Rapid belt tearing detection and early warning device and method based on artificial intelligence

Abstract

The invention provides a belt tearing rapid detection and early warning device and method based on artificial intelligence, and relates to the technical field of equipment detection. According to the device provided by the invention, a monitoring unit is arranged at a blanking port, and a high-speed CCD (Charge Coupled Device) camera mounted at a 45-degree elevation angle works cooperatively with a laser transmitter to capture a line laser image at the bottom of a belt in real time; the core technology of the rapid detection and early warning method comprises the following steps: 1) an improved Sobel enhancement operator is used for inhibiting dust and water stain interference while keeping a real edge; 2) performing multi-scale LBP feature analysis, and accurately extracting crack textures in combination with dynamic threshold segmentation; the method comprises the following steps of (1) building a deformation index DI model to realize grading early warning, (2) building a deformation index DI model to realize grading early warning, (3) capturing space-time continuity through a 3D convolution kernel and realizing longitudinal crack detection in cooperation with the curvature quantification model, (3) building a distortion index DI model to realize grading early warning, (4) iron ore field verification shows that the transverse crack detection rate and the longitudinal crack early warning accuracy are both higher than those in the prior art, the false alarm rate in a dust environment is as high as 7%, and material leakage and safety accidents are effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of equipment testing technology, and in particular to an artificial intelligence-based rapid detection and early warning device and method for belt tearing. Background Technology

[0002] In industrial production, belt conveyors are one of the important methods of material transportation; however, belts are prone to tearing during long-term operation. Once a belt tears, it can not only lead to material leakage and affect the normal operation of production, but also cause safety accidents, resulting in economic losses and environmental pollution.

[0003] Therefore, we need to effectively monitor belt tearing to promptly detect and maintain the belt's operating condition, ensuring the normal operation of related equipment. However, existing belt tear monitoring methods have many shortcomings, specifically: Some traditional monitoring methods rely on manual inspections, which are inefficient and susceptible to human error, making it difficult to detect early signs of belt tears in a timely manner. For example, in some large mines or factories, the belt conveyor lines are long, and the intervals between manual inspections are long, which may prevent timely detection of belt tears in their early stages.

[0004] Another type of monitoring technology uses simple sensors, such as contact sensors installed under the belt to detect material leakage and indirectly determine whether the belt has torn. However, this method can only detect the problem when the belt has already torn to a certain extent and material leakage has begun, failing to provide early warning and effectively prevent the serious consequences caused by belt tears. In addition, due to the complex on-site environment, sensors are easily affected by factors such as material accumulation and moisture, leading to inaccurate detection results.

[0005] Therefore, we need to provide an artificial intelligence-based rapid detection and early warning device and method for belt tearing to solve the technical problems of low efficiency, inability to provide early warning, and susceptibility to environmental interference in existing belt tearing monitoring methods. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide an artificial intelligence-based rapid detection and early warning device and method for belt tearing. It aims to solve the technical defects of traditional manual inspection, such as low efficiency and the inability of contact sensors to provide early warning. Workers can quickly inspect and repair the belt based on the alarm information, thus avoiding coal leakage and production interruption accidents caused by further expansion of belt tearing.

[0007] To achieve the above objectives, this application proposes an artificial intelligence-based rapid belt tear detection and early warning device, which is installed at the belt discharge port of the equipment to be inspected, and includes: A high-speed CCD optical camera is used to acquire image information of the bottom of the belt in the monitoring area; wherein, the high-speed CCD optical camera is installed at the bottom of the belt and at a 45° elevation angle to the horizontal plane, and the shooting center is aligned with the center line of the belt; A laser emitter is used to generate a linear laser beam and irradiate the bottom of the belt; wherein the linear laser beam is uniformly applied to the bottom of the belt, and the laser beam stably covers the monitoring area during belt movement. An AI edge processor is used to detect belt tears by analyzing the image information at the bottom of the belt; wherein the belt tear detection is based on AI image recognition algorithm processing and outputs the belt tear detection result; The monitoring terminal is used to continuously monitor the belt tear monitoring results, and when the belt shows signs of tearing at the bottom, it will issue a belt tear warning to the device under test and send an alarm signal and abnormal location information to the associated warning device. The communication power distribution module is electrically connected to the high-speed CCD optical camera, laser emitter, AI edge processor and monitoring terminal respectively, and establishes power supply link and communication link respectively.

[0008] As a further solution, the communication power distribution module establishes a communication link via a high-speed data line.

[0009] As a further solution, the laser emitter generates a line laser beam with an arc-shaped profile distributed at the bottom of the belt, and the line laser beam performs dynamic scanning as the belt moves.

[0010] As a further solution, the camera parameters of the high-speed CCD optical camera are adapted to the ambient light conditions and belt speed; wherein, the camera parameters include exposure time, camera frame rate and sensitivity.

[0011] On the other hand, the present invention also provides an artificial intelligence-based rapid detection and early warning method for belt tearing, applied to an artificial intelligence-based rapid detection and early warning device for belt tearing as described in any of the preceding claims, comprising: Step 1: Start the equipment to be tested and drive the belt to work, and obtain the ambient light conditions and belt speed; Step 2: Activate the laser emitter to generate a line laser beam, and after the belt running speed reaches the threshold, dynamically scan the bottom of the belt by irradiating it with laser light; Step 3: Set the camera parameters according to the ambient light conditions and belt speed; after completing the parameter configuration, start the high-speed CCD optical camera to acquire image information of the bottom of the belt in the monitoring area; Step 4: Use an AI edge processor to detect belt tearing in the bottom image information of the belt and output the belt tearing detection results; Step 5: The monitoring terminal continuously monitors the belt tear monitoring results and determines whether there are any signs of impending tearing at the bottom of the belt; when signs of impending tearing appear, a belt tear warning is issued for the equipment under test and Step 6 is executed. Step 6: Based on the correlation calculation between the camera frame rate and the belt running speed, obtain the abnormal location information, and send the alarm signal and abnormal location information to the associated early warning device.

[0012] As a further solution, belt tear monitoring can be performed based on the following steps: The image information of the bottom of the belt is obtained and used as input images, which are then input into the first branch and the second branch respectively; wherein, First branch: Based on laser morphology analysis, the characteristics of the first branch are obtained; Second branch: Based on texture anomaly detection, the features of the second branch are obtained; The first branch features and the second branch features are input into the feature fusion layer to obtain the crack feature vector; The distortion index DI is calculated based on the crack feature vector to obtain the distortion index DI. The alarm output is graded based on the obtained distortion index DI.

[0013] As a further solution, the laser morphology analysis process includes: Dynamic noise suppression of the input image is performed based on the Sobel edge enhancement operator; Laser center line extraction from input image based on Steger algorithm sub-pixel localization; Morphological distortion detection based on laser centerline; including: Curvature abrupt change point detection is performed to obtain the morphological distortion value k; Width anomaly detection is performed to obtain the width distortion value ΔW.

[0014] As a further solution, the texture anomaly detection process includes: The input image is processed based on LBP feature extraction to obtain multi-scale LBP features; The crack space is obtained by capturing the crack space continuity of the input image using 3D convolution; Based on multi-scale LBP features and crack confidence, the belt tearing probability score in the crack space is calculated. Crack orientation analysis was performed based on Hough transform to obtain the crack angle distribution.

[0015] As a further solution, the AI ​​image recognition algorithm adopts a training strategy based on adversarial generation and transfer learning. Crack samples are obtained through adversarial generation and used for model training in transfer learning, resulting in an AI image recognition model that takes crack feature vectors as input and generates heat maps of the corresponding crack locations as output.

[0016] As a further solution, the distortion index DI is calculated using the following formula: DI=ω_1·(ΔW / W_norm)+ω_2·|K|+ω_3·Crack_score; Where ω_1, ω_2, and ω_3 are the weight coefficients of the corresponding items, W_norm represents the baseline laser width under normal and undamaged belt conditions, and Crack_score represents the belt tear probability score.

[0017] Compared with related technologies, the process method for open-circuit cable assemblies and their assembly and testing provided by the present invention has the following advantages: 1. This invention installs an intelligent monitoring unit at a fixed point at the material inlet. Utilizing AI image recognition technology, it can detect early signs of belt tearing in a timely manner, greatly improving the accuracy and timeliness of monitoring. In contrast, traditional methods rely heavily on manual inspections or simple sensors, which cannot achieve accurate early monitoring and are highly dependent on sensor performance. 2. This invention uses line laser scanning and high-speed camera image acquisition to obtain information about the bottom of the conveyor belt in a non-contact manner, which is unaffected by environmental factors such as material accumulation and humidity, thus ensuring the stability and reliability of monitoring; while traditional contact sensors are easily affected by environmental interference. 3. This invention can immediately alarm when an anomaly is detected and record the distortion location and the corresponding crack location thermal map, which facilitates monitoring personnel to quickly locate and handle the problem and effectively avoid the serious consequences caused by belt tearing; while traditional monitoring methods cannot alarm in time after a problem is detected and it is difficult to accurately record the fault location. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0020] Figure 1A schematic diagram of a rapid belt tear detection and early warning device based on artificial intelligence provided by the present invention; Figure 2 A schematic diagram illustrating the steps of a rapid detection and early warning method for belt tearing based on artificial intelligence provided by the present invention; Figure 3 This is a schematic diagram of the belt tear monitoring steps provided by the present invention.

[0021] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0023] Example 1 Please see Figure 1 This application provides an artificial intelligence-based rapid belt tear detection and early warning device, which is installed at the belt discharge port of the equipment to be detected, and includes: A high-speed CCD optical camera is used to acquire image information of the bottom of the belt in the monitoring area; wherein, the high-speed CCD optical camera is installed at the bottom of the belt and at a 45° elevation angle to the horizontal plane, and the shooting center is aligned with the center line of the belt; A laser emitter is used to generate a linear laser beam and irradiate the bottom of the belt; wherein the linear laser beam is uniformly applied to the bottom of the belt, and the laser beam stably covers the monitoring area during belt movement. An AI edge processor is used to detect belt tears by analyzing the image information at the bottom of the belt; wherein the belt tear detection is based on AI image recognition algorithm processing and outputs the belt tear detection result; The monitoring terminal is used to continuously monitor the belt tear monitoring results, and when the belt shows signs of tearing at the bottom, it will issue a belt tear warning to the device under test and send an alarm signal and abnormal location information to the associated warning device. The communication power distribution module is electrically connected to the high-speed CCD optical camera, laser emitter, AI edge processor and monitoring terminal respectively, and establishes power supply link and communication link respectively.

[0024] It should be noted that: In this embodiment, a belt tear rapid detection and early warning device is installed at the material drop point of the belt; the high-speed CCD optical camera is installed at an elevation angle of approximately 45° to the horizontal on both sides of the bottom of the belt. This angle is precisely designed to ensure that the camera can capture image information of the bottom of the belt to the maximum extent and avoid blind spots caused by the angle; the laser emitter is installed at a position that allows the line laser beam to be evenly hit on the bottom of the belt and ensures that the laser beam stably covers the monitoring area during the belt movement.

[0025] The AI ​​edge processor is connected to the high-speed CCD optical camera and laser emitter via high-speed data cables, responsible for receiving and processing image data. The communication and power distribution module provides a stable power supply to the entire monitoring unit and is responsible for transmitting data to the monitoring terminal. The wiring between the devices is carefully routed to avoid mutual interference and ensure the stable operation of the entire system.

[0026] A laser emitter continuously emits a linear laser beam, which, after striking the bottom of the belt, follows the belt's curved contour. As the belt moves at a certain speed, the linear laser beam dynamically scans the belt. A high-speed CCD optical camera operates at an extremely high frame rate, acquiring images of the belt's bottom with the linear laser beam in real time.

[0027] To ensure the clarity and accuracy of image acquisition, the camera's exposure time, ISO, and other parameters were optimized based on the ambient lighting conditions and belt speed. Simultaneously, the camera was equipped with a special optical lens that effectively filtered out stray light from the environment, highlighting the line laser beam and the texture features at the bottom of the belt, resulting in higher-quality images and providing a reliable data foundation for subsequent image analysis.

[0028] Image data acquired by a high-speed CCD optical camera is transmitted to the AI ​​edge processor with extremely low latency via a high-speed data transmission line in the communication power distribution module.

[0029] The AI ​​edge processor incorporates an image recognition algorithm specifically optimized for belt tear detection. This algorithm first preprocesses the acquired images, including noise reduction and contrast enhancement, to improve image quality and facilitate subsequent feature extraction. Next, by analyzing the texture features at the bottom of the belt, the algorithm can identify the normal shape and distortion of the linear laser beam.

[0030] When cracks, tears, or other signs of impending tearing are present at the bottom of the conveyor belt, the abnormal portion will be distorted when passing through a line laser. The algorithm will accurately identify these distortion features and compare them with preset normal image features. To improve the accuracy and efficiency of identification, the AI ​​edge processor uses deep learning technology, training with a large number of normal and abnormal conveyor belt image samples to continuously optimize the algorithm model, enabling it to adapt to various complex on-site working conditions.

[0031] If the AI ​​edge processor analyzes and determines that the image is abnormal, i.e., the belt may be at risk of tearing, it will immediately send an alarm signal to the monitoring terminal through the communication power distribution module.

[0032] Alarm signals can take various forms, such as audible and visual alarms, SMS notifications, and pop-up prompts in the monitoring system, to ensure that monitoring personnel receive information in a timely manner.

[0033] Meanwhile, the AI ​​edge processor records the location of the linear laser distortion and, by correlating it with parameters such as belt speed and camera frame rate, accurately determines the specific coordinates of the abnormal location on the belt. This alarm information and distortion location information are transmitted to monitoring personnel in real time via the communication network. Based on this information, monitoring personnel can quickly arrange for maintenance personnel to inspect and repair the belt, effectively preventing further tearing and ensuring normal production.

[0034] Example 2 Based on Embodiment 1, this embodiment also provides an artificial intelligence-based rapid detection and early warning method for belt tearing, applied to an artificial intelligence-based rapid detection and early warning device for belt tearing as described in any of the preceding embodiments, comprising: Step 1: Start the equipment to be tested and drive the belt to work, and obtain the ambient light conditions and belt speed; Step 2: Activate the laser emitter to generate a line laser beam, and after the belt running speed reaches the threshold, dynamically scan the bottom of the belt by irradiating it with laser light; Step 3: Set the camera parameters according to the ambient light conditions and belt speed; after completing the parameter configuration, start the high-speed CCD optical camera to acquire image information of the bottom of the belt in the monitoring area; Step 4: Use an AI edge processor to detect belt tearing in the bottom image information of the belt and output the belt tearing detection results; Step 5: The monitoring terminal continuously monitors the belt tear monitoring results and determines whether there are any signs of impending tearing at the bottom of the belt; when signs of impending tearing appear, a belt tear warning is issued for the equipment under test and Step 6 is executed. Step 6: Based on the correlation calculation between the camera frame rate and the belt running speed, obtain the abnormal location information, and send the alarm signal and abnormal location information to the associated early warning device.

[0035] It should be noted that traditional methods rely heavily on manual inspections or simple sensors, which cannot achieve accurate early monitoring and are highly dependent on sensor performance. This embodiment utilizes AI image recognition technology, which can detect early signs of belt tearing, greatly improving the accuracy and timeliness of monitoring.

[0036] Traditional monitoring methods cannot promptly trigger alarms after a problem is detected, nor can they accurately record the location of the fault. This technology can immediately trigger an alarm upon detecting an anomaly and record the location of the distortion and the corresponding crack location on a thermal map, facilitating monitoring personnel to quickly locate and handle the problem, effectively avoiding the serious consequences of belt tearing.

[0037] Furthermore, such as Figure 3 As shown, belt tear monitoring is performed based on the following steps: The image information of the bottom of the belt is obtained and used as input images, which are then input into the first branch and the second branch respectively; wherein, First branch: Based on laser morphology analysis, the characteristics of the first branch are obtained; Second branch: Based on texture anomaly detection, the features of the second branch are obtained; The first branch features and the second branch features are input into the feature fusion layer to obtain the crack feature vector; The distortion index DI is calculated based on the crack feature vector to obtain the distortion index DI. The alarm output is graded based on the obtained distortion index DI.

[0038] Specifically, the laser morphology analysis and processing workflow includes: Dynamic noise suppression of the input image is performed based on the Sobel edge enhancement operator; Laser center line extraction from input image based on Steger algorithm sub-pixel localization; Morphological distortion detection based on laser centerline; including: Curvature abrupt change point detection is performed to obtain the morphological distortion value k; Width anomaly detection is performed to obtain the width distortion value ΔW.

[0039] It should be noted that the Sobel edge enhancement operator is: Sobel_enhanced=|▽I|*exp(-(I-μ)^2 / 2σ^2); where |▽I| represents the pixel gradient magnitude, μ represents the mean gray level of the sliding window, σ is the standard deviation of the window gray level, and I represents the current pixel gray level.

[0040] The morphological distortion value k = |d²y / dx²| / (1+(dy / dx)^2)^(3 / 2); where dy / dx represents the slope of the stripe tangent (first derivative), d²y / dx² represents the rate of change of the slope (second derivative), and the denominator is the arc length correction factor; the width anomaly detection is based on dynamic threshold segmentation (OTSU improvement).

[0041] Specifically, the texture anomaly detection process includes: The input image is processed based on LBP feature extraction to obtain multi-scale LBP features; The crack space is obtained by capturing the crack space continuity of the input image using 3D convolution; Based on multi-scale LBP features and crack confidence, the belt tearing probability score in the crack space is calculated. Crack orientation analysis was performed based on Hough transform to obtain the crack angle distribution.

[0042] It should be noted that in belt tear detection, we use multi-scale LBP (i.e., different radii R and different numbers of points P) to capture texture features at different scales. By analyzing the LBP features of the belt surface image, cracked areas can be identified (because the texture pattern of cracked areas is different from that of normal areas); LBP feature extraction is performed using the following formula: LBP_{P,R}=∑_{p=0}^{P-1}s(g_p-g_c)·2^p; Where P is the number of pixels selected within a circular neighborhood of radius R (neighborhood points, e.g., P=30), R is the radius of the circle containing the neighboring pixels with the center pixel as the center (pixel radius, e.g., R=4), g_c is the gray value of the center pixel, g_p is the gray value of the p-th neighboring pixel on the circle of radius R, and s(*) is a threshold function (also called the sign function).

[0043] The 3D convolution kernel is 6×6×4 in size to capture the spatial continuity of the crack; then the severity of the crack is quantified by calculating the belt tear probability score (Crack_score); the belt tear probability score calculation formula is: Crack_score=∑(LBP_{diff}>τ) / Area; where LBP_diff: represents the difference of LBP features between the current image and the normal belt texture, τ: is the set threshold used to judge whether the difference is significant, and Area: is the area of ​​the crack space.

[0044] Finally, the Hough transform is used to detect the angular distribution, which can determine whether the crack is transverse or longitudinal.

[0045] Furthermore, the AI ​​image recognition algorithm adopts a training strategy based on adversarial generation and transfer learning. Crack samples are obtained through adversarial generation and used for model training in transfer learning, resulting in an AI image recognition model that takes crack feature vectors as input and generates heat maps of corresponding crack locations as output.

[0046] Specifically, crack sample generation can be performed using the following code: # Generator Structure (U-Net Improvement) defgenerator(): inputs=Input(shape=(256,256,1)) # Encoder: 4-layer downsampling (with residual connections) #Decoder: 4-layer upsampling (injecting crack feature vectors) #Output: Generated image + heatmap of crack location returnModel(inputs,[gen_img,heatmap]) # Discriminator: PatchGAN structure defdiscriminator(): returnSequential([ Conv2D(64,4,strides=2,padding='same'), LeakyReLU(0.2), ...#5 ​​layers of convolution Conv2D(1,4,padding='same') ]) Among them, different crack shapes are generated by controlling the latent space vector z_crack: such as: transverse crack: z_crack=[0.8,-0.3,0.1]; longitudinal crack: z_crack=[-0.2,0.9,0.4].

[0047] Transfer learning mechanism: The pre-trained model uses EfficientNet-B3 pre-trained on ImageNet as the backbone and performs domain adaptation; among which, Phase 1: Freeze the backbone network and train only the dual-branch heads. Phase 2: Unfreeze the last 3 layers and use cosine annealing learning rate scheduling. Loss function: Loss = α·MSE + β·Focal_Loss + γ·SSIM; where α·MSE is the mean squared error, β·Focal_Loss is the focal loss, and γ·SSIM is the structural similarity.

[0048] Furthermore, in this embodiment, the distortion index DI calculation quantifies the degree of belt damage into a comprehensive index (DI) by weighted fusion of three dimensions: laser deformation, curvature change, and texture anomaly, achieving graded and precise early warning from "observation to shutdown". Its core innovation lies in multi-feature collaboration, which not only overcomes the limitations of traditional single sensor technology, but also improves anti-interference and early recognition capabilities through algorithm optimization.

[0049] Specifically, the distortion index DI is calculated using the following formula: DI=ω_1·(ΔW / W_norm)+ω_2·|K|+ω_3·Crack_score; Where ω_1, ω_2, and ω_3 are the weight coefficients of the corresponding items, W_norm represents the baseline laser width under normal and undamaged belt conditions, and Crack_score represents the belt tear probability score.

[0050] It should be noted that when the belt shows signs of impending tearing (such as cracks or dents), surface deformation causes the laser beam to diffuse or focus, significantly increasing ΔW. This is a core indicator for identifying macroscopic deformation. Furthermore, belt cracks cause localized bending of the laser stripes (e.g., transverse cracks interrupt the stripes, longitudinal cracks distort them), and |K| can effectively quantify this degree of bending, used to capture the spatial continuity of the crack. Crack_score, based on texture feature analysis, quantifies the density and distribution range of microcracks, and is particularly adept at detecting microcracks that are difficult to detect with the naked eye.

[0051] The weight allocation can be flexibly adjusted according to different working conditions and actual needs. Parameter fusion can avoid the influence of environmental noise (such as dust and water mist) on a single feature. In a specific embodiment, ω_1:ω_2:ω_3=0.6:0.3:0.1. Among them, (ΔW / W_norm) is the most intuitive manifestation of tearing and is given the highest weight. |K| and crack texture (Crack_score) capture deformation and micro-damage respectively, and together improve the accuracy of comprehensive judgment.

[0052] Furthermore, this embodiment also establishes a graded early warning mechanism based on the DI value. By quantifying the degree of damage (DI value), it achieves graded responses, avoids false alarms that disrupt production, and proactively shuts down operations in the event of significant risks to ensure safety. In a specific embodiment, the response measures and alarm levels are compared as shown in Table 1 below: Table 1. DI Value - Classification and Early Warning Comparison Table This technical solution was piloted in the belt conveyor system of a large coal mine. The mine's belt conveyor line is approximately 2 kilometers long, transporting thousands of tons of coal daily. After installing the intelligent belt tear monitoring unit, and following one month of operation and monitoring, three instances of minute cracks at the bottom of the belt were successfully detected. Upon detecting linear laser distortion images, the monitoring unit promptly issued an alarm signal and accurately recorded the crack location. Based on the alarm information, staff quickly inspected and repaired the belt, preventing further expansion of the tear and potential coal leakage and production interruption. This application validated the effectiveness and reliability of this technical solution in a real industrial production environment.

[0053] The test results at a certain iron ore mine are shown in Table 2: Table 2 Comparison of Test Results *Test conditions: belt speed 4.5m / s, dust concentration 200mg / m³, ambient illuminance <50lux*.

[0054] The above are only some embodiments of this application and do not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A rapid belt tear detection and early warning device based on artificial intelligence, installed at the belt discharge port of the equipment to be detected, characterized in that, include: A high-speed CCD optical camera is used to acquire image information of the bottom of the belt in the monitoring area; wherein, the high-speed CCD optical camera is installed at the bottom of the belt and at a 45° elevation angle to the horizontal plane, and the shooting center is aligned with the center line of the belt; A laser emitter is used to generate a linear laser beam and irradiate the bottom of the belt; wherein the linear laser beam is uniformly applied to the bottom of the belt, and the laser beam stably covers the monitoring area during belt movement. An AI edge processor is used to detect belt tears by analyzing the image information at the bottom of the belt; wherein the belt tear detection is based on AI image recognition algorithm processing and outputs the belt tear detection result; The monitoring terminal is used to continuously monitor the belt tear monitoring results, and when the belt shows signs of tearing at the bottom, it will issue a belt tear warning to the device under test and send an alarm signal and abnormal location information to the associated warning device. The communication power distribution module is electrically connected to the high-speed CCD optical camera, laser emitter, AI edge processor and monitoring terminal respectively, and establishes power supply link and communication link respectively.

2. The method for rapid detection and early warning of belt tearing based on artificial intelligence according to claim 1, characterized in that, The communication power distribution module establishes a communication link via a high-speed data line.

3. The method for rapid detection and early warning of belt tearing based on artificial intelligence according to claim 1, characterized in that, The line laser beam generated by the laser emitter is distributed in an arc shape at the bottom of the belt, and the line laser beam performs dynamic scanning as the belt moves.

4. The method for rapid detection and early warning of belt tearing based on artificial intelligence according to claim 1, characterized in that, The camera parameters of the high-speed CCD optical camera are adapted to the ambient light conditions and belt speed; the camera parameters include exposure time, camera frame rate and sensitivity.

5. A method for rapid detection and early warning of belt tearing based on artificial intelligence, applied to a rapid detection and early warning device for belt tearing based on artificial intelligence as described in any one of claims 1 to 4, characterized in that, include: Step 1: Start the equipment to be tested and drive the belt to work, and obtain the ambient light conditions and belt speed; Step 2: Activate the laser emitter to generate a line laser beam, and after the belt running speed reaches the threshold, dynamically scan the bottom of the belt by irradiating it with laser light; Step 3: Set the camera parameters according to the ambient light conditions and belt speed; after completing the parameter configuration, start the high-speed CCD optical camera to acquire image information of the bottom of the belt in the monitoring area; Step 4: Use an AI edge processor to detect belt tearing in the bottom image information of the belt and output the belt tearing detection results; Step 5: The monitoring terminal continuously monitors the belt tear monitoring results and determines whether there are any signs of impending tearing at the bottom of the belt; when signs of impending tearing appear, a belt tear warning is issued for the equipment under test and Step 6 is executed. Step 6: Based on the correlation calculation between the camera frame rate and the belt running speed, obtain the abnormal location information, and send the alarm signal and abnormal location information to the associated early warning equipment.

6. The method for rapid detection and early warning of belt tearing based on artificial intelligence according to claim 5, characterized in that, Belt tear monitoring is performed based on the following steps: The image information of the bottom of the belt is obtained and used as input images, which are then input into the first branch and the second branch respectively; wherein, First branch: Based on laser morphology analysis, the characteristics of the first branch are obtained; Second branch: Based on texture anomaly detection, the features of the second branch are obtained; The first branch features and the second branch features are input into the feature fusion layer to obtain the crack feature vector; The distortion index DI is calculated based on the crack feature vector to obtain the distortion index DI. The alarm output is graded based on the obtained distortion index DI.

7. The method for rapid detection and early warning of belt tearing based on artificial intelligence according to claim 6, characterized in that, The laser morphology analysis and processing flow includes: Dynamic noise suppression of the input image is performed based on the Sobel edge enhancement operator; Laser center line extraction from input image based on Steger algorithm sub-pixel localization; Morphological distortion detection based on laser centerline; including: Curvature abrupt change point detection is performed to obtain the morphological distortion value k; Width anomaly detection is performed to obtain the distorted width value ΔW.

8. The method for rapid detection and early warning of belt tearing based on artificial intelligence according to claim 6, characterized in that, The texture anomaly detection process includes: The input image is processed based on LBP feature extraction to obtain multi-scale LBP features; The crack space is obtained by capturing the crack space continuity of the input image using 3D convolution; Based on multi-scale LBP features and crack confidence, the belt tearing probability score in the crack space is calculated. Crack orientation analysis was performed based on Hough transform to obtain the crack angle distribution.

9. A rapid detection and early warning method for belt tearing based on artificial intelligence according to claim 6, characterized in that, The AI ​​image recognition algorithm adopts a training strategy based on adversarial generation and transfer learning. Crack samples are obtained through adversarial generation and used for model training in transfer learning, resulting in an AI image recognition model that takes crack feature vectors as input and generates heat maps of the corresponding crack locations as output.

10. A rapid detection and early warning method for belt tearing based on artificial intelligence according to any one of claims 7 and 8, characterized in that, The distortion index (DI) is calculated using the following formula: DI=ω_1·(ΔW / W_norm)+ω_2·|K|+ω_3·Crack_score; Where ω_1, ω_2, and ω_3 are the weight coefficients of the corresponding items, W_norm represents the baseline laser width under normal and undamaged belt conditions, and Crack_score represents the belt tear probability score.

Images