Control system of laser cladding processing device

Abstract

The invention innovatively provides a laser cladding processing device control system which comprises the following steps: a target extraction step: collecting to-be-processed workpiece image data acquired by a visual sensor to obtain a total data set, and performing similarity analysis on the total data set in a rough classification strategy to obtain a target detection data set; an image analysis step: training the target detection data set to obtain a plurality of to-be-processed workpiece target detection models, obtaining a dimension reduction plane through point projection according to the to-be-processed workpiece target detection models, and constructing the dimension reduction plane into a plurality of grid matrixes; a defect extraction step: according to the grid matrix, a change grid is obtained through change ratio analysis in a fine analysis strategy, and the change grid reflects a change point of the workpiece to be machined; and a defect judgment step: obtaining the change grid and judging whether the workpiece to be machined has defects or not through a state analysis strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of defect detection, and more specifically to a control system for a laser cladding processing device. Background Art

[0002] In the laser cladding process, cameras, visual sensors and other equipment are currently used for inspection, and the human eye is used to identify whether there are defects. Human eye recognition takes a lot of time, and the screening accuracy is relatively low, and it is easy to miss the screening; the recognition is not used for screening, that is, the AI ​​recognition of the workpiece to be processed is used to detect the defects of the workpiece to be processed through pictures or videos. Due to the small number of AI training model samples, the recognition rate and accuracy of AI recognition are relatively low. . Summary of the invention

[0003] In view of the deficiencies in the prior art, an object of the present invention is to provide a control system for a laser cladding processing device, so as to overcome the above-mentioned defects in the prior art.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A laser cladding processing device control system comprises the following steps:

[0006] In a target extraction step, the image data of the workpiece to be processed acquired by the visual sensor are collected to obtain a data set, and the data set is subjected to similarity analysis in a rough classification strategy to obtain a target detection data set;

[0007] In the image analysis step, the target detection data set is trained to obtain a number of target detection models for the workpiece to be processed, and a dimension reduction plane is obtained by point projection according to the target detection model for the workpiece to be processed, and the dimension reduction plane is divided into a number of grid matrices;

[0008] A defect extraction step, obtaining a changed grid by analyzing the change ratio in the fine analysis strategy according to the grid matrix, wherein the changed grid reflects the change points of the workpiece to be processed;

[0009] The defect judgment step is to obtain the changed grid and judge whether the workpiece to be processed is defective through a state analysis strategy.

[0010] Preferably, the target monitoring data set is obtained according to the depth of field recognition strategy and the depth of field of each pixel is calculated, the depth of field value is obtained by comparing the depth of field historical parameters of the pixel point, and the depth of field value and the depth of field threshold are compared with a preset depth of field threshold. If the depth of field value is greater than the depth of field threshold, the pixel point is marked as the target detection model of the workpiece to be processed.

[0011] Preferably, the detailed analysis strategy includes a mesh division step, a mesh node change step and a mesh comparison step.

[0012] A grid division step, establishing a training network according to the grid matrix, and evenly dividing the network into a number of grid nodes, and obtaining coordinate information of the grid nodes;

[0013] A grid node change step, judging whether it is a changed node according to the coordinate information of the grid node, and if so, performing a grid comparison step;

[0014] In the grid comparison step, all changed nodes are obtained, and if the number of calculated changed nodes is greater than the node threshold, it is determined to be a changed grid.

[0015] Preferably, the method further comprises a defective model component step, wherein the defective model component step is used to collect the change node information in the detailed analysis strategy and the coordinate information corresponding to the change node, and use the change node to construct a defect three-dimensional model and output it.

[0016] Preferably, the state analysis strategy includes a change cluster determination step and a defect judgment step;

[0017] a change cluster determination step, obtaining change nodes in the change grid and defining external graphics of the change nodes as change clusters;

[0018] The defect judgment step calculates the proportion of the change cluster in the dimensionality reduction plane as the change value, and presets a defect threshold, judges the change value and the defect threshold, and if the change value is greater than the defect threshold, it is judged as defective.

[0019] Preferably, the state analysis strategy includes a cluster enveloping step, and the cluster enveloping step is located between the change cluster determining step and the defect judging step;

[0020] The cluster enveloping step includes obtaining all the change clusters in the current dimensionality reduction plane and marking the boundary points of all the change clusters. If the boundary points are on the same contour line, the defect judgment step is performed.

[0021] Preferably, a cluster envelope sub-step is also provided in the cluster envelope step, and the cluster envelope sub-step includes: if the boundary points are not located on the same contour line, they are defined as a first contour line and a second contour line, the areas of the first contour line and the second contour line are compared, the contour line with the larger area is defined as a variation cluster, and a defect judgment step is performed.

[0022] Preferably, a cluster enveloping sub-step is further provided in the cluster enveloping step, and the cluster enveloping sub-step includes, if the boundary points are not located on the same contour line, obtaining the closest boundary points between two adjacent groups of contour lines, and automatically connecting the two adjacent groups of contour lines as a new contour line through the boundary points, defining the contour line as a change cluster, and performing a defect judgment step.

[0023] Preferably, the target monitoring data set obtains pixel points through structured light technology or deep learning.

[0024] The beneficial effects of the present invention are as follows: a visual sensor is used to fix a shooting angle to shoot the workpiece to be processed, so as to realize a before-and-after comparison of the shooting pictures and realize a rapid analysis of the situation of the workpiece to be processed; at the same time, a target extraction step is adopted to extract the picture shot by the visual sensor, and the current shot image is compared with the historical image through the similarity analysis in the coarse classification strategy, and the lace with differences is extracted and used as a target detection data set; an image analysis step is used to analyze the images in the target detection data set in sequence, and the images are constructed into a plurality of grid matrices for facilitating analysis thereof; a defect extraction step is used to analyze the grid matrices respectively, and if the number of grid node variables in the grid matrix is ​​greater than a preset number, it is identified as a changed grid; a defect judgment step is used to obtain the number of all changed grids, and the proportion of the number of changed grids is calculated to determine whether the workpiece to be processed has defects; through multi-step analysis, the accuracy of defect judgment is improved, and small defects can also be quickly identified and obtained, thereby improving the accuracy of defect judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a defect detection step diagram of the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] It should be noted that when a component is referred to as being "fixed to" another component, it may be directly on the other component or there may also be a component centered. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may also be a component centered. When a component is considered to be "set on" another component, it may be directly set on the other component or there may also be a component centered. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0029] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings:

[0030] like Figure 1 As shown, the present invention provides a control system for a laser cladding processing device, comprising the following steps:

[0031] In the target extraction step, the image data of the workpiece to be processed acquired by the visual sensor are collected to obtain a data set, and the data set is subjected to a similarity analysis in a rough classification strategy to obtain a target detection data set; the workpiece to be processed is photographed by a visual sensor, and the route and shooting angle of the visual sensor are fixed, so as to facilitate the comparison between the pictures taken by the visual sensor and the historical pictures; the image data of the workpiece to be processed photographed by the visual sensor are collected to obtain a data set of the situation of the workpiece to be processed, and the data set is subjected to a similarity analysis in a rough classification strategy to compare the similarity between the data set and the historical data. If the similarity is lower than a preset value, the area is used as target detection data, and target detection data is obtained through phase velocity analysis in a rough classification strategy, and the target detection data is collected to obtain a target detection data set; the fixed route and shooting angle of the visual sensor are used to facilitate the photographing of the workpiece to be processed, and at the same time, it is convenient to realize the comparison of the historical data of the workpiece to be processed. At the same time, the target extraction step is used to facilitate the preliminary screening of abnormal conditions of the workpiece to be processed, avoid manual screening, improve the screening accuracy, and reduce the manpower loss of screening;

[0032] In the image analysis step, the target detection data set is trained to obtain several target detection models of the workpiece to be processed, and the dimension reduction plane is obtained by point projection according to the target detection model of the workpiece to be processed, and the dimension reduction plane is divided into several grid matrices; the target monitoring data set is obtained according to the depth of field recognition strategy and the depth of field of each pixel point is calculated, and the depth of field value is obtained by comparing the depth of field historical parameters of the pixel point, and the depth of field value and the depth of field threshold are compared with the preset depth of field threshold. If the depth of field value is greater than the depth of field threshold, the pixel point is marked as the target detection model of the workpiece to be processed; the target monitoring data set obtains pixel points through structured light technology or deep learning; the target training set is obtained through deep learning or structured light technology. And calculate the depth of field of each pixel point, compare the historical parameters of the pixel point to obtain the depth of field value, the depth of field value is the difference between the existing pixel point and the historical parameters, compare the depth of field value and the depth of field threshold, the depth of field threshold is set to 5mm, if the difference between the pixel before and after change is greater than 5mm, then the pixel point is marked, all the marked pixel points are obtained, and all the pixel points are collected into the target detection model of the workpiece to be processed, and the dimension reduction plane is obtained by point projection according to the target detection model of the workpiece to be processed, and the dimension reduction plane is composed of several 5*5mm grid matrices, and 25 grid nodes are evenly arranged on the grid matrix. The grid matrix is ​​used to realize the division of the workpiece to be processed and improve the accuracy of the calculation

[0033] In the defect extraction step, a change grid is obtained by analyzing the change ratio in the fine analysis strategy according to the grid matrix, and the change grid reflects the change point of the workpiece to be processed; according to the grid nodes on the grid matrix, a fine analysis strategy is used to determine whether it is a change grid node. If the number of change grid nodes is greater than 10, the grid is determined to be a change grid; the fine analysis strategy includes a grid division step, a grid node change step, and a grid comparison step. In the grid division step, a training network is established according to the grid matrix, and a number of grid nodes are evenly divided, and the coordinate information of the grid nodes is obtained. The grid matrix is ​​evenly divided, and the grid matrix is ​​divided into 25 grid nodes, and the coordinate information of the corresponding grid nodes is respectively obtained. Mark information, that is, the depth information of the grid node; a grid node change step, obtaining the coordinate information of the grid node, and comparing the historical coordinate information of the grid node according to the coordinate information of the grid node, and judging whether the grid node is a changed node through the coordinate information of the grid node and the historical coordinate information, if so, performing a grid comparison step, if not, ending the step; a grid comparison step, obtaining all the changed nodes in the grid matrix, and summing up the total number of all the changed nodes in the grid matrix, and calculating the number of changed nodes is greater than the node threshold, then it is judged to be a changed grid, the node threshold is preset to 10, and when the number of changed nodes is greater than or equal to 10, the grid is judged to be a changed grid.

[0034] The defect judgment step obtains the changed mesh and determines whether the workpiece to be processed is defective through the state analysis strategy. All the changed meshes in the current target detection model are obtained, and the state analysis strategy is used to determine whether the model is defective, thereby realizing automatic judgment of defects, improving the accuracy of defect judgment, and achieving the recognition rate of small defects.

[0035] The state analysis strategy includes a change cluster determination step and a defect judgment step; in the change cluster determination step, the change nodes in the change grid are obtained, and the external graphics of the change nodes are delineated as the change cluster; in the defect judgment step, the external graphics of the change cluster are obtained, and the area of ​​the change cluster is obtained according to the external graphics. At the same time, according to the number of change grids, the area of ​​the dimensionality reduction plane is calculated, and the proportion of the change cluster in the dimensionality reduction plane is calculated as the change value. The change value is the proportion of the change cluster in the dimensionality reduction plane, and a defect threshold is preset. The defect threshold is preset to 20%. The change value and the defect threshold are compared to judge the change value and the defect threshold. If the change value is greater than the defect threshold, it is judged to be defective. When the change value is greater than 20%, it is judged to be defective at that location, and the defect shape is. The state analysis strategy includes a cluster envelope step, which is located between the change cluster determination step and the defect judgment step; a cluster envelope step is set between the change cluster determination step and the defect judgment step, and the change clusters are screened and identified using the cluster envelope step to reduce the existence of errors; the cluster envelope step includes obtaining all change clusters in the current dimensionality reduction plane, and marking the boundary points of all change clusters, and determining whether the boundary points are on the same contour line. If they are on the same contour line, a defect judgment step is performed; by obtaining the boundary points of the change clusters and determining whether all the boundary points are on the same contour line, the number of change clusters is determined. If the number of change clusters is one, a defect judgment step is performed on the change cluster corresponding to the contour line.

[0036] A cluster envelope sub-step is also provided in the cluster envelope step, and the cluster envelope sub-step includes: if the boundary points are not located on the same contour line, they are defined as the first contour line and the second contour line, and the areas of the first contour line and the second contour line are compared, and the contour line with the larger area is defined as the change cluster, and a defect judgment step is performed; when the boundary points are not located on the same contour line, there are at least two groups of contour lines, and the contour lines are defined as the first contour line, the second contour line and the third contour line respectively, and the areas of the first contour line, the second contour line and the third contour line are calculated respectively, and the contour line with the largest area is defined as the change cluster, and defect judgment is performed. By screening different contour lines, the detection accuracy is improved. When the area corresponding to one group of contour lines in multiple groups of contour lines is much larger than that of other contour lines or the distance between two adjacent groups of contour lines is too large, the cluster envelope sub-step is used to screen the contour lines to reduce the influence of other factors on defect judgment.

[0037] In another embodiment, a cluster enveloping sub-step is further provided in the cluster enveloping step, and the cluster enveloping sub-step includes: if the boundary points are not located on the same contour line, obtaining the closest boundary points between two adjacent groups of contour lines, and automatically connecting the two adjacent groups of contour lines as a new contour line through the boundary points, defining the contour line as a change cluster, and performing a defect judgment step; when the boundary points are not located on the same contour line, determining the boundary points on two adjacent groups of contour lines, and determining the closest boundary points, automatically connecting the closest boundary points of the two adjacent groups of contour lines to form a new contour line, and merging them in turn according to the number of contour lines to a group of contour lines, defining the contour line as a change cluster, and performing defect judgment, when the area difference of multiple groups of contour lines is not obvious or the spacing between two adjacent groups of contour lines is small, the cluster including sub-step is used to merge the contour lines to avoid missing selection and affecting the accuracy of defect judgment.

[0038] It also includes a defective model component step, which is used to collect the change node information in the detailed analysis strategy and the coordinate information corresponding to the change node, and use the change node to generate a three-dimensional model of the component defects, and output it; through the defective model component step, all change nodes are collected, and the coordinate information of all change nodes is obtained, and based on the coordinate information of all change nodes, a three-dimensional model of the component defects is generated and output, and the three-dimensional model is used to observe and analyze the causes of the defects, and provide a preliminary maintenance plan for the defects.

[0039] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A laser cladding processing device control system, characterized in that: The steps include: In a target extraction step, the image data of the workpiece to be processed acquired by the visual sensor are collected to obtain a data set, and the data set is subjected to similarity analysis in a rough classification strategy to obtain a target detection data set; In the image analysis step, the target detection data set is trained to obtain a number of target detection models for the workpiece to be processed, and a dimension reduction plane is obtained by point projection according to the target detection model for the workpiece to be processed, and the dimension reduction plane is divided into a number of grid matrices; A defect extraction step, obtaining a changed grid by analyzing the change ratio in the fine analysis strategy according to the grid matrix, wherein the changed grid reflects the change points of the workpiece to be processed; The defect judgment step is to obtain the changed grid and judge whether the workpiece to be processed is defective through a state analysis strategy.

2. A laser cladding processing device control system according to claim 1, characterized in that: According to the depth of field recognition strategy, the target monitoring data set is obtained and the depth of field of each pixel is calculated. The depth of field historical parameters of the pixel point are compared to obtain the depth of field value, and the depth of field value and the depth of field threshold are compared with the preset depth of field threshold. If the depth of field value is greater than the depth of field threshold, the pixel point is marked as the target detection model of the workpiece to be processed.

3. A laser cladding processing device control system according to claim 1, characterized in that: The detailed analysis strategy includes a mesh division step, a mesh node change step, and a mesh comparison step. A grid division step, establishing a training network according to the grid matrix, and evenly dividing the network into a number of grid nodes, and obtaining coordinate information of the grid nodes; A grid node change step, judging whether it is a changed node according to the coordinate information of the grid node, and if so, performing a grid comparison step; In the grid comparison step, all changed nodes are obtained, and if the number of calculated changed nodes is greater than the node threshold, it is determined to be a changed grid.

4. A laser cladding processing device control system according to claim 3, characterized in that: The method also includes a defective model component step, wherein the defective model component step is used to collect the change node information and the coordinate information corresponding to the change node in the detailed analysis strategy, and use the change node to construct a defect three-dimensional model and output it.

5. A laser cladding processing device control system according to claim 3, characterized in that: The state analysis strategy includes a change cluster determination step and a defect judgment step; a change cluster determination step, obtaining change nodes in the change grid and defining external graphics of the change nodes as change clusters; The defect judgment step calculates the proportion of the change cluster in the dimensionality reduction plane as the change value, and presets a defect threshold, judges the change value and the defect threshold, and if the change value is greater than the defect threshold, it is judged as defective.

6. A laser cladding processing device control system according to claim 5, characterized in that: The state analysis strategy includes a cluster enveloping step, and the cluster enveloping step is located between the change cluster determining step and the defect judging step; The cluster enveloping step includes obtaining all the change clusters in the current dimensionality reduction plane and marking the boundary points of all the change clusters. If the boundary points are on the same contour line, the defect judgment step is performed.

7. A laser cladding processing device control system according to claim 6, characterized in that: The cluster enveloping step also includes a cluster enveloping sub-step, which includes defining the boundary points as a first contour line and a second contour line if they are not located on the same contour line, comparing the areas of the first contour line and the second contour line, defining the contour line with the larger area as a variation cluster, and performing a defect judgment step.

8. A laser cladding processing device control system according to claim 6, characterized in that: The cluster enveloping step is also provided with a cluster enveloping sub-step, and the cluster enveloping sub-step includes, if the boundary points are not located on the same contour line, obtaining the closest boundary points between two adjacent groups of contour lines, and automatically connecting the two adjacent groups of contour lines as a new contour line through the boundary points, defining the contour line as a change cluster, and performing a defect judgment step.

9. A laser cladding processing device control system according to claim 2, characterized in that: The target monitoring data set obtains pixel points through structured light technology or deep learning.

Images