Laser cladding method

Abstract

The present invention discloses a laser cladding method, which belongs to the technical field of laser cladding process optimization. The present invention realizes intelligent control and automatic optimization of process parameters by collecting cladding images and performing HSV color space analysis, combined with a multi-channel process prediction model, avoiding the defects of relying on manual experience and repeated experiments in traditional methods. The method simultaneously considers the two dimensions of molten pool quality and cladding layer quality, and realizes accurate optimization of process parameters by introducing a quality gap coefficient. The present invention quantifies the quality of the cladding process through image processing, thereby improving the control accuracy of process parameters.

Description

Technical Field

[0001] The invention relates to the technical field of laser cladding process optimization, and in particular to a laser cladding method. Background Art

[0002] As an advanced surface treatment and additive manufacturing process, laser cladding technology is widely used in aerospace, automobile manufacturing, mold manufacturing and other fields. This technology uses a high-energy laser beam to melt metal powder and form a metallurgical bond with the base material, which can significantly improve the wear resistance, corrosion resistance and service life of the part surface.

[0003] However, the laser cladding process involves multiple process parameters such as laser power, scanning speed, powder feed rate, etc. There is a complex coupling relationship between these parameters, and they have a direct impact on the cladding quality. At present, the selection of process parameters mainly depends on the operator's experience and repeated tests. This method is not only time-consuming and labor-intensive, but also difficult to ensure the control accuracy of process parameters. It is easy to cause defects such as cracks, pores, and poor bonding in the cladding layer, resulting in material waste. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, a laser cladding method provided by the present invention solves the problems of the prior art being time-consuming and labor-intensive and having difficulty in ensuring the control accuracy of process parameters.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is: a laser cladding method, comprising the following steps:

[0006] S1, collecting cladding images during the laser cladding process;

[0007] S2, projecting the cladding image into the HSV color space, segmenting the cladding image into the molten pool area and the cladding layer area based on the V channel value, and obtaining the molten pool quality coefficient and the cladding layer quality coefficient;

[0008] S3, subtracting the molten pool quality target coefficient from the molten pool quality coefficient to obtain the molten pool quality gap coefficient;

[0009] S4, subtracting the cladding layer quality target coefficient from the cladding layer quality coefficient to obtain the cladding layer quality gap coefficient;

[0010] S5. Input the molten pool quality coefficient, the cladding layer quality coefficient, the molten pool quality gap coefficient and the cladding layer quality gap coefficient into each channel of the multi-channel process prediction model respectively to obtain process parameters, wherein the process parameters include: laser power, scanning speed and powder supply rate.

[0011] Further, S2 includes the following sub-steps:

[0012] S21, projecting the cladding image into the HSV color space, extracting the V channel value, and forming a V channel image;

[0013] S22, calculating the brightness change value of each pixel according to the V channel image;

[0014] S23, when the brightness change value is greater than the brightness change threshold, the pixel point is a segmentation point;

[0015] S24, dividing the cladding image into a molten pool area and a cladding layer area according to each segmentation point;

[0016] S25. Calculate the molten pool quality coefficient according to the shape characteristics of the molten pool area;

[0017] S26. Calculate the quality coefficient of the cladding layer according to the shape characteristics of the cladding layer area.

[0018] Furthermore, the formula for calculating the brightness change value of each pixel in S22 is: , where V c is the brightness change value of the pixel at (x, y), x is the horizontal coordinate of the pixel, y is the vertical coordinate of the pixel, V (x,y+1) is the V channel value of the pixel at (x,y+1), V (x,y-1) is the V channel value of the pixel at (x, y-1), V (x+1,y) is the V channel value of the pixel at (x+1,y), V (x-1,y) is the V channel value of the pixel at (x-1, y), and || is the absolute value operation.

[0019] Further, S24 includes the following sub-steps:

[0020] S241, eliminating isolated segmentation points, retaining continuous segmentation points, and obtaining segmentation boundaries;

[0021] S242, segmenting the cladding image according to the segmentation boundary to obtain various connected regions;

[0022] S243, extracting the V channel mean and pixel mean of the connected area to form a feature sequence;

[0023] S244, calculating the similarity between the feature sequence of each connected region and the stored molten pool feature sequence, and taking the connected region with the greatest similarity as the molten pool region;

[0024] S245, calculating the similarity between the feature sequence of each connected region and the stored cladding layer feature sequence, and taking the connected region with the greatest similarity as the cladding layer region.

[0025] Further, S25 includes the following sub-steps:

[0026] S251, dividing the molten pool area into multiple local areas, and calculating the surface smoothness value;

[0027] S252, calculating the edge smoothing value of the molten pool area according to the curvature value of each edge pixel point on the molten pool area;

[0028] S253. Calculate the molten pool quality coefficient according to the surface smoothness value and the edge smoothness value.

[0029] Furthermore, the formula for calculating the surface smoothness value in S251 is: , where S f is the surface smoothness value, P i,j is the pixel value of the jth pixel in the i-th local area on the molten pool area, is the pixel mean of the i-th local area on the molten pool area, N is the number of pixels in the local area on the molten pool area, M is the number of local areas on the molten pool area, and i and j are positive integers;

[0030] The formula for calculating the edge smoothing value in S252 is: , where S b is the edge smoothing value, C i,j is the curvature value of the jth pixel point in the i-th local edge of the molten pool area, is the mean curvature of the ith local edge, L is the number of pixels in the ith local edge, and K is the number of local edges segmented in the molten pool area;

[0031] The formula for calculating the molten pool quality coefficient in S253 is: , where M p is the molten pool quality coefficient, e is a natural constant, S f,th is the threshold of the surface smoothness value, S b,th is the threshold of edge smoothing value.

[0032] Further, S26 includes the following sub-steps:

[0033] S261, calculating the texture feature value for each local area on the cladding layer area;

[0034] S262, calculating texture uniformity according to the texture feature values ​​of each local area;

[0035] S263. Calculate the quality coefficient of the cladding layer based on the texture uniformity.

[0036] Furthermore, the formula for calculating the texture feature value in S261 is: , where θ i is the texture characteristic value of the i-th local area on the cladding layer area, e is a natural constant, X i,jis the pixel value of the jth pixel point in the i-th local area on the cladding layer area, is the pixel mean of the i-th local area on the cladding layer area, Z is the number of pixel points in the i-th local area on the cladding layer area, i and j are positive integers, and | | is the absolute value operation;

[0037] The formula for calculating texture uniformity in S262 is: , where μ is the texture uniformity and R is the number of local areas on the cladding layer area;

[0038] The formula for calculating the quality coefficient of the cladding layer in S263 is: , where M F is the quality coefficient of the cladding layer, μ th is the threshold of texture uniformity.

[0039] Further, the multi-channel process prediction model in S5 includes: a first feature enhancement input channel, a second feature enhancement input channel, a third feature enhancement input channel, a fourth feature enhancement input channel, a first hidden layer, a second hidden layer, a first fully connected layer, a second fully connected layer, and a third fully connected layer;

[0040] The input end of the first hidden layer is respectively connected to the output end of the first feature enhancement input channel, the output end of the second feature enhancement input channel, the output end of the third feature enhancement input channel and the output end of the fourth feature enhancement input channel, and its output end is connected to the input end of the second hidden layer; the output end of the second hidden layer is respectively connected to the input end of the first fully connected layer, the input end of the second fully connected layer and the input end of the third fully connected layer; the first fully connected layer is used to output laser power; the second fully connected layer is used to output scanning speed; the third fully connected layer is used to output powder supply rate.

[0041] Further, the first feature enhancement input channel is used to input the molten pool quality coefficient, the second feature enhancement input channel is used to input the cladding layer quality coefficient, the third feature enhancement input channel is used to input the molten pool quality gap coefficient, and the fourth feature enhancement input channel is used to input the cladding layer quality gap coefficient;

[0042] In the first feature enhancement input channel and the second feature enhancement input channel, the expressions of the feature enhancement input channels are both: , in the third feature enhancement input channel and the fourth feature enhancement input channel, the expressions of the feature enhancement input channels are: ,in, is the output of the feature enhancement input channel, x is the input of the feature enhancement input channel, tanh is the hyperbolic tangent activation function, ω is the weight in the feature enhancement input channel, b is the bias in the feature enhancement input channel, and e is a natural constant.

[0043] The beneficial effects of the present invention are as follows: by collecting cladding images and performing HSV color space analysis, combined with a multi-channel process prediction model, the present invention realizes intelligent control and automatic optimization of process parameters, avoiding the defects of relying on manual experience and repeated experiments in traditional methods. The method considers the two dimensions of molten pool quality and cladding layer quality at the same time, and realizes accurate optimization of process parameters by introducing the quality gap coefficient. In addition, due to the accurate control and optimization of process parameters, the number of tests and material waste are reduced, the product qualification rate is improved, and the production cost is effectively reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flow chart of a laser cladding method;

[0045] Figure 2 Schematic diagram of the structure of the multi-channel process prediction model. DETAILED DESCRIPTION

[0046] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0047] like Figure 1 As shown, a laser cladding method comprises the following steps:

[0048] S1, collecting cladding images during the laser cladding process;

[0049] S2, projecting the cladding image into the HSV color space, segmenting the cladding image into the molten pool area and the cladding layer area based on the V channel value, and obtaining the molten pool quality coefficient and the cladding layer quality coefficient;

[0050] S3, subtracting the molten pool quality target coefficient from the molten pool quality coefficient to obtain the molten pool quality gap coefficient;

[0051] S4, subtracting the cladding layer quality target coefficient from the cladding layer quality coefficient to obtain the cladding layer quality gap coefficient;

[0052] S5. Input the molten pool quality coefficient, the cladding layer quality coefficient, the molten pool quality gap coefficient and the cladding layer quality gap coefficient into each channel of the multi-channel process prediction model respectively to obtain process parameters, wherein the process parameters include: laser power, scanning speed and powder supply rate.

[0053] The present invention projects the cladding image into the HSV color space and extracts the V channel value. The V channel value is the brightness and can significantly reflect the difference between the molten pool and the cladding layer. Therefore, it is convenient to segment the molten pool area and the cladding layer area in the cladding image to obtain the molten pool quality coefficient and the cladding layer quality coefficient.

[0054] Subtract the molten pool quality target coefficient from the molten pool quality coefficient, and subtract the cladding layer quality target coefficient from the cladding layer quality coefficient. The quality gap coefficient obtained reflects the gap with the target, which makes it easier for the model to better capture the scale that needs to be adjusted. At the same time, it has more input data to improve the accuracy of process optimization.

[0055] When the molten pool quality coefficient is greater than or equal to the molten pool quality target coefficient and the cladding layer quality coefficient is greater than or equal to the cladding layer quality target coefficient, the corresponding process parameters are the required optimal parameters.

[0056] In this embodiment, S2 includes the following sub-steps:

[0057] S21, projecting the cladding image into the HSV color space, extracting the V channel value, and forming a V channel image;

[0058] S22, calculating the brightness change value of each pixel according to the V channel image;

[0059] S23, when the brightness change value is greater than the brightness change threshold, the pixel point is a segmentation point;

[0060] S24, dividing the cladding image into a molten pool area and a cladding layer area according to each segmentation point;

[0061] S25. Calculate the molten pool quality coefficient according to the shape characteristics of the molten pool area;

[0062] S26. Calculate the quality coefficient of the cladding layer according to the shape characteristics of the cladding layer area.

[0063] In this embodiment, the formula for calculating the brightness change value of each pixel in S22 is: , where V c is the brightness change value of the pixel at (x, y), x is the horizontal coordinate of the pixel, y is the vertical coordinate of the pixel, V (x,y+1) is the V channel value of the pixel at (x,y+1), V (x,y-1) is the V channel value of the pixel at (x, y-1), V (x+1,y) is the V channel value of the pixel at (x+1,y), V (x-1,y) is the V channel value of the pixel at (x-1, y), and || is the absolute value operation.

[0064] The present invention projects the cladding image into the HSV color space and extracts the V channel value to better reflect the brightness information of the image, facilitate the distinction between the molten pool and the cladding layer, and reflects the edge and contour in the image according to the brightness change value of the pixel point. When the brightness change value is greater than the brightness change threshold, the pixel point is determined as the segmentation point, thereby segmenting the molten pool area and the cladding layer area. According to the shape characteristics of the molten pool area, the molten pool quality coefficient is calculated to reflect the quality of the formed molten pool. According to the shape characteristics of the cladding layer area, the cladding layer quality coefficient is calculated to reflect the quality of the formed cladding layer.

[0065] In this embodiment, the brightness change threshold is specifically set according to experiments or requirements.

[0066] In this embodiment, S24 includes the following sub-steps:

[0067] S241, eliminating isolated segmentation points, retaining continuous segmentation points, and obtaining segmentation boundaries;

[0068] S242, segmenting the cladding image according to the segmentation boundary to obtain various connected regions;

[0069] S243, extracting the V channel mean and pixel mean of the connected area to form a feature sequence;

[0070] S244, calculating the similarity between the feature sequence of each connected region and the stored molten pool feature sequence, and taking the connected region with the greatest similarity as the molten pool region;

[0071] S245, calculating the similarity between the feature sequence of each connected region and the stored cladding layer feature sequence, and taking the connected region with the greatest similarity as the cladding layer region.

[0072] For a split point, If there are no other segmentation points in the neighborhood, the segmentation point is an isolated segmentation point. If there are other segmentation points in the neighborhood of a segmentation point, the segmentation point is a continuous segmentation point. According to the segmentation boundary, the cladding image is segmented to obtain each connected area, and the V channel mean (obtained from the V channel image) and pixel mean (obtained from the cladding image) of each connected area are extracted. The molten pool area and cladding layer area are determined from the brightness value and pixel value.

[0073] The stored molten pool feature sequence is a sequence consisting of the stored V channel mean and pixel mean of the molten pool, and the stored cladding layer feature sequence is a sequence consisting of the stored V channel mean and pixel mean of the cladding layer.

[0074] In this embodiment, S25 includes the following sub-steps:

[0075] S251, dividing the molten pool area into multiple local areas, and calculating the surface smoothness value;

[0076] S252, calculating the edge smoothing value of the molten pool area according to the curvature value of each edge pixel point on the molten pool area;

[0077] S253. Calculate the molten pool quality coefficient according to the surface smoothness value and the edge smoothness value.

[0078] In this embodiment, the formula for calculating the surface smoothness value in S251 is: , where S f is the surface smoothness value, P i,j is the pixel value of the jth pixel in the i-th local area on the molten pool area, is the pixel mean of the i-th local area on the molten pool area, N is the number of pixels in the local area on the molten pool area, M is the number of local areas on the molten pool area, and i and j are positive integers.

[0079] The present invention divides the molten pool area into multiple local areas, which is equivalent to shearing the molten pool area into multiple local areas. In each local area, the difference between the pixel value and the pixel mean is calculated to obtain the smoothness of a local area. By combining various local areas, a surface smoothness value is obtained, which can characterize the uniformity and stability inside the molten pool area. Corresponding to the temperature distribution and melting state of the molten pool, a lower surface smoothness value means that the temperature distribution inside the molten pool is uniform, indicating that the process parameters (such as laser power, scanning speed) are well matched. A higher surface smoothness value indicates that there are overheated or underheated areas inside the molten pool, which may lead to the formation of defects such as pores and inclusions.

[0080] In this embodiment, the formula for calculating the edge smoothing value in S252 is: , where S b is the edge smoothing value, C i,j is the curvature value of the jth pixel point in the i-th local edge of the molten pool area, is the mean curvature of the i-th local edge, L is the number of pixels in the i-th local edge, and K is the number of local edges segmented in the molten pool area.

[0081] In the molten pool area, internal pixels are removed and edge pixels are retained to obtain edge contours, which are then divided into fixed lengths to obtain local edges.

[0082] The edge smoothness value characterizes the regularity and stability of the molten pool boundary, which corresponds to the bonding state between the molten pool and the substrate. A regular and smooth molten pool edge means good metallurgical bonding, while an irregular edge will lead to poor bonding.

[0083] In this embodiment, the formula for calculating the molten pool quality coefficient in S253 is: , where M p is the molten pool quality coefficient, e is a natural constant, Sf,th is the threshold of the surface smoothness value, S b,th is the threshold of edge smoothing value.

[0084] The smaller the surface smoothing value and the smaller the edge smoothing value, the larger the molten pool quality coefficient. When the surface smoothing value and the edge smoothing value are close to 0, the molten pool quality coefficient is the highest. Considering the surface smoothing value and the edge smoothing value at the same time, a comprehensive evaluation of the molten pool quality is achieved.

[0085] In this embodiment, the threshold of the surface smoothness value and the threshold of the edge smoothness value are used to adjust the range of the molten pool quality coefficient and are set according to experiments and requirements.

[0086] In this embodiment, S26 includes the following sub-steps:

[0087] S261, calculating the texture feature value for each local area on the cladding layer area;

[0088] S262, calculating texture uniformity according to the texture feature values ​​of each local area;

[0089] S263. Calculate the quality coefficient of the cladding layer based on the texture uniformity.

[0090] In this embodiment, the formula for calculating the texture feature value in S261 is: , where θ i is the texture characteristic value of the i-th local area on the cladding layer area, e is a natural constant, X i,j is the pixel value of the jth pixel point in the i-th local area on the cladding layer area, is the pixel mean of the i-th local area on the cladding layer area, Z is the number of pixels in the i-th local area on the cladding layer area, i and j are positive integers, and | | is the absolute value operation.

[0091] The present invention can capture the texture characteristics of the surface of the cladding layer more finely by dividing the cladding layer area into multiple local areas (equivalent to cutting into multiple local areas), which is helpful to find local defects or unevenness. The cladding layer itself has a fine texture, so an exponential function is used to amplify this texture to obtain a texture characteristic value of a local area.

[0092] In this embodiment, the formula for calculating texture uniformity in S262 is: , where μ is the texture uniformity and R is the number of local areas on the cladding layer area.

[0093] Subtract the texture eigenvalue of each local area from the mean of the texture eigenvalue to obtain the deviation of each local area relative to the mean. When the texture uniformity is smaller, the texture on the cladding layer area is more uniform. When the texture uniformity is larger, the texture uniformity on the cladding layer area is worse, the more defective locations there are, and the worse the overall quality is.

[0094] In this embodiment, the formula for calculating the quality coefficient of the cladding layer in S263 is: , where M F is the quality coefficient of the cladding layer, μ th is the threshold of texture uniformity.

[0095] In this embodiment, the threshold of texture uniformity is used to adjust the range of the quality coefficient of the cladding layer and is set according to experiments or requirements.

[0096] The smaller the texture uniformity, the larger the quality coefficient of the cladding layer. When the quality coefficient of the cladding layer is close to 0, the quality coefficient of the molten pool is the highest.

[0097] In this embodiment, if Figure 2 As shown, the multi-channel process prediction model in S5 includes: a first feature enhancement input channel, a second feature enhancement input channel, a third feature enhancement input channel, a fourth feature enhancement input channel, a first hidden layer, a second hidden layer, a first fully connected layer, a second fully connected layer and a third fully connected layer;

[0098] The input end of the first hidden layer is respectively connected to the output end of the first feature enhancement input channel, the output end of the second feature enhancement input channel, the output end of the third feature enhancement input channel and the output end of the fourth feature enhancement input channel, and its output end is connected to the input end of the second hidden layer; the output end of the second hidden layer is respectively connected to the input end of the first fully connected layer, the input end of the second fully connected layer and the input end of the third fully connected layer; the first fully connected layer is used to output laser power; the second fully connected layer is used to output scanning speed; the third fully connected layer is used to output powder supply rate.

[0099] In this embodiment, the first feature enhancement input channel is used to input the molten pool quality coefficient, the second feature enhancement input channel is used to input the cladding layer quality coefficient, the third feature enhancement input channel is used to input the molten pool quality gap coefficient, and the fourth feature enhancement input channel is used to input the cladding layer quality gap coefficient.

[0100] In this embodiment, the number of nodes in the first hidden layer is 8, and the number of nodes in the second hidden layer is 4.

[0101] In this embodiment, the expression of the fully connected layer is: , where H is the output of the fully connected layer, h k is the output of the kth node on the second hidden layer, ωk h k The weight of b k h k , k is a positive integer, and f is the activation function.

[0102] The present invention inputs the molten pool quality coefficient, the cladding layer quality coefficient, the molten pool quality gap coefficient and the cladding layer quality gap coefficient through each channel, and performs feature enhancement in each channel, so that the model has a clearer perception of quality, and then through two hidden layers, the model can learn more complex nonlinear relationships, improve the prediction ability of process parameters, and output corresponding types of process parameters through three fully connected layers.

[0103] In the first feature enhancement input channel and the second feature enhancement input channel, the expressions of the feature enhancement input channels are both: , in the third feature enhancement input channel and the fourth feature enhancement input channel, the expressions of the feature enhancement input channels are: ,in, is the output of the feature enhancement input channel, x is the input of the feature enhancement input channel, tanh is the hyperbolic tangent activation function, ω is the weight in the feature enhancement input channel, b is the bias in the feature enhancement input channel, and e is a natural constant.

[0104] The weights and biases in the multi-channel process prediction model can be trained by gradient descent method.

[0105] When the molten pool quality coefficient and the cladding layer quality coefficient are larger, the attention received in the feature enhancement input channel is smaller. When the molten pool quality coefficient and the cladding layer quality coefficient are smaller, the attention received in the feature enhancement input channel is greater. When the molten pool quality gap coefficient and the cladding layer quality gap coefficient are larger, the molten pool quality and the cladding layer quality are worse. Therefore, the greater the attention received in the feature enhancement input channel, the adaptive adjustment of the size of the eigenvalue of the input hidden layer is achieved. According to the size of the input quality coefficient and the gap coefficient, the model can automatically adjust the attention to different features and realize intelligent feature extraction. By automatically adjusting the attention to different features, the adaptive adjustment of the size of the eigenvalue input to the hidden layer is realized, and different process parameters are set for molten pools and cladding layers with different quality conditions.

[0106] The present invention realizes intelligent control and automatic optimization of process parameters by collecting cladding images and performing HSV color space analysis, combined with a multi-channel process prediction model, avoiding the defects of relying on manual experience and repeated experiments in traditional methods. The method considers the two dimensions of molten pool quality and cladding layer quality at the same time, and realizes accurate optimization of process parameters by introducing the quality gap coefficient. The present invention quantifies the quality of the cladding process through image processing, thereby improving the control accuracy of process parameters.

[0107] In addition, due to the precise control and optimization of process parameters, the number of tests and material waste are reduced, while the product qualification rate is improved, effectively reducing production costs.

[0108] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A laser cladding method, characterized in that: The following steps are involved: S1, collecting cladding images during the laser cladding process; S2, projecting the cladding image into the HSV color space, segmenting the cladding image into the molten pool area and the cladding layer area based on the V channel value, and obtaining the molten pool quality coefficient and the cladding layer quality coefficient; S3, subtracting the molten pool quality target coefficient from the molten pool quality coefficient to obtain the molten pool quality gap coefficient; S4, subtracting the cladding layer quality target coefficient from the cladding layer quality coefficient to obtain the cladding layer quality gap coefficient; S5, inputting the molten pool quality coefficient, the cladding layer quality coefficient, the molten pool quality gap coefficient and the cladding layer quality gap coefficient into each channel of the multi-channel process prediction model respectively to obtain process parameters, wherein the process parameters include: laser power, scanning speed and powder supply rate; The S2 comprises the following sub-steps: S21, projecting the cladding image into the HSV color space, extracting the V channel value, and forming a V channel image; S22, calculating the brightness change value of each pixel according to the V channel image; S23, when the brightness change value is greater than the brightness change threshold, the pixel point is a segmentation point; S24, dividing the cladding image into a molten pool area and a cladding layer area according to each segmentation point; S25. Calculate the molten pool quality coefficient according to the shape characteristics of the molten pool area; S26, calculating the quality coefficient of the cladding layer according to the shape characteristics of the cladding layer area; The S25 comprises the following sub-steps: S251, dividing the molten pool area into multiple local areas, and calculating the surface smoothness value; S252, calculating the edge smoothing value of the molten pool area according to the curvature value of each edge pixel point on the molten pool area; S253, calculating the molten pool quality coefficient according to the surface smoothness value and the edge smoothness value; The S26 comprises the following sub-steps: S261, calculating the texture feature value for each local area on the cladding layer area; S262, calculating texture uniformity according to the texture feature values ​​of each local area; S263. Calculate the quality coefficient of the cladding layer according to the texture uniformity; The multi-channel process prediction model in S5 includes: a first feature enhancement input channel, a second feature enhancement input channel, a third feature enhancement input channel, a fourth feature enhancement input channel, a first hidden layer, a second hidden layer, a first fully connected layer, a second fully connected layer and a third fully connected layer; The input end of the first hidden layer is respectively connected to the output end of the first feature enhancement input channel, the output end of the second feature enhancement input channel, the output end of the third feature enhancement input channel and the output end of the fourth feature enhancement input channel, and its output end is connected to the input end of the second hidden layer; the output end of the second hidden layer is respectively connected to the input end of the first fully connected layer, the input end of the second fully connected layer and the input end of the third fully connected layer; the first fully connected layer is used to output laser power; the second fully connected layer is used to output scanning speed; the third fully connected layer is used to output powder supply rate; The first feature enhancement input channel is used to input the molten pool quality coefficient, the second feature enhancement input channel is used to input the cladding layer quality coefficient, the third feature enhancement input channel is used to input the molten pool quality gap coefficient, and the fourth feature enhancement input channel is used to input the cladding layer quality gap coefficient; In the first feature enhancement input channel and the second feature enhancement input channel, the expressions of the feature enhancement input channels are both: , in the third feature enhancement input channel and the fourth feature enhancement input channel, the expressions of the feature enhancement input channels are: ,in, is the output of the feature enhancement input channel, x is the input of the feature enhancement input channel, tanh is the hyperbolic tangent activation function, ω is the weight in the feature enhancement input channel, b is the bias in the feature enhancement input channel, and e is a natural constant.

2. The laser cladding method according to claim 1, characterized in that: The formula for calculating the brightness change value of each pixel in S22 is: , where V c is the brightness change value of the pixel at (x, y), x is the horizontal coordinate of the pixel, y is the vertical coordinate of the pixel, V (x,y+1) is the V channel value of the pixel at (x,y+1), V (x,y-1) is the V channel value of the pixel at (x, y-1), V (x+1,y) is the V channel value of the pixel at (x+1,y), V (x-1,y) is the V channel value of the pixel at (x-1, y), and || is the absolute value operation.

3. The laser cladding method according to claim 1, characterized in that: The S24 comprises the following sub-steps: S241, eliminating isolated segmentation points, retaining continuous segmentation points, and obtaining segmentation boundaries; S242, segmenting the cladding image according to the segmentation boundary to obtain various connected regions; S243, extracting the V channel mean and pixel mean of the connected area to form a feature sequence; S244, calculating the similarity between the feature sequence of each connected region and the stored molten pool feature sequence, and taking the connected region with the greatest similarity as the molten pool region; S245, calculating the similarity between the feature sequence of each connected region and the stored cladding layer feature sequence, and taking the connected region with the greatest similarity as the cladding layer region.

4. The laser cladding method according to claim 1, characterized in that: The formula for calculating the surface smoothness value in S251 is: , where S f is the surface smoothness value, P i,j is the pixel value of the jth pixel in the i-th local area on the molten pool area, is the pixel mean of the i-th local area on the molten pool area, N is the number of pixels in the local area on the molten pool area, M is the number of local areas on the molten pool area, and i and j are positive integers; The formula for calculating the edge smoothing value in S252 is: , where S b is the edge smoothing value, C i,j is the curvature value of the jth pixel point in the i-th local edge of the molten pool area, is the mean curvature of the ith local edge, L is the number of pixels in the ith local edge, and K is the number of local edges segmented in the molten pool area; The formula for calculating the molten pool quality coefficient in S253 is: , where M p is the molten pool quality coefficient, e is a natural constant, S f,th is the threshold of the surface smoothness value, S b,th is the threshold of edge smoothing value.

5. The laser cladding method according to claim 1, characterized in that: The formula for calculating the texture feature value in S261 is: , where θ i is the texture characteristic value of the i-th local area on the cladding layer area, e is a natural constant, X i,j is the pixel value of the jth pixel point in the i-th local area on the cladding layer area, is the pixel mean of the i-th local area on the cladding layer area, Z is the number of pixels in the i-th local area on the cladding layer area, i and j are positive integers, and | | is the absolute value operation; The formula for calculating texture uniformity in S262 is: , where μ is the texture uniformity and R is the number of local areas on the cladding layer area; The formula for calculating the quality coefficient of the cladding layer in S263 is: , where M F is the quality coefficient of the cladding layer, μ th is the threshold of texture uniformity.

Images