Intelligent backhaul scraping device signal acquisition method and device
By dividing the return section into multiple sub-segments and dynamically adjusting the scraper parameters using multiple sensors and deep learning algorithms, the problem of low automation in the return scraping device was solved, achieving efficient material cleaning and conveyor belt wear assessment, and improving production stability.
Patent Information
- Application Number
- CN202611267170.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-20
- Publication Date
- 2026-11-17
AI Technical Summary
The existing return scraping device has a low degree of automation and cannot dynamically adjust the scraping strategy, resulting in uneven material residue, which affects conveyor belt wear and production continuity.
The return section is divided into a coarse scraping section, a fine scraping section, and an evaluation section. Data is collected collaboratively by multiple sensors, and deep learning algorithms are used to identify the type and density level of residual materials. The scraper angle and pressure are dynamically adjusted, and combined with conveyor belt loss assessment, intelligent scraping is achieved.
It significantly improves the scraping effect, reduces material residue, extends the service life of the conveyor belt, avoids equipment failure, and ensures production continuity.
Smart Images

Figure CN123115894A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of material conveying equipment technology, specifically to a signal acquisition method and device for an intelligent return scraper. Background Technology
[0002] In the field of industrial material conveying, cleaning residual materials on the return section of the conveyor belt is a key link to ensure production efficiency. Traditional return scraping devices generally use fixed parameters, such as constant scraper angle, pressure and conveyor belt running speed, to perform a one-size-fits-all cleaning of the entire return section. Moreover, the start and stop of the cleaning device are all controlled manually, resulting in low automation and low cleaning efficiency. In actual production, due to various factors such as fluctuations in working conditions and uneven material distribution, the fallen return material often cannot be cleaned up and recycled into the head hopper in time, causing the material to accumulate continuously in the return section, which in turn leads to the overload start-up of the cleaning device, causing equipment shutdown failure and disrupting the normal production rhythm. Currently, there are belt conveyor operation status acquisition devices developed based on the Hall sensor principle, which can detect the operating speed and running time of the conveyor in real time and support setting parameters such as operating alarm limits according to the on-site working conditions. However, such operation status monitoring has not yet been deeply integrated with the control logic of the return scraper, and cannot provide effective support for the dynamic adjustment of the cleaning strategy. In addition, existing return scraping devices still have the following shortcomings in terms of cleaning accuracy and the operation and maintenance of the conveyor belt throughout its entire life cycle: Firstly, the existing equipment does not have functional zoning for the return section, making it impossible to adjust the scraping strategy according to the needs of different stages such as coarse scraping, fine scraping, and loss assessment. For example, local high pressure is not applied to areas with dense residual material, resulting in stubborn material residue. High-intensity scraping is still used for areas with scattered residue, which aggravates unnecessary wear on the conveyor belt. Secondly, traditional devices do not integrate multiple sensors, making it impossible to obtain the type and density level of residual materials in real time. This results in the scraper angle and pressure not being able to be dynamically adjusted according to the material characteristics. For example, when dealing with sticky materials, a scraper with a fixed angle can easily cause the material to stick to the conveyor belt. When processing lumpy materials, insufficient pressure can cause incomplete crushing, resulting in low scraping efficiency. Third, there is a lack of monitoring of conveyor belt loss. Existing technology does not conduct a systematic assessment of the conveyor belt during the scraping process, which makes it impossible to identify minor damage to the conveyor belt in the early stage and to include areas of moderate loss in planned maintenance, which seriously affects the continuity of production. Therefore, there is an urgent need for an intelligent return scraping device and signal acquisition method that integrates staged precise scraping, dynamic parameter adjustment, and conveyor belt status assessment to solve the above-mentioned technical problems. Summary of the Invention
[0003] To address the shortcomings of existing technologies, this invention provides a signal acquisition method and device for an intelligent return scraping device, which solves the problems of poor scraping effect, high conveyor belt wear, and inability to assess wear in a timely manner during the return stroke of the conveyor belt.
[0004] To achieve the above objectives, the present invention provides the following technical solution: a signal acquisition method for an intelligent return scraper device, comprising: Step 1: Divide the return section into coarse scraping section L1, fine scraping section L2, and evaluation section L3. Use the conveyor belt circumference as a reference to divide fixed areas at equal intervals as target areas. Step 2: The target area enters segment L1. The area number is verified by both the encoder and RFID tag, which are composed of a magnetic probe switch device. This triggers the linear array camera and vibration sensor to collect data. Multispectral image and vibration signal features are extracted by CNN and LSTM respectively, and then spliced into a feature vector and input into FCN to obtain the residue type and density level. Based on this, the scraper angle is corrected and the pressure base value is calculated. After coarse scraping, the remaining residue position is marked as the key area for fine scraping and transmitted to segment L2. The magnetic probe switch device includes a switch probe and a magnetic induction module. The switch probe is installed on the side of the conveyor roller, and the magnetic induction module is installed directly opposite the probe. When the conveyor is running, the magnetic probe senses a magnetic field signal and generates a current signal once for each rotation of the roller. This signal is input into the electrical control system to drive the intelligent return scraping device to run synchronously with the conveyor. Step 3: Confirm the real-time position of the target area entering the L2 section through the encoder, calculate the proportion of the key area for fine scraping (α), and dynamically adjust the belt speed and scraper pressure according to priority. Step 4: After the target area enters the L3 segment, obtain the thickness reduction rate, defect area ratio, and stress concentration coefficient, calculate the loss severity index (DSI), classify the loss level according to the DSI, and take corresponding measures.
[0005] As a further aspect of the present invention, the residue types include lumps, powders, and viscous substances, represented by the numbers 1, 2, and 3 respectively, and the density grades are divided into L, M, H, and VH grades, represented by the numbers 0, 1, 2, and 3 respectively.
[0006] As a further aspect of the present invention, the specific steps for correcting the scraper angle according to the type of residue are as follows: The initial angle of the scraper is preset to 45 degrees; If the residue is identified as lumpy material, reduce the scraper angle by 10 degrees. If the material is sticky, reduce the scraper angle by 15 degrees; If the material is in powder form, increase the scraper angle by 5 degrees.
[0007] As a further aspect of the present invention, based on density levels, a fuzzy logic controller is used to calculate the scraper pressure base value Pbase for each target area.
[0008] As a further aspect of the present invention, a sliding window is used to calculate the gray-scale mean of the multispectral image, and then the residue area is dynamically segmented using the Otsu algorithm to generate a density heat map D. If a certain area D>1.5*D', a pressure of 20%-50% needs to be superimposed on the base value, where D' is the average density heat value.
[0009] As a further aspect of the present invention, it is verified whether the modified angle and pressure exceed the safety threshold; if they do, the boundary value is taken.
[0010] As a further embodiment of the present invention, if α > αmax, it represents high priority, and local high-pressure fine scraping and belt speed reduction are required; if αmin <= α <= αmax, it represents medium priority, and conventional fine scraping and speed adaptation are required; if α < αmin, it represents low priority, and rapid fine scraping and belt speed increase are required. Here, αmin and αmax are percentage thresholds.
[0011] As a further aspect of the present invention, the specific steps for dynamically adjusting the belt speed and scraper pressure according to priority are as follows: The conveyor belt speed is dynamically adjusted according to the formula v=vbase*(1+k*(1-ɑ / ɑmax)), where vbase is the system default speed, k is the speed adjustment coefficient, k=0.2 for high priority, k=0 for medium priority, and k=0.15 for low priority, and the speed range is [vmin,vmax]. If the speed adjustment exceeds the speed range, then the corresponding boundary value will be used; A segmented flexible scraper is used to apply local high pressure Plocal=Pbase*(1+ɑ / ɑmax) to the scraper segment corresponding to the key area of fine scraping; After the fine scraping is completed, α is calculated again. If α > 20%, the area is marked as an ultra-stubborn residue area, triggering an immediate stop command, and the residue in the area is manually removed.
[0012] As a further aspect of the present invention, the specific steps for determining the severity level of conveyor belt wear are as follows: A judgment matrix is constructed based on mutual information entropy, and the eigenvalues wh, wa, and wt of the judgment matrix are obtained. The actual thickness reduction rate, defect area ratio, and stress concentration factor are normalized to obtain Gh, Ga, and Kt. The DSI is calculated using the formula DSI=wh*Gh+wa*Ga+wt*Kt; The specific method for constructing the judgment matrix based on mutual information entropy is as follows: Create an indicator set {thickness reduction rate X1, defect area rate X2, stress concentration factor X3}, and a maintenance type Y={1,2,3}, where 1, 2, and 3 represent mild, moderate, and severe respectively; Collect several groups of labeled {X1, X2, X3, Y} samples; For each indicator Xi, according to the formula Calculate the mutual information value MIi with Y, where p(xi,y) is the joint probability of indicator Xi and maintenance type Y, and p(xi) and p(y) are marginal probabilities; Generate matrix elements aij according to the mutual information ratio, which satisfies MIi / MIj and aji=1 / aij, where i∈[1,3], j∈[1,3]; If DSI<=DSImin, the area is defined as mild wear; if DSImin<DSI<=DSImax, the area is defined as moderate wear; if DSI>DSImax, the area is defined as severe wear.
[0013] An intelligent return scraper signal acquisition device, comprising: an area division identification module, a rough scraping control module, a fine scraping control module, and an evaluation and maintenance management module; The area division identification module divides the return section into a rough scraping section L1, a fine scraping section L2, and an evaluation section L3, and divides fixed areas at equal intervals based on the circumference of the conveyor belt as the target area data acquisition and processing module; When the target area enters the L1 section, the rough scraping control module performs double verification of the area number through the encoder and the RFID tag, triggers the line scan camera and the vibration sensor to collect data, extracts the features of the multispectral image and the vibration signal through CNN and LSTM respectively, splices them into a feature vector and inputs it into FCN to obtain the residue type and density level, corrects the scraper angle and calculates the pressure base value accordingly, marks the position of the remaining residue as the key fine scraping area after rough scraping, and transmits it to the L2 section; The fine scraping control module confirms the real-time position of the target area entering the L2 section through the encoder, calculates the proportion ɑ of the key fine scraping area, and dynamically adjusts the belt speed and the scraper pressure according to the priority; After the target area enters the L3 section, the evaluation and maintenance management module obtains the thickness reduction rate, the defect area rate and the stress concentration factor, calculates the damage severity index DSI, divides the damage level according to DSI and takes corresponding measures.
[0014] The present invention provides an intelligent return scraper signal acquisition method and device, which has the following beneficial effects compared with the prior art: (1) The present invention realizes intelligent adjustment of the scraper angle and pressure through multi-sensor collaborative data collection and using deep learning algorithms to identify the residual material type and density level, which significantly improves the scraping effect and reduces material residue; (2) The present invention monitors and evaluates the conveyor belt in segments, which can keep track of the wear status of the conveyor belt in real time, and take reasonable maintenance measures according to different wear levels, thereby avoiding sudden equipment failure and extending the service life of the conveyor belt. Attached Figure Description
[0015] Figure 1 This is a flowchart of the steps of the present invention; Figure 2 This is a system framework diagram of the present invention. Detailed Implementation
[0016] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0017] like Figure 1 This invention provides a signal acquisition method for an intelligent return scraper device, comprising: Step 1: Divide the return section into three equal-length segments: the first monitoring segment L1 is the coarse scraping segment, the second monitoring segment L2 is the fine scraping segment, and the third monitoring segment L3 is the evaluation segment. Install laser beam sensors at the starting point of each segment to define the boundaries of each segment. In the coarse scraping section L1, the main task is to deal with a large amount of accumulated residual material. For example, in the coal conveying scenario, coal may accumulate in a large amount in the return section during the conveying process. The coarse scraping section can use a large scraper pressure and a suitable angle to quickly remove most of the residual coal. If the material is not segmented and the fine scraping method is used uniformly to deal with a large amount of residual material, it will not only be inefficient, but will also lead to excessive wear of the scraper and the conveyor belt. In the fine scraping section L2, the focus is on cleaning the residual material left after the coarse scraping. For example, when conveying powdery materials such as flour, some powdery material may still be attached to the surface of the conveyor belt after the coarse scraping. The fine scraping section can precisely adjust the scraper pressure and speed according to the specific situation of the residual material to achieve fine cleaning and improve the scraping quality. In the evaluation section L3, a comprehensive assessment of the conveyor belt's wear condition is conducted. Multi-stage scraping may cause problems such as wear and scratches on the conveyor belt. The evaluation section can detect changes in these wear indicators in a timely manner, accurately judge the degree of wear of the conveyor belt, and take corresponding maintenance measures in a timely manner to ensure the safe and stable operation of the conveyor belt. Laser beam sensors utilize the linear propagation characteristics of lasers. When the conveyor belt is running, if an object blocks the laser beam path, the sensor can quickly determine the starting position of each segment. It is not affected by common interference factors such as light and dust, and can work stably in complex industrial environments, providing a reliable position reference for subsequent precise control of material scraping and monitoring operations of each segment. Based on the circumference of the conveyor belt, divide it into several fixed areas at equal intervals. Each fixed area is used as the target area. For example, divide the area into 1 area every 2 meters and number them R001, R002, ..., R00N in sequence. In actual production, the residue situation in different areas varies greatly when conveying materials. For example, in a production line that conveys food raw materials, due to the different contact and influence of various components on the conveyor belt on the materials, some areas may be more prone to material residue. After dividing the area, the system can accurately locate the amount of residual material in different areas, providing accurate location information for subsequent targeted scraping operations. An encoder is installed on the drive roller. The encoder counts in real time to locate the area position. An RFID electronic tag is attached to the front end of each area. For example, tag A is attached to the front end of area R001 and tag B is attached to the front end of area R002. When different areas of the conveyor belt enter the three-section entrance, the RFID reader reads the tag ID quickly within 0.1 seconds to confirm the current processing area and avoid area identification errors due to encoder counting errors. The encoder records the number of rotations of the drive roller and then calculates the position information of each area based on the pre-set conveyor belt circumference and area division spacing. In actual operation, the encoder may be affected by vibration, electrical interference, etc., resulting in counting errors. At this time, RFID electronic tags play a key auxiliary role. For example, when an area passes through the L1 section entrance, the RFID reader quickly reads the tag ID to confirm the current processing area. The two work together to greatly improve the accuracy of area positioning.
[0018] Step 2: Install a line scan camera and vibration sensor in section L1. When the target area enters section L1, the encoder and RFID tag, which are composed of a magnetic probe switch device, verify the area number. Then, the line scan camera and vibration sensor are triggered to start collecting data and establish a mapping relationship between the area number and the real-time position. The magnetic probe switching device includes a switch probe and a magnetic induction module. The switch probe is installed on the side of the belt conveyor roller, and the magnetic induction module is installed directly opposite the probe. When the belt conveyor is running, the magnetic probe senses a magnetic field signal and generates a current signal once for each rotation of the roller. The signal is input into the electrical control system to drive the intelligent return scraper to run synchronously with the belt conveyor. The aforementioned switching device, through the cooperation of the switch probe and the magnetic induction module, converts the rotational motion of the belt conveyor roller into a periodic electrical pulse signal: when the roller rotates once, the switch probe installed on its side aligns with the magnetic induction module fixed directly opposite it once. The probe then senses the change in the magnetic field and generates a current pulse. After the pulse signal is input into the electronic control system, it directly serves as the synchronous clock reference for the operation of the intelligent return scraping device, thereby driving the scraping device to strictly follow the rotation rhythm of the roller and realize the real-time synchronous linkage between the scraping process and the belt conveyor. The line scan camera scans the surface of the conveyor belt in section L1 to acquire multispectral images. Different types of residual materials will exhibit unique color, texture and other characteristics in the multispectral images. For example, in the scenario of conveying non-ferrous metal ores, the spectral reflectance characteristics of copper ore residue and iron ore residue are different, and the colors and textures reflected in the images are also different. Vibration sensors are used to capture vibration signals generated when a scraper comes into contact with residual material. Factors such as the hardness, shape, and distribution of the residual material will affect the frequency and amplitude of the vibration. For example, when encountering blocky and unevenly distributed residual material, the vibration signal generated by the scraper at the moment of contact with the material will fluctuate significantly. Multispectral vectors are obtained by extracting features from multispectral images using a convolutional neural network (CNN). Vibration vectors are obtained by extracting features from vibration signals using a long short-term memory network (LSTM). These two vectors are concatenated into a feature vector, which is then input into a pre-trained fully connected neural network (FCN) for analysis, thereby determining the type and density level of the residue at that time. The types of residues include lumps, powders, and sticky substances, represented by the numbers 1, 2, and 3 respectively. The density grades are divided into L, M, H, and VH grades, represented by the numbers 0, 1, 2, and 3 respectively. CNNs can automatically extract spatial features at different scales in images through the sliding scanning mechanism of convolution kernels. When processing multispectral images of residual materials on conveyor belts, for millimeter-sized fine particles and centimeter-sized blocky materials, CNNs can capture local texture features and overall shape features respectively by adjusting the convolution kernel size and pooling layer parameters. Vibration signals are essentially time-series data, containing rich information about time variations. The gating mechanism of LSTM can effectively capture the long-term and short-term dependencies in vibration signals. For example, during the contact between the scraper and the material, the initial high-frequency vibration reflects the hardness of the material, while the subsequent decaying vibration is related to the adhesion of the material. By learning these time-series patterns, LSTM can accurately determine the characteristics of the material. The combination of CNN and LSTM forms a complementary feature extraction system. CNN extracts spatial features from multispectral images to reflect the static properties of materials, while LSTM extracts temporal features from vibration signals to reveal the dynamic interaction characteristics of materials. After the two are integrated, the system makes a more comprehensive judgment on the properties of materials. The reason for choosing the fully connected neural network FCN as the subsequent output model is that it can efficiently integrate multimodal features and adapt to complex task requirements: on the one hand, FCN can seamlessly stitch together the combined features of CNN and LSTM to achieve deep fusion of multi-source information; on the other hand, through multi-layer nonlinear transformation, it can accurately fit the complex mapping relationship between residue type and density level, avoiding the expression limitations of linear models. The scraper angle is adjusted according to the type of residue. The initial scraper angle is set to 45 degrees. If the residue is identified as lumpy material, such as lumpy ore left after conveying ore, the scraper angle is reduced by 10 degrees to better crush and scrape it off. If it is sticky material, such as sticky substances left after conveying certain chemical raw materials, the scraper angle is reduced by 15 degrees to prevent material from sticking together. If it is powdery material, such as powder left after conveying flour, the scraper angle is increased by 5 degrees to collect the powder more effectively. When the scraper contacts the conveyor belt at a 45-degree angle, the vertical component of the force is equal to the horizontal component. This balance allows the scraper to effectively cut into the material layer and push the scraped material away from the conveyor belt in the horizontal direction, preventing the material from accumulating in front of the scraper. At the same time, the 45-degree angle has a good initial scraping effect on medium-hardness, non-sticky materials, such as sand and grains, and can be used as a benchmark angle for most conveying scenarios. For mixed residual materials of unknown type, this angle can provide a relatively balanced scraping ability. For lumpy materials, when the scraper angle is 35 degrees, the blade pressure is higher than that at 45 degrees, making it easier to crush hard materials. At a 35-degree angle, the contact point between the scraper and the material is closer to the root of the blade, reducing stress concentration at the tip of the blade and making the blade wear more even. For viscous materials, a 30-degree angle further reduces the contact area between the scraper and the conveyor belt, reducing the adhesion between the material and the scraper surface. The smaller angle makes the direction of the scraper's peeling force on the viscous material closer to horizontal, which is in line with the shearing and breaking characteristics of viscous materials. For powdery materials, a 50-degree angle creates a wider V-groove between the scraper and the conveyor belt, which facilitates the collection and guidance of the powdery material flow. At a 50-degree angle, the impact force of the scraper on the powder layer is more biased towards the surface of the conveyor belt, reducing the possibility of powder flying. Based on density levels, the base value Pbase of each target region is calculated using a fuzzy logic controller. Density grade is a global quantitative indicator of the overall density of residual material in the target area, which solves the global control problem of how much base pressure is needed for the entire area. For example, a high density grade corresponds to a high Pbase, ensuring that the overall scraping force is sufficient. The mean gray value of the multispectral image is calculated using a sliding window, and the residue area is dynamically segmented using the Otsu algorithm to generate a density heat map D. If D > 1.5 * D' in a certain area, it indicates that there is a lot of residual material in the area, and 20%-50% pressure needs to be added on the basis of the base value to enhance the scraping effect. Here, D' is the average density heat value. The density heat map D is a pixel-level local mapping of the distribution of residual materials in the target area. It is used to achieve local pressure compensation and solve the problem of fine control of which local locations in the area need additional pressurization. For example, if the overall density level of a certain area is medium, but one sub-area has D=2.0×D', then the pressure of that sub-area will be increased. Check whether the pressure and angle exceed the safety threshold. If they do, take the boundary value to ensure the safe operation of the equipment. After the coarse scraping is completed, the location of the remaining residue in each area is marked, and these areas are designated as key areas for fine scraping. The specific coordinates are then sent to the second monitoring area L2.
[0019] Step 3: Confirm the real-time position of the target area entering segment L2 by counting encoder pulses, ensure that the coordinates of the key area for fine scraping are accurately mapped to the position of the scraper, and calculate the proportion α of the key area for fine scraping in each area to the target area. Determining the proportion of key areas for fine scraping allows for the quantification of processing priorities. α directly reflects the impact of each area on the target accuracy, laying a quantitative foundation for weight calculation and process parameter optimization, reducing reliance on subjective experience. At the same time, based on the α value, high-value processing areas can be accurately located, guiding resources to concentrate on key areas, avoiding ineffective processing, and significantly improving fine scraping efficiency. If α > αmax, it represents high priority, indicating dense and stubborn residue, which requires local high-pressure fine scraping and a reduction in belt speed. If αmin <= α <= αmax, it represents medium priority, indicating scattered residue, which requires conventional fine scraping and speed adaptation. If α < αmin, it represents low priority, indicating sporadic residue, which can be quickly fine scraped and the belt speed increased. Here, αmin can be taken as 5% and αmax can be taken as 20%. The specific formula for dynamic adjustment of conveyor belt speed is: v=vbase*(1+k*(1-ɑ / ɑmax)), where vbase is the system default speed, k is the speed adjustment coefficient, k=0.2 for high priority, k=0 for medium priority, and k=0.15 for low priority. The speed range is [vmin, vmax]. If the speed adjustment exceeds the speed range, the corresponding boundary value is taken. A segmented flexible scraper is used. Local high pressure Plocal=Pbase*(1+ɑ / ɑmax) is applied to the scraper segment corresponding to the key area of fine scraping, while the base pressure Pbase is maintained in the non-key area. Pbase is also transmitted from segment L1. After the fine scraping is completed, α is calculated again. If α > 20%, it indicates that the fine scraping of the target area has failed. The area is marked as an ultra-stubborn residue area, triggering an immediate stop command. The residue in the area is then manually removed. If α > 20% after fine scraping, it indicates that there is serious residue in this area. Continuing to rely on automatic scraping equipment will not be able to remove it, wasting time and energy and affecting the overall scraping efficiency. At the same time, if the conveyor belt in the ultra-stubborn residue area continues to run, the residual material will increase the wear of the scraper and the conveyor belt, reduce the service life of the equipment, and increase maintenance costs. Therefore, it is necessary to stop the machine immediately and switch to manual handling to ensure the normal operation of the equipment and production.
[0020] Step 4: Install a laser thickness gauge, machine vision defect detection, and strain gauge sensor array in section L3. When the target area enters L3, the thickness reduction rate, defect area ratio, and stress concentration coefficient are obtained by the three sensors respectively. The loss severity index (DSI) of the conveyor belt in each target area is calculated based on the three-dimensional parameters. The specific steps for calculating the Loss Severity Index (DSI) of a conveyor belt are as follows: A judgment matrix is constructed based on mutual information entropy. By calculating the mutual information value MI between each loss index and maintenance type, the contribution of the index to the decision is reflected. The greater the mutual information, the more important the index. The eigenvalues wh, wa, and wt of the judgment matrix are obtained. The actual thickness reduction rate, defect area ratio, and stress concentration factor are normalized to obtain Gh, Ga, and Kt. The DSI is calculated using the formula DSI=wh*Gh+wa*Ga+wt*Kt; The specific method for constructing the judgment matrix is as follows: Create an index set {thickness reduction rate X1, defect area ratio X2, stress concentration factor X3}, and a repair type Y={1,2,3}, where 1,2,3 represent mild, moderate, and severe damage, respectively. Collect several sets of labeled (X1, X2, X3, Y) samples; For each index Xi, according to the formula Calculate the mutual information value Mii with Y, where p(xi,y) is the joint probability of index Xi and maintenance type Y, and p(xi) and p(y) are the marginal probabilities; Generate matrix elements aij based on the mutual information ratio, satisfying Mii / MIj and aji=1 / aij, where i∈[1,3] and j∈[1,3]. The nonlinear dependency relationship between indicators and targets is accurately captured using mutual information entropy, which is sensitive to low-probability high-risk scenarios and dynamically assigns higher weights to key indicators. A scientific judgment matrix is constructed starting from the internal correlations of the data to ensure the reliability of weight calculation; If DSI is less than or equal to DSImin, the area is defined as mild wear; if DSI is greater than DSImin and less than or equal to DSImax, the area is defined as moderate wear; if DSI is greater than DSImax, the area is defined as severe wear; If an area is detected as severe wear, an immediate shutdown instruction is triggered, and the specific position of the area on the conveyor belt is locked; Severe wear indicates that the conveyor belt structure is seriously damaged, with a risk of sudden breakage that may cause safety accidents. Immediate shutdown and position locking can eliminate the breakage hidden danger at the first time, and accurate positioning of the damaged area can greatly shorten the maintenance inspection time and reduce production losses caused by unplanned shutdowns; If an area is detected as moderate wear, it is included in the next planned shutdown maintenance queue, and work orders are generated in the order of area numbers; Moderate wear does not pose an immediate danger. Including it in the planned shutdown maintenance queue and generating work orders by area number allows overall arrangement with other production line maintenance tasks to avoid disrupting the production rhythm. Meanwhile, sequential processing can effectively curb the deterioration of wear, enable advance planning of spare parts and manpower scheduling, save certain costs compared with emergency maintenance, and realize optimal allocation of maintenance resources; If an area is detected as mild wear, it is marked as a daily maintenance focus, and preventive treatment is performed during off-peak hours; Mild wear is in the initial stage of damage. Marking it as a daily maintenance focus and processing it during off-peak hours can prevent damage spread through preventive measures such as coating repair, keeping maintenance costs at the lowest level. Intervention during production line downtime can avoid affecting production capacity, and tracking the wear trend combined with regular monitoring can extend the average service life of the conveyor belt, achieving the equipment maintenance goal of low cost and high reliability; After the maintenance is completed, the belt thickness database of the corresponding area is updated through the RFID reader-writer, for example, the thickness after mild wear repair is updated to 95% of the original thickness.
[0021] A signal acquisition device for an intelligent return scraper device, comprising: an area division identification module, a rough scraping control module, a fine scraping control module, and an assessment maintenance management module; The area division identification module divides the return section into a rough scraping section L1, a fine scraping section L2, and an assessment section L3, and divides fixed areas at equal intervals based on the circumference of the conveyor belt to serve as the target area data acquisition and processing module; In the coarse scraping control module, the target area enters the L1 segment. The area number is verified by both the encoder and the RFID tag. The linear array camera and vibration sensor are triggered to collect data. The multispectral image and vibration signal features are extracted by CNN and LSTM respectively, and then spliced into a feature vector and input into FCN to obtain the type and density level of the residue. Based on this, the scraper angle is corrected and the pressure base value is calculated. After coarse scraping, the remaining residue position is marked as the key area for fine scraping and transmitted to the L2 segment. The fine scraping control module confirms the real-time position of the target area entering the L2 section through the encoder, calculates the proportion of the key area for fine scraping (α), and dynamically adjusts the belt speed and scraper pressure according to priority. The assessment and maintenance management module obtains the thickness reduction rate, defect area ratio, and stress concentration coefficient after the target area enters the L3 section. It then calculates the damage severity index (DSI), classifies the damage level according to the DSI, and takes corresponding measures.
[0022] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.
[0023] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A signal acquisition method for an intelligent return scraper device, characterized in that, include: Step 1: Divide the return section into coarse scraping section L1, fine scraping section L2, and evaluation section L3. Use the conveyor belt circumference as a reference to divide fixed areas at equal intervals as target areas. Step 2: The target area enters segment L1. The area number is verified by both the encoder and RFID tag, which are composed of a magnetic probe switch device. This triggers the linear array camera and vibration sensor to collect data. Multispectral image and vibration signal features are extracted by CNN and LSTM respectively, and then spliced into a feature vector and input into FCN to obtain the residue type and density level. Based on this, the scraper angle is corrected and the pressure base value is calculated. After coarse scraping, the remaining residue position is marked as the key area for fine scraping and transmitted to segment L2. The magnetic probe switch device includes a switch probe and a magnetic induction module. The switch probe is installed on the side of the conveyor roller, and the magnetic induction module is installed directly opposite the probe. When the conveyor is running, the magnetic probe senses a magnetic field signal and generates a current signal once for each rotation of the roller. This signal is input into the electrical control system to drive the intelligent return scraping device to run synchronously with the conveyor. Step 3: Confirm the real-time position of the target area entering the L2 section through the encoder, calculate the proportion of the key area for fine scraping (α), and dynamically adjust the belt speed and scraper pressure according to priority. Step 4: After the target area enters the L3 segment, obtain the thickness reduction rate, defect area ratio, and stress concentration coefficient, calculate the loss severity index (DSI), classify the loss level according to the DSI, and take corresponding measures.
2. The signal acquisition method for an intelligent return scraper according to claim 1, characterized in that, The types of residues include lumps, powders, and sticky substances, represented by the numbers 1, 2, and 3, respectively. The density grades are divided into L, M, H, and VH grades, represented by the numbers 0, 1, 2, and 3, respectively.
3. The signal acquisition method for an intelligent return scraper according to claim 1, characterized in that, The specific steps for adjusting the scraper angle based on the type of residue are as follows: The initial angle of the scraper is preset to 45 degrees; If the residue is identified as lumpy material, reduce the scraper angle by 10 degrees. If the material is sticky, reduce the scraper angle by 15 degrees; If the material is in powder form, increase the scraper angle by 5 degrees.
4. The signal acquisition method for an intelligent return scraper according to claim 1, characterized in that, Based on density levels, the scraper pressure base value Pbase for each target area is calculated using a fuzzy logic controller.
5. The signal acquisition method for an intelligent return scraper according to claim 1, characterized in that, The mean grayscale value of the multispectral image is calculated using a sliding window, and the residue area is dynamically segmented using the Otsu algorithm to generate a density heat map D. If D > 1.5 * D' in a certain area, a pressure of 20%-50% needs to be superimposed on the base value, where D' is the average density heat value.
6. The signal acquisition method for an intelligent return scraper according to claim 1, characterized in that, Verify whether the modified angle and pressure exceed the safety threshold; if they do, use the boundary value.
7. The signal acquisition method for an intelligent return scraper according to claim 1, characterized in that, If α > αmax, it represents high priority, and local high-pressure fine scraping and reduced belt speed are required. If αmin <= α <= αmax, it represents medium priority, and regular fine scraping and speed adaptation are required. If α < αmin, it represents low priority, and rapid fine scraping and increased belt speed are required. Here, αmin and αmax are percentage thresholds.
8. The signal acquisition method for an intelligent return scraper according to claim 1, characterized in that, The specific steps for dynamically adjusting belt speed and scraper pressure according to priority are as follows: Dynamically adjust the conveyor belt speed according to the formula v=vbase*(1+k*(1-ɑ / ɑmax)), where vbase is the system default speed, k is the speed adjustment coefficient, k=0.2 for high priority, k=0 for medium priority, k=0.15 for low priority, and the speed range is [vmin, vmax]; If the adjusted speed exceeds the speed range, the corresponding boundary value is taken; A segmented flexible scraper is adopted, and a local high pressure Plocal=Pbase*(1+ɑ / ɑmax) is applied to the scraper segment corresponding to the fine scraping key area; After fine scraping is completed, ɑ is calculated again. If ɑ>20%, the area is marked as an extremely stubborn residual area, an immediate shutdown instruction is triggered, and the residual in this area is manually processed.
9. The signal acquisition method for an intelligent return scraper according to claim 1, characterized in that, The specific steps for judging the severity level of conveyor belt wear are as follows: Construct a judgment matrix based on mutual information entropy, and obtain the eigenvalues wh, wa, wt of the judgment matrix; Normalize the actually obtained thickness reduction rate, defect area rate, and stress concentration coefficient to obtain Gh, Ga, Kt; Calculate DSI according to the formula DSI=wh*Gh+wa*Ga+wt*Kt; The specific method for constructing the judgment matrix based on mutual information entropy is as follows: Create an indicator set {thickness reduction rate X1, defect area rate X2, stress concentration coefficient X3}, and a maintenance type Y={1,2,3}, where 1, 2, and 3 represent mild, moderate, and severe respectively; Collect several groups of labeled {X1,X2,X3,Y} samples; For each index Xi, according to the formula Calculate the mutual information value Mii with Y, where p(xi,y) is the joint probability of index Xi and maintenance type Y, and p(xi) and p(y) are the marginal probabilities; Generate matrix elements aij according to the mutual information ratio, which satisfies MIi / MIj and aji=1 / aij, where i∈[1,3], j∈[1,3]; If DSI<=DSImin, the area is defined as mild wear; if DSImin<DSI<=DSImax, the area is defined as moderate wear; if DSI>DSImax, the area is defined as severe wear.
10. A signal acquisition device for an intelligent return scraper, used to execute a signal acquisition method for an intelligent return scraper, characterized in that, It includes an area division identification module, a rough scraping control module, a fine scraping control module, and an evaluation and maintenance management module; The area division identification module divides the return section into a rough scraping section L1, a fine scraping section L2, and an evaluation section L3, and divides fixed areas at equal intervals based on the circumference of the conveyor belt to serve as the target area data acquisition and processing module; The rough scraping control module: when the target area enters the L1 section, the area number is double-verified by the encoder and the RFID tag, triggering the line-scan camera and the vibration sensor to collect data. The CNN and LSTM are used to extract the features of the multispectral image and the vibration signal respectively, which are spliced into a feature vector and input into the FCN to obtain the residue type and density level. Accordingly, the scraper angle is corrected, the pressure base value is calculated, and after rough scraping, the position of the remaining residue is marked as the fine scraping key area and transmitted to the L2 section; The fine scraping control module confirms the real-time position of the target area entering the L2 section through the encoder, calculates the proportion ɑ of the fine scraping key area, and dynamically adjusts the belt speed and scraper pressure according to the priority; After the target area enters the L3 section, the evaluation and maintenance management module obtains the thickness reduction rate, defect area rate, and stress concentration coefficient, calculates the wear severity index DSI, divides the wear level according to the DSI, and takes corresponding measures.
