Conveying belt tearing positioning and tracing system based on RFID and multi-sensor fusion

Abstract

The invention relates to a conveying belt tearing positioning and tracing system based on RFID and multi-sensor fusion, which effectively overcomes the influence of belt slipping and elastic deformation through multi-sensor fusion and motion compensation, and realizes accurate positioning of a tearing point. Meanwhile, through cross validation of vibration, vision and infrared data, the recognition accuracy is improved to 99.5% or above, false alarms caused by environmental factors (such as dust and moisture) are greatly reduced, stable operation in a severe industrial environment is ensured, and high reliability and low false alarm rate are achieved. And thirdly, the system supports long-term data tracing and intelligent operation and maintenance, and a user can analyze a tearing trend and identify a high-frequency fault area by storing historical data of at least 180 days and providing a visual tool, so that a maintenance plan is optimized, predictive maintenance is realized, and faults are reduced from the source.

Description

Technical Field

[0001] This invention relates to the field of industrial Internet of Things and positioning technology, specifically to a conveyor belt tear positioning and tracking system based on RFID and multi-sensor fusion. Background Technology

[0002] As a critical transportation equipment in industries such as coal mining, mining, and ports, the operational safety of belt conveyors directly impacts production efficiency and personnel safety. In actual operation, conveyor belts are prone to longitudinal tearing due to foreign object punctures, material blockages, or structural fatigue. If this damage is not detected and located in time, it can lead to the tear expanding, conveyor belt breakage, or even major safety accidents, causing significant economic losses. Currently, in the field of conveyor belt tear detection, existing technologies have significant shortcomings in rapid and accurate location, mainly because existing detection systems mostly adopt a single technology approach based on vision or radio frequency identification. Pure vision solutions (such as those described in patent CN111661590A) identify tears by analyzing light stripe deformation through image processing. While they can detect damage, they have the following location defects: Large response latency: It usually takes 200-500ms from image acquisition to processing completion. Coupled with communication latency, it is difficult to achieve second-level positioning. Large position estimation error: Position estimation is based solely on belt speed multiplied by time, without considering factors such as belt slippage and elastic deformation, resulting in positioning errors often exceeding 3-5 meters. Poor environmental adaptability: Dust and humidity in the mine affect image quality, leading to a higher positioning failure rate.

[0003] Furthermore, most existing systems only provide real-time alarms and lack long-term data storage and traceability capabilities. Historical data is typically retained for a short period (e.g., 7-30 days) and lacks integrated visualization and backtracking tools, making it impossible to analyze tearing trends, identify high-frequency failure areas, or optimize maintenance plans. Summary of the Invention

[0004] In view of the problems existing in the prior art, the purpose of this invention is to provide a conveyor belt tear positioning and traceability system based on RFID and multi-sensor fusion, which can achieve accurate positioning of tear points and support long-term data traceability.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A conveyor belt tear location and traceability system based on RFID and multi-sensor fusion, comprising a perception layer, a decision layer and an application layer; The sensing layer includes an RFID sensing module, a speed sensor, an infrared thermal imager, an industrial camera, and a positioning chip. The RFID sensing module employs at least one UHF RFID reader and multiple RFID tags; the UHF RFID reader is deployed at key points along the conveyor belt, located downstream of the industrial camera; the RFID tags are evenly spaced on the surface of the conveyor belt or embedded in the belt body, each RFID tag has a unique ID and is bound to the absolute position coordinates of the conveyor belt. The speed sensor is used to measure the running speed of the conveyor belt in real time and continuously; The infrared thermal imager and industrial camera are installed above the return conveyor belt; The positioning chip integrates a vibration sensor and a communication module, which serves as a collaborative triggering and auxiliary detection unit. There are multiple positioning chips, and each positioning chip is associated with its surrounding RFID tags. The decision-making layer includes an edge computing server, which receives data from the perception layer and performs corresponding analysis to determine whether a tear exists and its location. The edge computing server receives vibration information from a positioning chip, infrared image information from an infrared thermal imager, visible light image information from an industrial camera, tag information from an UHF RFID reader, and real-time speed information from a speed sensor. It then activates some RFID tags based on the vibration, infrared, and visible light image information, respectively. Simultaneously, it calculates the overall tear probability by fusing the infrared and visible light image information to determine if a longitudinal tear exists. When a longitudinal tear exists, it uses the high-precision timestamp T_capture from the visible and infrared images, along with the tag and speed information, to determine the tear location. The application layer includes a data storage module, a user interface module, and an alarm module. The data storage module uses a time-series database to store raw data and tear event records, retaining at least 180 days of historical data. The user interface module adopts a web-based graphical user interface, supporting real-time monitoring, historical data query, and visual backtracking. The alarm module supports audible and visual alarms.

[0006] The specific steps for the edge computing server to activate the RFID tag are as follows: After noise reduction processing of vibration information sent by all positioning chips, time-domain and frequency-domain features are extracted. Each time-domain and frequency-domain feature is used as a key feature, and a dynamic threshold is set for each key feature. When one or more key feature values ​​continuously and significantly exceed the corresponding dynamic threshold for a set time, it is determined that the vibration of the conveyor belt corresponding to the positioning chip is abnormal. When a vibration abnormality is determined, the edge computing server activates the RFID tag associated with the positioning chip corresponding to the vibration abnormality. After preprocessing the visible light images sent by the industrial camera, a convolutional neural network is used to extract visible light features from the visible light image data, and the visible light features are fused and probabilistically inferred to output a binary classification probability, i.e., the visible light tearing probability P_visual. When the visible light tearing probability P_visual is greater than or equal to P, the activation radius is calculated, and all RFID tags within the activation radius are activated with the industrial camera as the center. After preprocessing the infrared images sent by the infrared thermal imager, the Otsu threshold segmentation method is used to extract temperature anomaly regions from the infrared image data, calculate infrared features, and then input the infrared features into a pre-trained machine learning model to map the infrared features to probability values, thus obtaining the infrared tear probability P_thermal. When the visible light tear probability P_visual is greater than or equal to P, the activation radius is calculated, and all RFID tags within the activation radius are activated with the infrared thermal imager as the center.

[0007] The time-domain features include root mean square value and kurtosis, and the frequency-domain features include vibration frequency.

[0008] The activation radius R is calculated as follows: R = R_base + k × V; Where R_base is the set radius, K is determined according to the delay and is set to 0.5-1.5 seconds, and V is the speed of the conveyor belt.

[0009] The edge computing server calculates the overall tear probability as follows: Bayesian inference is used to fuse infrared and visible light data to calculate the overall tear probability. The calculation formula is: P_tear = w1 * P_thermal + w2 * P_visual; where the weights w1 and w2 are optimized through training. When the overall tear probability P_tear ≥ P, it is determined that there is a longitudinal tear in the conveyor belt, and a high-precision timestamp T_capture is generated based on the positioning chip time. At the same time, an alarm and tear location are triggered.

[0010] The edge computing server performs tear localization as follows: For each UHF RFID reader, when the tear area reaches the reader, the tag information of the activated RFID tag is read and the reading time T_tag is recorded. Then, the following processing is performed: First, tag information with signal strength exceeding the preset range is filtered out. Second, the tag information with the strongest signal strength is selected from the filtered tag information, and its position L_tag is determined based on the information of that tag. Finally, the tear position L_tear = L_tag + V × ΔT is calculated, where ΔT is the time difference, ΔT = T_tag - T_capture; V is the speed of the conveyor belt, measured by a speed sensor. The final tear position is obtained by weighting and averaging the calculated tear positions from all UHF RFID readers, with the weight values ​​set.

[0011] When calculating the final tear location, the edge computing server assigns a higher weight to the tear location calculated from data read by the UHF RFID reader closer to the industrial camera, and a lower weight to the tear location calculated from data read by data further away from the industrial camera.

[0012] By adopting the above solution, the present invention has the following beneficial effects: (1) By fusing visible light images and infrared thermal imaging, the present invention achieves complementary advantages. This can effectively reduce false alarms and missed alarms that may be generated by a single sensor, and can make extremely reliable diagnoses even in harsh environments, thus ensuring the reliability of the system.

[0013] (2) By combining the absolute position coordinates provided by the RFID tag, the real-time belt speed provided by the speed sensor, and the timestamp recorded by image recognition, the system can calculate the precise absolute location of the tear point. This means that maintenance personnel can go directly to the fault point, saving the time spent blindly searching on the conveyor belt which is several kilometers long, achieving a leap from "finding a needle in a haystack" to "precise positioning", and greatly improving the efficiency of operation and maintenance.

[0014] (3) The intelligent platform at the application layer acts as the "brain" of the system. It gathers all data and provides real-time monitoring, historical data tracing, health status assessment, and predictive maintenance recommendations. Managers can not only keep track of the system status in real time, but also predict component lifespan and optimize maintenance plans by analyzing historical data, thereby upgrading from passive emergency repairs to proactive maintenance and comprehensively improving the safety and economic benefits of production. Attached Figure Description

[0015] Figure 1 This is a flowchart of the processing flow of an edge computing server. Figure 2 A flowchart for tear localization on edge computing servers. Detailed Implementation

[0016] This invention discloses a conveyor belt tear location and traceability system based on RFID and multi-sensor fusion. The system adopts a three-layer architecture design: perception layer, decision layer and application layer to ensure efficient and reliable data flow.

[0017] The sensing layer includes an RFID sensing module, a speed sensor, an infrared thermal imager, an industrial camera, and a positioning chip.

[0018] The RFID sensing module employs at least one UHF RFID reader (supporting EPC Class 1 Gen 2 protocol) and multiple RFID tags. The UHF RFID reader is deployed at key points along the conveyor belt, typically downstream of the industrial camera. The number of UHF RFID readers is determined by the conveyor belt length. If the conveyor belt length is less than 200m, 2-3 UHF RFID readers are generally sufficient. As the conveyor belt length increases, the number of UHF RFID readers should be increased. When deploying multiple UHF RFID readers, it is essential to ensure that the coverage areas of adjacent UHF RFID readers overlap by at least 20%.

[0019] RFID tags are affixed to the surface of the conveyor belt or embedded within the belt body, specifically installed 20cm from the edge of the conveyor belt. One RFID tag is placed every 5 meters. Each RFID tag has a unique ID and is bound to the absolute position coordinates of the conveyor belt (such as the distance in meters from the machine head).

[0020] The speed sensor is mounted on the idler support of the conveyor belt and uses an encoder or laser tachometer to measure the conveyor belt's running speed (v) in real time and continuously. This running speed data is a key parameter for dynamic position compensation and calculating the actual location of tear points.

[0021] The infrared thermal imager is installed above the return conveyor belt. It uses an uncooled microbolometer infrared sensor with a resolution of ≥640×480 pixels, a frame rate of 30fps, a temperature measurement range of -20℃ to 150℃, and an accuracy of ±0.5℃.

[0022] The industrial camera is a high-definition camera (resolution ≥1920×1080, frame rate 30fps), mounted above the return conveyor belt at a 45° angle to the belt speed direction to capture surface images. It is also equipped with LED supplementary lighting to adapt to low-light environments. The infrared thermal imager and industrial camera can be integrated into a single device.

[0023] The positioning chip integrates a vibration sensor (such as an IMU) and a communication module, serving as a collaborative triggering and auxiliary detection unit. Multiple positioning chips are used, each associated with a nearby RFID tag. When an abnormal vibration is detected, a high-precision timestamp is immediately recorded, activating the associated RFID tag. The positioning chip provides a temporal reference and physical evidence, forming cross-validation with visual inspection.

[0024] The decision-making layer includes edge computing servers, which use industrial-grade computers (such as Intel Atom processors, 4GB RAM) and run Linux RT (real-time kernel). These edge computing servers are responsible for receiving data from the perception layer, performing corresponding analysis, and determining whether any gaps exist and their locations.

[0025] The application layer includes a data storage module, a user interface module, and an alarm module. The data storage module uses a time-series database (such as InfluxDB) to store raw data and tear event records, retaining at least 180 days of historical data. The user interface module uses a web-based graphical user interface, supporting real-time monitoring, historical data querying, and visual backtracking, displaying tear location, belt speed, alarm status, etc. The alarm module supports audible and visual alarms, issuing a warning signal when the overall tear probability P_tear ≥ 0.7. A shutdown is triggered when the overall tear probability P_tear ≥ 0.9 and persists for 3 sampling cycles (30ms).

[0026] like Figure 1 As shown, the edge computing server receives vibration information from the positioning chip, infrared image information from the infrared thermal imager, visible light image information from the industrial camera, and tag information from the UHF RFID reader. It then performs the following analysis and processing: (1) Activate the RFID tag.

[0027] After denoising the vibration information sent by all positioning chips, time-domain and frequency-domain features are extracted. Time-domain features generally include: RMS (Root Mean Square) value: Characterizes the average level of vibration energy. It is the most commonly used and stable indicator for monitoring vibration intensity. When the conveyor belt runs smoothly, the RMS value remains within a certain range. An abnormal increase often indicates imbalance, looseness, or continuous friction.

[0028] Kurtosis: Highly sensitive to the impact characteristics of vibration signals. It describes the sharpness of the signal distribution shape. Under normal operating conditions, the vibration signal distribution is close to a normal distribution, with a kurtosis value close to 3. When surface damage or periodic impacts occur, a large number of impact pulses appear in the signal, and the kurtosis value increases significantly.

[0029] Frequency domain characteristics include vibration frequencies. The vibration spectrum of a normal conveyor belt is usually regular and relatively stable. Tearing may change the system stiffness or generate periodic impacts, causing characteristic frequencies (such as natural frequencies and side frequencies) to appear or change.

[0030] Each time-domain feature and frequency-domain feature is used as a key feature, and a dynamic threshold is set for each key feature. This dynamic threshold is not fixed, but can be adaptively adjusted according to different operating speeds and load conditions of the conveyor belt.

[0031] When one or more key feature values ​​continuously and significantly exceed the corresponding dynamic threshold for a set time, it is determined that the vibration of the conveyor belt position corresponding to the positioning chip is abnormal.

[0032] When an abnormal vibration is detected, the edge computing server activates the RFID tag associated with the positioning chip corresponding to the abnormal vibration.

[0033] After preprocessing the visible light images sent by the industrial camera, a convolutional neural network (CNN) is used to extract visible light features (edge ​​features, texture features, and tear patterns) from the visible light image data. These features are then fused and probabilistically inferred to output a binary classification probability, i.e., the visible light tear probability P_visual. When the visible light tear probability P_visual is greater than or equal to P, the activation radius R = R_base + k × V is calculated, where R_base is the set radius, K is determined based on the delay (generally set to 0.5-1.5 seconds), and V is the conveyor belt speed. Using the industrial camera as the center, all RFID tags within the activation radius are activated.

[0034] After preprocessing the infrared images transmitted by the infrared thermal imager, the Otsu thresholding method is used to extract temperature anomaly regions from the infrared image data, and infrared features are calculated, namely the maximum temperature rise ΔT, the temperature rise gradient (ΔT / Δt), and the hotspot area. These infrared features are then input into a pre-trained machine learning model (such as logistic regression or support vector machine) to map them to probability values, thus obtaining the infrared tear probability P_thermal. When the visible light tear probability P_visual is greater than or equal to P, the activation radius is calculated (using the same method as described above), and all RFID tags within the activation radius are activated, with the infrared thermal imager as the center.

[0035] (2) Calculate the overall tear probability; The combined tear probability is calculated by fusing infrared and visible light data using Bayesian inference.

[0036] The calculation formula is: P_tear = w1 * P_thermal + w2 * P_visual; where the weights w1 and w2 are optimized through training (initial values ​​of 0.5 and 0.5 respectively).

[0037] When the overall tear probability P_tear ≥ P (generally set to 0.7), it is determined that there is a longitudinal tear in the conveyor belt, and a high-precision timestamp T_capture is generated based on the positioning chip time. At the same time, an alarm and tear location are triggered.

[0038] (3) Tear localization; For each UHF RFID reader, when the torn area reaches the reader's reading area, the tag information of the activated RFID tag is read, and the reading time T_tag is recorded. Then, the following processing is performed (e.g.) Figure 2 (as shown) First, tag information with signal strength exceeding the preset range is filtered out. Tags with weak signals (possibly due to distance from the reader or signal obstruction) or strong signals (possibly due to abnormal interference) are removed to ensure the quality of the data used.

[0039] Secondly, select the tag with the strongest signal strength from the filtered tag information, and determine its location L_tag based on the information of that tag.

[0040] Finally, the tear location is calculated as L_tear = L_tag + V × ΔT, where ΔT is the time difference, ΔT = T_tag - T_capture; and V is the speed of the conveyor belt, which is measured by a speed sensor.

[0041] The final tear location is obtained by weighted averaging the calculated tear locations from all UHF RFID readers. The weights are set values; when setting the weights, the tear locations calculated from data read by UHF RFID readers closer to the industrial camera have higher weights, while those farther away from the industrial camera have lower weights.

[0042] In summary, this invention achieves complementary advantages by fusing visible light imaging and infrared thermal imaging. This effectively reduces false alarms and missed alarms that may occur with a single sensor, and enables highly reliable diagnosis even in harsh environments, ensuring the system's reliability.

[0043] By combining the absolute position coordinates provided by RFID tags, the real-time belt speed provided by speed sensors, and the timestamps recorded by image recognition, the system can calculate the precise absolute location of the tear point. This means that maintenance personnel can go directly to the fault point, saving time spent blindly searching along a conveyor belt several kilometers long. This represents a leap from "finding a needle in a haystack" to "precise positioning," greatly improving maintenance efficiency.

[0044] The application-layer intelligent platform acts as the "brain" of the system, aggregating all data and providing real-time monitoring, historical data tracing, health status assessment, and predictive maintenance recommendations. Managers can not only monitor system status in real time but also predict component lifespan and optimize maintenance plans by analyzing historical data, thereby upgrading from reactive emergency repairs to proactive maintenance and comprehensively improving production safety and economic efficiency.

[0045] Furthermore, the complex industrial environment means that a single sensor may produce false alarms or temporary malfunctions due to momentary interference. This invention utilizes three independent detection methods—vibration, infrared, and visible light—to trigger RFID tag activation as soon as any of these methods detects an anomaly. This means that the RFID tag at the corresponding location is activated before a tear is actually confirmed. This provides triple protection for the positioning process. When a tear actually occurs, the torn area may deform or break quickly. Pre-activation ensures that the reader "captures" the moment the tear point passes, recording the crucial positioning data T_tag and L_tag, avoiding missing the data acquisition window due to slow response. By advancing the positioning preparation process, the detection delay is successfully eliminated from positioning errors, achieving the core effects of second-level response and meter-level positioning.

[0046] In addition, this invention only requires activating some RFID tags, so the RFID reader does not need to continuously scan all tags, but only needs to focus on the activated tags. This reduces the processing load of the reader and the probability of channel collisions, and improves communication efficiency.

[0047] This invention effectively overcomes the effects of belt slippage and elastic deformation through multi-sensor fusion and motion compensation, achieving precise location of tear points. Simultaneously, by cross-validating vibration, visual, and infrared data, the invention improves the recognition accuracy to over 99.5%, significantly reducing false alarms caused by environmental factors (such as dust and humidity), ensuring stable operation in harsh industrial environments, and exhibiting high reliability and a low false alarm rate. Third, the system supports long-term data traceability and intelligent operation and maintenance. By storing at least 180 days of historical data and providing visualization tools, users can analyze tear trends and identify high-frequency fault areas, thereby optimizing maintenance plans and achieving predictive maintenance, reducing failures at their source. Finally, the system significantly improves operational efficiency, reducing fault response time from hours to seconds, reducing unplanned downtime, ensuring production continuity, and lowering costs through precise maintenance, thus improving overall production safety and economic benefits.

[0048] The present invention also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist alone and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by an electronic device, cause the electronic device to perform the methods described in the above embodiments. It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units. From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the method according to the embodiments of this disclosure. Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein.

Claims

1. A conveyor belt tear location and traceability system based on RFID and multi-sensor fusion, characterized by: The system comprises a perception layer, a decision layer and an application layer. The perception layer comprises an RFID sensing module, a speed sensor, an infrared thermal imager, an industrial camera and a positioning chip. The RFID sensing module comprises at least one ultra-high frequency RFID reader and a plurality of RFID tags. The speed sensor is used to measure the running speed of the conveyor belt in real time and continuously. The infrared thermal imager and the industrial camera are installed above the return conveyor belt. The positioning chip is integrated with a vibration sensor and a communication module, which serve as a cooperative trigger and an auxiliary detection unit. The decision layer comprises an edge computing server, which is responsible for receiving data from the perception layer, performing corresponding analysis, determining whether there is a tear and determining the location of the tear. The edge computing server receives vibration information of the positioning chip, infrared image information of the infrared thermal imager, visible light image information of the industrial camera, tag information of the ultra-high frequency RFID reader, real-time speed information sent by the speed sensor, and activates part of the RFID tags according to the vibration information, the infrared image information and the visible light image information respectively. Meanwhile, the edge computing server fuses and calculates the comprehensive tear probability according to the infrared image information and the visible light image information to determine whether there is a longitudinal tear. When there is a longitudinal tear, the edge computing server determines the tear location by using the high-precision timestamp T_capture of the visible light infrared image, the tag information and the speed information.

2. The conveyor belt tear location and traceability system based on RFID and multi-sensor fusion according to claim 1, characterized in that: The application layer comprises a data storage module, a user interface module and an alarm module. The data storage module uses a time series database to store raw data and tear event records, and retains at least 180 days of historical data. The user interface module uses a Web-based graphical user interface to support real-time monitoring, historical data query and visual backtracking. The alarm module supports audible and visual alarms. The edge computing server activates the RFID tags as follows: After denoising the vibration information sent by all positioning chips, time domain features and frequency domain features are extracted. Each time domain feature and frequency domain feature is used as a key feature, and a dynamic threshold is set for each key feature. When one or more key feature values continuously and significantly exceed the corresponding dynamic threshold for a set time, it is determined that the vibration at the position of the conveyor belt corresponding to the positioning chip is abnormal. When it is determined that the vibration is abnormal, the edge computing server activates the RFID tags associated with the positioning chip corresponding to the vibration abnormality. After preprocessing the visible light image sent by the industrial camera, visible light features are extracted from the visible light image data using a convolutional neural network, and the visible light features are fused and probabilistically inferred to output a binary classification probability, i.e., a visible light tear probability P_visual; when the visible light tear probability P_visual is greater than or equal to P, the activation radius is calculated, and all RFID tags within the activation radius range are activated with the industrial camera as the center point; After preprocessing the infrared image sent by the infrared thermal imager, the temperature anomaly area is extracted from the infrared image data using the Otsu threshold segmentation method, the infrared features are calculated, and then the infrared features are input into the pre-trained machine learning model to map the infrared features to a probability value, i.e., an infrared tear probability P_thermal; when the visible light tear probability P_visual is greater than or equal to P, the activation radius is calculated, and all RFID tags within the activation radius range are activated with the infrared thermal imager as the center point.

3. The RFID and multi-sensor fusion based conveyor belt tear location and traceability system as claimed in claim 2, wherein: The time domain features include root mean square value and kurtosis, and the frequency domain features include vibration frequency.

4. The conveyor belt tear location and traceability system based on RFID and multi-sensor fusion of claim 2, wherein: The activation radius R is calculated as follows: R = R_base + k × V; where R_base is a set radius, K is determined according to the delay, and is set to 0.5-1.5 seconds, and V is the speed of the conveyor belt.

5. The RFID and multi-sensor fusion based conveyor belt tear location and traceability system as claimed in claim 1, wherein: The edge computing server calculates the comprehensive tear probability as follows: The infrared data and visible light data are fused using Bayesian inference to calculate the comprehensive tear probability, and the calculation formula is: P_tear = w1 * P_thermal + w2 * P_visual; where the weights w1 and w2 are optimized through training; When the comprehensive tear probability P_tear ≥ P, it is determined that the conveyor belt has a longitudinal tear, and a high-precision timestamp T_capture is generated according to the positioning chip time, and at the same time, an alarm and tear positioning are triggered.

6. The RFID and multi-sensor fusion based conveyor belt tear location and traceability system as claimed in claim 1, wherein: The edge computing server performs tear positioning as follows: For each ultra-high frequency RFID reader, when the tear area reaches the ultra-high frequency reader reading, the tag information of the activated RFID tag is read, and the reading time T_tag is recorded, and then the following processing is performed: first, filter out the tag information with signal strength exceeding the preset range, second, select the strongest signal strength tag information from the filtered tag information, and determine the position L_tag according to the tag information; finally, calculate the tear position L_tear = L_tag + V × ΔT, where ΔT is the time difference, ΔT = T_tag - T_capture; V is the speed of the conveyor belt, which is measured by a speed sensor; After weighting and averaging all the tear positions calculated by the ultra-high frequency RFID readers, the final tear position is obtained, and the weight value is a set value.

7. The RFID and multi-sensor fusion based conveyor belt tear location and traceability system as claimed in claim 6, wherein: When calculating the final tear position, the data calculated by the ultra-high frequency RFID reader close to the industrial camera has a high weight, and the data calculated by the ultra-high frequency RFID reader far from the industrial camera has a low weight.

Images