Belt conveyor fault identification and prediction method and system

Abstract

The invention discloses a belt conveyor fault identification and prediction method and system, and relates to the technical field of digital intelligent monitoring of belt conveyors, and the method comprises the steps: collecting the operation data of a belt conveyor in real time through various sensing devices arranged on the belt conveyor, analyzing the collected video data and multi-modal data through an AI algorithm, and carrying out the real-time monitoring of the operation data of the belt conveyor; analyzing the operation state and fault prior information of parts and materials of the conveyor, inputting the information into a dynamic simulation model, simulating the common influence of various faults on the belt conveyor through Monte Carlo simulation, and calculating the probability prediction values of the various faults; finally, an analysis result is displayed in real time in combination with modes of graphic display, probability curves and the like, and working personnel are timely notified of occurred and about-to-occur faults; meanwhile, fault probability prediction and operation data serve as training data again to conduct incremental learning on the AI algorithm, and the accuracy of the AI algorithm is continuously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital intelligent monitoring of belt conveyors, and specifically to a method and system for identifying and predicting faults of belt conveyors. Background Art

[0002] Belt conveyor is a continuous conveying equipment that is powered by a motor to drive the transmission roller to rotate, and then the transmission roller transmits power to the conveyor belt to drive it to operate, and then uses friction to drive the material to move. Due to the characteristics of low energy consumption, large transportation volume and strong versatility, belt conveyor has been widely used in various industries such as electricity, metallurgy, coal, chemical industry, mining, ports, etc.

[0003] With the advancement of technology and the increase in social demand, the conveying capacity of belt conveyors has gradually increased. At present, the conveying capacity of belt conveyors can reach thousands of tons or even tens of thousands of tons per hour, which greatly increases the load of the conveyor and puts higher requirements on its safety. Since belt conveyors have many parts and usually need to operate for a long time in harsh working environments such as rugged terrain and extreme weather, this greatly increases the probability of failure. However, the increase in social demand has also gradually increased the conveying distance of belt conveyors. The conveying distance of a single machine can reach several kilometers, more than ten kilometers or even more. By bridging multiple belt conveyors, a total continuous conveying distance of tens of kilometers can be achieved. The increase in conveying distance has increased the difficulty of fault monitoring and reliability maintenance of belt conveyors.

[0004] The traditional method of belt conveyor fault monitoring is for inspectors to go to the conveying site with various handheld detection equipment to detect abnormalities one by one on the motor, roller, roller and other parts of the belt conveyor. This method is not only less accurate and timely in the case of long distances, large volumes, and harsh working environments, but also has huge labor costs. Therefore, how to improve the accuracy and timeliness of fault monitoring and enhance the automation and unmanned level of monitoring has become a key issue in solving the current belt conveyor reliability engineering.

[0005] In recent years, many technologies have begun to apply intelligent methods to try to solve the above problems. For example, Chinese patent document CN116424768B discloses a centralized monitoring method and system for wheel-rail belt conveyors based on digital twin technology, and the application of digital twin technology focuses on the fault monitoring of the trolley and the conveyor belt. Chinese patent document CN118565828A discloses a cloud-edge collaborative belt conveyor roller bearing digital twin state monitoring method, focusing on the real-time monitoring and fault warning of the roller bearing. Chinese patent document CN117800039A discloses a belt conveyor belt deviation detection system, which automatically analyzes the optical image and thermal image of the roller operation through a computing device to achieve belt deviation detection and warning. The existing technology only uses data collected by sensors arranged at a single point to train machine learning algorithms or models. However, there are many parts and fault types of belt conveyors, and single-point data collection makes the training data set of the machine learning algorithm small, thereby reducing the effectiveness of the machine learning algorithm and resulting in inaccurate fault judgment and prediction results. Based on this, a belt conveyor fault identification and prediction method and system are designed, which integrates AI algorithm and dynamic simulation technology to systematically solve the problems of effective training of AI algorithm and prediction of component failures in belt conveyor operation and maintenance. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a belt conveyor fault identification and prediction method and system, which integrates AI algorithm and dynamic simulation technology to systematically solve the problems of effective training of AI algorithm and prediction of component failures in belt conveyor operation and maintenance.

[0007] The object of the present invention is achieved through the following technical solution: a belt conveyor fault identification and prediction method, the method comprising the following steps:

[0008] Step S1, collecting operation data of the belt conveyor;

[0009] Step S2: Use the AI ​​algorithm to process the collected operation data of the belt conveyor, determine the operation status of the belt conveyor, create and calculate the prior information of the fault and record the analysis results;

[0010] Step S3, updating the dynamic simulation model of the belt conveyor by using the calculated prior information of the fault, constructing a single empirical distribution of various faults and fitting it into a joint distribution, performing simulation and calculating the probability prediction value of various faults;

[0011] Step S4: construct the operating status of the belt conveyor and the simulated calculated value into a key-value pair and use them as training samples to train the AI ​​algorithm through incremental learning; display the analysis results in real time, and issue an alarm when the operating status of the belt conveyor is abnormal or the probability prediction value is greater than the preset value.

[0012] In step S1, the start and end points of the time chain for collecting data and the time discretization accuracy are set, and the operation data chain of the belt conveyor is collected in real time according to the set start and end points of the time chain, and stored and marked with timestamps according to the time line.

[0013] The operation data chain includes vibration data, audio data, rotation speed data, force data, thermal imaging video image data and ordinary video image data, and the operation status of the belt conveyor includes noise information, vibration information, temperature information, rotation speed information, force information, physical entity type information, contour information, size information, angle information and motion trajectory.

[0014] In step S2, the collected operation data of the belt conveyor is processed by an AI algorithm, and the AI ​​algorithm includes a computer vision algorithm, a target detection and multi-classification algorithm, an edge detection algorithm, a RoI Transformer algorithm, a dense optical flow algorithm and a multimodal learning algorithm. It is determined whether the operation status of the belt conveyor under the current time chain is normal, and the change value of the operation status, the difference percentage between the current operation status and the benchmark operation status, and the timestamp are sorted as prior information of the fault and the analysis results are recorded according to the timestamp. The benchmark operation status is obtained by the computer vision algorithm and the multimodal learning algorithm.

[0015] The steps to determine whether the operating status of the belt conveyor under the current time chain is normal are as follows: process the time series data in the operating data chain through a recurrent neural network or a convolutional network, and calculate the normal or abnormal binary classification probability value of each component and material at each moment in the set timeline.

[0016] In step S3, the operating status and prior information calculated in step S2 are used as input to update the dynamic simulation model of the belt conveyor, and a single empirical distribution of various faults is constructed according to the operating status and prior information of various components and materials to determine whether the empirical distribution obeys the Gaussian distribution. If so, a mixed Gaussian model is used to fit the empirical distribution of various faults into a joint distribution; if not, a kernel density estimation method is used to fit the empirical distribution of various faults into a joint distribution. For each type of fault, two combinations are made to determine whether there is or may be a physical connection. If not, the correlation coefficient is corrected to zero, and a Monte Carlo simulation is performed on the updated dynamic simulation model. A sampling analysis is performed on the simulated future operating status, and the probability prediction of each type of fault is calculated and output as a probability prediction value.

[0017] In step S4, the operating status of the belt conveyor and the calculation results of the Monte Carlo simulation are constructed as key-value pairs and used as training samples to train the AI ​​algorithm through incremental learning; if the AI ​​algorithm shows that a component or material has failed, the staff is notified for maintenance; if the fault type is serious, the shutdown operation is directly triggered and the fault type and fault point are displayed.

[0018] The fault conditions of various parts and materials of the belt conveyor are displayed in real time using charts and three-dimensional images; the changes in the probability prediction values ​​of various faults are displayed using probability curves; and an early warning is issued when the fault probability prediction value is greater than the preset value.

[0019] A belt conveyor fault identification and prediction system, the system includes a perception device module, an AI algorithm module, a dynamic simulation model and a result display module, the perception device module is connected to the AI ​​algorithm module, the AI ​​algorithm module is respectively connected to the dynamic simulation model and the result display module, the dynamic simulation model is connected to the result display module, the AI ​​algorithm module includes a computer vision algorithm and a multimodal learning algorithm, the computer vision algorithm is used to process ordinary video image data and thermal imaging video data, and transmit the processing results to the dynamic simulation model and the result display module; the multimodal learning algorithm is used to process a multimodal data chain, and transmit the processing results to the dynamic simulation model and the result display module, the dynamic simulation model includes a mechanism simulation model and a parameterized analysis model, and the mechanism simulation model is used to exhaustively All types of possible failures of each type of component or material, determine the physical connection status between different failures, and perform Monte Carlo simulation; the parameterized analysis model includes a Gaussian mixture model and a kernel density estimation algorithm; it is used to calculate the joint probability distribution of various failures and the Monte Carlo simulation results of the computer simulation model, and transmit the calculation results to the result display module; the result display module includes a warning system, a curve module, a chart display module and a three-dimensional image module. The warning system is used to quickly notify the staff to perform maintenance operations when the state is abnormal or the probability of failure is greater than the preset value; the curve module is used to display the changes in the probability prediction value of the failure in real time; the chart display module is used to display the statistical information of the belt conveyor parts and failures; the three-dimensional image module is used to directly present the operating status of the belt conveyor.

[0020] The sensing device module includes a high-speed camera, a thermal imaging camera, a noise sensor, a vibration sensor, a rotation speed sensor and a force sensor.

[0021] The beneficial effects of the present invention are:

[0022] 1. The present invention acquires a data chain containing elements such as images, vibration, noise, temperature, force, and rotation speed by setting a timeline and arranging various types of sensors. Compared with the prior art, it collects composite data of various parts and materials in an all-round way, providing comprehensive support for fault analysis of belt conveyors.

[0023] 2. The present invention utilizes AI algorithms to process video images and data links, and can accurately obtain information such as the type, outline, size, orientation, motion trajectory, temperature, etc. of parts and materials through video images, realize multi-angle judgment of the operating status of parts and materials, and provide reliable prior information for the simulation analysis of belt conveyors.

[0024] 3. The present invention uses dynamic simulation technology to simulate the mechanism and perform parametric analysis of the belt conveyor, thereby integrating the operating status and prior fault information of all components and materials, and further calculates the probability value of the occurrence of faults through Monte Carlo simulation, thereby ensuring the accuracy of fault prediction.

[0025] 4. The present invention combines the operating status and the Monte Carlo simulation results and uses the reinforcement learning technology to continuously update the AI ​​algorithm, so that the effectiveness of the AI ​​algorithm can continue to increase with the operation of the belt conveyor.

[0026] 5. The present invention displays the analysis results of AI algorithms and dynamic simulation models in a variety of ways, which can interact with staff in a more effective way and inform the operating status of the belt conveyor in real time. Comprehensive monitoring of faults through probability prediction values ​​can reduce false alarms and missed alarms and improve the accuracy of alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic flow chart of the belt conveyor fault identification and prediction method of the present invention;

[0028] Figure 2 This is a structural diagram of the belt conveyor fault identification and prediction system. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions 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, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0030] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0031] In one embodiment of the present application:

[0032] like Figure 1 to Figure 2 As shown, a belt conveyor fault identification and prediction method comprises the following steps:

[0033] Step S1, timeline parameter setting and data chain collection: set the start and end points of the time chain for collecting data and the time discretization accuracy, and collect the operation data chain of the belt conveyor in real time according to the set time chain start and end points. The monitoring time span is 12 hours, starting at 8 o'clock in the morning and ending at 8 o'clock in the evening. The timeline length is set to 10 seconds, that is, within the time span, a total of 4320 operation data chains are collected, and the time discretization accuracy of each data chain is 0.1 second. The data is stored and marked with a timestamp according to the preset timeline. The operation data chain includes vibration data, audio data, speed data, force data, thermal imaging video image data and ordinary video image data. The operation status of the belt conveyor includes noise information, vibration information, temperature information, speed information, force information, physical entity type information, contour information, size information, angle information, etc. Information and motion trajectory, while the running data chain is divided into a multimodal data chain and a video image data chain. The multimodal data chain includes audio data, vibration data, speed data and force data, which are collected by multiple noise sensors, vibration sensors, speed sensors and force sensors arranged at the drive frame, frame and tensioning car respectively. The collection interval is 0.1 seconds, and every 10 seconds accumulated, it is stored as a multimodal data chain. The video image data chain includes ordinary video image data and thermal imaging video image data. These data are obtained by thermal imaging cameras arranged next to the drive frame and outside the guide trough and ordinary cameras arranged every 250 meters along the conveyor. The image frame extraction interval is 0.1 seconds, and the length of each cycle is 10 seconds. The video image data of the belt conveyor operation collected in real time is extracted and stored according to the preset timeline and marked with a timestamp.

[0034] Step S2, data processing and real-time analysis: determine the operating status of the belt conveyor, create and calculate the prior information of the fault and record the analysis results;

[0035] The collected belt conveyor operation data is processed using AI algorithms, including computer vision algorithms, target detection and multi-classification algorithms, edge detection algorithms, RoI Transformer algorithms, dense optical flow algorithms, and multimodal learning algorithms. Computer vision algorithms are used to analyze ordinary video image frames in real time, and the YOLOV8 algorithm (i.e., target detection and multi-classification algorithm) is used to obtain the types of parts and materials. The Laplacian algorithm (i.e., edge detection algorithm) is used to obtain the contour and size of the target object. RoI The Transformer algorithm processes the contour of the target object, analyzes the target rotation angle, uses the dense optical flow algorithm to process the contours in two consecutive frames of images, calculates the motion trajectory of the target object, and determines whether the operating status of the belt conveyor is normal under the current time chain. The change value of the operating status, the difference percentage between the current operating status and the visual benchmark operating status, and the timestamp are sorted as the prior information of the fault and the analysis results are recorded according to the timestamp. The timestamp marking time interval is 0.1 seconds, and the length of each cycle is 10 seconds. The visual benchmark operating state is obtained by training the computer vision algorithm. The operating state includes: noise information, vibration information, temperature information, speed information, force information, physical entity type information, contour information, size information, orientation information and motion trajectory. The computer vision algorithm is then used to analyze the thermal imaging video image frame in real time, and the YOLOV8 algorithm is used to obtain the types of parts and materials in the thermal imaging video frame. Through Laplacian an algorithm obtains the outline and size of the target object, grayscales the thermal imaging video frame, obtains the temperature value of each pixel through linear mapping, records the maximum, minimum and average temperatures of all pixels within the target object outline as the temperature data of the target object, integrates the temperature information of the components in the continuous image frames into a data chain according to the timestamp, and adds it to the multimodal data chain. The length of each temperature data chain is 10 seconds, with a data point interval of 0.1 seconds. The multimodal learning algorithm is used to analyze the multimodal data chain in real time. The time series data in the multimodal data chain is processed by the GRU algorithm (i.e., recurrent neural network) or convolutional network, and the normal or abnormal binary classification probability value of each component and material at each moment in the set timeline is calculated. The change value of the operating state, the difference percentage between the current operating state and the multimodal benchmark operating state, and the timestamp are organized as prior information. The multimodal benchmark operating state is obtained by training the multimodal learning algorithm.

[0036] Step S3, mechanism simulation and parameterized calculation: the operating state and prior information calculated in step S2 are used as input to update the dynamic simulation model of the belt conveyor, and a single empirical distribution of various faults is constructed according to the operating state and prior information of various parts and materials to determine whether the empirical distribution obeys the Gaussian distribution. If so, a mixed Gaussian model is used to fit the empirical distribution of various faults to a joint distribution; if not, a kernel density estimation method is used to fit the empirical distribution of various faults to a joint distribution, wherein the kernel function of the kernel density estimation method adopts a triangular kernel, and the bandwidth parameter is set to the Scott method. For various faults, two combinations are used to determine whether there is or may be a physical connection. If not, the correlation coefficient is corrected to zero, and Monte Carlo simulation is performed on the updated dynamic simulation model. The simulated future operating state is sampled and analyzed, and the probability of occurrence of various faults is calculated and output as a probability prediction value. The number of Monte Carlo sampling is set to 500 times, the average single sampling time is 0.02 seconds, and the simulation time is 10 seconds in total. The probability prediction value is obtained by dividing the number of fault occurrences by 500.

[0037] Step S4, incremental learning and result output: The operating status of the belt conveyor and the calculation results of the Monte Carlo simulation are constructed as key-value pairs and used as training samples. The AI ​​algorithm is trained through incremental learning. The incremental learning algorithm is an online random forest algorithm with 10 trees, a learning rate of 0.01, and a subsampling rate of 0.5. The analysis results are displayed in real time. If the AI ​​algorithm shows that a component or material is faulty, the staff is notified for maintenance. If the fault type is serious, the shutdown operation is directly triggered and the fault type and fault location are displayed.

[0038] The fault conditions of various parts and materials of the belt conveyor are displayed in real time using charts and three-dimensional images; the changes in the probability prediction values ​​of various faults are displayed using probability curves; and an early warning is issued when the fault probability prediction value is greater than the preset value.

[0039] In another embodiment of the present application:

[0040] A belt conveyor fault identification and prediction system, the system includes a perception device module, an AI algorithm module, a dynamic simulation model and a result display module, the perception device module is respectively connected to the belt conveyor and the AI ​​algorithm module, the perception device module includes a high-speed camera, a thermal imaging camera, a noise sensor, a vibration sensor, a speed sensor and a force sensor, which are used to collect the belt conveyor operation data chain and transmit the data chain to the AI ​​algorithm module, the high-speed camera is arranged along the conveyor at intervals of 250mm, and the viewing angle is 45° downward to the center line of the conveyor belt, the thermal imaging camera is arranged on the driving side and next to the driven roller, the noise sensor, the vibration sensor, the speed sensor, and the force sensor are arranged on the driving frame, the frame and all the rollers; the AI ​​algorithm module is respectively connected to the dynamic simulation model and the result display module, the dynamic simulation model is connected to the result display module, the AI ​​algorithm module includes a computer vision algorithm and a multimodal learning algorithm, which are used to process the data chain, calculate the real-time operation status of parts and materials and the prior information of various faults, and transmit the processing results to the dynamic simulation model and the result display module, the computer vision algorithm includes the YOLOV8 algorithm, the Laplacian algorithm, the RoI Transformer algorithm and dense optical flow algorithm, computer vision algorithm is used to process ordinary video image data and thermal imaging video data, and transmit the processing results to the dynamic simulation model and result display module, multimodal learning algorithm includes GRU algorithm and online random forest algorithm, dynamic simulation model includes mechanism simulation model and parameterized analysis model, which are used to simulate the impact of various faults on the overall belt conveyor, and calculate the probability of each type of fault through Monte Carlo simulation, and transmit the calculation results to the result display module, mechanism simulation model is used to enumerate all types of faults that may occur in each type of parts or materials, judge the physical connection status between different faults, perform Monte Carlo simulation, parameter The quantitative analysis model includes a kernel density estimation algorithm, which is used to calculate the joint probability distribution of various faults and the Monte Carlo simulation results of the computer simulation model, and transmit the calculation results to the result display module. The types of parts are divided into six categories: motors, reducers, drive rollers, conveyor belts, rollers, and driven rollers. The materials are divided into three categories: ordinary bulk materials, bulk materials, and abnormal materials. Among them, there is a physical connection between the motor and the reducer, between the reducer and the drive roller, and between the conveyor belt and the drive roller, roller, and driven roller respectively. The possible faults are: motor failure, reducer failure, drive roller bearing damage, conveyor belt deviation, conveyor belt tearing, roller bearing damage, and driven roller bearing damage;The result display module includes a warning system, a curve module, a chart display module and a three-dimensional image module, which are used to receive data signals transmitted by the AI ​​algorithm module and the dynamic simulation model, and display the running status of the belt conveyor. The warning system is used to quickly notify the staff to perform maintenance operations when the state is abnormal or the probability of failure is greater than the preset value (the preset value is set by the staff according to their own needs, such as 1%). The curve module is used to display the changes in the probability prediction value of the failure in real time. The chart display module is used to display the statistical information of the belt conveyor parts and failures. The three-dimensional image module is used to directly present the running status of the belt conveyor. The three-dimensional image is realized by virtual reality technology (VR). In the three-dimensional image, the abnormal conditions and fault predictions of each component are distinguished by different colors. Red means urgent maintenance, yellow means manual inspection is required, and green means normal operation. ;

[0041] The above is only an embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A belt conveyor fault identification and prediction method, characterized in that: The following steps are involved: Step S1, collecting operation data of the belt conveyor; Step S2: Use the AI ​​algorithm to process the collected operation data of the belt conveyor, determine the operation status of the belt conveyor, create and calculate the prior information of the fault and record the analysis results; Step S3, updating the dynamic simulation model of the belt conveyor by using the calculated prior information of the fault, constructing a single empirical distribution of various faults and fitting it into a joint distribution, performing simulation and calculating the probability prediction value of various faults; Step S4: construct the operating status of the belt conveyor and the simulated calculated value into a key-value pair and use them as training samples to train the AI ​​algorithm through incremental learning; display the analysis results in real time, and issue an alarm when the operating status of the belt conveyor is abnormal or the probability prediction value is greater than the preset value.

2. The belt conveyor fault identification and prediction method according to claim 1, characterized in that: In step S1, the start and end points of the time chain for collecting data and the time discretization accuracy are set, and the operation data chain of the belt conveyor is collected in real time according to the set start and end points of the time chain, and stored and marked with timestamps according to the time line.

3. The belt conveyor fault identification and prediction method according to claim 2 is characterized in that: The operation data chain includes vibration data, audio data, rotation speed data, force data, thermal imaging video image data and ordinary video image data, and the operation status of the belt conveyor includes noise information, vibration information, temperature information, rotation speed information, force information, physical entity type information, contour information, size information, angle information and motion trajectory.

4. The belt conveyor fault identification and prediction method according to claim 2 is characterized in that: In step S2, the collected operation data of the belt conveyor is processed by an AI algorithm, and the AI ​​algorithm includes a computer vision algorithm, a target detection and multi-classification algorithm, an edge detection algorithm, a RoI Transformer algorithm, a dense optical flow algorithm and a multimodal learning algorithm. It is determined whether the operation status of the belt conveyor under the current time chain is normal, and the change value of the operation status, the difference percentage between the current operation status and the benchmark operation status, and the timestamp are sorted as prior information of the fault and the analysis results are recorded according to the timestamp. The benchmark operation status is obtained by the computer vision algorithm and the multimodal learning algorithm.

5. The belt conveyor fault identification and prediction method according to claim 4 is characterized in that: The steps to determine whether the operating status of the belt conveyor under the current time chain is normal are as follows: process the time series data in the operating data chain through a recurrent neural network or a convolutional network, and calculate the normal or abnormal binary classification probability value of each component and material at each moment in the set timeline.

6. The belt conveyor fault identification and prediction method according to claim 1, characterized in that: In step S3, the operating status and prior information calculated in step S2 are used as input to update the dynamic simulation model of the belt conveyor, and a single empirical distribution of various faults is constructed according to the operating status and prior information of various components and materials to determine whether the empirical distribution obeys the Gaussian distribution. If so, a mixed Gaussian model is used to fit the empirical distribution of various faults into a joint distribution; if not, a kernel density estimation method is used to fit the empirical distribution of various faults into a joint distribution. For each type of fault, two combinations are made to determine whether there is or may be a physical connection. If not, the correlation coefficient is corrected to zero, and a Monte Carlo simulation is performed on the updated dynamic simulation model. A sampling analysis is performed on the simulated future operating status, and the probability prediction of each type of fault is calculated and output as a probability prediction value.

7. The belt conveyor fault identification and prediction method according to claim 1, characterized in that: In step S4, the operating status of the belt conveyor and the calculation results of the Monte Carlo simulation are constructed as key-value pairs and used as training samples to train the AI ​​algorithm through incremental learning; if the AI ​​algorithm shows that a component or material has failed, the staff is notified for maintenance; if the fault type is serious, the shutdown operation is directly triggered and the fault type and fault point are displayed.

8. The belt conveyor fault identification and prediction method according to claim 7, characterized in that: The fault conditions of various parts and materials of the belt conveyor are displayed in real time using charts and three-dimensional images; the changes in the probability prediction values ​​of various faults are displayed using probability curves; and an early warning is issued when the fault probability prediction value is greater than the preset value.

9. A belt conveyor fault identification and prediction system, characterized in that: It includes a perception device module, an AI algorithm module, a dynamic simulation model and a result display module. The perception device module is respectively connected to the belt conveyor and the AI ​​algorithm module. The AI ​​algorithm module is respectively connected to the dynamic simulation model and the result display module. The dynamic simulation model is connected to the result display module. The AI ​​algorithm module includes a computer vision algorithm and a multimodal learning algorithm. The dynamic simulation model includes a mechanism simulation model and a parameterized analysis model. The result display module includes a warning system, a curve module, a chart display module and a three-dimensional image module.

10. The belt conveyor fault identification and prediction system according to claim 9, characterized in that: The sensing device module includes a high-speed camera, a thermal imaging camera, a noise sensor, a vibration sensor, a rotation speed sensor and a force sensor.

Images