Hydraulic support welding seam polishing and slag removing method and system

Abstract

The invention relates to the technical field of weld joint polishing autonomous tracking, in particular to a hydraulic support weld joint polishing slag removal method and system, and the method is achieved based on a polishing robot and a control center in communication connection with the polishing robot. Acquiring welding seam surface state data of the first target welding seam, and acquiring an original image corresponding to the first target welding seam based on a preset image collector; the control center inputs the weld surface state data and the original image into a trained defect recognition model to obtain a corresponding defect recognition result; and generating a polishing instruction according to the defect identification result, and sending the polishing instruction to a polishing robot to control the polishing robot to polish the target defect. The device has the beneficial effects that the surface defects of the welding seam of the hydraulic support are automatically identified and accurately polished, the slag removal efficiency and quality of the welding seam are improved, and the labor cost and potential safety hazards are reduced.

Description

Technical Field

[0001] The invention relates to the technical field of weld grinding autonomous tracking, and in particular to a hydraulic support weld grinding and slag removal method and system. Background Art

[0002] Hydraulic supports are important support equipment in underground projects such as coal mines. The quality of their welds is directly related to the stability and safety of the supports. Traditional weld slag removal is mainly done manually, and workers need to use tools such as angle grinders to grind the welds to remove welding slag and uneven parts.

[0003] With the rapid development of robotics technology, more and more industries are beginning to introduce automated equipment for production operations. In the welding industry, robotic welding technology has been widely used and has shown the advantages of high efficiency, high precision and high reliability. However, at present, the slag removal of welds usually adopts a combination of pneumatic slag removal guns and manual labor, using the wind source in the workplace to quickly remove welding residues and spatter. However, this method is still highly dependent on manual participation, which is inefficient and difficult to guarantee quality.

[0004] Therefore, developing a robot system that can automatically complete the slag removal operation of hydraulic support welds has important practical significance and application value. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a hydraulic support weld grinding and slag removal method and system, which solves the technical problems that traditional weld slag removal and grinding operations are highly dependent on manual labor, resulting in low efficiency, unstable quality and great safety hazards.

[0007] (II) Technical solution

[0008] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0009] In a first aspect, an embodiment of the present invention provides a method for grinding and removing slag from a weld of a hydraulic support, the method being implemented based on a grinding robot and a control center that is communicatively connected to the grinding robot, the method comprising:

[0010] S11, the grinding robot obtains weld surface state data of a first target weld based on a preset weld surface detection sensor, and obtains an original image corresponding to the first target weld based on a preset image collector;

[0011] S12, the control center inputs the weld surface state data and the original image into the trained defect recognition model to obtain a defect recognition result corresponding to the first target weld;

[0012] Generating a polishing instruction according to the defect recognition result and sending it to the polishing robot to control the polishing robot to polish the target defect;

[0013] The polishing instruction is used to control the polishing robot to perform polishing according to the corresponding polishing trajectory, and to control the polishing robot to select the corresponding polishing tool for each defect during the polishing process.

[0014] Optionally, before S11, the step further includes:

[0015] S10-1, the grinding robot collects first contour data of the second weld based on a preset line scanning laser sensor; the second weld includes a first target weld and a background; the first contour data includes a height value corresponding to any position in the second weld;

[0016] S10-2, the control center performs median filtering on the first contour data based on a preset filtering window size;

[0017] S10-3, the control center inputs the first contour data after median filtering into the trained weld inflection point extraction model to obtain the inflection point data of the first target weld;

[0018] The weld inflection point extraction model includes a model obtained by training based on known weld profile data and corresponding inflection point data;

[0019] S10-4, the control center obtains the center point position of the first target weld according to the inflection point data of the first target weld;

[0020] S10-5, the control center determines whether there is a deviation between the center point of the first target weld and the center point of the line scan laser sensor according to the center point position of the first target weld and the pre-set center point position of the line scan laser sensor;

[0021] When there is a deviation, an adjustment instruction is sent to the polishing robot;

[0022] The adjustment instruction is used to control the grinding robot to adjust its own position so that the center point of the first target weld is aligned with the center point of the line scanning laser sensor.

[0023] Optionally, the S10-3 includes:

[0024] Obtain at least one area to be compared by traversing the first contour data after median filtering according to a preset sliding window;

[0025] According to the preset formula 1, the matching degree between each area to be compared and the preset weld template is obtained; the formula 1 is:

[0026]

[0027] Among them, NCC(T,Z) is the matching degree between the area to be compared and the preset weld template, t i is the height value corresponding to the position indexed by i in the weld template, is the average height value of all positions in the weld template, z i is the height value corresponding to the position with index i in the area to be compared, is the average height value of all locations in the area to be compared;

[0028] Performing region segmentation on the first contour data according to the matching degree between each to-be-compared region and a preset weld template and a preset matching degree threshold, to obtain second contour data;

[0029] The second contour data is contour data corresponding to the first target weld;

[0030] Fitting corresponding contour curves according to the height values ​​of any two adjacent positions in the second contour data to obtain the slopes corresponding to all contour curves in the first target weld;

[0031] According to the preset slope threshold and the slopes corresponding to all contour curves in the first target weld, the key change point on the first target weld, that is, the inflection point data of the first target weld, is obtained.

[0032] Optionally, the defect recognition model includes a first defect type detection sub-model and a second defect position detection sub-model;

[0033] Then, the S12 includes:

[0034] S12-1, the control center inputs the original image into a preset second defect position detection sub-model to obtain the defect position and defect range corresponding to each defect on the first target weld;

[0035] S12-2, the control center performs data segmentation on the weld surface state data corresponding to the first target weld according to the defect position and defect range corresponding to each defect, and obtains first defect state data corresponding to each defect on the first target weld;

[0036] S12-3, the control center inputs the first defect state data corresponding to each defect into a preset first defect type detection sub-model to obtain the defect type corresponding to each defect on the first target weld;

[0037] S12-4, the control center obtains a grinding tool corresponding to each defect on the first target weld according to the defect type corresponding to each defect on the first target weld and a preset mapping relationship table;

[0038] The mapping relationship table is a corresponding relationship table between defect types and corresponding grinding tools;

[0039] S12-5. The control center generates a grinding instruction based on the defect position and defect range corresponding to each defect on the first target weld and the grinding tool corresponding to each defect on the first target weld, and sends it to the grinding robot to control the grinding robot to grind the target defect.

[0040] Optionally, the S12-1 includes:

[0041] The control center performs grayscale processing on the original image based on a preset formula 2 to obtain a corresponding grayscale image; the formula 2 is:

[0042] Gray t =0.2989R t +0.5870G t +0.1140B t ;

[0043] Among them, R, G, B are the values ​​of the red, green, and blue channels respectively, t is the index of the pixel point, and Gray is the output grayscale value;

[0044] Performing smoothing on the grayscale image;

[0045] Performing planar convolution on the smoothed grayscale image according to a preset Sobel operator to obtain horizontal and vertical brightness difference approximations corresponding to the grayscale image, respectively; the horizontal matrix and vertical matrix of the Sobel operator are both 3×3;

[0046] Based on the horizontal and vertical brightness difference approximations corresponding to the grayscale image, the gradient amplitude corresponding to each pixel position in the smoothed grayscale image is obtained;

[0047] Based on the gradient amplitude corresponding to each pixel position in the smoothed grayscale image, and the pre-set high gradient threshold and low gradient threshold, edge segmentation is performed on the smoothed grayscale image to generate a defect detection image with at least one edge frame;

[0048] Based on the preset Hough line detection algorithm, the edge point coordinates corresponding to each edge frame are obtained;

[0049] Based on the edge point coordinates corresponding to each edge frame, the defect position and defect range corresponding to each defect in the original image are obtained.

[0050] Optionally, the S12-1, based on the edge point coordinates corresponding to each edge frame, obtains the defect position and defect range corresponding to each defect in the original image, including:

[0051] Based on the edge point coordinates corresponding to each edge frame and the preset formula three, the center point coordinates corresponding to each edge frame are fitted, that is, the defect position corresponding to each defect; the formula three is:

[0052] A*f=b;

[0053] Among them, A is the parameter matrix, f is the unknown matrix, and b is the result matrix;

[0054] and

[0055]

[0056] Among them, (x r ,y r ) is the edge point coordinate, r is the index, (x c ,y c ) is the coordinate of the center point corresponding to the edge box, and L is the distance between the two farthest edge points on the edge box;

[0057] Take the center point coordinates corresponding to the edge box as the center of the circle, The circle with a radius of is the defect range corresponding to each defect in the initial image.

[0058] Optionally, the S12-3 includes:

[0059] The control center preprocesses the received first defect status data; the preprocessing includes abnormal value identification, missing value filling, data standardization and data verification;

[0060] Performing feature extraction on the preprocessed first defect state data to obtain feature parameters corresponding to the first defect state data;

[0061] Based on the preset attention mechanism, the attention matrix corresponding to the feature parameters is obtained through the feature parameters and the preset formula 2; the formula 4 is:

[0062]

[0063] Among them, Q is the query matrix, K is the key matrix, V is the value matrix, T is the transpose of the matrix, d k is the dimension of the key vector, O is the attention matrix corresponding to the feature parameters;

[0064] The output attention matrix is ​​input into a preset fully connected layer, and feature transformation is performed based on a preset weight matrix; the feature transformation formula is:

[0065] A = W·O;

[0066] Among them, W is the pre-set weight matrix, and A is the transformation result of the attention matrix;

[0067] Based on the transformation result of the attention matrix and the preset attention score matrix, obtaining the attention score corresponding to the feature parameter;

[0068] The attention score is input into a preset classifier to obtain a classification result corresponding to the first defect state data, that is, the defect type corresponding to the first defect state data.

[0069] Optionally, the second defect position detection sub-model includes a position detection model obtained by training a YOLO model using the first training set;

[0070] The first training set includes at least one weld image annotated with a defect position and a defect range, and each weld image includes at least one defect;

[0071] The first defect type detection sub-model includes a type detection model obtained by training the LSTM model using the second training set;

[0072] The second training set includes at least one set of second defect state data labeled with defect type detection results.

[0073] Optionally, the step S12-1 of smoothing the grayscale image includes:

[0074] The grayscale image is smoothed based on a preset Gaussian filter formula; the Gaussian filter formula is:

[0075]

[0076] Where m and n are the coordinate offsets relative to the center point, and σ is the standard deviation of the Gaussian distribution.

[0077] In a second aspect, an embodiment of the present invention provides a hydraulic support weld grinding and slag removal system, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the above-mentioned hydraulic support weld grinding and slag removal method.

[0078] (III) Beneficial effects

[0079] The beneficial effects of the present invention are as follows: the hydraulic support weld grinding and slag removal method of the present invention, due to the use of high-precision sensors, image collectors and control centers, realizes automatic identification and precise grinding of surface defects of the hydraulic support welds compared to the prior art, which not only improves the efficiency and quality of weld slag removal, but also reduces labor costs and safety hazards. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1A schematic diagram of a flow chart of a hydraulic support weld grinding and slag removal method provided by an embodiment of the present invention;

[0081] Figure 2 A structural block diagram of a hydraulic support weld grinding and slag removal method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0082] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0083] A method for grinding and deslagging the weld of a hydraulic support proposed in an embodiment of the present invention adopts high-precision sensors, image collectors and control centers. Compared with the prior art, it realizes automatic recognition and precise grinding of surface defects of the weld of the hydraulic support, which not only improves the efficiency and quality of weld deslagging, but also reduces labor costs and safety hazards.

[0084] In order to better understand the above technical solution, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0085] Example 1

[0086] This embodiment provides a hydraulic support weld grinding and slag removal method, the method is based on a grinding robot and a control center connected to the grinding robot. Figure 1 As shown, including:

[0087] S11, the grinding robot obtains weld surface state data of a first target weld based on a preset weld surface detection sensor, and obtains an original image corresponding to the first target weld based on a preset image collector;

[0088] S12, the control center inputs the weld surface state data and the original image into the trained defect recognition model to obtain a defect recognition result corresponding to the first target weld;

[0089] Generating a polishing instruction according to the defect recognition result and sending it to the polishing robot to control the polishing robot to polish the target defect;

[0090] The polishing instruction is used to control the polishing robot to perform polishing according to the corresponding polishing trajectory, and to control the polishing robot to select the corresponding polishing tool for each defect during the polishing process.

[0091] The present embodiment provides a method for grinding and removing slag from the weld of a hydraulic support. Through high-precision sensors, image collectors and a control center, the method realizes automatic identification and precise grinding of surface defects of the weld of the hydraulic support, which not only improves the efficiency and quality of weld slag removal, but also reduces labor costs and safety hazards.

[0092] Example 2

[0093] This embodiment provides a hydraulic support weld grinding and slag removal system, including: a grinding robot, a sensor group, an image collector, a grinding tool and a control center;

[0094] Among them, the sensor group includes: line scanning laser sensor, touch sensor, optical sensor and flatness sensor, etc. The sensor group is used to obtain information on the weld surface, including defects such as welding slag and uneven parts, and can be loaded on the grinding robot.

[0095] Polishing robots, usually six-axis or seven-axis industrial robots, have the characteristics of high precision, fast response and flexibility, and can adapt to various complex working environments.

[0096] Grinding tools, according to the defect position and defect range (shape) of the identified defects, select appropriate grinding tools (such as grinding wheels, grinding discs, etc.) for grinding operations. They are usually loaded on a grinding robot and controlled / selected by the grinding robot.

[0097] The image collector is used to collect the original image of the weld. The original image can be used to automatically identify and locate the surface defects of the weld through algorithms such as image filtering, edge detection and pattern matching.

[0098] The control center is used to optimize and adjust the robot's trajectory through control algorithms based on feedback from the sensor group and image collector, so as to achieve precise grinding of the weld.

[0099] The hydraulic support weld grinding and slag removal system provided in this embodiment can achieve 24-hour uninterrupted operation in terms of efficiency, greatly improving the efficiency of weld slag removal. Compared with the traditional manual grinding method, the system can significantly reduce labor costs and improve production efficiency; in terms of precise control, through high-precision sensors, image collectors and control centers, it can achieve millimeter-level precise control to ensure the quality and consistency of weld slag removal, which not only improves the stability and safety of the hydraulic support, but also improves the overall quality of the product; in terms of safety and environmental protection, it can effectively reduce the safety hazards caused by manual operation and reduce the probability of work-related accidents. At the same time, it can also collect and treat dust during the grinding process by equipping dust suction equipment and other means, thereby reducing pollution to the environment and harm to workers' health.

[0100] Example 3

[0101] This embodiment provides a hydraulic support weld grinding and slag removal method, the method is implemented based on a hydraulic support weld grinding and slag removal system described in Example 2, and the method process is as follows: Figure 2 As shown, including:

[0102] S11, the grinding robot obtains weld surface state data of a first target weld based on a preset weld surface detection sensor, and obtains an original image corresponding to the first target weld based on a preset image collector;

[0103] S12, the control center inputs the weld surface state data and the original image into the trained defect recognition model to obtain a defect recognition result corresponding to the first target weld;

[0104] Generating a polishing instruction according to the defect recognition result and sending it to the polishing robot to control the polishing robot to polish the target defect;

[0105] The polishing instruction is used to control the polishing robot to perform polishing according to the corresponding polishing trajectory, and to control the polishing robot to select the corresponding polishing tool for each defect during the polishing process.

[0106] Furthermore, before S11, it also includes:

[0107] S10-1, the grinding robot collects first contour data of the second weld based on a preset line scanning laser sensor; the second weld includes a first target weld and a background; the first contour data includes a height value corresponding to any position in the second weld;

[0108] S10-2, the control center performs median filtering on the first contour data based on a preset filtering window size;

[0109] S10-3, the control center inputs the first contour data after median filtering into the trained weld inflection point extraction model to obtain the inflection point data of the first target weld;

[0110] The inflection point data extraction model includes an extraction model obtained by training based on known weld profile data and its corresponding inflection point data;

[0111] S10-4, the control center obtains the center point position of the first target weld according to the inflection point data of the first target weld;

[0112] S10-5, the control center determines whether there is a deviation between the center point of the first target weld and the center point of the line scan laser sensor according to the center point position of the first target weld and the pre-set center point position of the line scan laser sensor;

[0113] When there is a deviation, an adjustment instruction is sent to the polishing robot;

[0114] The adjustment instruction is used to control the grinding robot to adjust its own position so that the center point of the first target weld coincides with the center point of the line scanning laser sensor.

[0115] Furthermore, the weld inflection point extraction model is used to:

[0116] Obtain at least one area to be compared by traversing the first contour data after median filtering according to a preset sliding window;

[0117] According to the preset formula 1, the matching degree between each area to be compared and the preset weld template is obtained; the formula 1 is:

[0118]

[0119] Among them, NCC(T,Z) is the matching degree between the area to be compared and the preset weld template, t i is the height value corresponding to the position indexed by i in the weld template, is the average height value of all positions in the weld template, z i is the height value corresponding to the position with index i in the area to be compared, is the average height value of all locations in the area to be compared;

[0120] Performing region segmentation on the first contour data according to the matching degree between each to-be-compared region and a preset weld template and a preset matching degree threshold, to obtain second contour data;

[0121] The second contour data is contour data corresponding to the first target weld;

[0122] Fitting a corresponding contour curve according to the height values ​​of any two adjacent positions in the second contour data to obtain the slopes corresponding to all contour curves in the first target weld;

[0123] According to the preset slope threshold and the slopes corresponding to all contour curves in the first target weld, the key change point on the first target weld, that is, the inflection point data of the first target weld, is obtained.

[0124] Further, the defect recognition model includes a first defect type detection sub-model and a second defect position detection sub-model;

[0125] Then the S12 includes:

[0126] S12-1, the control center inputs the original image into a preset second defect position detection sub-model to obtain the defect position and defect range corresponding to each defect on the first target weld;

[0127] S12-2, the control center performs data segmentation on the weld surface state data corresponding to the first target weld according to the defect position and defect range corresponding to each defect, and obtains first defect state data corresponding to each defect on the first target weld;

[0128] Any defect status data includes the corresponding sequence length and sequence dimension;

[0129] S12-3, the control center inputs the first defect state data corresponding to each defect into a preset first defect type detection sub-model to obtain the defect type corresponding to each defect on the first target weld;

[0130] S12-4, the control center obtains a grinding tool corresponding to each defect on the first target weld according to the defect type corresponding to each defect on the first target weld and a pre-set mapping relationship table; the mapping relationship table is a correspondence relationship table between defect types and corresponding grinding tools;

[0131] S12-5. The control center generates a grinding instruction based on the defect position and defect range corresponding to each defect on the first target weld and the grinding tool corresponding to each defect on the first target weld, and sends it to the grinding robot to control the grinding robot to grind the target defect.

[0132] Furthermore, the S12-1 includes:

[0133] The control center performs grayscale processing on the original image based on a preset formula 2 to obtain a corresponding grayscale image; the formula 2 is:

[0134] Gray t =0.2989R t +0.5870G t +0.1140B t ;

[0135] Among them, R, G, B are the values ​​of the red, green, and blue channels respectively, t is the index of the pixel point, and Gray is the output grayscale value;

[0136] The grayscale image is smoothed based on a preset Gaussian filter formula; the Gaussian filter formula is:

[0137]

[0138] Where m and n are the coordinate offsets relative to the center point, and σ is the standard deviation of the Gaussian distribution.

[0139] Performing planar convolution on the smoothed grayscale image according to a preset Sobel operator to obtain horizontal and vertical brightness difference approximations corresponding to the grayscale image, respectively; the horizontal matrix and vertical matrix of the Sobel operator are both 3×3;

[0140] Based on the horizontal and vertical brightness difference approximations corresponding to the grayscale image, the gradient amplitude corresponding to each pixel position in the smoothed grayscale image is obtained;

[0141] Based on the gradient amplitude corresponding to each pixel position in the smoothed grayscale image, and the pre-set high gradient threshold and low gradient threshold, edge segmentation is performed on the smoothed grayscale image to generate a defect detection image with at least one edge frame;

[0142] Based on the preset Hough line detection algorithm, the edge point coordinates corresponding to each edge frame are obtained;

[0143] Based on the edge point coordinates corresponding to each edge frame, the defect position and defect range corresponding to each defect in the original image are obtained.

[0144] Wherein, based on the edge point coordinates corresponding to each edge frame, the defect position and defect range corresponding to each defect in the original image are obtained, including:

[0145] Based on the edge point coordinates corresponding to each edge frame and the preset formula three, the center point coordinates corresponding to each edge frame are fitted, that is, the defect position corresponding to each defect; the formula three is:

[0146] A*f=b;

[0147] Among them, A is the parameter matrix, f is the unknown matrix, and b is the result matrix;

[0148] and

[0149]

[0150]

[0151] Among them, (x r ,y r ) is the edge point coordinate, r is the index, (x c ,y c ) is the coordinate of the center point corresponding to the edge box, and L is the distance between the two farthest edge points on the edge box;

[0152] Take the center point coordinates corresponding to the edge box as the center of the circle, The circle with a radius of is the defect range corresponding to each defect in the initial image.

[0153] Further, the S12-3 includes:

[0154] The control center preprocesses the received first defect status data; the preprocessing includes abnormal value identification, missing value filling, data standardization and data verification;

[0155] Performing feature extraction on the preprocessed first defect state data to obtain feature parameters (feature vectors) corresponding to the first defect state data;

[0156] Based on the preset attention mechanism, the attention matrix corresponding to the feature parameters is obtained through the feature parameters and the preset formula 2; the formula 4 is:

[0157]

[0158] Among them, Q is the query matrix, K is the key matrix, V is the value matrix, T is the transpose of the matrix, d k is the dimension of the key vector, and O is the attention matrix corresponding to the feature parameter; specifically, when confirming the defect type, local feature extraction is crucial because these local features may include key information of the defect type features. In order to enhance the local features in the attention mechanism and enable the model to better capture local patterns, convolution operations are usually used to generate query, key, and value matrices. As shown in the following formula:

[0159] Q=Conv1D(X), K=Conv1D(X), V=Conv1D(X);

[0160] The key is to use local perception to extract features from adjacent time steps, Q c , K c and V c The calculation process is similar to that of Q c Take this as an example to show the calculation process:

[0161]

[0162] Among them, u is the size of the convolution kernel, W convk is the weight of the u-th convolution kernel, x i+u is the i+uth time step of the input sequence.

[0163] Next, the attention matrix is ​​calculated using convolutional projection;

[0164]

[0165] The output attention matrix is ​​input into a preset fully connected layer, and feature transformation is performed based on a preset weight matrix; the feature transformation formula is:

[0166] A = W·O;

[0167] Among them, W is the pre-set weight matrix, and A is the transformation result of the attention matrix;

[0168] Based on the transformation result of the attention matrix and the preset attention score matrix, obtaining the attention score corresponding to the feature parameter;

[0169] The attention score is input into a preset classifier to obtain a classification result corresponding to the first defect state data, that is, the defect type corresponding to the first defect state data.

[0170] The second defect position detection sub-model includes a position detection model obtained by training the YOLO model using the first training set;

[0171] The first training set includes at least one weld image annotated with a defect position and a defect range, and each weld image includes at least one defect;

[0172] The first defect type detection sub-model includes a type detection model obtained by training the LSTM model using the second training set;

[0173] The second training set includes at least one set of second defect state data labeled with defect type detection results.

[0174] The present embodiment provides a method for grinding and removing slag from the weld of a hydraulic support. Through high-precision sensors, image collectors and a control center, the method realizes automatic identification and precise grinding of surface defects of the weld of the hydraulic support, which not only improves the efficiency and quality of weld slag removal, but also reduces labor costs and safety hazards.

[0175] Example 4

[0176] This embodiment provides a hydraulic support weld grinding and slag removal system, including a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to implement a hydraulic support weld grinding and slag removal method described in the above-mentioned embodiment 1 or embodiment 3.

[0177] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0178] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0179] In the present invention, unless otherwise clearly specified and limited, when a first feature is “on” or “below” a second feature, it may be that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Moreover, when a first feature is “above”, “above” or “above” a second feature, it may be that the first feature is directly above or obliquely above the second feature, or it may simply mean that the first feature is higher in level than the second feature. When a first feature is “below”, “below” or “below” a second feature, it may be that the first feature is directly below or obliquely below the second feature, or it may simply mean that the first feature is lower in level than the second feature.

[0180] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0181] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may alter, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for grinding and removing slag from the weld of a hydraulic support, characterized in that: The method is implemented based on a grinding robot and a control center that is communicatively connected to the grinding robot, and the method includes: S11, the grinding robot obtains weld surface state data of a first target weld based on a preset weld surface detection sensor, and obtains an original image corresponding to the first target weld based on a preset image collector; S12, the control center inputs the weld surface state data and the original image into the trained defect recognition model to obtain a defect recognition result corresponding to the first target weld; Generating a polishing instruction according to the defect recognition result and sending it to the polishing robot to control the polishing robot to polish the target defect; The polishing instruction is used to control the polishing robot to perform polishing according to the corresponding polishing trajectory, and to control the polishing robot to select the corresponding polishing tool for each defect during the polishing process.

2. The method for grinding and removing slag from the weld of a hydraulic support according to claim 1, characterized in that: The S11 and before also include: S10-1, the grinding robot collects first contour data of the second weld based on a preset line scanning laser sensor; the second weld includes a first target weld and a background; the first contour data includes a height value corresponding to any position in the second weld; S10-2, the control center performs median filtering on the first contour data based on a preset filtering window size; S10-3, the control center inputs the first contour data after median filtering into the trained weld inflection point extraction model to obtain the inflection point data of the first target weld; The weld inflection point extraction model includes a model obtained by training based on known weld profile data and corresponding inflection point data; S10-4, the control center obtains the center point position of the first target weld according to the inflection point data of the first target weld; S10-5, the control center determines whether there is a deviation between the center point of the first target weld and the center point of the line scan laser sensor according to the center point position of the first target weld and the pre-set center point position of the line scan laser sensor; When there is a deviation, an adjustment instruction is sent to the polishing robot; The adjustment instruction is used to control the grinding robot to adjust its own position so that the center point of the first target weld is aligned with the center point of the line scanning laser sensor.

3. The method for grinding and removing slag from the weld of a hydraulic support according to claim 2, characterized in that: The S10-3 includes: Obtain at least one area to be compared by traversing the first contour data after median filtering according to a preset sliding window; According to the preset formula 1, the matching degree between each area to be compared and the preset weld template is obtained; the formula 1 is: Among them, NCC(T,Z) is the matching degree between the area to be compared and the preset weld template, t i is the height value corresponding to the position indexed by i in the weld template, is the average height value of all positions in the weld template, z i is the height value corresponding to the position with index i in the area to be compared, is the average height value of all locations in the area to be compared; Performing region segmentation on the first contour data according to the matching degree between each to-be-compared region and a preset weld template and a preset matching degree threshold, to obtain second contour data; The second contour data is contour data corresponding to the first target weld; Fitting corresponding contour curves according to the height values ​​of any two adjacent positions in the second contour data to obtain the slopes corresponding to all contour curves in the first target weld; According to the preset slope threshold and the slopes corresponding to all contour curves in the first target weld, the key change point on the first target weld, that is, the inflection point data of the first target weld, is obtained.

4. The method for grinding and removing slag from the weld of a hydraulic support according to claim 1, characterized in that: The defect recognition model includes a first defect type detection sub-model and a second defect position detection sub-model; Then, the S12 includes: S12-1, the control center inputs the original image into a preset second defect position detection sub-model to obtain the defect position and defect range corresponding to each defect on the first target weld; S12-2, the control center performs data segmentation on the weld surface state data corresponding to the first target weld according to the defect position and defect range corresponding to each defect, and obtains first defect state data corresponding to each defect on the first target weld; S12-3, the control center inputs the first defect state data corresponding to each defect into a preset first defect type detection sub-model to obtain the defect type corresponding to each defect on the first target weld; S12-4, the control center obtains a grinding tool corresponding to each defect on the first target weld according to the defect type corresponding to each defect on the first target weld and a preset mapping relationship table; The mapping relationship table is a corresponding relationship table between defect types and corresponding grinding tools; S12-5. The control center generates a grinding instruction based on the defect position and defect range corresponding to each defect on the first target weld and the grinding tool corresponding to each defect on the first target weld, and sends it to the grinding robot to control the grinding robot to grind the target defect.

5. The method for grinding and removing slag from the weld of a hydraulic support according to claim 4, characterized in that: The S12-1 includes: The control center performs grayscale processing on the original image based on a preset formula 2 to obtain a corresponding grayscale image; the formula 2 is: Gray t =0.2989R t +0.5870G t +0.1140B t ; Among them, R, G, B are the values ​​of the red, green, and blue channels respectively, t is the index of the pixel point, and Gray is the output grayscale value; Performing smoothing on the grayscale image; Performing planar convolution on the smoothed grayscale image according to a preset Sobel operator to obtain horizontal and vertical brightness difference approximations corresponding to the grayscale image, respectively; the horizontal matrix and vertical matrix of the Sobel operator are both 3×3; Based on the horizontal and vertical brightness difference approximations corresponding to the grayscale image, the gradient amplitude corresponding to each pixel position in the smoothed grayscale image is obtained; Based on the gradient amplitude corresponding to each pixel position in the smoothed grayscale image, and the pre-set high gradient threshold and low gradient threshold, edge segmentation is performed on the smoothed grayscale image to generate a defect detection image with at least one edge frame; Based on the preset Hough line detection algorithm, the edge point coordinates corresponding to each edge frame are obtained; Based on the edge point coordinates corresponding to each edge frame, the defect position and defect range corresponding to each defect in the original image are obtained.

6. The method for grinding and removing slag from the weld of a hydraulic support according to claim 5, characterized in that: The step S12-1, based on the edge point coordinates corresponding to each edge frame, obtains the defect position and defect range corresponding to each defect in the original image, including: Based on the edge point coordinates corresponding to each edge frame and the preset formula three, the center point coordinates corresponding to each edge frame are fitted, that is, the defect position corresponding to each defect; the formula three is: A*f=b; Among them, A is the parameter matrix, f is the unknown matrix, and b is the result matrix; and Among them, (x r ,y r ) is the edge point coordinate, r is the index, (x c ,y c ) is the coordinate of the center point corresponding to the edge box, and L is the distance between the two farthest edge points on the edge box; Take the center point coordinates corresponding to the edge box as the center of the circle, The circle with a radius of is the defect range corresponding to each defect in the initial image.

7. The method for grinding and removing slag from the weld of a hydraulic support according to claim 4, characterized in that: The S12-3 includes: The control center preprocesses the received first defect status data; the preprocessing includes abnormal value identification, missing value filling, data standardization and data verification; Performing feature extraction on the preprocessed first defect state data to obtain feature parameters corresponding to the first defect state data; Based on the preset attention mechanism, the attention matrix corresponding to the feature parameters is obtained through the feature parameters and the preset formula 2; the formula 4 is: Among them, Q is the query matrix, K is the key matrix, V is the value matrix, T is the transpose of the matrix, d k is the dimension of the key vector, O is the attention matrix corresponding to the feature parameters; The output attention matrix is ​​input into a preset fully connected layer, and feature transformation is performed based on a preset weight matrix; the feature transformation formula is: A = W·O; Among them, W is the pre-set weight matrix, and A is the transformation result of the attention matrix; Based on the transformation result of the attention matrix and the preset attention score matrix, obtaining the attention score corresponding to the feature parameter; The attention score is input into a preset classifier to obtain a classification result corresponding to the first defect state data, that is, the defect type corresponding to the first defect state data.

8. The method for grinding and removing slag from the weld of a hydraulic support according to claim 4, characterized in that: The second defect position detection sub-model includes a position detection model obtained by training the YOLO model using the first training set; The first training set includes at least one weld image annotated with a defect position and a defect range, and each weld image includes at least one defect; The first defect type detection sub-model includes a type detection model obtained by training the LSTM model using the second training set; The second training set includes at least one set of second defect state data labeled with defect type detection results.

9. The method for grinding and removing slag from the weld of a hydraulic support according to claim 5, characterized in that: The step S12-1, performing smoothing processing on the grayscale image, comprises: The grayscale image is smoothed based on a preset Gaussian filter formula; the Gaussian filter formula is: Where m and n are the coordinate offsets relative to the center point, and σ is the standard deviation of the Gaussian distribution.

10. A hydraulic support weld grinding and slag removal system, comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement a hydraulic support weld grinding and slag removal method as described in any one of claims 1 to 9.

Images