A Shield Machine Tool Wear Monitoring Method and System

Abstract

The present invention relates to the technical field of material detection. Specifically, it relates to a method and system for monitoring the wear of shield machine cutters. The present invention constructs a linear equation by obtaining the deviation parameters between the cutter edge profiles of the shield machine at each time point, calculates the wear amount of the cutter edge at each time point, and can obtain the wear condition of the cutter edge in real time. By constructing a wear trace recognition model, the wear area on the cutter body surface at each time point is calculated. Based on the wear amount of the cutter edge and the wear area on the cutter body surface at each time point, the wear degree of the cutters of the shield machine at each time point is evaluated. By combining time series, a curve showing the change of the wear degree of the shield machine cutters over time is drawn, and the wear trend of the shield machine cutters is analyzed. It can quickly understand the wear law of the cutters, predict the future wear condition of the cutters, ensure the construction progress, and improve the construction efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of material detection, and in particular to a shield machine tool wear monitoring method and system. Background Art

[0002] With the acceleration of urbanization and the vigorous development of infrastructure construction, tunnel engineering plays an increasingly critical role in transportation, water conservancy, energy and other fields. As an efficient, safe and adaptable tunnel boring equipment for various complex geological conditions, shield machines are widely used in various tunnel construction projects. During the operation of the shield machine, the cutter is in direct contact with the rock and soil medium, and is subjected to huge cutting forces, friction and impact loads. Its wear directly affects the shield machine's excavation efficiency, construction costs and project quality. Therefore, the development of accurate and reliable shield machine cutter wear monitoring is of vital practical significance for ensuring the smooth progress of tunnel construction and improving construction efficiency.

[0003] At present, a variety of technical means have been developed for shield machine tool wear monitoring, but the existing technology still has many limitations. For example, the existing Chinese patent application number 202211449879.9 discloses a shield machine tool wear prediction method. This scheme is based on the collaborative working mechanism of minimizing the wear of tools of different combination types. Under the conditions of simulating different sand and gravel formations, a variety of tool combinations are selected for testing. The wear of the tool blocks in each tool combination is calculated by formula, and the overall excavation capacity of different tool combinations is uniformly quantitatively evaluated. The tool combination with the best adaptability to the excavation capacity under specific formation conditions is reasonably selected. It can better simulate the wear of multiple types of tools under collaborative cutting under real conditions and achieve the purpose of tool wear prediction.

[0004] For example, the existing Chinese patent application number 202411371810.8 discloses a tool wear detection device and method for a shield machine without opening the chamber. This solution uses an acoustic sensor to obtain acoustic signals of tools of different types and states, compares the tool signals collected after work with the initial database, and determines the degree of tool wear in real time. It can promptly detect the wear of the tool and provide a basis for tool maintenance and replacement.

[0005] However, the above patents all have the following problems: First, the solution only evaluates tool wear by calculating the wear of the cutting blade, which does not fully reflect the overall wear of the tool and cannot detect the wear of other parts of the tool (such as subtle wear marks on the surface of the tool body), which is not conducive to a comprehensive judgment of the remaining service life of the tool.

[0006] Second, this solution aims to predict tool wear by simulating the wear amount under the collaborative cutting of multiple types of tools in real - world scenarios. However, it does not involve the analysis of tool wear trends, and it is impossible to predict future wear degrees and wear rates based on the current wear situation. It is difficult to formulate reasonable tool maintenance plans and replacement strategies in advance, which may lead to equipment failures caused by excessive tool wear, affecting the construction progress and increasing maintenance costs. Summary of the Invention

[0007] To overcome the drawbacks in the background art, the embodiments of the present invention provide a method and system for monitoring the wear of shield machine tools, which can effectively solve the problems involved in the above - mentioned background art.

[0008] The object of the present invention can be achieved through the following technical solutions: The present invention provides a method for monitoring the wear of shield machine tools. The specific steps of the monitoring method are as follows: S1. Collect the blade images and tool body surface images of the shield machine at each time point through a high - definition camera.

[0009] S2. Extract the blade edge contours of the blade images at each time point through edge detection technology, and match them with the standard blade edge contours through the shape context matching algorithm to obtain the deviation parameters between the blade contours of the shield machine at each time point.

[0010] S3. Construct a linear equation based on the actual wear amount of the blade corresponding to the deviation parameters between the blade contours of the shield machine at each time point, and calculate the wear amount of the blade at each time point.

[0011] S4. Build a wear trace recognition model based on a convolutional neural network, identify and locate the wear traces in the tool body surface images at each time point, and then calculate the wear area of the tool body surface at each time point.

[0012] S5. Evaluate the wear degree of the shield machine tools at each time point based on the wear amount of the blade and the wear area of the tool body surface at each time point, draw a curve showing the change of the wear degree of the shield machine tools over time in combination with the time series, and analyze the wear trend of the shield machine tools.

[0013] Preferably, the step S2 includes the steps: S21. After pre - processing the blade images at each time point, extract the blade edge contours of the blade images at each time point through double - threshold detection and edge connection.

[0014] S22. Extract the standard blade edge contours from the management database, establish the corresponding relationship between the blade edge contours and the standard blade edge contours based on the shape context matching algorithm, and calculate the deviation parameters between the blade contours of the shield machine at each time point.

[0015] Preferably, S21 includes: converting the collected blade images at each time point into grayscale images, respectively calculating their gradient components in the X and Y directions, and calculating the gradient magnitude and gradient direction through trigonometric functions based on the gradient components in the X and Y directions. Traverse each pixel point. If the gradient magnitude of a certain pixel point is not the local maximum in its gradient direction, set the gradient magnitude of this pixel point to 0.

[0016] Set an upper limit and a lower limit for the gradient magnitude. Mark the pixel points with a gradient magnitude greater than the upper limit as strong edge pixels, mark the pixel points with a gradient magnitude between the upper limit and the lower limit as weak edge pixels, and suppress the pixel points with a gradient magnitude less than the lower limit as non-edge pixels. Through the edge connection algorithm, obtain the blade edge contour of the blade images at each time point.

[0017] Preferably, the specific steps of the shape context matching algorithm include: for the blade edge contours of the blade images at each time point and the pre-stored standard blade edge contours, select points at a fixed interval distance on the edge contours to form a point set. Construct a polar coordinate system with each point in the point set as the center point, set the angle range and radius range of the polar coordinate system, and divide them into several regions on average respectively to form a grid structure composed of each region. Denote each region as each grid region, and each grid region corresponds to a cell in the histogram.

[0018] Under the polar coordinate system of each center point, calculate the number of points in the point set other than the center point that fall into each grid region, and record the number of points contained in each grid region into the corresponding histogram cell, so as to construct a complete histogram. The complete histogram contains the distribution information of the points around each center point under the polar coordinate system, and use it as the shape context descriptor of each center point.

[0019] Preferably, establishing the correspondence between the blade edge contour and the standard blade edge contour in step S22 includes the steps of: taking the number of points on the blade edge contour as the number of rows of the matrix and the number of points on the standard blade edge contour as the number of columns of the matrix to construct a cost matrix. For each point on the blade edge contour and each point on the standard blade edge contour, measure the similarity between the distributions represented by their corresponding shape context descriptors through the Bhattacharyya distance, and use the similarity as the matching cost to form each element in the cost matrix.

[0020] In the cost matrix, find a set of corresponding point pairs from the sets of edge contour points of the cutting edge at each time point and the standard edge contour points of the cutting edge respectively. The set of corresponding point pairs consists of multiple point pairs, such that the sum of the matching costs corresponding to this set of corresponding point pairs in the cost matrix is minimized. Based on the sets of point pairs with the minimum sum of matching costs found, establish the correspondence between the edge contour of the cutting edge image at each time point and the standard edge contour of the cutting edge.

[0021] Preferably, the specific calculation method of the deviation parameter between the cutting edge contours of the shield machine is as follows: According to the correspondence between the edge contour of the cutting edge image at each time point and the standard edge contour of the cutting edge, calculate the distance between each corresponding point pair through the Euclidean distance formula, and obtain the pixel distance deviation between the cutting edge contours of the shield machine by taking the average value, thereby obtaining the pixel distance deviation between the cutting edge contours in the cutting edge image at each time point.

[0022] For each corresponding point pair, obtain the coordinates of the points on the edge contour of the cutting edge and the standard edge contour of the cutting edge at the corresponding time point, respectively select its adjacent points to calculate the coordinate difference, and based on the coordinate difference, use the arctangent function to calculate the tangent direction angle at each point, and then calculate the angle deviation between each corresponding point pair. Obtain the angle deviation between the cutting edge contours of the shield machine by taking the average value of the angle deviations between each corresponding point pair, thereby obtaining the angle deviation between the cutting edge contours in the cutting edge image at each time point.

[0023] Preferably, S3 includes: According to the time interval corresponding to the deviation parameter between the cutting edge contours of the shield machine at each time point, perform 3D modeling on the tool images of the shield machine at each time point, obtain the volume difference between it and the initial tool model through comparison, and then obtain the actual wear amount of the cutting edge at the corresponding time point. By fitting the linear equation between the deviation parameter and the wear amount of the cutting edge, calculate the wear amount of the cutting edge at each time point.

[0024] Preferably, the specific operation method of step S4 is: Obtain a large amount of image data of the tool body surface of the shield machine, construct a wear trace recognition model based on the convolutional neural network, input the tool body surface images of the shield machine at each time point into the wear trace recognition model for recognition, and recognize and locate the wear traces in the tool body surface images at each time point.

[0025] According to the pixel coordinates of the wear traces in the image, extract the wear trace texture on the tool body surface, count the number of pixels covered by the texture edge contour, and obtain the product of it and the actual area represented by each pixel, thereby obtaining the wear area on the tool body surface at each time point.

[0026] Preferably, the specific analysis method in step S5 is as follows: taking the period between adjacent time points as one stage, by calculating the ratio of the change in the wear degree of the shield machine cutter in each stage to the time interval in the curve of the wear degree of the shield machine cutter changing with time, the wear rate of each stage of the shield machine is obtained.

[0027] If the wear rates of all stages are lower than the set wear rate reference value, it is determined that the shield machine cutter is in the normal wear stage and can continue to work normally. On the contrary, compare the growth rate of the wear rates of each stage. If the growth rate of the wear rate in a certain stage is positive, it is determined that the cutter wear in this stage is accelerating, and a warning is given for this stage.

[0028] Set a wear degree threshold, predict the time point when the wear degree of the shield machine cutter reaches the set wear degree threshold according to the curve of the wear degree of the shield machine cutter changing with time, and feedback the wear trend analysis situation to the system.

[0029] Preferably, the present invention provides a shield machine cutter wear monitoring system, including: a cutter image acquisition module for acquiring the blade image and the cutter body surface image of the shield machine at each time point through a high-definition camera.

[0030] A blade contour deviation analysis module for extracting the blade edge contour of the blade image at each time point through edge detection technology, and matching it with the standard blade edge contour through the shape context matching algorithm to obtain the deviation parameters between the blade contours of the shield machine at each time point.

[0031] A blade wear quantification module for constructing a linear equation according to the actual wear amount of the blade corresponding to the deviation parameters between the blade contours of the shield machine, and calculating the wear amount of the blade at each time point.

[0032] A wear trace positioning module for constructing a wear trace recognition model based on a convolutional neural network, identifying and positioning the wear traces in the cutter body surface image at each time point, and then calculating the wear area of the cutter body surface at each time point.

[0033] A wear trend analysis module for evaluating the wear degree of the shield machine cutter at each time point based on the wear amount of the blade and the wear area of the cutter body surface at each time point, combining time series to draw a curve of the wear degree of the shield machine cutter changing with time, and analyzing the wear trend of the shield machine cutter.

[0034] A management database for storing the standard blade edge contour, the deviation parameters between the blade contours of the shield machine, the cutter body surface image data of the shield machine, the cutter body surface image of the shield machine at each time point, and the deviation parameter values between the blade contours of the shield machine at each time point.

[0035] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: First, the present invention extracts the edge contours of the cutter blade images at each time point through edge detection technology, and matches them with the standard edge contours of the cutter blade to obtain the deviation parameters between the cutter blade contours of the shield machine at each time point. The combination of edge detection and shape context matching algorithm reduces the interference of external factors on the detection of cutter blade wear, and improves the reliability and stability of the detection results.

[0036] Second, the present invention constructs a linear equation through the deviation parameters between the cutter blade contours of the shield machine at each time point, calculates the wear amount of the cutter blade at each time point, can obtain the wear situation of the cutter blade in real time, and is convenient for timely grasping the tool state.

[0037] Third, the present invention constructs a wear trace recognition model to identify and locate the wear traces in the tool body surface images at each time point, and calculates the wear area of the tool body surface at each time point, which helps to timely discover the early damage signs of the tool, take corresponding measures, and prevent the wear from further expanding.

[0038] Fourth, the present invention evaluates the wear degree of the shield machine tools at each time point based on the wear amount of the cutter blade and the wear area of the tool body surface at each time point, combines with the time series to draw a curve of the wear degree of the shield machine tools changing with time, analyzes the wear trend of the shield machine tools, and can quickly understand the wear law of the tools and predict the future wear situation of the tools by observing the trend of the curve. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0040] Figure 1 It is a flowchart of a shield machine tool wear monitoring system.

[0041] Figure 2 is Figure 1 The flowchart of step S2 in

[0042] Figure 3 It is a module connection diagram of a shield machine tool wear monitoring method. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0044] Please refer to Figure 1 As shown in the figure, a method for monitoring the wear of a shield machine cutter, the specific steps of the monitoring method are as follows: S1. Collect the blade images and the surface images of the cutter body of the shield machine at each time point through a high-definition camera.

[0045] S2. Extract the blade edge contours of the blade images at each time point through edge detection technology, and match them with the standard blade edge contours through the shape context matching algorithm to obtain the deviation parameters between the blade contours of the shield machine at each time point.

[0046] It should be noted that in the shape context matching algorithm, the shape context is a descriptor based on the object contour. By sampling points on the contour and constructing a polar coordinate system centered on these points to describe the shape, for each sampling point, the distribution of other points around it in the polar coordinates is statistically calculated to form a histogram, and this histogram is used as the shape context descriptor of this point.

[0047] The core idea of this algorithm is that for two similar shapes, the shape contexts of their corresponding points should have similar distributions. By comparing the differences between the shape context descriptors of the corresponding points on different shapes, the similarity between the shapes can be measured.

[0048] Please refer to Figure 2 As shown in the figure, the step S2 includes the steps: S21. After preprocessing the blade images at each time point, extract the blade edge contours of the blade images at each time point through double-threshold detection and edge connection.

[0049] S22. Extract the standard blade edge contours from the management database, establish the corresponding relationship between the blade edge contours and the standard blade edge contours based on the shape context matching algorithm, and calculate the deviation parameters between the blade contours of the shield machine at each time point.

[0050] The S21 includes: converting the collected blade images at each time point into grayscale images, respectively calculating their gradient components in the X and Y directions, and calculating the gradient magnitude and gradient direction through trigonometric functions based on the gradient components in the X and Y directions. Traversing each pixel point, if the gradient magnitude of a certain pixel point is not the local maximum in its gradient direction, the gradient magnitude of this pixel point is set to 0; converting the collected blade image into a grayscale image can reduce the complexity of the image, focus the attention on the grayscale features of the blade, and reduce the interference brought by color information. By calculating the gradient components and directions in the X and Y directions, the gradient changes at the blade edge can be accurately captured. Using non-maximum suppression, that is, setting the gradient magnitude of the pixel points whose gradient magnitude is not the local maximum to 0, can eliminate those pixel points that are not real edges and make the edge clearer and more accurate.

[0051] Exemplarily, the coordinates of the pixel point are , and its neighborhood pixel values are , where , , then the gradient component of this point in the X direction is: , where is the template element of the Sobel operator in the X direction.

[0052] Similarly, the gradient component of this point in the Y direction is: , where is the template element of the Sobel operator in the Y direction.

[0053] After obtaining the gradient components in the X direction and the Y direction, the gradient direction is calculated through the formula.

[0054] It should be noted that the formula for the gradient magnitude is .

[0055] The specific calculation methods for the gradient component and the gradient direction are: using the Sobel operator to calculate the gradient component. The template of the Sobel operator in the X direction is: When calculating, the template is slid point by point on the grayscale image. For each pixel point, the template is convolved with the pixel values of this point and its neighborhood.

[0056] The specific operation method for judging the local maximum is: 0 direction (horizontal direction): If the gradient direction of the pixel point is close to 0, check the gradient magnitudes of the left and right adjacent pixel points of this pixel point in the horizontal direction. If the gradient magnitude of this pixel point is less than any one of the left and right adjacent pixel points, the gradient magnitude of this pixel point is set to 0.

[0057] 45 - direction: When the gradient direction of a pixel is close to 45, check the gradient magnitudes of the adjacent pixels (upper - right and lower - left) of this pixel in the 45 - direction. If the gradient magnitude of this pixel is less than either of these two adjacent pixels, set its gradient magnitude to 0.

[0058] 90 - direction (vertical direction): If the gradient direction of a pixel is close to 90, check the gradient magnitudes of the adjacent pixels (above and below) of this pixel in the vertical direction. If the gradient magnitude of this pixel is less than either of the adjacent pixels above and below, set its gradient magnitude to 0.

[0059] 135 - direction: When the gradient direction of a pixel is close to 135, check the gradient magnitudes of the adjacent pixels (upper - left and lower - right) of this pixel in the 135 - direction. If the gradient magnitude of this pixel is less than either of these two adjacent pixels, set its gradient magnitude to 0.

[0060] Set an upper limit and a lower limit for the gradient magnitude. Mark the pixels with gradient magnitudes greater than the upper limit as strong - edge pixels, mark the pixels with gradient magnitudes between the upper limit and the lower limit as weak - edge pixels, and suppress the pixels with gradient magnitudes less than the lower limit as non - edge pixels. Through the edge - connection algorithm, obtain the edge contour of the blade image at each time point.

[0061] It should be noted that the edge - connection algorithm connects weak - edge pixels to strong - edge pixels. If a weak - edge pixel is connected to a strong - edge pixel, it is retained as an edge pixel; otherwise, it is suppressed as a non - edge pixel. Thus, obtain the edge contour of the blade image at each time point.

[0062] It should be noted that the strong - edge pixels usually represent positions in the image where the gray - level change is very drastic, which are relatively definite edge positions.

[0063] The weak - edge pixels represent positions where the gray - level change is relatively small but may still be an edge. Although the gradient magnitudes of these pixels are not as high as those of the strong - edge pixels, they may be part of the real edge. However, due to the influence of noise, uneven illumination, or other factors, the gray - level change is not as obvious. The weak - edge pixels do not have values that are unrelated or negatively correlated with the edge, but are pixels with relatively weak gray - level change degrees but still having certain edge characteristics.

[0064] The non - edge pixels represent regions in the image where the gray - level change is very gentle and are usually considered not to be edge positions.

[0065] It should be noted that the specific operating steps of the edge connection algorithm are: starting from a strong edge pixel using the region growing method, using it as a seed point, checking the pixels in the neighborhood (e.g., a 4-neighborhood or an 8-neighborhood) of various sub-points, if the pixel in the neighborhood is a weak edge pixel, then the weak edge pixel is connected to the seed point and marked as a processed edge pixel.

[0066] The newly connected weak edge pixel is used as a new seed point, and the pixels in its neighborhood are continued to be checked, and the above process is repeated until there are no new weak edge pixels that can be connected.

[0067] It should be noted that in actual applications, edges may not be connected. The main reasons are as follows and the corresponding processing methods are as follows: Noise interference: The noise in the image may cause the gradient amplitude of weak edge pixels to be unstable, making it difficult to connect with strong edge pixels. The image can be denoised before edge detection to reduce the impact of noise on edge detection.

[0068] Severe edge breakage: If the blade edge is severely broken due to wear, image quality, or other reasons, the distance between strong edge pixels and weak edge pixels may be too large to connect them through conventional neighborhood search or morphological operations. You can try adjusting the parameters of the edge connection algorithm, such as expanding the neighborhood range or increasing the number of morphological operations, to improve the success rate of connection.

[0069] Uneven lighting: Uneven lighting can cause inconsistent grayscale changes in different areas of the image, reducing the gradient amplitude of some real edge pixels and causing them to be misjudged as non-edge pixels or weak edge pixels, thus affecting edge connections. Image enhancement technology can be used to adjust the grayscale distribution of the image and improve the contrast of edge pixels.

[0070] The specific steps of the shape context matching algorithm include: for the blade edge contour of the blade image at each time point and the pre-stored standard blade edge contour, points are selected on the edge contour at a fixed interval to form a point set, a polar coordinate system is constructed with each point in the point set as the center point, the angle range and radius range of the polar coordinate system are set, and it is evenly divided into several areas to form a grid structure composed of various areas, each area is recorded as a grid area, and each grid area corresponds to a unit in the histogram; points are selected on the blade edge contour and the standard edge contour at fixed intervals to form a point set, and the contour can be standardized to facilitate subsequent analysis.

[0071] It should be noted that the method for setting the angle range and radius range of the polar coordinate system is: the complete angle covered by one rotation around the center point is taken as the angle range, and the distance from the center point to the farthest point on the edge contour is recorded as the radius range.

[0072] It should be noted that through the angle and radius parameters in polar coordinates, the positions of points on the edge contour of the blade relative to the center point can be accurately characterized, and the local shape characteristics of the contour can be presented in detail. The shape context descriptor formed in polar coordinates can quantify the differences between the blade edge contours at different time points and the standard contour, clearly showing the degree and specific location of blade wear.

[0073] The edge contour of the blade is usually an irregular curve. Polar coordinates use the center point as a reference and describe the position of points with angles and radii, which fits the curve characteristics of the blade edge contour and can naturally and accurately express its shape. It is more suitable for describing such complex curves than rectangular coordinates. At the same time, in practical applications, the blade image may be rotated or scaled due to factors such as the acquisition angle and equipment. Polar coordinates have good invariance to rotation and scaling. When rotating, the angle changes but the radius remains unchanged, and when scaling, the radius changes proportionally while the angle remains unchanged. This enables unified feature description and comparison based on polar coordinates in blade images with different poses, effectively avoiding the interference of image transformation on the analysis results.

[0074] In the polar coordinate system of each center point, calculate the number of points in the point set (excluding the center point) that fall into each grid area, and record the number of points contained in each grid area into the corresponding histogram unit, thereby constructing a complete histogram. The complete histogram contains the distribution information of points around each center point in the polar coordinate system, and this is used as the shape context descriptor of each center point. Constructing a polar coordinate system with a point as the center, dividing the grid area and counting the number of points to form a shape context descriptor can comprehensively describe the distribution information of points around each point, providing rich features for subsequent matching.

[0075] It should be noted that the number of shape context descriptors depends on the number of center points. For the edge contours of the blade images at each time point and the pre-stored standard edge contour of the blade, if N center points are selected on the edge contour of the blade, then for each blade edge contour (including the blade images at each time point and the pre-stored standard edge contour of the blade), there will be N shape context descriptors.

[0076] If there are M time points of blade images, plus the pre-stored standard edge contour of the blade, there will be a total of shape context descriptors.

[0077] It should be noted that the shape context descriptor can be regarded as an approximation of a probability distribution, so the Bhattacharyya distance can be used to measure the similarity between two probability distributions.

[0078] Exemplarily, let two shape context descriptors be 、 ,where are respectively the values of the th histogram unit, is the total number of histogram units, and the Bhattacharyya distance is calculated by the formula: .

[0079] Establishing the correspondence between the edge contour of the blade and the standard edge contour of the blade in step S22 includes the steps of: taking the number of points on the edge contour of the blade as the number of rows of the matrix, taking the number of points on the standard edge contour of the blade as the number of columns of the matrix, constructing a cost matrix, for each point on the edge contour of the blade and each point on the standard edge contour of the blade, measuring the similarity between the distributions represented by their corresponding shape context descriptors by the Bhattacharyya distance, and taking the similarity as the matching cost to form each element in the cost matrix; measuring the similarity of the shape context descriptors by the Bhattacharyya distance and constructing a cost matrix can accurately reflect the matching cost between different contour points.

[0080] In the cost matrix, find a set of corresponding point pairs from the point sets of the edge contour of the blade and the standard edge contour of the blade at each time point respectively. The set of corresponding point pairs is composed of multiple point pairs, so that the sum of the matching costs corresponding to this set of corresponding point pairs in the cost matrix is the smallest. According to the combination of the sets of point pairs with the smallest sum of matching costs found, establish the correspondence between the edge contour of the blade image at each time point and the standard edge contour of the blade; finding the combination of point pairs with the smallest sum of matching costs in the cost matrix can establish the optimal correspondence between the edge contour of the blade and the standard edge contour and provide a basis for deviation calculation.

[0081] It should be noted that the correspondence between the edge contour of the blade image at each time point and the standard edge contour of the blade, that is, a one-to-one correspondence is established between the points on the edge contour of the blade that participate in forming the point pair combination and the points in the corresponding point pair on the standard edge contour of the blade.

[0082] It should be noted that the specific steps for finding a set of corresponding point pairs such that the sum of the matching costs corresponding to this set of corresponding point pairs in the cost matrix is the smallest are as follows: row reduction: for each row of the cost matrix, find the minimum value in that row, and then subtract this minimum value from each element in that row.

[0083] Column reduction: for each column of the matrix after row reduction, find the minimum value in that column, and then subtract this minimum value from each element in that column.

[0084] Finding independent zero elements: In the matrix after row reduction and column reduction, find as many independent zero elements as possible (i.e., there are no other zero elements in the rows and columns where these zero elements are located). If the number of independent zero elements is equal to the order of the matrix, the optimal match is found; otherwise, it is necessary to perform covering of zero elements.

[0085] Covering zero elements: Cover all zero elements in the matrix with the fewest straight lines (row lines and column lines), find the smallest element not covered by the straight lines, subtract this minimum value from all the uncovered elements, and add this minimum value to the elements covered by the intersection of two straight lines. Repeat the operations of finding independent zero elements and covering zero elements until the optimal match is found.

[0086] It should be noted that the deviation parameters between the blade profiles of the shield machine include pixel distance deviation and angle deviation.

[0087] The specific calculation method of the deviation parameters between the blade profiles of the shield machine is as follows: According to the corresponding relationship between the blade edge profiles of the blade images at each time point and the standard blade edge profile, calculate the distance between each corresponding point pair through the Euclidean distance formula, and obtain the pixel distance deviation between the blade profiles of the shield machine by taking the average value, thereby obtaining the pixel distance deviation between the blade profiles in the blade images at each time point; based on the corresponding relationship, calculate the distance between the corresponding point pairs using the Euclidean distance formula and take the average value to obtain the pixel distance deviation between the blade profiles, which can intuitively reflect the difference in the position of the blade profiles.

[0088] Exemplarily, the coordinates of the points on the blade edge profile at each time point are , and the corresponding coordinates on the standard blade edge profile are , then the Euclidean distance between these two points is calculated by the formula: .

[0089] For each corresponding point pair, obtain the coordinates of the points on the blade edge profile and the standard blade edge profile at the corresponding time point, respectively select their adjacent points to calculate the coordinate differences, based on the coordinate differences, use the arctangent function to calculate the tangent direction angles at each point, and then calculate the angle deviation between each corresponding point pair. Obtain the angle deviation between the blade profiles of the shield machine by taking the average value of the angle deviations between each corresponding point pair, thereby obtaining the angle deviation between the blade profiles in the blade images at each time point.

[0090] Exemplarily, for each corresponding point and its adjacent points, calculate the coordinate differences respectively. For the point on the blade edge profile, its coordinate differences from the adjacent points are respectively: , let be the tangent direction angle at the point, , similarly, obtain the points on the standard edge contour of the blade The tangent direction angle , calculate the angular deviation between them .

[0091] S3. Construct a linear equation based on the actual wear amount of the blade at the corresponding time points according to the deviation parameters between the blade contours of the shield machine, and calculate the wear amount of the blade at each time point.

[0092] It should be noted that the blade wear amount is the shape change of the blade contour. In a two-dimensional plane, the coordinate of a certain point on the original blade contour moves to another position after a period of use due to wear. The wear amount of this point is quantified by calculating the distance between the two points.

[0093] The step S3 includes: according to the time interval corresponding to the deviation parameters between the blade contours of the shield machine at each time point, perform 3D modeling on the tool images of the shield machine at each time point, obtain the volume difference between it and the initial tool model by comparison, and then obtain the actual wear amount of the blade at the corresponding time point. By fitting the linear equation between the deviation parameters and the blade wear amount, calculate the wear amount of the blade at each time point; the wear amount of the blade can be predicted according to the real-time deviation parameters, realizing early warning of the blade wear situation and avoiding shield machine failures caused by excessive blade wear.

[0094] It should be noted that the specific analysis method for the wear amount of the blade at each time point is: fit the equations by the least squares method, and substitute the deviation parameters between the blade contours of the shield machine into the equations, so as to obtain the linear equation between the deviation parameters and the blade wear amount. Substitute the deviation parameter values between the blade contours of the shield machine at each time point into the linear equation, and calculate the wear amount of the blade at each time point.

[0095] It should be noted that the detailed steps of the linear equation between the deviation parameters and the blade wear amount are as follows: assume that there are time points, and the time points are respectively , and the deviation parameters between the corresponding blade contours of the shield machine are , and the actual wear amount of the blade at the corresponding time point obtained by 3D modeling comparison is , establish a linear equation assuming that there is a linear relationship between the deviation parameters and the blade wear amount , where are the coefficients to be determined.

[0096] Use the least squares method to fit the linear equation to minimize the sum of squared errors , respectively take the partial derivatives of S with respect to and set the partial derivatives equal to 0 to obtain the following system of equations: , solve this system of equations to obtain values to determine the linear equation , represents the number of the th time point, respectively represent the actual wear amount of the cutter blade at the th time point and the deviation parameter between the cutter blade profiles of the shield machine.

[0097] Substitute the deviation parameter values between the cutter blade profiles of the shield machine at each time point into the linear equation to obtain the predicted values of the cutter blade wear amount at each time point.

[0098] Substitute the deviation parameter between the cutter blade profiles of the shield machine at the prediction time into the linear equation to obtain the predicted value of the cutter blade wear amount at the prediction time. At the same time, obtain the actual wear amount of the cutter blade at the prediction time through 3D modeling comparison, and calculate the error between the predicted value and the actual value.

[0099] If the error is within the pre-set threshold, it means that the current linear equation can better describe the relationship between the deviation parameter and the cutter blade wear amount, and the equation is normal. Otherwise, it is necessary to re-check the data or consider using a more complex model.

[0100] A feasible simulation process. Assuming that the set threshold is 0.5, the simulation results can be obtained. For details, refer to Table 1, in which some representative data are listed.

[0101] Table 1. Part of the collected wear data

[0102]

[0103] Through the data in Table 1, the prediction error at the prediction time can be calculated as 5.3 - 5 = 0.3, which is less than the set threshold, indicating that the current linear equation is normal.

[0104] S4. Construct a wear trace recognition model based on a convolutional neural network to identify and locate the wear traces in the cutter body surface images at each time point, and then calculate the wear area of the cutter body surface at each time point.

[0105] The specific operation method of step S4 is as follows: Obtain a large amount of cutter body surface image data of the shield machine from the data records in the previous construction process, construct a wear trace recognition model based on a convolutional neural network, input the cutter body surface images of the shield machine at each time point into the wear trace recognition model for recognition, and identify and locate the wear traces in the cutter body surface images at each time point; Use a convolutional neural network to identify and locate the wear traces on the cutter body surface, and then calculate the wear area, which can evaluate the wear condition of the cutter body from another dimension. Combining with the cutter blade wear amount, the overall wear condition of the cutter can be understood more comprehensively and accurately.

[0106] According to the pixel coordinates of the wear marks in the image, the wear mark texture on the tool body surface is extracted, the number of pixels covered by the texture edge contour is counted, and the product of the number and the actual area represented by each pixel is obtained, thereby obtaining the wear area on the tool body surface at each time point; the detection of the wear area on the tool body surface can discover some subtle wear that is difficult to detect in the detection of the edge contour of the cutting edge, such as surface abrasions and scratches, which helps to timely discover potential problems of the tool and take corresponding measures.

[0107] S5. Evaluate the tool wear degree of the shield machine at each time point based on the wear amount of the cutting edge and the wear area on the tool body surface at each time point, and draw a curve showing the change of the tool wear degree of the shield machine over time in combination with the time series to analyze the wear trend of the shield machine tool.

[0108] In a preferred embodiment of the present invention, the specific method for the tool wear degree of the shield machine at each time point is as follows: extract the wear amount of the cutting edge and the wear area on the tool body surface at each time point, compare them with the maximum allowable values of the preset wear amount of the cutting edge and the wear area on the tool body surface, and then perform a weighted summation calculation to obtain the tool wear degree of the shield machine at each time point.

[0109] Exemplarily, the weights corresponding to the wear amount of the cutting edge and the wear area on the tool body surface are .

[0110] It should be noted that the specific method of step S5 is: arrange the tool wear degree of the shield machine at each time point in chronological order to form a time series of the tool wear degree, use time as the abscissa and the tool wear degree as the ordinate to draw a curve showing the change of the tool wear degree of the shield machine over time.

[0111] The specific analysis method of the wear trend analysis is: take the interval between adjacent time points as a stage, and obtain the wear rate of each stage of the shield machine by calculating the ratio of the change amount of the wear degree in each stage to the time interval in the curve showing the change of the tool wear degree of the shield machine over time.

[0112] If the wear rates of all stages are lower than the set wear rate reference value, it is determined that the shield machine tool is in the normal wear stage and can continue to work normally. On the contrary, compare the growth amplitudes of the wear rates of each stage. If the growth amplitude of the wear rate of a certain stage is positive, it is determined that the tool wear in this stage is accelerating and a warning is given for this stage; as a clear quantitative index, the wear rate can more accurately and scientifically reflect the wear state of the tool in different stages compared with simply observing the wear degree curve.

[0113] Set a wear degree threshold, predict the time point when the wear degree of the shield machine cutter reaches the set wear degree threshold according to the curve of the wear degree of the shield machine cutter changing with time, and feedback the wear trend analysis situation to the system; setting the wear degree threshold and predicting the time point when the cutter reaches this threshold according to the curve can plan the cutter replacement time in advance, reduce the downtime caused by sudden cutter failure, and improve the working efficiency of the shield machine.

[0114] Please refer to Figure 3 As shown. In addition, the present invention provides a shield machine cutter wear monitoring system, which includes a cutter image acquisition module for acquiring the blade image and the cutter body surface image of the shield machine at each time point through a high-definition camera.

[0115] A blade contour deviation analysis module for extracting the blade edge contour of the blade image at each time point through edge detection technology, and matching it with the standard blade edge contour through a shape context matching algorithm to obtain the deviation parameters between the blade contours of the shield machine at each time point.

[0116] A blade wear quantification module for constructing a linear equation according to the actual wear amount of the blade corresponding to the deviation parameters between the blade contours of the shield machine at each time point, and calculating the blade wear amount at each time point.

[0117] A wear trace positioning module for constructing a wear trace recognition model based on a convolutional neural network, identifying and positioning the wear traces in the cutter body surface image at each time point, and then calculating the wear area of the cutter body surface at each time point.

[0118] A wear trend analysis module for evaluating the wear degree of the cutter of the shield machine at each time point based on the blade wear amount and the wear area of the cutter body surface at each time point, combining time series to draw a curve of the wear degree of the shield machine cutter changing with time, and analyzing the wear trend of the shield machine cutter.

[0119] A management database for storing the standard blade edge contour, the deviation parameters between the blade contours of the shield machine, the cutter body surface image data of the shield machine, the cutter body surface image of the shield machine at each time point, and the deviation parameter values between the blade contours of the shield machine at each time point.

[0120] The management database is connected to the cutter image acquisition module, the blade contour deviation analysis module, the blade wear quantification module, the wear trace positioning module, and the wear trend analysis module. The blade contour deviation analysis module is connected to the cutter image acquisition module, the blade wear quantification module, and the wear trace positioning module. The wear trend analysis module is connected to the blade wear quantification module and the wear trace positioning module. The cutter image acquisition module is connected to the wear trace positioning module.

[0121] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention, and still be covered by the protection scope of the present invention.

Claims

1. A method for monitoring the wear of a shield machine cutter, characterized in that, The specific steps of the monitoring method are as follows: S1. Collect the blade images and the surface images of the cutter body of the shield machine at each time point through a high-definition camera; S2. Extract the blade edge contours of the blade images at each time point through edge detection technology, and match them with the standard blade edge contours through the shape context matching algorithm to obtain the deviation parameters between the blade contours of the shield machine at each time point; S3. Construct a linear equation based on the actual wear amount of the blade corresponding to the deviation parameters between the blade contours of the shield machine at each time point, and calculate the wear amount of the blade at each time point; S4. Build a wear trace recognition model based on a convolutional neural network, identify and locate the wear traces in the surface images of the cutter body at each time point, and then calculate the wear area of the surface of the cutter body at each time point; S5. Evaluate the wear degree of the cutter of the shield machine at each time point based on the wear amount of the blade and the wear area of the surface of the cutter body at each time point, draw a curve of the wear degree of the cutter of the shield machine changing with time in combination with the time series, and analyze the wear trend of the cutter of the shield machine; The specific calculation method of the deviation parameters between the blade contours of the shield machine is as follows: According to the corresponding relationship between the blade edge contours of the blade images at each time point and the standard blade edge contours, calculate the distance between each corresponding point pair through the Euclidean distance formula, and obtain the pixel distance deviation between the blade contours of the shield machine by taking the average value, thereby obtaining the pixel distance deviation between the blade contours in the blade images at each time point; For each corresponding point pair, obtain the coordinates of the points on the blade edge contour and the standard blade edge contour at the corresponding time point, respectively select its adjacent points to calculate the coordinate difference, and based on the coordinate difference, use the arctangent function to calculate the tangent direction angle at each point, and then calculate the angle deviation between each corresponding point pair. Obtain the angle deviation between the blade contours of the shield machine by taking the average value of the angle deviations between each corresponding point pair, thereby obtaining the angle deviation between the blade contours in the blade images at each time point; According to the time interval corresponding to the deviation parameters between the blade contours of the shield machine at each time point, perform 3D modeling on the cutter images of the shield machine at each time point, obtain the volume difference between it and the initial cutter model by comparison, and then obtain the actual wear amount of the blade at the corresponding time point. By fitting the linear equation between the deviation parameters and the blade wear amount, calculate the wear amount of the blade at each time point.

2. The method for monitoring the tool wear of a shield machine according to claim 1, characterized in that: The step S2 includes the steps: S21. After preprocessing the blade images at each time point, extract the blade edge contours of the blade images at each time point through double-threshold detection and edge connection; S22. Extract the standard blade edge contours from the management database, establish the corresponding relationship between the blade edge contours and the standard blade edge contours based on the shape context matching algorithm, and calculate the deviation parameters between the blade contours of the shield machine at each time point.

3. The method for monitoring the wear of a shield machine cutter according to claim 2, characterized in that: The S21 includes: Convert the collected blade images at each time point into grayscale images, calculate their gradient components in the X and Y directions respectively, and calculate the gradient amplitude and gradient direction through trigonometric functions according to the gradient components in the X and Y directions. Traverse each pixel point. If the gradient amplitude of a certain pixel point is not the local maximum in its gradient direction, set the gradient amplitude of this pixel point to 0; Set the upper limit and lower limit of the gradient magnitude. Pixel points with a gradient magnitude greater than the upper limit of the gradient magnitude are marked as strong edge pixels, pixel points with a gradient magnitude between the upper limit and the lower limit of the gradient magnitude are marked as weak edge pixels, and pixel points less than the lower limit of the gradient magnitude are suppressed as non-edge pixels. Through the edge connection algorithm, the edge contour of the blade image at each time point is obtained.

4. A method for monitoring the wear of a shield machine cutter according to claim 2, characterized in that: The specific steps of the shape context matching algorithm include: For the edge contour of the blade image at each time point and the pre-stored standard edge contour of the blade, points are selected on the edge contour at a fixed interval distance to form a point set. A polar coordinate system is constructed with each point in the point set as the center point. Set the angular range and radius range of the polar coordinate system, and divide them into several regions on average respectively to form a grid structure composed of each region. Each region is recorded as each grid region, and each grid region corresponds to a unit in the histogram. Under the polar coordinate system of each center point, calculate the number of points in the point set other than the center point that fall into each grid region, and record the number of points contained in each grid region into the corresponding histogram unit, thereby constructing a complete histogram. The complete histogram contains the distribution information of the points around each center point under the polar coordinate system, and it is used as the shape context descriptor of each center point.

5. A shield machine tool wear monitoring method according to claim 2, characterized in that: The establishment of the correspondence between the edge contour of the blade and the standard edge contour of the blade in step S22 includes the steps: Construct a cost matrix with the number of points on the edge contour of the blade as the number of rows of the matrix and the number of points on the standard edge contour of the blade as the number of columns of the matrix. For each point on the edge contour of the blade and each point on the standard edge contour of the blade, measure the similarity between the distributions represented by their corresponding shape context descriptors through the Bhattacharyya distance, and use the similarity as the matching cost to form each element in the cost matrix. In the cost matrix, find a set of corresponding point pairs from the point sets of the edge contour of the blade and the standard edge contour of the blade at each time point. The set of corresponding point pairs consists of multiple point pairs, so that the sum of the matching costs corresponding to this set of corresponding point pairs in the cost matrix is the smallest. Based on the combination of the sets of point pairs with the smallest sum of matching costs found, establish the correspondence between the edge contour of the blade image at each time point and the standard edge contour of the blade.

6. The shield machine tool wear monitoring method according to claim 1, characterized in that: The specific operation method of step S4 is: Obtain a large amount of image data of the cutter body surface of the shield machine from the data records in the previous construction process. Build a wear trace recognition model based on the convolutional neural network. Input the image of the cutter body surface of the shield machine at each time point into the wear trace recognition model for recognition, and recognize and locate the wear traces in the image of the cutter body surface at each time point. According to the pixel coordinates of the wear traces in the image, extract the wear trace texture on the cutter body surface, count the number of pixels covered by the texture edge contour, and obtain the product of it and the actual area represented by each pixel, thereby obtaining the wear area of the cutter body surface at each time point.

7. A method for monitoring the tool wear of a shield machine according to claim 1, characterized in that: The specific analysis method of step S5 is: Taking the time interval between adjacent time points as a stage, by calculating the ratio of the change in the wear degree of the shield machine cutter in each stage to the time interval in the curve of the wear degree of the shield machine cutter changing with time, the wear rate of each stage of the shield machine is obtained; If the wear rates of all the above stages are lower than the set wear rate reference value, it is determined that the cutter of the shield machine is in the normal wear stage. On the contrary, the growth amplitudes of the wear rates of each stage are compared. If the growth amplitude of the wear rate of a certain stage is positive, it is determined that the cutter wear in this stage is accelerating, and an early warning is given for this stage; A set wear degree threshold is set, and based on the curve of the wear degree of the shield machine cutter changing with time, the time point when the wear degree of the shield machine cutter reaches the set wear degree threshold is predicted, and the wear trend analysis situation is fed back to the system.

8. A cutter wear monitoring system for a shield machine, which is used to execute the steps in the shield machine cutter wear monitoring method according to any one of claims 1-7, characterized in that, This system specifically includes the following modules: The cutter image acquisition module is used to acquire the blade image and the cutter body surface image of the shield machine at each time point through a high-definition camera; The blade contour deviation analysis module is used to extract the blade edge contour of the blade image at each time point through edge detection technology, and match it with the standard blade edge contour through the shape context matching algorithm to obtain the deviation parameters between the blade contours of the shield machine at each time point; The blade wear quantification module is used to construct a linear equation according to the actual wear amount of the blade corresponding to the deviation parameters between the blade contours of the shield machine at each time point, and calculate the wear amount of the blade at each time point; The wear trace positioning module is used to construct a wear trace recognition model based on a convolutional neural network, identify and locate the wear traces in the cutter body surface image at each time point, and then calculate the wear area of the cutter body surface at each time point; The wear trend analysis module is used to evaluate the wear degree of the cutter of the shield machine at each time point based on the wear amount of the blade and the wear area of the cutter body surface at each time point, and draw a curve of the wear degree of the shield machine cutter changing with time in combination with the time series to analyze the wear trend of the shield machine cutter; The management database is used to store the standard blade edge contour, the deviation parameters between the blade contours of the shield machine, the cutter body surface image data of the shield machine, the cutter body surface image of the shield machine at each time point, and the deviation parameter values between the blade contours of the shield machine at each time point.

Images