Inclined basal plane laser cladding cross section size prediction method and system

Abstract

The invention relates to the technical field of laser cladding, in particular to a method and a system for predicting the size of a laser cladding cross section of an inclined base plane. Aiming at the problems of fracture, holes, noise and the like of the cladding layer boundary in a deep learning model prediction result, the invention provides an end point extension repair algorithm based on graph structure analysis, and boundary defects are effectively repaired through the steps of morphological operation, connected domain analysis, path reverse extension and the like. In order to improve the cladding layer section prediction precision, a Pix2Pix model is applied to cladding layer section form prediction for the first time, and perception loss is introduced to enhance the perception ability of the model to boundary details and a global structure. Besides, structural similarity indexes are introduced, molten pool images with high similarity are extracted from molten pool videos under different parameter conditions to serve as an extended training set, and the generalization ability and robustness of the model are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of laser cladding, and particularly relates to a laser cladding cross-section size prediction method and system for an inclined base surface. BACKGROUND

[0002] As an important additive manufacturing method in the field of advanced manufacturing, laser cladding technology has been widely used in the fields of aerospace, automobile manufacturing, mold repair and mechanical manufacturing in recent years, because it can realize high-precision and high-quality metal surface repair and functional coating preparation. The technology melts and deposits metal powder or wire on the surface of the substrate through a high-energy laser beam to form a cladding layer that is metallurgically combined with the substrate, thereby significantly improving the wear resistance, corrosion resistance and service life of the workpiece.

[0003] In the laser cladding process, the cladding layer cross-section size (such as width, height, area, etc.) is an important parameter for measuring the cladding quality, and is directly related to the mechanical properties of the cladding layer and the subsequent processing quality. The traditional cladding size prediction method mainly relies on empirical formula or numerical simulation technology based on physical mechanism, mainly using numerical simulation methods such as finite element analysis (FEA) and computational fluid dynamics (CFD) to model the heat conduction, molten pool flow, solidification shrinkage and other physical processes in the laser cladding process. These methods can accurately predict the cladding layer cross-section size in theory, but have the disadvantages of long calculation time, high calculation complexity, strong parameter dependence, high cost of parameter acquisition, poor adaptability of the model to changes in working conditions, etc.

[0004] With the rapid development of artificial intelligence, especially deep learning, traditional convolutional neural networks (CNNs), transformers, and generative adversarial networks (GANs) have been introduced into the field of cladding section prediction, achieving high prediction accuracy. For example, some studies have adopted a hybrid CNN-Transformer architecture, using ResNet for spatial feature extraction and Transformer for time series modeling. Through a self-attention mechanism, the spatiotemporal features of the molten pool surface thermal image sequence are effectively captured, thereby accurately predicting the two-dimensional depth contour of the molten pool. Other studies have combined an improved PredNet for long-term molten pool image prediction (140ms in advance) and used a ResNet-34 (SERes network) with an added SE channel attention module for regression analysis, achieving accurate prediction of the weld reinforcement height using molten pool image sequences and welding speed inputs. One study employed a GRU-GAN model, which combines a gated recurrent unit (GRU) with a conditional generative adversarial network (CGAN). The GRU model was used to model time-series dependencies, and the CGAN was used to generate high-fidelity images of the future molten pool. Combined with welding speed sequences, this enabled dynamic prediction and adjustment of the molten pool morphology, assisting human operators in making prediction-based model-predictive control (MPC) decisions.

[0005] Although the aforementioned deep learning models can leverage data-driven advantages to achieve a certain level of accuracy in predicting the cross-sectional dimensions of cladding sections, they are prone to defects such as boundary breaks, incomplete closure, and noisy regions when directly outputting predicted cross-sectional images. This is because most deep learning models tend to lose feature information at the turning and abrupt change regions of the cross-sectional boundary when extracting local features, resulting in breakage during boundary generation. While some deep learning models can model global correlations, the distinction between background and boundary features in the molten pool image is low. The model is prone to misjudging irrelevant background information as boundary features or missing some weak boundary signals, leading to incomplete boundary closure. This results in significant differences between the predicted results and the actual cross-section in terms of dimensional accuracy and morphological consistency, making it difficult to meet the accuracy and stability requirements of industrial applications for predicting the cross-sectional dimensions of inclined laser cladding surfaces. Summary of the Invention

[0006] Therefore, the technical problem to be solved by the present invention is to overcome the shortcomings of existing deep learning model-based methods for predicting the cross-sectional size of cladding sections, which are prone to boundary breaks, incomplete closures and noise areas when directly outputting the predicted image, resulting in poor accuracy in predicting the cross-sectional size of cladding sections.

[0007] To address the aforementioned technical problems, this invention provides a method for predicting the cross-sectional dimensions of laser cladding on an inclined datum, comprising: The initial cladding layer boundary map is obtained by using the cladding cross-section size prediction model to predict the image of the molten pool to be predicted. The areas of all connected regions in the initial cladding layer boundary map are sorted, and the connected region with the largest area is taken as the main boundary region, while the remaining regions are taken as non-main regions. Only the pixels of the main boundary region are retained to obtain the processed cladding layer boundary map. For each endpoint pixel in the processed cladding layer boundary map, starting from the endpoint pixel, extend in the opposite direction to the adjacent pixel belonging to the main boundary region until the extension path of the endpoint pixel coincides with the pixel of the main boundary region or reaches the boundary of the processed cladding layer boundary map. If an endpoint pixel's extension path reaches the boundary of the processed cladding layer boundary map, then starting from the boundary of the cladding layer boundary map, delete pixel by pixel in the opposite direction of the endpoint pixel's extension direction until the intersection of the endpoint pixel's extension path and the main boundary region is reached, thus obtaining the target cladding layer boundary map. Based on the target cladding layer boundary map, obtain the size parameters of the molten pool image to be predicted.

[0008] Preferably, it further includes: The initial cladding layer boundary map of the target is obtained by performing morphological opening and closing operations on the initial cladding layer boundary map.

[0009] Preferably, it further includes: Calculate the distance between each non-main region and the geometric center of the initial main boundary region. Merge non-main regions with a distance less than a set threshold with the main boundary region to form the target main boundary region.

[0010] Preferably, the method for determining the intersection point includes: Determine if there are 3 adjacent pixels belonging to the main boundary region in the 8-neighborhood of the current pixel. If so, determine if the extension paths formed by the current pixel and each of its adjacent pixels do not overlap. If they do not overlap, the current pixel is an intersection point.

[0011] Preferably, the prediction model for the cladding cross-section size is the Pix2Pix model.

[0012] Preferably, the training process of the Pix2Pix model includes: Obtain the training set of molten pool images, input the molten pool images into the generator of the Pix2Pix model, and output the initial cladding layer boundary map of the molten pool images; The initial cladding layer boundary map of the molten pool image is processed through a pre-trained feature extraction network to extract predicted features from multiple levels of the feature extraction network. The labeled image of the molten pool image is passed through a pre-trained feature extraction network to extract real features from multiple levels of the feature extraction network; The first confidence level is obtained by passing the molten pool image and its initial cladding layer boundary map through the discriminator of the Pix2Pix model; The second confidence level is obtained by passing the molten pool image and its label image through the discriminator of the Pix2Pix model; The Pix2Pix model is trained based on the conditional adversarial loss between the first and second confidence levels, the pixel-level L1 loss between the initial cladding layer boundary map of the molten pool image and the label image, and the perceptual loss between the predicted features and the real features of multiple layers of the feature extraction network to obtain the target Pix2Pix model.

[0013] Preferably, the formula for the perceptual loss between the predicted features and the true features at multiple levels of the feature extraction network is: , in, In order to perceive loss, This represents the number of layers in the feature extraction network. It is a molten pool image. It is a label image. It is random noise. This represents the initial cladding layer boundary diagram. Represents a generator. For the feature extraction network Predictive features of the layer For the feature extraction network The true characteristics of the layer For the feature extraction network The height of the layer's features For the feature extraction network The width of the layer's features, For the feature extraction network The number of channels in the layer's features. For layer index, It is an L2 norm.

[0014] Preferably, the process of obtaining the molten pool image training set includes: Acquire molten pool videos under different parameter conditions, select any frame of molten pool image from the molten pool videos under different parameter conditions, and the cross-sectional image of the cladding layer under that parameter condition as label images, and use them as samples in the molten pool image training set. Data augmentation was performed on the molten pool images selected under different parameter conditions to obtain data-augmented molten pool images under different parameter conditions; In the molten pool video under different parameter conditions, the molten pool images other than the selected molten pool images are used as candidate images under that parameter condition; Calculate the structural similarity index between each candidate image and the data-enhanced cladding pool image under different parameter conditions, sort them in descending order, and use the candidate images of the previous preset number of frames and the cross-sectional images of the cladding layer under the same parameter conditions as label images, which are then used as augmented samples in the cladding pool image training set.

[0015] Preferably, the size parameters of the molten pool image to be predicted include: the cross-sectional area, width, and height of the cladding layer.

[0016] The present invention also provides a system for predicting the cross-sectional dimensions of laser cladding on an inclined datum, comprising: A memory for storing computer programs; a processor for executing the computer programs to implement the steps of the above-described method for predicting the cross-sectional dimensions of laser cladding on an inclined datum.

[0017] Compared with the prior art, the above-described technical solution of the present invention has the following advantages: The present invention discloses a method and system for predicting the cross-sectional dimensions of laser cladding on an inclined datum. This invention determines the largest connected region in the initial cladding layer boundary map as the main boundary region and retains its pixels. This directly filters out small non-main regions caused by noise interference during model prediction, such as isolated noise pixels or fragmented pseudo-boundaries, reducing the interference of irrelevant regions on subsequent dimension calculations from the source. For endpoint pixels in the processed cladding layer boundary map, starting from that endpoint pixel, the extension path extends in the opposite direction to the adjacent pixels belonging to the main boundary region until the extension path of the endpoint pixel coincides with or does not coincide with the pixels of the main boundary region. Reaching the boundary of the processed cladding layer boundary map can effectively repair the boundary breakage problem caused by insufficient feature learning during model prediction, allowing the broken boundaries to reconnect and form a continuous structure. Finally, for the endpoints of the extension path that reach the image boundary, pixel-by-pixel deletion is performed in reverse from the boundary until the intersection of the extension path of the pixel reaching the endpoint and the main boundary region. This cleans up invalid extension pixels at the edge of the boundary map, further ensuring the closure integrity and morphological accuracy of the main boundary. Finally, the size parameters are calculated based on the regular, continuous, and noise-free target cladding layer boundary map, effectively improving the accuracy and reliability of predicting the cross-sectional size of laser cladding on inclined datum surfaces.

[0018] Traditional deep learning models such as convolutional neural networks and Transformers cannot accurately capture the implicit mapping relationship between typical cross-domain images with no direct correlation, such as molten pool images and cladding layer cross-sectional images. To solve this problem, this invention introduces the Pix2Pix model as the prediction model for cladding cross-sectional dimensions. This model inherently has the ability to handle the conversion between input and output images of different domains. Through adversarial training between the generator and discriminator, it can gradually learn the potential mapping rules between cross-domain images. To further improve the conversion accuracy, this invention additionally introduces a perceptual loss function and uses a pre-trained feature extraction network to capture the deep semantic features of the images. This allows the Pix2Pix model to not only focus on pixel-level surface differences but also understand the deep feature association between the two types of images under the process logic. This effectively enhances the model's understanding and generation accuracy of cross-domain feature mapping between molten pool images and cladding layer cross-sectional images, thereby improving the prediction accuracy of the cross-sectional dimensions of laser cladding on inclined base surfaces.

[0019] Furthermore, to address the issues of high acquisition costs, long acquisition cycles, and cumbersome cross-sectional annotation processes in existing inclined cladding data, which limit the size of high-quality training datasets, this invention selects any frame of the cladding pool image from a video under different parameter conditions and the corresponding real cross-sectional image of the cladding layer as initial samples. Combined with data augmentation, it generates data-augmented cladding pool images under those parameter conditions. Then, it sets the remaining cladding pool images from the same parameter conditions (excluding the initial frame) as candidate images. The structural similarity index between each candidate image and the data-augmented cladding pool image under those parameter conditions is calculated. Images with a preset number of frames and corresponding cross-sectional images are selected as supplementary samples. Data augmentation expands the feature coverage of a single initial frame, while the structural similarity index ensures consistency between candidate images and initial samples in the core features of the cladding pool. This eliminates the need for additional cross-sectional image acquisition or annotation for supplementary samples, significantly reducing data acquisition and annotation costs, enriching the sample quantity and diversity of the cladding pool image training set, and thus effectively improving the prediction accuracy of the cladding cross-sectional size prediction model. Attached Figure Description

[0020] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a schematic flowchart of a method for predicting the cross-sectional dimensions of laser cladding on an inclined base surface according to the present invention.

[0021] Figure 2 It is a flowchart for endpoint extension repair based on a graph structure. Figure 2 (a) in the diagram shows the process of obtaining the processed cladding layer boundary map. Figure 2(b) in the diagram describes the process of obtaining the boundary map of the target cladding layer.

[0022] Figure 3 This is a structural diagram of the Pix2Pix model used in the cladding cross-section size prediction model of this invention.

[0023] Figure 4 This is a schematic diagram illustrating the process of acquiring the training set of molten pool images.

[0024] Figure 5 These are experimental comparison figures showing the prediction models for different cladding cross-section sizes and the method of this invention. Figure 5 (a) is a schematic diagram comparing the prediction models for different cladding cross-sectional dimensions with the target cladding layer boundary map extracted from the first molten pool image by the method of this invention. Figure 5 (b) is a schematic diagram comparing the prediction models for different cladding cross-sectional dimensions with the target cladding layer boundary map extracted from the second molten pool image by the method of this invention. Figure 5 The diagram in (c) shows a comparison between the prediction models for different cladding cross-section sizes and the target cladding layer boundary map extracted from the third molten pool image by the method of this invention. Figure 5 The diagram shows a comparison between the prediction models for different cladding cross-sectional dimensions in (d) and the target cladding layer boundary map extracted from the fourth molten pool image by the method of this invention.

[0025] Figure 6 This diagram illustrates a comparison between prediction models for different cladding cross-section sizes and prediction results for cladding layer size using the method of this invention. Figure 6 (a) is a schematic diagram comparing the prediction models for different cladding cross-sectional dimensions with the method of this invention for predicting the cross-sectional area of ​​the cladding layer. Figure 6 (b) is a schematic diagram comparing the predicted width of different cladding cross-section size prediction models with that predicted by the method of this invention. Figure 6 The diagram shows a comparison between the prediction models for different cladding cross-section dimensions and the height predicted by the method of this invention, with behavior (c) in the figure. Detailed Implementation

[0026] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0027] Existing deep learning-based methods for predicting cladding cross-section dimensions often suffer from boundary breaks, incomplete closure, and noisy regions when directly outputting predicted images. To address this issue, those skilled in the art typically optimize the model itself. This can be achieved by increasing the number of training samples, adjusting the network structure (e.g., increasing the number of network layers or introducing an attention mechanism), or optimizing the loss function (e.g., increasing the loss weight for boundary regions). These methods aim to help the model learn boundary features better during training to reduce prediction defects. Alternatively, simple image post-processing techniques can be employed, such as using fixed threshold filtering to remove obviously noisy pixels or using morphological single operations (e.g., closing operations) to fill in small gaps.

[0028] However, these conventional approaches have obvious flaws. Increasing the number of samples or adjusting the model structure not only increases the cost of data collection and computation, but also makes it difficult to fundamentally solve the problem of insufficient model fitting of boundary details in small sample scenarios, and it is still easy to have breaks in complex contour areas. Although optimizing the loss function can enhance the model's attention to the boundary, it cannot avoid the interference of noise features during training, and fragmented pseudo-boundaries will still remain in the prediction results. Simple post-processing methods lack specificity. Fixed threshold filtering may accidentally delete effective boundary pixels, and single morphological operations cannot simultaneously take into account noise removal and boundary repair, let alone solve the problems of boundary closure and invalid edge extension.

[0029] Therefore, existing methods struggle to completely resolve the aforementioned shortcomings, resulting in significant discrepancies between the predicted results and the actual cross-section in terms of dimensional accuracy and morphological consistency, failing to meet the accuracy and stability requirements of industrial applications. This invention addresses this technological gap by proposing a graph-based endpoint extension and repair method that employs progressive post-processing techniques, including main boundary region filtering, endpoint reverse extension, intersection point determination, and edge cleanup. This method precisely overcomes the difficulties of existing methods in noise differentiation, boundary repair integrity, and invalid edge handling, effectively solving the core problem of cladding section boundary prediction. The specific solution is as follows: Reference Figure 1 As shown, this embodiment provides a method for predicting the cross-sectional dimensions of laser cladding on an inclined datum, including: like Figure 2 As shown, Figure 2 It is a flowchart for endpoint extension repair based on a graph structure. Figure 2 (a) in the diagram shows the process of obtaining the processed cladding layer boundary map. Figure 2 (b) in the diagram describes the process of obtaining the boundary map of the target cladding layer.

[0030] Step S1: Obtain the initial cladding layer boundary map by using the cladding cross-sectional size prediction model to predict the image of the molten pool to be predicted; In this embodiment, preferably, the initial cladding layer boundary map is obtained by performing morphological opening and closing operations on the initial cladding layer boundary map. The opening operation (erosion followed by dilation) can accurately remove tiny noise pixels and fragmented protruding structures attached to the boundary region in the initial boundary map. This type of noise often originates from the misjudgment of local interference features during deep learning model prediction (such as background noise in training data and pseudo-boundary protrusions caused by model fitting deviation). The opening operation can eliminate these irrelevant interferences without destroying the overall shape of the main boundary, thus avoiding the inclusion of noise into the effective boundary during subsequent main boundary region screening. Closing operations (dilation followed by erosion) effectively fill tiny holes inside the main boundary in the initial boundary map, while also connecting adjacent boundary pixel gaps caused by insufficient model feature learning. These tiny holes and gaps are significant contributing factors to increased difficulty in subsequent boundary extension and repair, and deviations in size calculations. Closing operations, by regularizing the local boundary structure, enable the main boundary to initially form a more continuous and complete contour. The combined initial cladding layer boundary map has a smoother and more continuous boundary shape compared to the initial cladding layer boundary map, reducing noise interference in subsequent steps and laying a more regular and clear boundary foundation for accurate selection of the main boundary region and endpoint extension repair. This provides a guarantee for improving the accuracy of the final cladding cross-section size prediction from the preprocessing stage.

[0031] Step S2: Sort the areas of all connected regions in the initial cladding layer boundary map, take the connected region with the largest area as the main boundary region, and the remaining regions as non-main regions. Only the pixels of the main boundary region are retained to obtain the processed cladding layer boundary map. Methods for determining connected components include: If two pixels in the initial cladding layer boundary map are adjacent in the horizontal, vertical, or diagonal direction of their 8-neighborhood, they are considered to be connected to each other. A connected region is the largest set of all interconnected pixels, that is, any two pixels in the region can be connected through a series of adjacent pixels, and no boundary pixel outside the region can be connected to the pixels in the region through the 8-neighborhood relationship.

[0032] The area of ​​a connected region refers to the total number of pixels contained within that region, such as... Figure 2 As shown in (a) in the figure, Figure 2 In the initial cladding boundary map (a), there are three connected regions. The regions are sorted by their areas, and only the pixels of the main boundary region are retained to obtain the processed cladding boundary map.

[0033] In this embodiment, preferably, the distance between each non-main region and the geometric center of the initial main boundary region is calculated, and non-main regions with a distance less than a set threshold are merged with the main boundary region to form the target main boundary region.

[0034] By calculating the distance between each non-primary region and the primary boundary region, non-primary regions with a distance less than a set threshold are merged into the primary boundary region as the target primary boundary region. This effectively avoids the loss of effective boundaries caused by simply filtering by area. In the prediction of cladding section boundaries, some non-primary regions (such as small branches near the primary boundary, local protrusions, or short broken boundaries) may have small areas, but they are actually organic components of the cladding layer outline. If they are deleted as noise simply because of their small area, the boundary structure will be incomplete, affecting the accuracy of subsequent dimension calculations. However, by determining and merging nearby non-primary regions through distance thresholds, these effective local structures closely related to the primary boundary can be accurately preserved. This not only eliminates isolated noise regions far from the primary boundary but also ensures the integrity and continuity of the primary boundary region, making the processed boundary map more closely match the actual outline characteristics of the cladding layer.

[0035] In this embodiment, the distance used to determine the correlation between the non-main region and the main boundary region can be calculated using various methods, such as Euclidean distance, Manhattan distance, and Chebyshev distance.

[0036] In this embodiment, the minimum Euclidean distance between each non-main region and the main region is calculated. If the distance is less than a preset threshold, the small region is retained and merged into the main boundary region.

[0037] Choosing the minimum Euclidean distance (i.e., the straight-line distance between the two closest points between the non-main region and the main boundary region) as the criterion has significant advantages: Firstly, the minimum Euclidean distance can most intuitively reflect the spatial proximity of two regions. Compared with the center distance, it can better capture the local connection relationship between the non-main region and the main boundary (such as a certain endpoint of the non-main region being close to the main boundary), avoiding the accidental deletion of locally connected effective structures due to the overall distance of the region's center. Secondly, the Euclidean distance is more sensitive to subtle spatial changes in the boundary, and can accurately distinguish between "effective small regions that are locally close to the main boundary" and "completely isolated noise regions". It is especially suitable for scenarios where there may be local protrusions, depressions or subtle branches at the boundary of the cladding layer, thereby maximizing the elimination of noise while retaining effective structures, and further improving the accuracy of the target main boundary region.

[0038] To avoid accidental deletion of small regions on the boundary path, the minimum Euclidean distance between each non-main region and the main region is further calculated. If the distance is less than a preset threshold, the small region is retained and merged into the main boundary region.

[0039] Step S3: For each endpoint pixel in the processed cladding layer boundary map, starting from the endpoint pixel, extend in the opposite direction to the adjacent pixels belonging to the main boundary region until the extension path of the endpoint pixel coincides with the pixels of the main boundary region or reaches the boundary of the processed cladding layer boundary map. In this embodiment, specifically, the determination rule for endpoint pixels includes two cases: (1) This pixel has only one neighboring pixel in its 8-neighborhood; (2) The pixel has two adjacent pixels in its 8-neighborhood, and these two adjacent pixels are connected only in a single direction, that is, they do not form a closed or multi-directional extension structure.

[0040] A pixel is an endpoint pixel if it meets any of the above conditions.

[0041] Step S4: If an endpoint pixel's extension path reaches the boundary of the processed cladding layer boundary map, then starting from the boundary of the cladding layer boundary map, delete pixel by pixel in the opposite direction of the endpoint pixel's extension direction until the intersection of the endpoint pixel's extension path and the main boundary region is reached, thus obtaining the target cladding layer boundary map; wherein, the intersection satisfies that the pixel has three adjacent pixels in its 8-neighborhood, and each adjacent pixel has an extension direction independent of the main path. Such pixels constitute the intersection position in the path structure, and the extension and deletion operations are terminated based on this condition.

[0042] like Figure 2 As shown in (b), endpoint pixel 4 extends in the opposite direction to its adjacent pixel that belongs to the main boundary region until the extension path of the endpoint pixel coincides with the pixel of the main boundary region. Endpoint pixels 1, 2, 3, and 5 extend to the boundary of the processed cladding layer boundary map and need further processing. Starting from the boundary of the cladding layer boundary map, each pixel is deleted in the opposite direction to the extension direction of endpoint pixels 1, 2, 3, and 5 until the intersection of the extension path of endpoint pixels 1, 2, 3, and 5 with the main boundary region is reached.

[0043] In this embodiment, the method for determining the intersection point specifically includes: Determine if there are 3 adjacent pixels belonging to the main boundary region in the 8-neighborhood of the current pixel. If so, determine if the extension paths formed by the current pixel and each of its adjacent pixels do not overlap. If they do not overlap, the current pixel is an intersection point.

[0044] Through the above steps, the present invention can effectively repair boundary breaks and missing parts caused by model prediction errors, ensure the continuity and geometric accuracy of the cladding layer boundary, and thus improve the reliability of subsequent size measurement and feature extraction.

[0045] Step S5: Based on the target cladding layer boundary map, obtain the size parameters of the molten pool image to be predicted.

[0046] In this embodiment, the size parameters of the molten pool image to be predicted specifically include: the cross-sectional area, width, and height of the cladding layer.

[0047] While traditional convolutional neural networks and Transformers have demonstrated excellent performance in computer vision tasks such as image classification and object detection, their feature extraction and modeling logic have inherent limitations. These models are typically designed based on the assumption of a high visual correlation between the input image and the output. For example, in image classification, the input is a clearly defined image of the target, and the output is the corresponding category label. In detection tasks, both the input and output revolve around the target's location information within the same image, essentially remaining within the scope of features associated within the same domain. Therefore, when faced with typical cross-domain image conversion requirements, such as molten pool images and cladding layer cross-sectional images, traditional models struggle to overcome the fundamental differences between the two types of images in the visual domain (molten pool images present a dynamic high-temperature region morphology, while cladding layer cross-sectional images show a static cross-sectional contour structure). They cannot accurately capture the implicit process mapping relationship between the two, naturally making it difficult to achieve effective conversion from molten pool features to cladding layer cross-sectional features.

[0048] To address this core issue, this invention employs the Pix2Pix model based on Conditional Generative Adversarial Network (CGAN) as its core model framework, combined with a perceptual loss function. This model inherently possesses the ability to handle the conversion between input and output images of different domains. Through adversarial training between the generator and discriminator, it can gradually learn the potential mapping rules between cross-domain images. Furthermore, to further improve conversion accuracy, this invention introduces an additional perceptual loss function, utilizing a pre-trained feature extraction network (such as the VGG network) to capture the deep semantic features of the images. This allows the model to not only focus on pixel-level surface differences but also understand the deep feature relationships between the two types of images under the process logic, ultimately significantly enhancing the model's understanding and generation accuracy of cross-domain feature mapping between molten pool images and cladding layer cross-sectional images.

[0049] like Figure 3 As shown, Figure 3 This is a structural diagram of the Pix2Pix model used in the cladding cross-section size prediction model of this invention.

[0050] In this embodiment, preferably, the training process of the Pix2Pix model includes: Obtain the training set of molten pool images, input the molten pool images into the generator of the Pix2Pix model, and output the initial cladding layer boundary map of the molten pool images; The Pix2Pix model's generator uses a U-Net (UnetGenerator) architecture with 3 input channels and 3 output channels, outputting an RGB format cladding layer boundary map. The base number of channels is 64, and a total of 8 downsampling layers (num_down=8) are used. During encoding and decoding, batch normalization (BN) layers are used to accelerate convergence, and dropout operations are introduced in some layers to enhance the model's generalization ability. The generator weights are initialized using a normal distribution with a mean of 0 and a variance of 0.022.

[0051] The U-Net structure extracts multi-scale features through the encoder and uses skip connections to fuse high-resolution spatial information during the decoding process, thereby preserving the structural features of the input image.

[0052] The initial cladding layer boundary map of the molten pool image is processed through a pre-trained feature extraction network to extract predicted features from multiple levels of the feature extraction network. The labeled image of the molten pool image is passed through a pre-trained feature extraction network to extract real features from multiple levels of the feature extraction network; In this embodiment, specifically, the feature extraction network adopts the VGG network, and the predicted features and real features of the 4th, 9th and 16th layers of the feature extraction network are extracted.

[0053] The first confidence level is obtained by passing the molten pool image and its initial cladding layer boundary map through the discriminator of the Pix2Pix model; The second confidence level is obtained by passing the molten pool image and its label image through the discriminator of the Pix2Pix model; The Pix2Pix model's discriminator uses a PatchGAN structure (PatchDiscriminator). It has 6 input channels (composed of a 3-channel melt pool image (input) and a 3-channel target boundary map (label / generated result) concatenated along the channel dimension), a base of 64 channels, and 3 convolutional layers. Batch normalization is used for feature standardization. PatchGAN uses local regions (patches) of the image as discriminative units, independently determining the realism of each patch, thus improving the fidelity of texture and detail.

[0054] The Pix2Pix model is trained based on the conditional adversarial loss between the first and second confidence levels, the pixel-level L1 loss between the initial cladding layer boundary map of the molten pool image and the label image, and the perceptual loss between the predicted features and the real features of multiple layers of the feature extraction network to obtain the target Pix2Pix model.

[0055] The formula for the conditional adversarial loss (GAN Loss) between the first and second confidence levels is: , in, In order to counter the loss, For generator, For discriminator, It is the source domain image (melt pool image). It is the target domain image (label image). It is random noise. For mathematical expectation, As the second confidence level, As the first confidence level, This is the initial cladding layer boundary diagram. This means the generator should minimize the conditional adversarial loss. This indicates that the discriminator should maximize the conditional resistance to loss.

[0056] The generator relies not only on random noise vectors when generating target data, but also on pre-existing conditional information in the dataset. This allows it to establish a mapping between input and output. The discriminator aims to maximize the probability difference between real and generated samples, while the generator aims to minimize this difference.

[0057] This invention introduces a pixel-wise L1 loss between the initial cladding layer boundary map of the molten pool image and the label image to constrain the pixel-level difference between the generated image and the real image, making the generated image closer to the target domain image. The loss weight is set to 100.0 to ensure structural consistency. The formula is as follows: , in, For pixel-level L1 loss, It is an L1 norm.

[0058] To compensate for the structural differences between the source and target domains and improve the similarity of the generated results in the high-level semantic feature space, this invention introduces a perceptual loss based on the original pix2pix loss function. The perceptual loss is calculated based on the multi-layer feature representations (layers 4, 9, and 16) of the pre-trained feature extraction network (VGG network), using the mean squared error (MSE) form. The formula for the perceptual loss between the predicted features and the true features at multiple layers of the feature extraction network is as follows: , in, In order to perceive loss, This represents the number of layers in the feature extraction network. It is a molten pool image. It is a label image. It is random noise. This represents the initial cladding layer boundary diagram. Represents a generator. For the feature extraction network Predictive features of the layer For the feature extraction network The true characteristics of the layer For the feature extraction network The height of the layer's features For the feature extraction network The width of the layer's features, For the feature extraction network The number of channels in the layer's features. For layer index, It is an L2 norm.

[0059] In actual industrial production, acquiring data on inclined basal cladding is costly and time-consuming, and the cross-sectional annotation process is cumbersome, resulting in a limited number of high-quality datasets available for training. Existing deep learning-based prediction methods generally require large-scale data to achieve good generalization ability. However, when datasets are scarce, these methods are prone to overfitting, leading to a significant decline in prediction performance under new working conditions or with new materials.

[0060] To address the aforementioned issues, this invention employs various data augmentation techniques (such as image inversion, brightness / contrast adjustment, cropping, and channel transformation), boundary transformation, and an augmentation method based on the Structural Similarity Index (SSIM) to increase the data size by approximately tenfold, thereby significantly alleviating the small sample size problem. The specific solution is as follows: like Figure 4 As shown, Figure 4 This is a schematic diagram illustrating the process of obtaining the training set of molten pool images.

[0061] In this embodiment, preferably, the process of obtaining the molten pool image training set includes: Acquire molten pool videos under different parameter conditions, select any frame of molten pool image from the molten pool videos under different parameter conditions, and the cross-sectional image of the cladding layer under that parameter condition as label images, and use them as samples in the molten pool image training set. In this embodiment, specifically, different parameter conditions refer to the core process parameters that affect the cladding process (such as laser cladding, arc cladding and other thermal processing processes), such as laser power, powder feeding rate and scanning speed.

[0062] In this embodiment, in each video of the molten pool obtained under different parameter conditions, a single frame of the molten pool image is first selected as the input image, and the cross-sectional image of the cladding layer under that condition is found as the label image to form a set of training samples. Since the process of obtaining the molten pool image and the cross-sectional image is costly and time-consuming, this experiment finally obtained only 108 sets of sample data. After being divided into training set and test set, the number of samples that can be used for training is even more limited.

[0063] Data augmentation was performed on the molten pool images selected under different parameter conditions to obtain data-augmented molten pool images under different parameter conditions; To address the small sample size problem and improve the model's generalization ability, this invention employs the following measures in the data augmentation stage: Random rotation: The image is randomly rotated to a certain angle (range [-degrees, +degrees]) to simulate different shooting directions; where degrees is the maximum offset of the rotation angle; Random cropping: randomly selects a region of a specified size in an image for cropping, thereby changing the local information of the input image; Random style transfer: Convert the image to grayscale with a certain probability (e.g., 30%) while preserving the RGB channel structure to enhance the model's robustness to changes in brightness and texture; Color perturbation: Randomly adjusts the brightness, contrast, saturation, and hue of an image to simulate different lighting environments.

[0064] Since the prediction task of this invention only targets the outer contour of the cladding layer cross section, during the enhancement process, the boundary line with a width of about 5 pixels on the outermost edge of the cladding layer cross section in the label image corresponding to the data-enhanced molten pool image under different parameter conditions is extracted to form an independent cladding layer boundary map; then the cladding layer boundary map is stitched together with the corresponding data-enhanced molten pool image to generate a data-enhanced molten pool image with stronger structural information.

[0065] To fully utilize video information, this invention extracts multiple frames from each melt pool video segment as additional samples. However, to avoid excessive differences between different frames within the same video leading to model learning bias, this invention employs SSIM for filtering. Melt pool images other than the selected melt pool images from melt pool videos under different parameter conditions are used as candidate images under that parameter condition. Calculate the structural similarity index between each candidate image and the data-enhanced cladding pool image under different parameter conditions, sort them in descending order, and use the candidate images of the previous preset number of frames and the cross-sectional images of the cladding layer under the same parameter conditions as label images, which are then used as augmented samples in the cladding pool image training set.

[0066] In this embodiment, based on the originally selected melt pool image, the preset number of frames is set to 9, and the 9 most similar frames with the highest structural similarity index are selected for data augmentation. Through the above processing, the original dataset of only 108 groups is expanded to about 10 times the size, which effectively increases the number of effective samples for model training and enhances the diversity and representativeness of the data.

[0067] In this embodiment, to obtain laser cladding pool and cladding layer cross-sectional data for different inclined substrates, the present invention first fixes the metal substrate on a worktable with an adjustable tilt angle. By adjusting the tilt angle of the worktable, the substrate can be made to present different tilt states. At the same time, the posture angle of the robotic arm is adjusted to ensure that the coaxial powder feeding nozzle is perpendicular to the substrate surface under any substrate tilt condition, thereby ensuring powder feeding stability and forming quality.

[0068] During the experiment, four substrate tilt angles were selected for laser cladding tests: 0°, 30°, 60°, and 90°. For each tilt angle, the following process parameter combinations were set sequentially: Laser power: 1200 W, 1000 W, 800 W; Scanning speeds: 3 mm / s, 3.5 mm / s, 7 mm / s; Powder feeding rates: 4 g / min, 6 g / min, 8 g / min.

[0069] During laser cladding, a high-speed camera is used to capture images of the molten pool from the side. The camera is held at a 35° angle to the substrate, with a shooting distance of 20 cm. The equipment used is the Acuteye V4.0 high-speed welding camera, which can capture clear images at high speeds, ensuring the fidelity of details in the dynamic process of the molten pool.

[0070] After cladding is completed, the single-layer cladding is wire-cut to obtain a cross-sectional sample. The sample is then mounted and polished to ensure a flat and smooth cross-section for easy microscopic observation. Subsequently, cross-sectional images of the cladding layer are captured using an MX6R optical microscope as label data required for subsequent model training.

[0071] This embodiment is built on the MMgeneration environment and performed on an NVIDIA RTX3090 GPU. The training process of the cladding cross-section size prediction model uses Adam optimizers for both the generator and discriminator, with a learning rate of [missing information]. Momentum parameters in the Adam optimizer The exponential decay rate of the second moment estimate of the Adam optimizer The total number of training iterations was 300,000. Training logs were recorded every 100 iterations, and model checkpoints were saved and performance evaluated every 10,000 iterations. Evaluation metrics included Average Surface Distance (ASD) and Hausdorff distance. The evaluation sample size was 214, and the image resolution was 512×512. To improve training efficiency and stability, distributed data parallelism (NCCL backend) was employed, cudnn_benchmark was enabled to optimize convolution calculations, OpenCV multithreading was disabled, and multi-processing was started using the fork method. Additionally, the generated results were visualized and saved every 5000 iterations to monitor the model's generation quality.

[0072] To verify the effectiveness of the proposed method for predicting the cross-sectional size of laser cladding on an inclined base surface, this embodiment compares the proposed method with several existing mainstream models. The selected mainstream models include Unet and DeepLabV3+ in convolutional neural networks, ViT based on the Transformer structure, and the Pix2Pix model in CGAN.

[0073] The dataset was divided into training and testing sets in an 8:2 ratio. The experimental results are shown in Table 1, which illustrates the comparative experimental results of different models.

[0074] Table 1

[0075] The experimental results show that the method of the present invention outperforms other comparative models in both the two key indicators of average surface distance (ASD) and Hausdorff distance.

[0076] The average surface distance (ASD) measures the average shortest distance between the predicted boundary and the true boundary, reflecting the accuracy of the overall boundary fitting. A smaller ASD indicates that the predicted result is closer to the true contour. Its formula is: , in, for and The average surface distance between them For the boundary pixel set of the target cladding layer boundary map, This is the set of boundary pixels of the actual cladding layer cross-section. The predicted profile in the target cladding layer boundary map. This is the true outline in the cross-sectional view of the actual cladding layer. for The elements in for The elements in For a single pixel arrive The shortest distance between all pixels in the array; in this embodiment, the distance is the Euclidean distance. For a single pixel arrive The shortest distance between all pixels in the array. The total number of pixels in the boundary pixel set of the target cladding layer boundary map. This represents the total number of pixels in the boundary pixel set of the actual cladding layer cross-sectional image.

[0077] The Hausdorff distance measures the maximum distance between two sets of boundary points, focusing on the farthest deviation on the boundary and reflecting extreme error conditions. A smaller Hausdorff distance indicates a stronger fit to outliers or marginal regions. Its formula is: , in, for and The distance between Hausdorf and [other locations] It is a norm.

[0078] Traversing point sets Each pixel in (i.e., for any) ), in point set Find the pixel The closest pixel in space (i.e., determine) And satisfy the distance Minimum), for Each pixel in Find the nearest points to form a pair; calculate the distance between all such pairs, and select the largest distance, denoted as . Similarly, we can obtain ,Pick , The larger value in the middle is used as .

[0079] like Figure 5 As shown, Figure 5 These are experimental comparison figures showing the prediction models for different cladding cross-section sizes and the method of this invention. Figure 5 (a) is a schematic diagram comparing the prediction models for different cladding cross-sectional dimensions with the target cladding layer boundary map extracted from the first molten pool image by the method of this invention. Figure 5 (b) is a schematic diagram comparing the prediction models for different cladding cross-sectional dimensions with the target cladding layer boundary map extracted from the second molten pool image by the method of this invention. Figure 5 The diagram in (c) shows a comparison between the prediction models for different cladding cross-section sizes and the target cladding layer boundary map extracted from the third molten pool image by the method of this invention. Figure 5 The diagram shows a comparison between the prediction models for different cladding cross-sectional dimensions in (d) and the target cladding layer boundary map extracted from the fourth molten pool image by the method of this invention.

[0080] Figure 5 The visualization comparison results of the method of this invention with several existing mainstream models are presented. It is clearly observed from the comparison figures that the method of this invention outperforms other methods in terms of the integrity of the cladding layer cross-sectional profile, boundary continuity, and detail reproduction, demonstrating higher prediction accuracy and superior visual effects.

[0081] like Figure 6 As shown, Figure 6 This diagram illustrates a comparison between prediction models for different cladding cross-section sizes and prediction results for cladding layer size using the method of this invention. Figure 6 (a) is a schematic diagram comparing the prediction models for different cladding cross-sectional dimensions with the method of this invention for predicting the cross-sectional area of ​​the cladding layer. Figure 6 (b) is a schematic diagram comparing the predicted width of different cladding cross-section size prediction models with that predicted by the method of this invention. Figure 6 The diagram shows a comparison between the prediction models for different cladding cross-section dimensions and the height predicted by the method of this invention, with behavior (c) in the figure.

[0082] Figure 6 This paper presents a detailed comparison of the best-performing ViT model, Pix2Pix model, and the method of this invention in three dimensions: cross-sectional area, width, and height of the cladding layer. The evaluation metrics include the coefficient of determination (COP). The mean absolute error (MAE) and mean squared error (MSE) of the method are clearly observed in the scatter plot. The prediction results of the method of the present invention are closely distributed near the ideal fitting line (45-degree line) for these three key parameters, indicating that its fitting accuracy to the actual size is significantly better than the comparison model, and the prediction results are more accurate and stable.

[0083] Among them, the coefficient of determination ( : represents the goodness of fit between the predicted and actual values, with a numerical range of [−∞,1], where 1 indicates a perfect fit, 0 indicates that the model has no predictive ability, negative values ​​indicate that the model performs worse than the simple mean, and higher values ​​indicate that the model performs better than the simple mean. The value indicates that the model has a strong ability to explain changes in the data.

[0084] Mean Absolute Error (MAE): This is the average of the absolute values ​​of the errors between the predicted and actual values, reflecting the average magnitude of the prediction error. The smaller the MAE, the more accurate the prediction and the more concentrated the error.

[0085] Mean Squared Error (MSE): This is the average of the squares of the prediction errors. It emphasizes the penalty effect of larger errors. The smaller the MSE, the smaller the overall error of the model and the better the prediction stability.

[0086] As shown in Table 2, Table 2 illustrates the performance of different methods and the method of the present invention in predicting the cladding layer size on the test set.

[0087] Table 2

[0088] As shown in Table 3, Table 3 presents the ablation experimental results of the method of the present invention.

[0089] Table 3

[0090] This invention verifies the impact of perceptual loss, the enhancement method based on structural similarity index (SSIM), and the endpoint extension repair algorithm based on graph structure analysis on the prediction performance of cladding cross-sectional dimensions.

[0091] This second embodiment provides a system for predicting the cross-sectional dimensions of laser cladding on an inclined base surface, including: A memory for storing computer programs; a processor for executing the computer programs to implement the steps of the above-described method for predicting the cross-sectional dimensions of laser cladding on an inclined datum.

[0092] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0093] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0094] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0095] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 Figure 1 The steps of the function specified in one or more boxes.

[0096] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for predicting the cross-sectional dimensions of laser cladding on an inclined datum, characterized in that, include: The initial cladding layer boundary map is obtained by using the cladding cross-section size prediction model to predict the image of the molten pool to be predicted. The areas of all connected regions in the initial cladding layer boundary map are sorted, and the connected region with the largest area is taken as the main boundary region, while the remaining regions are taken as non-main regions. Only the pixels of the main boundary region are retained to obtain the processed cladding layer boundary map. For each endpoint pixel in the processed cladding layer boundary map, starting from the endpoint pixel, extend in the opposite direction to the adjacent pixel belonging to the main boundary region until the extension path of the endpoint pixel coincides with the pixel of the main boundary region or reaches the boundary of the processed cladding layer boundary map. If an endpoint pixel's extension path reaches the boundary of the processed cladding layer boundary map, then starting from the boundary of the cladding layer boundary map, delete pixel by pixel in the opposite direction of the endpoint pixel's extension direction until the intersection of the endpoint pixel's extension path and the main boundary region is reached, thus obtaining the target cladding layer boundary map. Based on the target cladding layer boundary map, obtain the size parameters of the molten pool image to be predicted.

2. The method for predicting the cross-sectional dimensions of laser cladding on an inclined datum surface according to claim 1, characterized in that, Also includes: The initial cladding layer boundary map of the target is obtained by performing morphological opening and closing operations on the initial cladding layer boundary map.

3. The method for predicting the cross-sectional dimensions of laser cladding on an inclined datum surface according to claim 1, characterized in that, Also includes: Calculate the distance between each non-main region and the geometric center of the initial main boundary region. Merge non-main regions with a distance less than a set threshold with the main boundary region to form the target main boundary region.

4. The method for predicting the cross-sectional dimensions of laser cladding on an inclined datum surface according to claim 1, characterized in that, The methods for determining intersections include: Determine if there are 3 adjacent pixels belonging to the main boundary region in the 8-neighborhood of the current pixel. If so, determine if the extension paths formed by the current pixel and each of its adjacent pixels do not overlap. If they do not overlap, the current pixel is an intersection point.

5. The method for predicting the cross-sectional dimensions of laser cladding on an inclined datum surface according to claim 1, characterized in that, The prediction model for cladding cross-section size is the Pix2Pix model.

6. The method for predicting the cross-sectional dimensions of laser cladding on an inclined datum surface according to claim 5, characterized in that, The training process of the Pix2Pix model includes: Obtain the training set of molten pool images, input the molten pool images into the generator of the Pix2Pix model, and output the initial cladding layer boundary map of the molten pool images; The initial cladding layer boundary map of the molten pool image is processed through a pre-trained feature extraction network to extract predicted features from multiple levels of the feature extraction network. The labeled image of the molten pool image is passed through a pre-trained feature extraction network to extract real features from multiple levels of the feature extraction network; The first confidence level is obtained by passing the molten pool image and its initial cladding layer boundary map through the discriminator of the Pix2Pix model; The second confidence level is obtained by passing the molten pool image and its label image through the discriminator of the Pix2Pix model; The Pix2Pix model is trained based on the conditional adversarial loss between the first and second confidence levels, the pixel-level L1 loss between the initial cladding layer boundary map of the molten pool image and the label image, and the perceptual loss between the predicted features and the real features of multiple layers of the feature extraction network to obtain the target Pix2Pix model.

7. The method for predicting the cross-sectional dimensions of laser cladding on an inclined datum surface according to claim 6, characterized in that, The formula for the perceptual loss between the predicted features and the true features at multiple levels of the feature extraction network is as follows: ， in, In order to perceive loss, This represents the number of layers in the feature extraction network. It is a molten pool image. It is a label image. It is random noise. This represents the initial cladding layer boundary diagram. Represents a generator. For the feature extraction network Predictive features of the layer For the feature extraction network The true characteristics of the layer For the feature extraction network The height of the layer's features For the feature extraction network The width of the layer's features, For the feature extraction network The number of channels in the layer's features. For layer index, It is an L2 norm.

8. The method for predicting the cross-sectional dimensions of laser cladding on an inclined datum surface according to claim 6, characterized in that, The process of obtaining the training set of molten pool images includes: Acquire molten pool videos under different parameter conditions, select any frame of molten pool image from the molten pool videos under different parameter conditions, and the cross-sectional image of the cladding layer under that parameter condition as label images, and use them as samples in the molten pool image training set. Data augmentation was performed on the molten pool images selected under different parameter conditions to obtain data-augmented molten pool images under different parameter conditions; In the molten pool video under different parameter conditions, the molten pool images other than the selected molten pool images are used as candidate images under that parameter condition; Calculate the structural similarity index between each candidate image and the data-enhanced cladding pool image under different parameter conditions, sort them in descending order, and use the candidate images of the previous preset number of frames and the cross-sectional images of the cladding layer under the same parameter conditions as label images, which are then used as augmented samples in the cladding pool image training set.

9. The method for predicting the cross-sectional dimensions of laser cladding on an inclined datum surface according to claim 1, characterized in that, The dimensional parameters of the molten pool image to be predicted include: the cross-sectional area, width, and height of the cladding layer.

10. A system for predicting the cross-sectional dimensions of laser cladding on an inclined datum, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the method for predicting the cross-sectional dimensions of a tilted datum laser cladding as described in any one of claims 1 to 9.

Images