustness of the model, a variety of data enhancement strategies are implemented in the YoloV5 training stage, such as random cropping, rotation, flipping and color transformation, to cope with complex and changeable actual environments. Its loss function design takes into account the three dimensions of position, confidence and classification to ensure that the model can accurately locate and classify targets to improve the accuracy of large pieces of coal recognition.

[0093] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should all be included in the scope of protection of the present invention.

Claims

1. A scraper conveyor, provided with a controller for controlling the scraper conveyor, characterized in that: It also includes a large piece of coal automatic control system arranged on the scraper conveyor, the large piece of coal automatic control system includes a data acquisition module, a data processing module running the YoloV5 model and a data analysis module for optimizing the YoloV5 model, the data acquisition module is communicatively connected with the data processing module, the data processing module is communicatively connected with the data analysis module and the controller, the data acquisition module is used to collect characteristic information of the coal transported on the scraper conveyor, and transmit the characteristic information to the data processing module, after the data processing module receives the characteristic information, the characteristic information is judged according to the YoloV5 model and an action signal is generated, the data processing module sends the action signal to the controller, the controller responds to the action signal and adjusts the working state of the scraper conveyor, the data processing module sends the characteristic information and the corresponding action signal to the data analysis module, after the data analysis module receives the characteristic information and the corresponding action signal, the optimization information is generated according to the characteristic information and the corresponding action signal, the data analysis module sends the optimization information to the data processing module, and the data processing module corrects the YoloV5 model according to the optimization information.

2. The scraper conveyor according to claim 1, characterized in that: The data acquisition module includes at least three groups of high-definition cameras, infrared sensors and weight sensors. The characteristic information includes image information, temperature information and weight information. The three groups of high-definition cameras are respectively arranged at the entrance, exit and middle section of the scraper conveyor, and are used to capture the image information of each piece of coal transported on the scraper conveyor in real time. The infrared sensor is used to detect the temperature information of each piece of coal transported on the scraper conveyor in real time. The weight sensor is used to detect the weight information of each piece of coal transported on the scraper conveyor in real time.

3. The scraper conveyor according to claim 1, characterized in that: The data analysis module includes a database, a server and a user end that is communicatively connected to the controller. The database is used to receive and store the feature information and the corresponding action signal. The server is used to call the feature information and the corresponding action signal in the database and generate the optimization information and analysis information based on them. The user end is used to query the analysis information.

4. A large lump coal automatic control system, which is applied to a scraper conveyor, characterized in that: It also includes a data acquisition module, a data processing module running the YoloV5 model, and a data analysis module for optimizing the YoloV5 model. The data acquisition module is communicatively connected to the data processing module, and the data processing module is communicatively connected to the data analysis module and the controller. The data acquisition module is used to collect characteristic information of the coal transported on the scraper conveyor, and transmit the characteristic information to the data processing module. After receiving the characteristic information, the data processing module judges the characteristic information according to the YoloV5 model and generates an action signal. The data processing module sends the action signal to the controller. The controller responds to the action signal and adjusts the working state of the scraper conveyor. The data processing module sends the characteristic information and the corresponding action signal to the data analysis module. After receiving the characteristic information and the corresponding action signal, the data analysis module generates optimization information according to the characteristic information and the corresponding action signal. The data analysis module sends the optimization information to the data processing module, and the data processing module corrects the YoloV5 model according to the optimization information.

5. A method for automatic control of large lump coal, characterized in that: The following steps are involved: S10. Pre-store the large coal range in the YoloV5 model; S20. Real-time capture of characteristic information of each piece of coal transported on the scraper conveyor; S30. The characteristic information is transmitted to the YoloV5 model, and the YoloV5 model judges the coal size according to the characteristic information. If the coal size is judged to be lower than the large coal range, the scraper conveyor operates normally. If the coal size is judged to be within the large coal range or exceeds the large coal range, S40 is executed. S40. If it is determined that the coal size is within the range of large pieces of coal, the YoloV5 model sends a first action signal to the controller, and the controller controls the scraper conveyor to slow down in response to the first action signal. If it is identified that the coal size exceeds the range of large pieces of coal, the YoloV5 model sends a second action signal to the data analysis module, and the controller controls the scraper conveyor to stop in response to the second action signal, and processes the coal whose size exceeds the range of large pieces of coal; S50. The scraper conveyor resumes normal operation.

6. The automatic control method for large lump coal according to claim 5, characterized in that: The following steps are also included: S41. Generate first optimization information according to the characteristic information corresponding to the size of the coal located in the range of the large piece of coal and the first action signal; S42. Optimize the YoloV5 model according to the first optimization information and optimize the judgment strategy.

7. The automatic control method for large lump coal according to claim 6, characterized in that: The following steps are also included: S43. Generate second optimization information according to the characteristic information corresponding to the coal size exceeding the range of the large coal and the second action signal; S44. Optimize the YoloV5 model according to the second optimization information to optimize the judgment strategy.

8. The automatic control method for large lump coal according to claim 5, characterized in that: The step S40 further includes: after the scraper conveyor stops, controlling the controller to perform a crushing operation on the large pieces of coal on the scraper conveyor that exceed the range of the large pieces of coal.

9. The automatic control method for large lump coal according to claim 5, characterized in that: The total loss function of the YoloV5 model is the weighted sum of the position loss function, the confidence loss function and the classification loss function. The total loss function is expressed as: L total =L loc +L conf +L cls Where, L total is the total loss function, L loc is the position loss function, L conf is the confidence loss function, L cls is the classification loss function.

10. The automatic control method for large lump coal according to claim 9, characterized in that: The mathematical formula of the position loss function is expressed as: Where S is the number of grids into which the image is divided; B is the number of bounding boxes predicted for each grid; is an indicator variable, which is 1 when the jth bounding box in the i-th grid is responsible for predicting an object, otherwise it is 0; coord is the weight factor of the position loss, which is used to adjust the impact of the position loss; pred x , pred y The center coordinates of the predicted bounding box; true x , true y is the actual bounding box center coordinate; pred w , pred h is the predicted bounding box width and height; true w , true h is the actual bounding box width and height; The mathematical formula of the confidence loss function is expressed as: In the formula, λ noobj ,λ cbj are the weights of the confidence loss in the case of no object and with object, respectively; BCE is the binary cross entropy loss function; IOU(pred ij ,true ij ) is the intersection-over-union ratio between the predicted bounding box and the true bounding box; The mathematical formula of the classification loss function is expressed as: In the formula, is the predicted class probability vector; is the true category label vector.

Images