Belt conveyor belt defect detection device and defect detection and positioning method

Abstract

The invention discloses a belt conveyor belt defect detection device and a defect detection and positioning method, the belt conveyor belt defect detection device comprises three sets of detection sensors, two sets of detection sensors are arranged below a conveying belt supported by a supporting roller, and two side non-working surface section images collected by the two sets of detection sensors are mutually intersected; and the other set of detection sensor is arranged above the conveying belt to detect a complete transverse working face image of the conveying belt. The 3D camera is used as a detection sensor to obtain the point cloud data of the belt surface of the belt conveyor, belt tearing and belt surface defects can be detected, the belt butt joint is used as a reference point to quickly position the relative position of the belt tearing and the belt surface defects, a reference basis is provided for belt maintenance, and the detection accuracy is improved. The maintenance efficiency is improved, help is provided for belt operation and maintenance, and a guarantee is provided for belt safety and production safety.

Description

TECHNICAL FIELD

[0001] The present application relates to a belt conveyor belt defect detection device and a defect detection and positioning method. BACKGROUND

[0002] In the production process of the steel industry, the belt conveyor is the main equipment for conveying lump and bulk materials such as coal, ore, and mineral powder. In addition, the belt conveyor is also widely used in mining, ports, chemical industry, thermal power plants, and other fields. The belt conveyor has the characteristics of fast running speed, long transmission distance, and continuous transportation.

[0003] In the production process, the belt is prone to transverse rupture due to aging and longitudinal tearing due to sharp object puncture. If the belt abnormal working condition detection is not timely, especially the belt tearing, once the tearing accident occurs, it will cause several meters to several tens of meters of damage or even the whole belt to be scrapped within a few seconds or even a few minutes, which will cause tens of thousands or even hundreds of thousands of direct economic losses, and more indirect economic losses caused by affecting normal production. Therefore, how to accurately detect the belt defect and how to accurately locate the repair after detecting the belt defect are problems that people are constantly exploring and urgently need to solve. SUMMARY

[0004] In view of the above problems, the purpose of the present application is to provide a belt conveyor belt defect detection device and a defect detection and positioning method, which uses a 3D camera and a laser emitter as a detection sensor to obtain 3D point cloud data of the running belt surface, detects and analyzes the point cloud data, identifies whether there is a belt defect on the belt surface and whether there is a belt tearing, and uses the belt joint as a reference point to quickly locate the belt defect using a timestamp.

[0005] In order to achieve the above purpose, the scheme of the present application is: A belt conveyor belt defect detection device, comprising a support frame and a detection sensor, the detection sensor comprising a 3D image collector and a laser emitter, the support frame sequentially and spacedly provided with horizontal support rollers and oblique support rollers oppositely arranged on both sides of the horizontal support rollers, and guide rollers arranged at both longitudinal ends of the support frame, and a conveying belt is wrapped around the guide rollers at both ends along the surfaces of the horizontal support rollers and the oblique support rollers on both sides, wherein: the detection sensor has three sets, two sets of detection sensors in the three sets are arranged below the conveying belt supported by the support rollers, and the two sets of detection sensors are respectively directed towards the non-working surface sections on both sides of the bottom surface of the conveying belt, the images of the non-working surface sections on both sides of the conveying belt collected by the two sets of detection sensors intersect with each other, and the other set of detection sensors in the three sets is arranged above the conveying belt and directed towards the horizontal working surface of the conveying belt, and the image collected by the other set of detection sensors is a complete image of the horizontal working surface section of the conveying belt, and a speed sensor is further arranged on the support frame for measuring the speed of the conveying belt, and the detection sensor and the speed sensor are connected with an analysis control server.

[0006] The scheme further comprises: a plurality of silos are sequentially arranged above the conveying belt, the silos are used to supply materials to the conveying belt, and the conveying belt is taken as the front in the advancing direction, the two sets of detection sensors are arranged below the conveying belt on the front end side of the plurality of silos, and the two sets of detection sensors are staggered front and back, and the other set of detection sensors is arranged vertically above the wound conveying belt on the rear end side of the plurality of silos and opposite the rear end guide roller.

[0007] The scheme further comprises: the detection sensor is fixed in a closed shell, the side end face of the shell facing the conveying belt is a perspective window end face with transparent glass, the detection sensor faces the conveying belt through the perspective window, the shell is placed below or above the conveying belt through a support, the support comprises a stand column, a horizontal arm is arranged at the upper end of the stand column, hanging plates are connected and fixed downward on both sides of the front end of the horizontal arm, an angle-adjustable platform is connected and arranged on the hanging plates, and the shell is fixed on the platform.

[0008] The scheme further comprises: a dust removal mechanism is arranged on the shell, the dust removal mechanism comprises a gas cylinder and a guide shaft arranged on the side wall of the shell, a straight arm is driven by the telescopic arm of the gas cylinder to reciprocate along the guide shaft, a brush is arranged on the straight arm and contacts the transparent glass, and the reciprocating straight arm drives the brush to remove dust from the outer surface of the transparent glass.

[0009] A belt conveyor belt defect detection and positioning method is based on the belt defect detection device, the method is for each set of detection sensor defect detection and positioning method, the method includes belt defect detection method and belt defect positioning method, a mark is determined on the surface of the conveying belt as the positioning origin, the belt is started, the belt running speed is determined, the laser emitter emits laser irradiation to the surface of the conveying belt, the 3D image collector continuously acquires the image of the laser irradiation to determine the positioning origin, then, taking the positioning origin as the timing origin, each frame of image of the laser irradiation on the surface of the conveying belt is cyclically acquired; The belt defect detection method is: Step one, the continuous frame image data forms the point cloud data of the belt surface, each point in the point cloud data contains X, Y, Z three-dimensional coordinates, wherein the Z coordinate represents the distance between the belt surface and the 3D camera; Step two, using 3D detection algorithm to detect the point cloud data of the belt surface to obtain the current surface texture feature of the fluctuation change of the belt surface, according to the Z coordinate, whether the belt has defects is judged from the current surface texture feature; The belt defect positioning method is: Step one, each frame of image acquired after the timing origin is attached with a timestamp with the timing origin as the starting timing; Step two, when the belt has defects, the distance from the defect to the positioning origin is calculated according to the frame image timestamp of the defect and the belt running speed, and the relative position of the defect on the conveying belt is determined.

[0010] The scheme further is that the positioning origin is the butt joint of the belt forming a ring.

[0011] The scheme further is that the belt defects are divided into three levels according to the size, including: A level defect: the length is more than 30mm or the depth is more than 5mm, which belongs to the serious level; B level defect: the length is more than 20mm or the depth is more than 3mm, and does not reach A level, which belongs to the warning level; C level defect: the length is more than 10mm or the depth is more than 2mm, and does not reach B level, which belongs to the attention level.

[0012] The scheme is further: the current surface texture feature of the fluctuation change of the belt surface obtained by detecting the point cloud data of the belt surface by using the 3D detection algorithm is: the point cloud data is projected and converted into a two-dimensional gray image, including: taking the x coordinate of each data point as the column coordinate of the image pixel, taking the y coordinate as the row coordinate of the image pixel, and taking the z coordinate as the gray value of the image pixel, all data points of the point cloud data jointly constitute a two-dimensional depth distribution map of the belt surface, the partial derivative along the x axis and the y axis is calculated for each pixel point of the depth distribution map, the depth change rate is calculated to obtain a gradient distribution map to form the texture feature of the belt surface.

[0013] The scheme is further: the method further comprises determining the belt deviation, and setting a maximum allowed position distance of the belt edge, and when the actual edge of the belt exceeds the maximum allowed position distance, the belt deviation is determined.

[0014] The scheme is further: the belt defect detection further comprises judging whether the belt has defects from the surface texture feature, and comparing the current surface texture feature with a surface texture feature model to judge whether the belt has defects. The surface texture feature model is obtained by using the 3D detection algorithm to detect the point cloud data of the belt surface after the new belt is put on, and then using quantifiable feature indexes to describe and train the belt surface texture feature model, and the quantifiable feature indexes refer to different defects on the belt surface.

[0015] The application has the advantages that: the 3D camera is used as a detection sensor to obtain the 3D point cloud data of the belt surface, the belt tear and the belt surface defects can be detected by detecting the texture feature of the belt surface, the relative position of the belt tear and the belt defect is quickly positioned by using the belt joint as a reference point, a reference basis is provided for belt maintenance, the efficiency of maintenance is improved, help is provided for belt operation and maintenance, and safety of the belt and production safety are ensured.

[0016] The utility model will be described in detail in connection with the drawings and embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 It is the device layout schematic diagram of the application; Figure 2 It is the device installation structure schematic diagram of the application; Figure 3 It is the device support structure schematic diagram of the application; Figure 4 It is the device closed shell structure schematic diagram of the application; Figure 5 It is the point cloud data visualization image obtained by the 3D camera of the application; Figure 6is a schematic diagram of a belt defect.

[0018] The numbers in it respectively represent: support frame 1, horizontal support roller 101, inclined support roller 102, guide roller 103, detection sensor 2, two sets of detection sensor 201, another set of detection sensor 202, conveying belt 3, speed sensor 4, silo 5, shell 6, perspective window 601, transparent glass 7, support 8, column 801, cross arm 802, hanging plate 803, 803-1, platform 804, bolt 805, dust removal mechanism 9, dust removal mechanism support 901, air cylinder 902, sliding block 903, guide shaft 904, right angle arm 905, brush 906, communication interface 10. DETAILED DESCRIPTION

[0019] Example 1 A belt conveyor belt defect detection device, as shown in Figures 1 to 4 The belt defect detection device includes a support frame 1 and a detection sensor 2, the detection sensor 2 includes a 3D image collector and a laser emitter, the support frame is sequentially and spacedly provided with horizontal support rollers 101 and inclined support rollers 102 oppositely arranged on both sides of the horizontal support rollers, and guide rollers 103 are arranged at the longitudinal ends of the support frame, one of the guide rollers is a motor-driven belt drive roller, and a conveying belt 3 is wrapped around the two end guide rollers 103 along the surfaces of the horizontal support rollers 101 and the inclined support rollers 102 on both sides. Among them: the detection sensor 2 has three sets, two sets of detection sensors 201 in the three sets are arranged below the conveying belt 3 supported by the support rollers, and the two sets of detection sensors face the non-working surface sections of the bottom surface of the conveying belt laterally on both sides, respectively, the images of the non-working surface sections of the conveying belt on both sides collected by the two sets of detection sensors intersect with each other, which ensures that the complete non-working surface is collected, another set of detection sensors 202 in the three sets is arranged above the conveying belt 3 supported by the support rollers and faces the lateral working surface of the conveying belt, the image collected by the other set of detection sensors 202 is a complete surface image of the lateral working section of the conveying belt, a speed sensor 4 is further arranged on the support frame 1 for measuring the speed of the conveying belt 3, and the detection sensor 2 and the speed sensor 4 are connected with an analysis control server (not shown).

[0020] A plurality of silos 5 are sequentially arranged above the conveying belt 3 supported by the supporting rollers, the lower part of the silos 5 is provided with a discharge port for feeding the conveying belt 3, the conveying direction of the conveying belt 3 is indicated by arrow a, the two sets of detection sensors 201 are arranged below the conveying belt at the front end side of the discharge port of the last silo of the plurality of silos 5, the two sets of detection sensors 201 are staggered front and back, the other set of detection sensors 202 is arranged above the conveying belt 3 at the rear end side of the plurality of silos 5, and is vertically above the rear end guiding roller at the rear end side of the plurality of silos. The conveying belt on the rear end guiding roller is taut and flat, and the image obtained above it can fully reflect the surface state of the conveying belt. Of course, it can also be slightly inclined upward, in any case, it is ensured that it is above the flat conveying belt.

[0021] Due to the poor working environment of the conveying belt, in order to ensure the accuracy of detection and prolong the service life of the sensor, as shown in Figure 3 and Figure 4 The detection sensor 2 is fixed in a closed shell 6, the side end face of the shell 6 facing the conveying belt is provided with a transparent glass 7, the detection sensor 2 is opposite to the conveying belt 3 through the transparent window 601, the shell 6 is placed below or above the conveying belt 3 through a support 8, the support 8 includes a vertical column 801, a horizontal arm 802 is arranged at the upper end of the vertical column, hanging plates 803 are connected and fixed on both sides of the front end of the horizontal arm 802 downward, and an angle-adjustable platform 804 is connected and arranged on the hanging plate 803, and the shell 6 is fixed on the platform 804.

[0022] In the embodiment, a dust removal mechanism 9 is arranged on the shell 6, the dust removal mechanism 9 includes a gas cylinder 902 and a guide shaft 904 arranged on the side wall of the shell 6 through a dust removal mechanism support 901, a sliding block 903 is connected to the telescopic arm of the gas cylinder 902, the sliding block 903 is sleeved on the guide shaft 904 to drive a right-angle arm 905 to reciprocate along the guide shaft 904, a brush 906 in contact with the transparent glass 7 is arranged on the right-angle arm 905, the reciprocating right-angle arm 905 drives the brush 906 to remove the dust on the outer surface of the transparent glass 7, a communication interface 10 is arranged on the side wall of the shell 6, and the signal of the detection sensor 2 is connected to an analysis control server through the communication interface 10.

[0023] The hanging plate 803 is provided with at least two vertical grooves 803-1, the platform 804 is connected and fixed with the grooves 803-1 through bolts 805, the adjustment of the bolts 805 in the upper and lower fixed positions of the two grooves 803-1 forms the adjustment of the inclination and vertical direction of the platform 804, and then the position and angle adjustment of the shell transparent window 601 towards the belt 3 are realized.

[0024] Embodiment 2: A belt defect detection and positioning method based on the belt defect detection device of embodiment 1, the method is a defect detection and positioning method for each set of detection sensors 2, the content of embodiment 1 is applicable to this embodiment, the method includes a belt defect detection method and a belt defect positioning method, a mark is determined on the surface of the conveying belt as the positioning origin, this embodiment uses the butt joint formed in a ring shape of the belt as the positioning origin, when there are multiple butt joints, the last detected butt joint mark is used as the positioning origin.

[0025] Start the belt running, determine the belt running speed, the laser emitter emits laser irradiation to the surface of the conveying belt, the 3D image collector continuously acquires the image of the laser irradiation to determine the positioning origin, and then each frame of image of the laser irradiation on the surface of the conveying belt is cyclically acquired with the positioning origin as the timing origin; The belt defect detection method is: Step one, form the point cloud data of the belt surface from the continuous frame image data, each point in the point cloud data contains X, Y, Z three-dimensional coordinates, wherein the Z coordinate represents the distance between the belt surface and the 3D camera; Step two, use the 3D detection algorithm to detect the point cloud data of the belt surface to obtain the current surface texture feature of the high and low fluctuation of the belt surface, and judge whether the belt has defects according to the Z coordinate from the current surface texture feature; The belt defect positioning method is: Step one, attach a timestamp with the timing origin as the starting time to each frame of image continuously acquired after the timing origin; Step two, when the belt has defects, calculate and determine the distance from the defect to the positioning origin according to the frame image timestamp when the defect occurs and the belt running speed, and further determine the relative position of the defect on the conveying belt.

[0026] For example, the timing origin is recorded as t0, the frame image timestamp when the belt defect occurs is recorded as t1, the longitudinal coordinate position of the belt defect in the length direction of the conveying belt is recorded as Px, the transverse coordinate of the belt defect in the width direction of the conveying belt is recorded as Py, and the speed of the belt is recorded as v. Then the position of the defect on the conveying belt relative to the positioning origin is Px = v * (t1– t0). In the transverse direction of the belt (i.e. the width direction), the center line of the belt width is used as the reference point (i.e. 0 point), and the positive direction is from left to right, so that the transverse coordinate (recorded as Py) of the belt defect can be obtained.

[0027] In the embodiment, the current surface texture feature of the fluctuation of the belt surface is obtained by detecting the point cloud data of the belt surface using a 3D detection algorithm, which is: projecting and converting the point cloud data into a two-dimensional gray image, including: taking the x coordinate of each data point as the column coordinate of the image pixel, taking the y coordinate as the row coordinate of the image pixel, and taking the z coordinate as the gray value of the image pixel, all data points of the point cloud data together constitute a two-dimensional depth distribution map of the belt surface, and the partial derivative of each pixel point of the depth distribution map along the x axis and the y axis is calculated to obtain the gradient distribution map to form the texture feature of the belt surface.

[0028] The method further comprises determining the belt edge: taking the transverse width of the belt as w, and setting the maximum allowed offset position of the left and right edges of the belt with reference to half of the belt width, i.e. w / 2. The detection of the two edges of the belt by another set of detection sensors on the upper side of the conveying belt forms the deviation detection of the belt, and therefore the method further comprises determining the deviation of the belt: setting the maximum allowed offset position of the belt edge, and when the actual edge of the belt exceeds the maximum allowed offset position, it is considered that the belt is deviated.

[0029] The following is a further description of the detection and identification of the depth distribution map and the gradient distribution map by the 3D detection algorithm (belt defect detection algorithm): Step 1: Detecting the perforation of the belt surface, the part of the belt transparent to the 3D camera cannot be imaged, which is represented as zero in the depth map, and the gradient value of the edge position of the perforation in the gradient distribution map is large, so the perforation of the belt can be detected according to these two characteristics; Step 2: Detecting the deviation of the belt, mainly for another set of detection sensors 202, the position of the left and right edges of the conveying belt is not disturbed by the amount of material on the belt, and the maximum allowed offset position of the belt edge is set in advance for the other set of detection sensors 202. The belt near the lower surface detection device floats up by tens of centimeters when there is little or no material, and each 3D camera can only capture the single side edge of the belt, which cannot reliably detect the deviation. Since the height of the belt near the tail guide roller 103 basically does not change, according to the reasonable range of the working distance between the detection sensor 202 and the belt surface, the point cloud data collected from the detection sensor 202 can be filtered to exclude interference data other than the belt surface, and the actual position of the belt edge can be obtained. Comparing the actual position of the belt edge with the maximum allowed offset position of the belt edge can determine whether the belt is deviated; Step 3: Detecting the crack on the surface of the belt: a mask is set in advance, the size of the mask should not be less than the minimum crack width to be detected; first, median filtering is performed on the depth distribution map to obtain a smooth map; the difference between the depth distribution map and the smooth map is obtained, and the area with a larger difference value is the concave part of the belt surface; the difference between the smooth map and the depth distribution map is obtained, and the area with a larger difference value is the convex part of the belt surface. Secondly, the maximum allowed value of gradient change is set in advance, and then threshold segmentation is performed on the gradient distribution map, so that the area with a larger gradient value can be obtained, which is the uneven part of the belt surface. All the defect parts calculated above are separated into separate defect areas; Step 4: Defect classification, according to the length, width and maximum depth of the belt defect, descending order is divided into three levels A, B and C; each defect area calculated in the above step is calculated one by one, if the depth is greater than the A-level depth standard, and the length or width is greater than the A-level length standard, it is classified as A-level defect; if the depth is greater than the B-level depth standard, and the length or width is greater than the B-level length standard, it is classified as B-level defect; the rest is classified as C-level defect. The perforation defect is classified as the A-level defect; Step 5: Finding the belt joint, the belt joint is the positioning origin, which is used for relative positioning of the belt defect. The belt joint generally has several ways such as mechanical joint, hot melt joint, cold melt joint, etc. In the application scenario of the embodiment, the belt joint adopts the form of hot melt and layered lap joint, and the point cloud data of this kind of joint has obvious texture characteristics. The belt joint at the belt joint presents a slanting straight line with a width of several centimeters, and the inclination directions of the belt joints on the upper and lower surfaces of the belt are consistent. If there are multiple joints on the belt, the last detected joint is taken as the positioning origin.

[0030] Since each group of point cloud data is provided with a time stamp, the moment when the belt joint is detected is recorded as the timing origin, which is denoted as t0. t 0 The X coordinate of the belt joint in the point cloud data is recorded as x0. X j The moment when the belt defect is detected from the point cloud data is recorded as t1. t 1 The relative X coordinate of the belt defect is recorded as x1. X e The relative Y coordinate of the belt defect is recorded as y1. Y e The relative distance of the belt defect from the positioning origin is recorded as d1. L The speed of the belt is recorded as v. V belt According to the following formula, the relative distance of the belt defect can be calculated: L = X j-X e +(t 1 -t 0 )·V belt If the positioning origin is identified again, the timestamp of the last positioning origin is taken as the new timing origin t0.

[0031] The depth profile is subjected to frequency domain transformation using Fourier transform method to obtain a frequency domain profile, and then inverse Fourier transform is performed on the frequency domain profile to obtain the position coordinates of the belt joint X j .

[0032] Step 6: Detecting belt tear, according to step 3, further detect the cracks on the surface of the belt. First, a set of belt tear determination criteria is set in advance, which includes length, width and depth set values of the cracks. If the length of the crack is not less than the length set value, and the width of the crack is not less than the width set value, and the depth of the crack is not less than the depth set value, the crack is determined as a belt tear. According to step 3, detect each crack to determine whether it belongs to a belt tear.

[0033] Figure 5 The point cloud data visualization image obtained by the 3D camera of the upper side belt detection sensor 202 is shown, Figure 6 The output result of the belt defect detection algorithm is shown, which is a two-dimensional image, i.e. the depth profile, which is the result of removing the height information (i.e. Z-axis coordinates) of the three-dimensional belt point cloud data. Two defects B and C appear on the surface of the belt. The belt defect detection algorithm detects the belt point cloud data in real time and identifies the defects on the surface of the belt, including scratches, peeling, cracks and edge wear. The defect detection algorithm can accurately calculate the three-dimensional size of the above defects.

[0034] Typically, the conveyor belt operates at a speed of approximately 0.5 m / s to 2.0 m / s. The 3D camera acquires one contour data point in less than 1 millisecond. Setting the camera's detection frequency to 1000 Hz means the camera can acquire 1000 contour lines (i.e., belt point cloud data) per second. Assuming a belt speed of 2.0 m / s, the distance between two adjacent contour lines is 2.0 mm. These 1000 contour lines are treated as a set of data (i.e., point cloud data) for detection and display (equivalent to one detection per second). This allows us to obtain information such as whether belt defects exist, the number of defects, the 3D dimensions of each defect, its relative position, and its severity level. The system software, which calls the belt defect detection algorithm, then automatically stores these detection results in a database. Figure 5 This illustrates a set of point cloud data.

[0035] In addition to the methods described above, the method of determining whether a belt has a defect based on surface texture features also includes determining whether a belt has a defect based on surface texture features. This involves comparing the current surface texture features with a surface texture feature model to determine whether a belt has a defect. The surface texture feature model is obtained by using a 3D detection algorithm to detect point cloud data of the belt surface after a new belt is fitted, and then using quantifiable feature indicators to describe and train the belt surface texture feature model. The quantifiable feature indicators refer to the different defects that simulate the belt surface.

[0036] In this embodiment: Based on the external dimensions (including depth, corresponding to the Z coordinate) of the belt defects, the belt defects are conventionally divided into three levels according to their external dimensions, including: Class A defects: those exceeding 30 mm in length or 5 mm in depth are classified as severe. Grade B defects: those exceeding 20 mm in length or 3 mm in depth, and failing to meet Grade A standards, are classified as warning defects. Grade C defects: those exceeding 10 mm in length or 2 mm in depth, but not reaching Grade B, are considered to be of concern.

Claims

1. A belt conveyor belt defect detection device, comprising a support frame and detection sensors, the detection sensors including a 3D image acquisition unit and a laser emitter, wherein horizontal support rollers and oblique support rollers arranged opposite each other on both sides of the horizontal support rollers are sequentially spaced on the support frame, and guide rollers are respectively arranged at both longitudinal ends of the support frame, the conveyor belt wraps around the guide rollers at both ends along the surfaces of the horizontal support rollers and the two oblique support rollers, characterized in that, There are three sets of detection sensors. Two of the three sets are set below the conveyor belt supported by the support rollers. The two sets of detection sensors face the non-working sections on both sides of the bottom surface of the conveyor belt. The images of the non-working sections on both sides of the conveyor belt collected by the two sets of detection sensors intersect. The third set of detection sensors is set above the conveyor belt and faces the transverse working surface of the conveyor belt. The image collected by the third set of detection sensors is a complete image of the transverse working section of the conveyor belt. A speed sensor is also set on the support frame to measure the speed of the conveyor belt. The detection sensors and the speed sensor are connected to an analysis and control server.

2. The belt defect detection device according to claim 1, characterized in that, Multiple silos are arranged sequentially above the conveyor belt. The silos are used to feed material to the conveyor belt. With the direction of the conveyor belt's travel as the front, two sets of detection sensors are set below the conveyor belt on the front side of the multiple silos. The two sets of detection sensors are arranged alternately. Another set of detection sensors is set on the rear side of the multiple silos, directly above the rear guide rollers, facing the winding conveyor belt.

3. The belt defect detection device according to claim 1, characterized in that, The detection sensor is fixed in a closed housing. The end face of the housing facing the conveyor belt is a viewing window with transparent glass. The detection sensor faces the conveyor belt through the viewing window. The housing is placed below or above the conveyor belt by a bracket. The bracket includes a column, with a horizontal arm extending from the upper end of the column. Hanging plates are fixed downward on both sides of the front end of the horizontal arm. An adjustable-angle platform is connected to the hanging plate, and the housing is fixed on the platform.

4. The belt defect detection device according to claim 3, characterized in that, The housing is equipped with a dust removal mechanism, which includes a cylinder and a guide shaft mounted on the side wall of the housing. The telescopic arm of the cylinder drives a right-angle arm to reciprocate along a guide shaft. The right-angle arm is equipped with a brush that contacts the transparent glass. The reciprocating right-angle arm drives the brush to remove dust from the outer surface of the transparent glass.

5. A method for detecting and locating defects in a belt conveyor, which is a defect detection and location method based on the belt defect detection device described in claim 1, characterized in that, The method is a defect detection and location method for each set of the detection sensors. The method includes a belt defect detection method and a belt defect location method. A mark is determined on the surface of the conveyor belt as the location origin. The belt is started and the belt running speed is determined. A laser emitter emits a laser to irradiate the surface of the conveyor belt. A 3D image acquisition device continuously acquires the laser-irradiated images to determine the location origin. Then, the location origin is used as the timing origin to cyclically acquire each frame of the laser-irradiated image on the surface of the conveyor belt. The belt defect detection method is as follows: Step 1: The continuous frame image data is used to form point cloud data of the belt surface. Each point in the point cloud data contains coordinates in three dimensions: X, Y, and Z. The Z coordinate represents the distance between the belt surface and the 3D camera. Step 2: Use a 3D detection algorithm to detect the point cloud data of the belt surface to obtain the current surface texture features of the belt surface with varying heights and undulations. Based on the Z coordinate, determine whether the belt has defects from the current surface texture features. The belt defect location method is as follows: Step 1: Attach a timestamp, with the timing origin as the starting point, to each frame of image acquired after the timing origin; Step 2: When a defect occurs on the conveyor belt, the distance from the defect to the positioning origin is calculated based on the timestamp of the frame image where the defect occurs and the belt running speed, thereby determining the relative position of the defect on the conveyor belt.

6. The method according to claim 5, characterized in that, The positioning origin is the annular joint seam formed by the belt.

7. The method according to claim 5, characterized in that, The belt defects are classified into three levels according to their external dimensions, including: Class A defects: those exceeding 30 mm in length or 5 mm in depth are classified as severe. Grade B defects: those exceeding 20 mm in length or 3 mm in depth, and failing to meet Grade A standards, are classified as warning defects. Grade C defects: those exceeding 10 mm in length or 2 mm in depth, but not reaching Grade B, are considered to be of concern.

8. The method according to claim 5, characterized in that, The current surface texture features of the belt surface, which are obtained by using a 3D detection algorithm to detect point cloud data of the belt surface, are: The point cloud data is projected and converted into a two-dimensional grayscale image, including: using the x-coordinate of each data point as the column coordinate of the image pixel, the y-coordinate as the row coordinate of the image pixel, and the z-coordinate as the grayscale value of the image pixel. All data points of the point cloud data together constitute a two-dimensional depth distribution map of the belt surface. The partial derivatives along the x-axis and y-axis are calculated for each pixel of the depth distribution map, and the depth change rate is calculated to obtain a gradient distribution map to form the texture features of the belt surface.

9. The method according to claim 5, characterized in that, The method also includes the determination of belt misalignment: a maximum allowable position of the belt edge is set, and belt misalignment is defined as the actual edge of the belt exceeding the maximum allowable position.

10. The method according to claim 6, characterized in that, The belt defect detection also includes judging whether the belt has defects from the surface texture features. This is done by comparing the current surface texture features with the surface texture feature model to determine whether the belt has defects. The surface texture feature model is obtained by using a 3D detection algorithm to detect the point cloud data of the belt surface after a new belt is put on, and then using quantifiable feature indicators to describe and train the belt surface texture feature model. The quantifiable feature indicators refer to the different defects that appear on the belt surface.

Images