A method and device for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill.

By generating a full-domain 3D model before and after wear using a handheld 3D scanner and constructing a remaining life prediction model, the problem of accurately assessing the wear state of the grinding rollers and grinding discs of coal mills was solved, thus improving the stability of coal mill equipment operation and predicting its life.

CN123118475APending Publication Date: 2026-11-17CHANGCHUN POWER GENERATION EQUIP PLANT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611215655.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-12
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In the existing technology, it is difficult to assess the wear condition of the grinding rollers and grinding discs of coal mills in a timely and accurate manner, which affects the stability of equipment operation and lacks an effective equipment maintenance cycle, increasing the risk of unplanned downtime.

Method used

A handheld 3D scanner was used to collect dimensional data before and after wear, generating a full-domain 3D model before and after wear. Wear distribution data was obtained through difference comparison calculation, and a logical prediction model for remaining life was constructed for full life cycle assessment.

Benefits of technology

It improves the accuracy of wear condition assessment, enhances equipment operational stability, provides full lifecycle remaining life assessment and visual prompts, and reduces the risk of unplanned downtime.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118475A_ABST
    Figure CN123118475A_ABST
Patent Text Reader

Abstract

This invention discloses a method and device for monitoring the entire life cycle of grinding rollers and discs in a medium-speed coal mill, relating to the field of coal mill monitoring technology. The method includes: using a handheld 3D scanner to collect pre-wear and post-wear dimensional data of the grinding rollers and discs of a medium-speed coal mill; performing 3D modeling to generate a pre-wear and post-wear global 3D model; performing a difference comparison calculation between the pre-wear and post-wear global 3D models to obtain global wear distribution data; and using a remaining life logical prediction model to predict and analyze the global wear distribution data, determining the remaining life assessment result for the entire life cycle and providing visual notifications. This invention solves the technical problem in existing technologies where the wear state of coal mill grinding rollers and discs is difficult to assess in a timely and accurate manner, leading to an impact on equipment operational stability. It achieves the technical effect of improving the accuracy of wear state assessment and enhancing equipment operational stability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of coal mill monitoring technology, specifically to a method and device for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill. Background Technology

[0002] As a key pulverizing device in coal-fired power generating units, the medium-speed coal mill operates under high load, high impact, and strong abrasion conditions for extended periods. With increasing operating time, the surfaces of the grinding rollers and discs continuously wear down, leading to changes in grinding gaps, decreased coal powder preparation efficiency, and fluctuations in equipment operating parameters. Therefore, accurately assessing the wear condition of the grinding rollers and discs is crucial for ensuring stable operation of the coal mill and developing reasonable maintenance plans. Currently, the condition monitoring of coal mill grinding rollers and discs mainly relies on manual inspections, shutdown measurements, or indirect judgments based on changes in operating parameters. Due to the uneven spatial distribution and significant local differences in wear, traditional detection methods struggle to comprehensively capture the actual changes in the wear area, easily resulting in insufficient detection accuracy and inaccurate assessment of wear degree. Furthermore, existing methods typically rely on single-inspection results for condition evaluation, lacking continuous analysis of the wear process, making it difficult to accurately predict the remaining service life of the equipment. This leads to a lack of effective basis for setting equipment maintenance cycles, increasing the risk of unplanned downtime and affecting the long-term stable operation of the coal mill. Summary of the Invention

[0003] This application provides a method and device for monitoring the entire life cycle of grinding rollers and grinding discs in medium-speed coal mills, which solves the technical problem that the wear status of grinding rollers and grinding discs in coal mills is difficult to assess in a timely and accurate manner, thus affecting the stability of equipment operation.

[0004] The first aspect of this application provides a method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill, the method comprising: A fixed handheld 3D scanner is calibrated, and the scanner is used to collect the pre-wear and post-wear dimensions of the grinding rollers and discs of a medium-speed coal mill. Based on these dimensions, 3D models are created to generate a pre-wear global 3D model and a post-wear global 3D model. The pre-wear and post-wear global 3D models are compared using a difference calculation to obtain global wear distribution data. A remaining life logical prediction model is constructed, and this model is used to predict and analyze the global wear distribution data to determine the full lifecycle remaining life assessment results and provide visual notifications.

[0005] A second aspect of this application provides a full life-cycle monitoring device for the grinding rollers and grinding discs of a medium-speed coal mill, the device comprising: Data Acquisition Module: calibrates a fixed handheld 3D scanner and uses it to acquire pre-wear and post-wear dimensional data of the grinding rollers and discs of a medium-speed coal mill; 3D Modeling Module: performs 3D modeling based on the pre-wear and post-wear dimensional data to generate a pre-wear global 3D model and a post-wear global 3D model; Comparison and Calculation Module: performs difference comparison calculations between the pre-wear and post-wear global 3D models to obtain global wear distribution data; Life Prediction Module: constructs a remaining life logical prediction model, uses this model to predict and analyze the global wear distribution data, determines the full life cycle remaining life assessment result, and provides visual push notifications.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, a fixed handheld 3D scanner was calibrated, and its dimensions before and after wear were collected from the grinding rollers and discs of a medium-speed coal mill. Next, 3D models were created based on these two data points, generating a pre-wear global 3D model and a post-wear global 3D model. Then, the pre-wear and post-wear global 3D models were compared using a difference calculation to obtain the global wear distribution data. Finally, a remaining life logical prediction model was constructed, and this model was used to predict and analyze the global wear distribution data, determining the full lifecycle remaining life assessment result and providing visual notifications. This approach solves the technical problem in existing technologies where the wear status of coal mill grinding rollers and discs is difficult to assess in a timely and accurate manner, thus affecting equipment operational stability. It achieves the technical effect of improving the accuracy of wear status assessment and enhancing equipment operational stability. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 A schematic diagram of a method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill, provided in an embodiment of this application; Figure 2 This is a schematic diagram of a full life cycle monitoring device for a medium-speed coal mill roller and grinding disc, provided in an embodiment of this application.

[0009] Figure labeling: Data acquisition module 11, 3D modeling module 12, comparison and calculation module 13, life prediction module 14. Detailed Implementation

[0010] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0011] Example 1, as Figure 1 As shown, this application provides a method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill, wherein the method includes: A fixed handheld 3D scanner was calibrated, and the handheld 3D scanner was used to collect the dimensional data of the grinding rollers and grinding discs of the medium-speed coal mill before and after wear.

[0012] In this embodiment, a handheld 3D scanner is installed at a preset fixed support position. By adjusting the relative distance and scanning angle between the scanner and the grinding roller and grinding disc of the medium-speed coal mill, the scanning range covers the outer circular surface and end face of the grinding roller and the working surface of the grinding disc. The scanner calibration program is started, and the spatial attitude parameters, scale parameters, and point cloud coordinate transformation parameters of the scanner are calibrated using a standard calibration plate or a calibration part of known size. This eliminates positional offset and scale error during the scanning process and establishes a spatial mapping relationship between the scanner coordinate system and the coal mill equipment coordinate system.

[0013] Furthermore, the handheld 3D scanner is used to collect the pre-wear and post-wear dimensional data of the grinding rollers and discs of the medium-speed coal mill, including: The handheld 3D scanner is used to perform circular motion scanning along the grinding rollers of the medium-speed coal mill before and after wear, respectively, to collect the shape data of the grinding rollers before and after wear. The handheld 3D scanner is clamped at the manhole inspection door frame, the grinding rollers of the medium-speed coal mill are raised, and the grinding disc is rotated using a rotating tool. After the grinding disc rotates once, the shape data of the grinding disc before and after wear is collected by the handheld 3D scanner. The shape data of the grinding rollers before and after wear, and the shape data of the grinding disc before and after wear are integrated to obtain the shape and dimension data before and after wear.

[0014] Specifically, a handheld 3D scanner performs a circular scan along the outer periphery of the grinding roller of a medium-speed coal mill, according to a preset scanning distance and angle, so that the scanning area covers the working surface, side wall area, and edge transition area of ​​the grinding roller. During the scanning process, the handheld 3D scanner continuously acquires point cloud data formed by reflections from the grinding roller surface, and performs spatial registration on the point cloud data acquired at different angles based on the position and posture information during scanning, forming 3D point cloud data of the grinding roller covering the entire outer surface of the grinding roller. Scanning is performed in the initial state before wear and in the worn state after a preset operating cycle, using the same scanning path, the same scanning range, and the same spatial reference, to obtain the shape data of the grinding roller before wear and the shape data of the grinding roller after wear, respectively.

[0015] When acquiring shape data for the grinding disc structure, a handheld 3D scanner is fixedly installed at the manhole inspection door frame of the medium-speed coal mill, maintaining a fixed spatial posture relative to the grinding disc. During the inspection process, the grinding rollers of the medium-speed coal mill are raised using a lifting mechanism, exposing the working area of ​​the grinding disc within the scanning field of view. Then, a rotating tool is used to drive the grinding disc to slowly rotate around its central axis. When the grinding disc completes one full rotation, the handheld 3D scanner simultaneously acquires surface point cloud data at different rotation angles. Based on the grinding disc rotation angle information, the continuously acquired point cloud data is stitched and fused to obtain 3D shape data covering the entire working surface of the grinding disc. The above scanning process is performed in both the pre-wear and post-wear states to obtain the pre-wear and post-wear shape data of the grinding disc.

[0016] Furthermore, the shape data of the grinding roller before wear is spatially correlated with the shape data of the grinding disc before wear, and the shape data of the grinding roller after wear is integrated with the shape data of the grinding disc after wear, so that the grinding roller area data and the grinding disc area data are under a unified equipment coordinate system, forming complete shape dimension data before wear and shape dimension data after wear, respectively. The shape dimension data before and after wear includes at least the spatial coordinate information, surface contour information, and geometric dimension information corresponding to each detection position on the grinding roller and grinding disc surfaces, which are used for subsequent construction of the full-domain 3D model before and after wear.

[0017] Based on the pre-wear and post-wear dimensional data, three-dimensional modeling is performed to generate a pre-wear global three-dimensional model and a post-wear global three-dimensional model.

[0018] Furthermore, based on the pre-wear and post-wear dimensional data, three-dimensional modeling is performed to generate a pre-wear global three-dimensional model and a post-wear global three-dimensional model, including: The pre-wear and post-wear dimensional data are filtered and denoised to obtain usable pre-wear and post-wear dimensional data. The usable pre-wear and post-wear dimensional data are then modeled and stitched together to obtain a basic pre-wear 3D model and a basic post-wear 3D model. Surface reconstruction is then performed on the pre-wear and post-wear 3D models to generate a pre-wear global 3D model and a post-wear global 3D model.

[0019] Specifically, the pre-wear and post-wear dimension data are converted into corresponding 3D point cloud datasets, where each point cloud dataset contains spatial coordinate information. For each sampling point in the 3D point cloud dataset, anomaly detection is performed using a filtering method based on neighborhood distance statistics. The average distance between the current point and its neighboring points is calculated. When the average distance deviates from the overall distance distribution range of the point cloud, the corresponding sampling point is identified as an abnormal noise point and removed. Furthermore, a neighborhood smoothing algorithm is used to smooth the retained point cloud. By calculating the mean of the coordinates of the neighboring points, the local point cloud position deviation is corrected, reducing the point cloud noise caused by equipment vibration, environmental reflection, and acquisition errors during the scanning process. This yields usable pre-wear and post-wear dimension data that can be used for 3D modeling. Subsequently, the available pre-wear and post-wear external dimension data are spatially divided. Based on the structural features of the grinding roller's outer surface, end face, and grinding disc working surface, the complete point cloud data is divided into multiple local point cloud regions. For each local point cloud region, point cloud connections are established according to the topological relationships between adjacent spatial points. The discrete point cloud is converted into a local 3D mesh model through triangulation. Furthermore, overlapping point cloud regions between adjacent local 3D mesh models are extracted. Spatial registration parameters are calculated based on the positional relationships of corresponding point clouds within the overlapping regions. These spatial registration parameters are then used to perform coordinate transformations on different local models, ensuring that each local model is placed in a unified spatial coordinate system. After spatial registration, the local 3D mesh models are stitched together and merged according to the correspondence between boundary points of adjacent regions, eliminating positional deviations at region connections, thus obtaining complete pre-wear and post-wear 3D models of the foundation. Further, surface reconstruction processing is performed on both the pre-wear and post-wear 3D models of the foundation. Specifically, the spatial coordinates of each vertex, the connection relationships of triangular faces, and the surface normal vector information of the basic 3D model are obtained. Continuous surface fitting constraints are established based on the spatial distribution relationships between adjacent vertices. A surface fitting algorithm is used to process the discrete mesh surface into a continuous surface, converting the triangular mesh model into a 3D surface model with continuous curvature changes. Simultaneously, the integrity of the reconstructed model is checked to identify hole regions, fracture regions, and regions with abrupt changes in curvature on the model surface. Surface compensation and repair are performed using the spatial variation trends of adjacent surface regions, ensuring that the reconstructed model maintains consistency with the surface morphology of the actual grinding roller / grinding disc. Through the above processing, a pre-wear global 3D model corresponding to the pre-wear state and a post-wear global 3D model corresponding to the post-wear state are generated, providing a foundation for subsequent calculation of wear distribution data based on the spatial differences between the two models.

[0020] Furthermore, the available pre-wear and post-wear dimensional data are modeled and stitched together to obtain a pre-wear 3D model and a post-wear 3D model of the foundation, including: Spatial region division and modeling are performed on the available pre-wear and post-wear dimensional data to generate a pre-wear 3D region model set and a post-wear 3D region model set; the pre-wear overlapping point cloud and post-wear overlapping point cloud of the pre-wear and post-wear 3D region model sets are identified and marked; the pre-wear 3D region model set and the post-wear 3D region model set are spliced ​​and fused based on the pre-wear overlapping point cloud and the post-wear overlapping point cloud to obtain a basic pre-wear 3D model and a basic post-wear 3D model.

[0021] First, the available pre-wear and post-wear dimensional data are spatially divided. Specifically, based on the actual structural morphology of the grinding roller and grinding disc, the grinding roller area is divided axially into the roller end area, the roller body working area, and the roller shoulder transition area; the grinding disc area is divided radially into the center area, the intermediate grinding area, and the outer edge area. Based on the spatial coordinate range of the point cloud corresponding to these structural areas, the available pre-wear and post-wear dimensional data are spatially indexed, and point cloud data within the same structural area are assigned to their respective areas, forming a pre-wear local point cloud region set and a post-wear local point cloud region set. Subsequently, triangular meshes are constructed for each local point cloud region set, connecting adjacent spatial points to generate triangular facets. Based on the topological connections between these triangular facets, a local three-dimensional region model is established, resulting in a pre-wear three-dimensional region model set and a post-wear three-dimensional region model set.

[0022] Further, overlapping point clouds are identified and marked in the pre-wear 3D region model set and the post-wear 3D region model set. Specifically, for the connection boundary of adjacent 3D region models, boundary point clouds located within a certain neighborhood of the region boundary are extracted, and the spatial distance between adjacent region boundary point clouds is calculated; when there is a point cloud correspondence between two region boundary point clouds that are spatially close and have the same normal direction, the corresponding point clouds are determined to be overlapping point clouds. Among them, for any first boundary point cloud... Calculate the Euclidean distance between it and each point in the second boundary point cloud set: ,in, This represents the spatial distance between the first boundary point cloud and the matched point cloud; Represents the spatial coordinates of the first boundary point cloud; The spatial coordinates of the candidate matching point cloud are represented. When the spatial distance is less than a set distance threshold and the angle between the normal vectors of the corresponding regions of the two point clouds is less than a set angle threshold, the two point clouds are determined to belong to the same overlapping region, and overlapping point cloud sets before wear and after wear are formed respectively.

[0023] Furthermore, the pre-wear and post-wear 3D region model sets are stitched and fused based on the pre-wear and post-wear overlapping point cloud sets. Specifically, the overlapping point clouds in adjacent region models are used as spatial registration references. Rigid body coordinate transformation is used to solve for the rotation matrix and translation vector between adjacent region models to minimize the spatial distance between corresponding overlapping point clouds. The rotation matrix and translation vector are iteratively optimized using the nearest point algorithm until the average distance error between corresponding point clouds meets the convergence condition. Subsequently, the spatially registered region models are transformed to a unified coordinate system, and the repeated point clouds in the overlapping regions are fused, retaining point cloud data with consistent spatial positions. The region boundary triangular faces are reconnected according to the model topology, and finally, complete pre-wear and post-wear 3D models of the foundation are obtained respectively.

[0024] Furthermore, identifying the pre-wear overlapping point cloud and post-wear overlapping point cloud of the pre-wear 3D region model set and the post-wear 3D region model set includes: Edge recognition is performed on the pre-wear 3D region model set and the post-wear 3D region model set to obtain the pre-wear region edge point cloud set and the post-wear region edge point cloud set; the similarity set of the region edge point cloud of the pre-wear region edge point cloud set and the post-wear region edge point cloud set are calculated respectively, and based on the similarity set of the region edge point cloud set, the pre-wear overlapping point cloud set and the post-wear overlapping point cloud set are screened and identified.

[0025] First, edge recognition processing is performed on the pre-wear and post-wear 3D region model sets respectively. Specifically, for each local 3D region model in the pre-wear and post-wear 3D region model sets, the vertex coordinate data, triangular facet connection relationships, and the number of adjacent faces corresponding to each vertex in the corresponding triangular mesh model are read; the local model boundary position is determined according to the mesh topology relationship, and vertices belonging to only a single triangular facet connection relationship, or vertices located in the outer contour region of the model and whose number of adjacent faces is less than a preset adjacency threshold, are identified as edge vertices; further, the 3D spatial coordinate information corresponding to the edge vertices is extracted to form edge point clouds of the pre-wear region and edge point clouds of the post-wear region respectively.

[0026] After obtaining the edge point cloud sets of the pre-wear region and the post-wear region, a matching analysis is performed on the edge point cloud features between each local 3D region model to determine areas with spatial overlap. Specifically, for the edge point cloud sets corresponding to two adjacent local 3D region models, edge point cloud feature description data is constructed. This edge point cloud feature description data includes the spatial distance relationship between edge points, the arrangement direction relationship of edge points, and local curvature change features. For a target point in any edge point cloud set, a nearest neighbor search method is used to determine the candidate matching point with the smallest spatial distance in another edge point cloud set, and the spatial distance error between the target point and the candidate matching point is calculated. Simultaneously, the consistency of edge extension direction is calculated based on the direction vectors between consecutive edge points, and the consistency of local curvature change is calculated based on the change in the normal vectors of adjacent points, to obtain the spatial feature matching results of the corresponding edge regions.

[0027] Further, a similarity set of region edge point clouds is calculated based on the spatial distance error, orientation consistency, and curvature consistency. Specifically, for any corresponding edge region, its edge point cloud similarity is calculated: ,in, This represents the similarity between the edge point cloud of the i-th region and the edge point cloud of the j-th region; This represents the distance similarity calculated based on the spatial distance between matching points; This represents the directional consistency similarity calculated based on the arrangement direction of edge points; This represents the curvature similarity calculated based on local curvature features. , , These represent the weight coefficients corresponding to the distance feature, direction feature, and curvature feature, respectively.

[0028] Among them, distance similarity It is determined based on the average distance error between matching points, specifically: ,in, This represents the coordinates of the k-th matching point in the first edge point cloud; This represents the coordinates of the k-th matching point in the second edge point cloud; N represents the number of matching points. This indicates the preset maximum allowable distance deviation.

[0029] Directional consistency similarity This is used to evaluate whether the extension directions of the edge point clouds of two regions are consistent. Specifically, for the two matched edge point cloud sets, the edge direction vector is calculated based on the spatial coordinate difference between adjacent matched points: , ,in, This represents the direction vector formed between adjacent matching points in the first edge point cloud; This represents the direction vector formed between corresponding adjacent matching points in the second edge point cloud; , Let represent the coordinates of the (k+1)th matching point. Further, calculate the cosine of the angle between the two direction vectors: ,in, Let represent the directional consistency evaluation value of the k-th matching edge point; the directional consistency similarity is obtained by averaging the directional consistency evaluation values ​​of all matching points. Where N-1 represents the number of direction vectors between adjacent points. The closer to 1, the more consistent the extension directions of the point clouds at the edges of the two regions.

[0030] Curvature similarity This is used to evaluate the consistency of local topographic changes at the edges of two regions. Specifically, local curvature features are first calculated based on the point cloud distribution within the neighborhood of each edge point. For any edge point... Obtain its neighborhood point set and calculate the neighborhood point covariance matrix: ,in, Let M represent the neighborhood covariance matrix corresponding to the k-th edge point; M represents the number of neighborhood points. This represents the coordinates of the m-th neighboring point; This represents the mean coordinates of the neighborhood points. Eigenvalue decomposition is performed on the covariance matrix to obtain three eigenvalues: The local curvature is determined based on the ratio of the minimum eigenvalue to the sum of eigenvalues: ,in, This represents the local curvature value at the k-th edge point; , , These represent the eigenvalues ​​corresponding to the covariance matrix. Further calculation of the curvature difference between the two corresponding edge point clouds is then performed. ,in, Indicates the first edge point cloud Local curvature values ​​at each matching point; This represents the local curvature value of the matching point corresponding to the second edge point cloud.

[0031] In this embodiment, distance feature weights are set based on the geometric features of the grinding roller and grinding disc scanning data and the model stitching accuracy requirements. directional feature weights Curvature feature weights This causes the calculation of region edge point cloud similarity to be simultaneously constrained by spatial location, structural orientation, and local topography. When the calculated region edge point cloud similarity... When the similarity exceeds the preset threshold of 0.85, it is determined that there is an effective overlapping area between the two corresponding 3D region models, and the corresponding matching point cloud is marked as an overlapping point cloud. The above processing is performed on the 3D region model set before wear and the 3D region model set after wear respectively, and the point cloud sets that meet the matching conditions are summarized to obtain the overlapping point cloud set before wear and the overlapping point cloud set after wear.

[0032] The global 3D model before and after wear is compared by difference calculation to obtain global wear distribution data.

[0033] Specifically, model reference points, boundary contour points, and structural feature points are extracted from both the pre-wear and post-wear global 3D models. A spatial correspondence is established between the two 3D models based on these reference points and feature points. Spatial coordinate transformation is used to correct the attitude of the post-wear global 3D model, ensuring that the pre-wear and post-wear global 3D models are in the same device coordinate system. Subsequently, using surface sampling points in the pre-wear global 3D model as reference points, matching points corresponding to their spatial positions are searched in the post-wear global 3D model. A correspondence is established between surface points before and after wear, and the spatial offset of each corresponding point along the normal direction of the model surface is calculated to obtain the local wear amount corresponding to each detection position. The formula for calculating the local wear amount is as follows: ,in, This represents the local wear amount corresponding to the i-th detection position; This represents the spatial coordinates of the i-th corresponding detection point in the global 3D model before wear. This represents the spatial coordinates of the i-th corresponding detection point in the global 3D model after wear. Further, based on the local wear amount corresponding to each detection position, the local wear amount is mapped to the 3D spatial surface of the grinding roller / disk. Spatial interpolation calculations are performed on the wear amount changes between adjacent detection positions to form a continuous wear amount distribution region. The wear amount in different spatial regions is then classified and labeled according to a preset wear degree classification rule, obtaining global wear distribution data including wear location, wear depth, wear range, and wear change trend. This global wear distribution data is used to characterize the spatial wear state of the grinding roller / disk of a medium-speed coal mill under different operating stages and serves as data input for subsequent remaining life logical prediction models for life assessment.

[0034] A residual lifetime logical prediction model is constructed, and the residual lifetime logical prediction model is used to predict and analyze the wear distribution data of the whole domain, determine the residual lifetime assessment results of the whole life cycle, and provide visual push notifications.

[0035] Furthermore, constructing a logical prediction model for remaining lifetime includes: Historical usage datasets of grinding rollers and grinding discs in medium-speed coal mills are collected. These datasets include data on wear-resistant layer thickness, wear rate, and corresponding service life before use. Sample association and annotation are performed on the historical usage datasets to obtain a sample set of grinding roller and grinding disc service life. Based on the sample set of grinding roller and grinding disc service life, a remaining service life prediction training is conducted to construct a logical prediction model for remaining service life.

[0036] When constructing the remaining service life logical prediction model, historical operating data of the grinding rollers and discs of a medium-speed coal mill were first collected, and a historical usage dataset was established according to the single usage cycle of the grinding rollers and discs. Specifically, the initial thickness of the wear-resistant layer before the grinding rollers and discs were put into use was obtained. Cumulative running time corresponding to each detection time The remaining wear-resistant layer thickness measured at each testing time. And the actual service life T corresponding to the time when the replacement standard is reached, and the wear rate of the corresponding stage is calculated based on the change in wear layer thickness at adjacent testing times: ,in, This represents the wear rate during the i-th operating phase; This represents the wear-resistant layer thickness at the (i-1)th detection time. This represents the thickness of the wear-resistant layer at the i-th detection time. This indicates the time interval between two consecutive detections.

[0037] Furthermore, the initial wear-resistant layer thickness, wear rate at different operating stages, and actual service life at which replacement conditions are met for the same grinding roller / grinding disc object are correlated to form a sample set of grinding roller / grinding disc service life. Specifically, each sample set is represented as: And use the corresponding actual service life as the sample label: ,in, This represents the j-th group of grinding roller and grinding disc life characteristic samples; This represents the initial wear-resistant layer thickness corresponding to the j-th sample group; This represents the wear rate of the j-th sample during the n-th operating phase; This indicates the actual lifespan of the corresponding sample.

[0038] When training the remaining life prediction based on the sample set of grinding roller and grinding disc service life, the wear-resistant layer thickness and wear rate parameters in each sample group are first normalized to eliminate the influence of different equipment specifications and measurement scales on the prediction results. Then, based on the variation law of wear rate with operating time in historical samples, a logical relationship between wear state characteristics and remaining life is established. Specifically, the initial wear-resistant layer thickness and current wear rate are used as model inputs, and the remaining operating time before reaching the preset wear limit thickness is used as the output target. By minimizing the error between the predicted life value and the actual life value, the life prediction parameters are iteratively corrected to obtain the remaining life logical prediction model. The prediction process is expressed as follows: ,in, This indicates the predicted remaining operating life; Indicates the initial thickness of the wear-resistant layer before use; Indicates the current wear rate; This represents the lifetime prediction function obtained by training based on historical samples.

[0039] After model training is complete, a portion of the historical data is used as validation samples. The initial wear-resistant layer thickness and wear rate from the validation samples are input into the remaining life logical prediction model to calculate the predicted remaining life. The error is then compared with the actual remaining life corresponding to the validation samples. When the prediction error meets the preset accuracy requirements, the model parameters are determined, and the remaining life logical prediction model is completed. Through this method, the remaining life logical prediction model can predict the remaining operating life of the grinding rollers and discs of a medium-speed coal mill based on the current wear rate and remaining wear-resistant layer status calculated from subsequently acquired global wear distribution data.

[0040] Furthermore, the remaining life logical prediction model is used to predict and analyze the global wear distribution data to determine the full life cycle remaining life assessment results, including: The current wear rate and the thickness of the wear-resistant layer before use are calculated based on the obtained global wear distribution data. The remaining life logic prediction model is used to predict and analyze the current wear rate and the thickness of the wear-resistant layer before use to determine the remaining life assessment result for the entire life cycle.

[0041] Specifically, based on the wear depth data corresponding to each spatial location in the global wear distribution data, the wear change value at the corresponding detection location between the current detection cycle and the previous detection cycle is obtained, and the current wear rate is calculated by combining the actual running time between the two detections. The current wear rate characterizes the degree of wear of the grinding roller / disc per unit time during the current operating phase, and its calculation method is as follows: ,in, Indicates the current wear rate; This indicates the cumulative wear depth corresponding to the current inspection cycle; This indicates the cumulative wear depth corresponding to the previous testing cycle; This represents the actual operating time interval between two adjacent detections. Further, based on the initial state information corresponding to the pre-wear dimensional data, the wear-resistant layer thickness data before the grinding roller / grinding disc is put into operation is obtained. Specifically, the wear-resistant layer region boundary is determined in the pre-wear global 3D model, the initial wear-resistant layer thickness is calculated based on the spatial distance between the outer surface of the wear-resistant layer and the substrate structure, and this initial wear-resistant layer thickness is used as the pre-use wear-resistant layer thickness parameter. Subsequently, the current wear rate and the pre-use wear-resistant layer thickness are input into the remaining lifespan logical prediction model for predictive analysis. Specifically, the remaining lifespan logical prediction model performs a matching analysis on the current wear state of the grinding roller / grinding disc based on the input current wear rate and pre-use wear-resistant layer thickness, determines the remaining operating time corresponding to the current wear state, and outputs the full lifespan remaining lifespan assessment result. Finally, the full lifespan remaining lifespan assessment result is correlated with the global wear distribution data, the remaining lifespan result is mapped to the corresponding wear area, and the remaining operating time, wear state, and maintenance prompts are sent to a visualization terminal for display, realizing the full lifespan operating status assessment of the medium-speed coal mill grinding roller / grinding disc.

[0042] In summary, the embodiments of this application have at least the following technical effects: First, a fixed handheld 3D scanner was calibrated, and its dimensions before and after wear were collected from the grinding rollers and discs of a medium-speed coal mill. Next, 3D models were created based on these two data points, generating a pre-wear global 3D model and a post-wear global 3D model. Then, the pre-wear and post-wear global 3D models were compared using a difference calculation to obtain the global wear distribution data. Finally, a remaining life logical prediction model was constructed, and this model was used to predict and analyze the global wear distribution data, determining the full lifecycle remaining life assessment result and providing visual notifications. This approach solves the technical problem in existing technologies where the wear status of coal mill grinding rollers and discs is difficult to assess in a timely and accurate manner, thus affecting equipment operational stability. It achieves the technical effect of improving the accuracy of wear status assessment and enhancing equipment operational stability.

[0043] Example 2, based on the same inventive concept as the method for monitoring the entire life cycle of grinding rollers and grinding discs in the aforementioned examples, such as... Figure 2 As shown, this application provides a full life-cycle monitoring device for the grinding rollers and grinding discs of a medium-speed coal mill, wherein the device includes: Data acquisition module 11: calibrates a fixed handheld 3D scanner and uses the handheld 3D scanner to acquire the pre-wear and post-wear dimensions of the grinding rollers and discs of a medium-speed coal mill; 3D modeling module 12: performs 3D modeling based on the pre-wear and post-wear dimensions to generate a pre-wear global 3D model and a post-wear global 3D model; Comparison and calculation module 13: performs difference comparison calculations between the pre-wear and post-wear global 3D models to obtain global wear distribution data; Life prediction module 14: constructs a remaining life logical prediction model, uses the remaining life logical prediction model to predict and analyze the global wear distribution data, determines the full life cycle remaining life assessment results, and provides visual push notifications.

[0044] Furthermore, the data acquisition module 11 is used to perform the following methods: The handheld 3D scanner is used to perform circular motion scanning along the grinding rollers of the medium-speed coal mill before and after wear, respectively, to collect the shape data of the grinding rollers before and after wear. The handheld 3D scanner is clamped at the manhole inspection door frame, the grinding rollers of the medium-speed coal mill are raised, and the grinding disc is rotated using a rotating tool. After the grinding disc rotates once, the shape data of the grinding disc before and after wear is collected by the handheld 3D scanner. The shape data of the grinding rollers before and after wear, and the shape data of the grinding disc before and after wear are integrated to obtain the shape and dimension data before and after wear.

[0045] Furthermore, the 3D modeling module 12 is used to perform the following methods: The pre-wear and post-wear dimensional data are filtered and denoised to obtain usable pre-wear and post-wear dimensional data. The usable pre-wear and post-wear dimensional data are then modeled and stitched together to obtain a basic pre-wear 3D model and a basic post-wear 3D model. Surface reconstruction is then performed on the pre-wear and post-wear 3D models to generate a pre-wear global 3D model and a post-wear global 3D model.

[0046] Furthermore, the 3D modeling module 12 is used to perform the following methods: Spatial region division and modeling are performed on the available pre-wear and post-wear dimensional data to generate a pre-wear 3D region model set and a post-wear 3D region model set; the pre-wear overlapping point cloud and post-wear overlapping point cloud of the pre-wear and post-wear 3D region model sets are identified and marked; the pre-wear 3D region model set and the post-wear 3D region model set are spliced ​​and fused based on the pre-wear overlapping point cloud and the post-wear overlapping point cloud to obtain a basic pre-wear 3D model and a basic post-wear 3D model.

[0047] Furthermore, the 3D modeling module 12 is used to perform the following methods: Edge recognition is performed on the pre-wear 3D region model set and the post-wear 3D region model set to obtain the pre-wear region edge point cloud set and the post-wear region edge point cloud set; the similarity set of the region edge point cloud of the pre-wear region edge point cloud set and the post-wear region edge point cloud set are calculated respectively, and based on the similarity set of the region edge point cloud set, the pre-wear overlapping point cloud set and the post-wear overlapping point cloud set are screened and identified.

[0048] Furthermore, the lifetime prediction module 14 is used to perform the following method: Historical usage datasets of grinding rollers and grinding discs in medium-speed coal mills are collected. These datasets include data on wear-resistant layer thickness, wear rate, and corresponding service life before use. Sample association and annotation are performed on the historical usage datasets to obtain a sample set of grinding roller and grinding disc service life. Based on the sample set of grinding roller and grinding disc service life, a remaining service life prediction training is conducted to construct a logical prediction model for remaining service life.

[0049] Furthermore, the lifetime prediction module 14 is used to perform the following method: The current wear rate and the thickness of the wear-resistant layer before use are calculated based on the obtained global wear distribution data. The remaining life logic prediction model is used to predict and analyze the current wear rate and the thickness of the wear-resistant layer before use to determine the remaining life assessment result for the entire life cycle.

[0050] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill, characterized in that, The method includes: A fixed handheld 3D scanner was calibrated, and the handheld 3D scanner was used to collect the external dimensional data of the grinding roller and grinding disc of the medium-speed coal mill before and after wear. Based on the pre-wear and post-wear dimensional data, three-dimensional modeling is performed to generate a pre-wear global three-dimensional model and a post-wear global three-dimensional model. The difference comparison calculation is performed between the pre-wear global 3D model and the post-wear global 3D model to obtain global wear distribution data. A residual lifetime logical prediction model is constructed, and the residual lifetime logical prediction model is used to predict and analyze the wear distribution data of the whole domain, determine the residual lifetime assessment results of the whole life cycle, and provide visual push notifications.

2. The method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill as described in claim 1, characterized in that, The handheld 3D scanner was used to collect the pre-wear and post-wear dimensional data of the grinding rollers and discs of a medium-speed coal mill, including: The handheld 3D scanner is used to perform circumferential scanning along the medium-speed coal mill roller before and after wear, respectively, to collect the roller's shape data before wear and after wear. The handheld 3D scanner is clamped in the manhole inspection door frame, the grinding roller of the medium-speed coal mill is raised, and the grinding disc is rotated by a rotating tool. After the grinding disc rotates once, the handheld 3D scanner is used to collect the shape data of the grinding disc before wear and the shape data of the grinding disc after wear. By integrating the shape data of the grinding roller before wear and the shape data of the grinding roller after wear, and the shape data of the grinding disc before wear and the shape data of the grinding disc after wear, the shape dimension data before wear and the shape dimension data after wear are obtained.

3. The method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill as described in claim 1, characterized in that, Based on the pre-wear and post-wear dimensional data, three-dimensional modeling is performed to generate a pre-wear global three-dimensional model and a post-wear global three-dimensional model, including: The pre-wear and post-wear dimensional data are filtered and denoised to obtain usable pre-wear and post-wear dimensional data. The available pre-wear and post-wear dimensional data are modeled and stitched together to obtain a three-dimensional model of the foundation before wear and a three-dimensional model of the foundation after wear. The three-dimensional models of the foundation before and after wear are reconstructed using surfaces to generate a global three-dimensional model before and after wear.

4. The method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill as described in claim 3, characterized in that, The available pre-wear and post-wear dimensional data are modeled and stitched together to obtain a pre-wear 3D model and a post-wear 3D model of the foundation, including: Spatial region division and modeling are performed on the available pre-wear and post-wear shape dimension data to generate a pre-wear three-dimensional region model set and a post-wear three-dimensional region model set. Identify and mark the pre-wear overlap point cloud and post-wear overlap point cloud of the pre-wear three-dimensional region model set and the post-wear three-dimensional region model set; Based on the pre-wear overlap point cloud and the post-wear overlap point cloud, the pre-wear three-dimensional region model set and the post-wear three-dimensional region model set are spliced ​​and fused to obtain the basic pre-wear three-dimensional model and the basic post-wear three-dimensional model.

5. The method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill as described in claim 4, characterized in that, Identifying and marking the pre-wear overlapping point cloud and post-wear overlapping point cloud of the pre-wear 3D region model set and the post-wear 3D region model set, including: Edge recognition is performed on the pre-wear three-dimensional region model set and the post-wear three-dimensional region model set to obtain the pre-wear region edge point cloud and the post-wear region edge point cloud; Calculate the similarity sets of the edge point clouds of the region before wear and the edge point clouds of the region after wear, respectively, and based on the similarity sets of the edge point clouds, filter and identify the overlapping point clouds of the region before wear and the overlapping point clouds of the region after wear.

6. The method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill as described in claim 1, characterized in that, Constructing a logical prediction model for remaining lifetime, including: Collect historical usage datasets of grinding rollers and grinding discs of medium-speed coal mills. The historical usage datasets include data on wear-resistant layer thickness before use, wear rate data, and corresponding service life data. The historical usage dataset is used to perform sample association annotation to obtain a sample set of grinding roller and grinding disc lifespan. Based on the sample set of grinding roller and grinding disc service life, a remaining life prediction training is performed to construct a remaining life logical prediction model.

7. The method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill as described in claim 1, characterized in that, The remaining life logical prediction model is used to predict and analyze the global wear distribution data to determine the full life cycle remaining life assessment results, including: The current wear rate and the thickness of the wear-resistant layer before use are calculated to obtain the global wear distribution data; The remaining life logic prediction model is used to predict and analyze the current wear rate and the thickness of the wear-resistant layer before use, and to determine the remaining life assessment result for the entire life cycle.

8. A monitoring device for the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill, characterized in that, The apparatus for implementing the method for monitoring the entire life cycle of grinding rollers and grinding discs in a medium-speed coal mill as described in any one of claims 1-7 comprises: Data acquisition module: calibrate a fixed handheld 3D scanner and use the handheld 3D scanner to acquire the external dimensional data of the grinding rollers and grinding discs of the medium-speed coal mill before and after wear; 3D modeling module: Based on the pre-wear and post-wear dimensional data, 3D modeling is performed to generate a pre-wear global 3D model and a post-wear global 3D model. Comparison and calculation module: Performs difference comparison calculation on the pre-wear global 3D model and the post-wear global 3D model to obtain global wear distribution data; Lifetime prediction module: Constructs a remaining lifetime logical prediction model, uses the remaining lifetime logical prediction model to predict and analyze the wear distribution data across the entire domain, determines the remaining lifetime assessment results for the entire life cycle, and provides visual push notifications.