A method and system for monitoring the welding process of composite plates

Abstract

This application relates to the field of image processing technology, and in particular to a method and system for monitoring the welding process of composite plates. The method includes: acquiring a thermal infrared image of the welding area of ​​the composite plate to be monitored, and determining the temperature assessment value corresponding to each pixel and initial multiple cluster centers; determining a first membership degree of the pixel relative to the cluster centers based on the difference in temperature assessment values ​​relative to the cluster centers and the initial membership degree determined by fuzzy C-means clustering; adjusting the first membership degree to obtain a second membership degree, and performing cluster iteration based on the second membership degree to obtain a clustering result that converges the iteration process; using the clustering result to determine the equivalent area of ​​the heat-affected zone on the metal side, thereby obtaining the monitoring result of the welding process. Through the above technical solution, the welding process of metal composite plates can be monitored more accurately.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a method and system for monitoring the welding process of composite plates. Background Technology

[0002] Metal composite panels are a new type of material made by combining two metal sheets with different physical and chemical properties through a specific process. The core structure of metal composite panels usually includes a base layer and a cladding layer: the base layer is made of metals with higher strength and lower cost, such as steel or iron, which mainly bears the structural support role and ensures the mechanical properties and stability of the material; the cladding layer is made of metals with special functions, such as titanium, copper, nickel or stainless steel, which endow the material with key properties such as corrosion resistance, high temperature resistance or electrical conductivity.

[0003] Titanium-steel composite plates or nickel-steel composite plates are often used to manufacture equipment such as reaction vessels or heat exchangers. The multilayered structure resists the corrosion of chemical media by resisting acid and alkali corrosion, while the high strength of the base layer ensures the structural safety of the equipment. In the field of marine engineering, metal composite plates can be used for ship hulls or offshore platform components to effectively cope with the strong corrosive environment of seawater. Metal composite plates can also become a key material for improving equipment reliability and reducing manufacturing costs.

[0004] The welding quality of metal composite plates directly affects their performance and safety. The differences in the properties of the two metals can lead to various abnormalities during the welding process. For example, the differences in the thermophysical properties of different metals can cause uneven distribution of the welding heat field, and heat can easily accumulate on one side, causing welding thermal stress, which in turn can lead to workpiece deformation or cracking, affecting the performance of the obtained metal composite plate. Therefore, it is necessary to monitor the welding process of metal composite plates. Summary of the Invention

[0005] To monitor the welding process of metal composite plates, this application provides a method and system for monitoring the welding process of composite plates.

[0006] According to a first aspect of the embodiments of this application, a method for monitoring the welding process of a composite plate is provided, comprising: acquiring a thermal infrared image of the welding area of ​​the composite plate to be monitored, and determining the temperature assessment value corresponding to each pixel; using the temperature assessment value to determine an initial plurality of cluster centers from the thermal infrared image; determining a first membership degree of the pixel relative to the cluster center based on the difference in temperature assessment values ​​of the pixel relative to the cluster center and an initial membership degree determined by fuzzy C-means clustering; adjusting the first membership degree to obtain a second membership degree based on the distribution of the first membership degrees of other pixels relative to the cluster center within the neighborhood window of the pixel; updating the temperature assessment value of the cluster center based on the second membership degree, and iterating the cluster centers of the thermal infrared image using the second membership degrees of different pixels relative to different cluster centers to obtain a clustering result that converges the iteration process; acquiring a pre-labeled metal interface line in the thermal infrared image, and using the membership degrees of the pixels in the clustering result relative to the cluster centers of different metal sides of the metal interface line to determine the equivalent area of ​​the heat-affected zone on different metal sides to obtain a monitoring result of the welding process.

[0007] This allows for monitoring of the welding process of metal composite plates.

[0008] Optionally, the temperature assessment value corresponding to a pixel is determined in the following way: for a target pixel in a thermal infrared image, the sum of the pixel values ​​of the target pixel in the red, green, and blue channels of the thermal infrared image is determined, and the ratio of the pixel value of the target pixel in the red channel of the thermal infrared image to the sum of the pixel values ​​is used as the temperature assessment value of the target pixel.

[0009] In this way, the long-wavelength components such as red light in the high-temperature radiation of welding can effectively highlight the characteristics of the high-temperature region. Using the proportion of the red channel as the temperature assessment value can better reflect the actual temperature information of the pixel value.

[0010] Optionally, multiple initial cluster centers are determined from the thermal infrared image using temperature assessment values, including: sorting all pixels in the thermal infrared image from largest to smallest according to temperature assessment values, and using the pixels corresponding to the multiple preset rankings as the initial multiple cluster centers corresponding to different welding influence regions; the different welding influence regions include the molten pool region, the base material region, and the heat-affected zone on different metal sides.

[0011] In this way, by selecting pixels of a specific sorting order as the initial center, we can avoid the problem of getting trapped in local optima caused by random initialization, and ensure that the cluster centers are distributed near key areas such as the melt pool, heat-affected zone and parent material in the initial stage of iteration, thus accelerating the convergence speed of the clustering process.

[0012] Optionally, the first membership degree of a pixel relative to the cluster center is determined in the following way: ,in, For the first For the pixel point The first membership degree of each cluster center, where exp is the natural exponential function. For the first Temperature evaluation value of each pixel For the first Temperature assessment values ​​for each cluster center To determine the first clustering step using the fuzzy C-means clustering algorithm The pixel and the The initial membership degree between cluster centers.

[0013] Optionally, the first membership degree can be adjusted to obtain the second membership degree, including: ,in, For the first For the pixel point The second membership degree of each cluster center This indicates normalization processing. For the first The number of other pixels within the spatial neighborhood window of a pixel, excluding itself. For the first The first pixel For the nth neighboring pixel point The first membership degree of each cluster center For the first The variance of the temperature evaluation values ​​of all pixels within the spatial neighborhood window of a pixel.

[0014] In this way, the membership degree of the center pixel is corrected by using the weighted sum of neighboring pixels, and the variance is... As a control term, the influence of the neighborhood is enhanced in areas with gentle temperature changes to smooth noise, while the influence of the neighborhood is weakened in edge areas to maintain edge sharpness, thus preserving the key contours of the heat-affected zone while denoising.

[0015] Optionally, the temperature assessment value of the cluster center is updated based on the second membership degree, including: ,in, For the first After the nth iteration Temperature assessment values ​​for each cluster center This represents the total number of pixels in the thermal infrared image. For the first For the pixel point The second membership degree of each cluster center For fuzzy coefficients, For the first Temperature assessment value for each pixel.

[0016] Optionally, the convergence of the iteration process can be determined by the following method: determining the change in the temperature assessment value of the cluster center in the iteration process relative to the temperature assessment value of the cluster center in the previous iteration process. If the change corresponding to two consecutive iteration processes is less than the preset convergence threshold, the iteration process is determined to be converged.

[0017] By using the condition that both changes in a row satisfy the convergence criterion, we can prevent the algorithm from oscillating near local extreme points and thus ensure the stability of the clustering results.

[0018] Optionally, the metal interface line in the thermal infrared image is pre-calibrated in the following way: a visible light image of the weld bevel of the composite plate to be monitored before welding is acquired using a visible light camera, and the position of the metal interface line in the spatial coordinate system of the workpiece is determined by image calibration; the position of the metal interface line in the spatial coordinate system of the workpiece is mapped to the pixel coordinate system of the thermal infrared image to obtain the position coordinates of the metal interface line in the thermal infrared image, so as to obtain the pre-calibrated metal interface line in the thermal infrared image.

[0019] Optionally, the equivalent area of ​​the heat-affected zone on the metal side is equal to the sum of the membership degrees of all pixels relative to the cluster center located on the metal side; the monitoring results of the welding process are obtained in the following way: based on the difference in the equivalent area of ​​the heat-affected zone on different metal sides, the thermal stress asymmetry index is determined, and when the thermal stress asymmetry index exceeds the preset safety threshold, an early warning signal is output to indicate the imbalance of welding heat input.

[0020] According to a second aspect of the present application, a composite plate welding process monitoring system is provided, comprising: a processor and a memory, wherein the memory stores computer program instructions, and the computer program instructions, when executed by the processor, implement the steps of the composite plate welding process monitoring method provided in the first aspect of the present application.

[0021] The technical solutions provided by the embodiments of this application may include the following beneficial effects: acquiring thermal infrared images of the welding area of ​​the composite plate to be monitored, determining the temperature assessment value corresponding to each pixel, and using the temperature assessment value to determine multiple initial cluster centers from the thermal infrared image; being able to iterate the clustering process based on the initial membership degree determined by fuzzy C-means clustering to obtain more accurate cluster centers and membership degrees; acquiring the pre-labeled metal interface line in the thermal infrared image, and using the membership degree of the pixel in the clustering result relative to the cluster centers of different metal sides of the metal interface line to determine the equivalent area of ​​the heat-affected zone on different metal sides; since the cluster centers and membership degrees of the obtained clustering results are more accurate, more accurate monitoring results of the welding process can be obtained.

[0022] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a method for monitoring the welding process of composite plates according to an exemplary embodiment;

[0024] Figure 2 This is a schematic diagram of the clustering results of different locations of the composite plate in the clustering space;

[0025] Figure 3 This is a schematic diagram comparing the thermal stress asymmetry index determined by the clustering results of the embodiments of this application with the thermal stress asymmetry index determined by the clustering results of fuzzy C-means clustering.

[0026] Figure 4 This is a schematic diagram of a composite plate welding process monitoring system according to an exemplary embodiment. Detailed Implementation

[0027] First, a brief introduction to the application scenarios of the embodiments of this application will be given. In the application scenarios of this application, the metal composite plate is composed of two metals (such as titanium and steel) with different physical properties, which makes the heat conduction during the welding process potentially have asymmetric and nonlinear characteristics.

[0028] In the process of welding two different metal plates to obtain a metal composite plate, if the heat input on one side is too large, it will cause the metal to overheat and grow brittle intermetallic compounds; if the heat input is insufficient, it will lead to incomplete fusion defects. This imbalance of thermal stress will reduce the mechanical properties of the metal composite plate. Therefore, it is necessary to monitor the welding process of the metal composite plate.

[0029] To address the aforementioned technical problems, embodiments of this application provide a method and system for monitoring the welding process of composite plates. Figure 1 This is a flowchart illustrating a method for monitoring the welding process of composite plates according to an exemplary embodiment, such as... Figure 1 As shown, the method includes the following steps.

[0030] In step S101, a thermal infrared image of the welding area of ​​the composite plate to be monitored is acquired, and the temperature assessment value corresponding to each pixel is determined. The initial multiple cluster centers are determined from the thermal infrared image using the temperature assessment value.

[0031] The application scenario of this application embodiment can be to achieve composite by welding two metal materials with different materials. The resulting metal composite plate can be, for example, a titanium steel composite plate or a nickel stainless steel composite plate.

[0032] Because the two base materials of titanium-steel composite plates have significant differences in melting point, thermal conductivity and specific heat capacity, the heat distribution on both sides of the interface during welding is naturally asymmetrical. If this heat distribution cannot be accurately monitored and controlled, it may lead to the formation of brittle intermetallic compounds or the generation of incomplete fusion defects, affecting the performance of the product obtained after welding.

[0033] In one embodiment, the temperature assessment value corresponding to a pixel is determined as follows: for a target pixel in a thermal infrared image, the sum of the pixel values ​​of the target pixel in the red, green, and blue channels of the thermal infrared image is determined, and the ratio of the pixel value of the target pixel in the red channel of the thermal infrared image to the sum of the pixel values ​​is used as the temperature assessment value of the target pixel.

[0034] Thermal imagers can be used to acquire and output thermal infrared images. For ease of visualization and general processing of thermal infrared images, the temperature information of the monitored area is usually mapped into a color RGB image.

[0035] For any target pixel in a thermal infrared image, the pixel values ​​in the red, green, and blue channels can be read. For example, if a pixel has a value of 220 in the red channel, 100 in the green channel, and 50 in the blue channel, then the temperature assessment value of the pixel is 220 / (220+100+50) = 0.59.

[0036] Following the steps for determining the temperature assessment value of the target pixel, the temperature assessment values ​​of other pixels in the thermal infrared image can be determined separately.

[0037] In welding monitoring scenarios, thermal infrared images are usually processed with pseudo-color. The temperature of higher-temperature areas in thermal infrared images is usually mapped to warm colors such as red or orange, while the temperature of lower-temperature areas in thermal infrared images is usually mapped to cool colors such as blue.

[0038] If a single-channel grayscale value is used directly as the temperature assessment value, it is easily affected by the overall brightness attenuation caused by smoke and dust. In the pseudo-color mapping logic of the infrared thermal imager, the high proportion of the red component stably corresponds to the high temperature area of ​​the welding core. Using the proportion of the red channel as the temperature assessment value can eliminate the interference caused by changes in ambient light intensity or fluctuations in the overall gain of the sensor.

[0039] Compared to using absolute grayscale values ​​directly, the proportion of red light can more sensitively capture the core shape of the molten pool and heat-affected zone, effectively filtering out background noise caused by the flickering of arc light, which makes the overall image appear bright and dark.

[0040] In one embodiment, determining initial multiple cluster centers from a thermal infrared image using temperature assessment values ​​includes: sorting all pixels in the thermal infrared image according to their temperature assessment values ​​from largest to smallest, and using the pixels corresponding to the multiple preset rankings as initial multiple cluster centers for different welding influence regions; the different welding influence regions include the molten pool region, the base material region, and the heat-affected zones on different metal sides.

[0041] For example, when welding titanium-steel composite plates, the welding area of ​​titanium-steel composite plates mainly includes four welding affected areas: the molten pool area, the heat-affected zone on the steel side, the heat-affected zone on the titanium side, and the base material area. The temperature of these four welding affected areas usually decreases in the order of the molten pool area, the heat-affected zone on the steel side, the heat-affected zone on the titanium side, and the base material area.

[0042] For example, the number of clusters can be set to 4, with different clusters corresponding to different welding-affected areas. This allows for the counting of the total number of pixels in the thermal infrared image. The temperature evaluation values ​​of all pixels are sorted in descending order.

[0043] For example, the ranking of the temperature assessment value could be selected from the top. , , as well as The pixels at each location are used as four different initial cluster centers.

[0044] K-means or FCM (Fuzzy C-Means) algorithms often use the method of randomly selecting initial centers. When the data distribution is complex, they may get stuck in local optima, leading to an increase in the number of iterations or even clustering failure.

[0045] The welding thermal field has a more distinct gradient distribution pattern, which makes the distribution of temperature information of pixels in infrared thermal images more regular. By selecting the initial cluster center by preset ranking, the prior knowledge of the temperature distribution in the welding area can be used to achieve a more accurate determination of the initial cluster center in the clustering process.

[0046] By identifying different initial cluster centers corresponding to different welding-affected areas, it is ensured that the initial cluster centers are numerically distributed near the typical values ​​of each region. This can shorten the convergence time of subsequent iterations and prevent class merging or empty classes caused by excessive deviation of initial values, which helps to achieve more accurate clustering of pixels in thermal infrared images.

[0047] In step S102, the first membership degree of a pixel relative to the cluster center is determined based on the difference in temperature evaluation values ​​of the pixel relative to the cluster center and the initial membership degree determined by fuzzy C-means clustering.

[0048] In one embodiment, the first membership degree of a pixel relative to the cluster center is determined in the following way: ,in, For the first For the pixel point The first membership degree of each cluster center It is a natural exponential function. For the first Temperature evaluation value of each pixel For the first Temperature assessment values ​​for each cluster center To determine the first clustering step using the fuzzy C-means clustering algorithm The pixel and the The initial membership degree between cluster centers.

[0049] First, the formula for fuzzy C-means clustering can be used to calculate the first... The pixel and the Initial membership degree between cluster centers ,For example Where K is the number of cluster centers. For the first The pixel and the Euclidean distance between cluster centers For the first The Euclidean distance between a pixel and the a-th cluster center The fuzzy index is a factor, which can be, for example, 2.

[0050] Fuzzy C-means clustering algorithm only relies on Membership calculation often results in excessive ambiguity at the boundaries of categories, meaning that the classification of edge pixels is ambiguous. In this embodiment, an exponential function is introduced to construct a decay model. When the difference between the pixel value and the cluster center in the temperature evaluation value is smaller, and the initial membership degree determined by the FCM algorithm is larger, the first membership degree of the pixel relative to the cluster center is larger.

[0051] By calculating the first membership degree of a pixel relative to its cluster center, not only is the global optimality of FCM preserved, but the contrast of membership degrees is also stretched through exponential mapping. This makes the first membership degree of pixels that actually belong to a certain cluster center more significant relative to the cluster center, while the first membership degree of pixels that do not actually belong to a certain cluster center decays faster relative to the cluster center. This helps to extract clearer contours from the blurred boundaries of the heat-affected zone.

[0052] In step S103, the first membership degree is adjusted to obtain the second membership degree by utilizing the distribution of the first membership degree of other pixels in the neighborhood window of the pixel relative to the cluster center.

[0053] In one embodiment, adjusting the first membership degree to obtain the second membership degree includes: ,in, For the first For the pixel point The second membership degree of each cluster center This indicates normalization processing. For the first The number of other pixels within the spatial neighborhood window of a pixel, excluding itself. For the first The first pixel For the nth neighboring pixel point The first membership degree of each cluster center For the first The variance of the temperature evaluation values ​​of all pixels within the spatial neighborhood window of a pixel.

[0054] For example, the spatial neighborhood window can be set as... A matrix where the center pixel is surrounded by 8 neighboring pixels. Extraction... The temperature evaluation values ​​of 9 pixels within the window are calculated, and the variance is calculated. If the spatial neighborhood window is located in a flat area such as inside the parent material, The value is smaller; conversely, if the spatial neighborhood window is located in an edge region such as the edge of the melt pool, The value is larger.

[0055] Infrared thermal images of the welding area during the welding process of metal composite plates may contain a large amount of speckle noise. If the classification is based solely on the membership degree of a single pixel, the noisy pixels may be misclassified as normal pixels.

[0056] If the All pixels surrounding a given pixel belong to the same category. , then the first This pixel belongs to the category. The probability will increase, introducing variance. As adaptive weights, they can utilize neighborhood information to achieve stronger filtering in smooth regions, and can also achieve better filtering in edge regions. Automatically reducing neighborhood weights to preserve edge details effectively removes isolated misclassified pixels caused by welding spatter or arc noise while preserving the geometric boundaries of the molten pool and heat-affected zone.

[0057] In step S104, the temperature evaluation value of the cluster center is updated according to the second membership degree, and the cluster centers of the thermal infrared image are iterated using the second membership degree of different pixels relative to different cluster centers to obtain the clustering result that makes the iteration process converge.

[0058] In one embodiment, updating the temperature assessment value of the cluster centers based on the second membership degree includes: ,in, For the first After the nth iteration Temperature assessment values ​​for each cluster center This represents the total number of pixels in the thermal infrared image. For the first For the pixel point The second membership degree of each cluster center For fuzzy coefficients, For the first Temperature assessment value for each pixel.

[0059] For the Each cluster center can be used to analyze all pixels in a thermal infrared image and assign a temperature assessment value to each pixel. Multiply by the pixel pair The membership degree of each cluster center The weighted average is calculated using the power of the power, and the denominator of the formula for calculating the temperature assessment value of the cluster center updated by the second membership degree is the sum of all weights.

[0060] The second membership degree enables the movement direction of the cluster center during subsequent iterations. Taking into account the spatial distribution of different pixels, the fuzziness coefficient m controls the degree of fuzziness in the clustering, which can prevent the clustering result from degenerating into hard clustering. It avoids the situation where some bright noise points pull the entire cluster center off course, ensuring that the final converged center point truly represents the average level of the temperature region.

[0061] In one embodiment, whether the iteration process has converged is determined by: determining the change in the temperature assessment value of the cluster center in the iteration process relative to the temperature assessment value of the cluster center in the previous iteration process; if the change corresponding to two consecutive iteration processes is less than a preset convergence threshold, the iteration process is determined to have converged.

[0062] If the changes in two consecutive iterations are both less than the preset convergence threshold, the iteration process is considered converged. This ensures that convergence is achieved when the clustering results are stable, and stops the iterative process of clustering pixels in the infrared thermal image.

[0063] In step S105, the pre-calibrated metal interface line in the thermal infrared image is obtained. The equivalent area of ​​the heat-affected zone on different metal sides is determined by using the membership degree of the pixel points in the clustering results relative to the cluster centers of different metal sides of the metal interface line, so as to obtain the monitoring results of the welding process.

[0064] When two metal materials are welded together, there is a metal interface line between them. The two sides of the metal interface line correspond to the two different metal regions to be welded. By acquiring the pre-marked metal interface line in the thermal infrared image, it is easy to monitor the welding process.

[0065] In one embodiment, the metal interface line in the thermal infrared image is pre-calibrated in the following way: a visible light image of the weld bevel of the composite plate to be monitored before welding is acquired using a visible light camera, and the position of the metal interface line in the spatial coordinate system of the workpiece is determined by image calibration; the position of the metal interface line in the spatial coordinate system of the workpiece is mapped to the pixel coordinate system of the thermal infrared image to obtain the position coordinates of the metal interface line in the thermal infrared image, so as to obtain the pre-calibrated metal interface line in the thermal infrared image.

[0066] Industrial cameras and infrared thermal imagers can be mounted on the welding stand. A calibration board can be used to determine the external parameters of the industrial camera and infrared thermal imager. External parameters can be, for example, rotation matrices and translation vectors.

[0067] Before welding begins, such as before the root pass, an industrial camera can be used to photograph the bevel. For example, when welding titanium and steel to obtain a titanium-steel composite plate, due to the color and texture differences between titanium and steel or the physical steps of the bevel processing, the physical coordinates of the titanium-steel bonding line can be extracted from the visible light image using an edge detection algorithm.

[0068] Using a calibrated coordinate transformation matrix, the coordinates of the bonding line in the visible light image can be transformed into thermal infrared image coordinates, and the titanium-steel bonding line can be stored in memory as a fixed geometric mask for use during welding.

[0069] Before welding metal materials, their positions are usually fixed so that the position of the metal interface line during welding is the same as that before welding. However, due to thermal diffusion, the physical interfaces of different metals in infrared thermography are often covered by thermal radiation and become blurred, making it difficult to extract accurate physical boundaries directly from the infrared image. In contrast, visible light images can clearly distinguish material differences in a cold state. Through multi-sensor fusion calibration, the advantages of visible light positioning and infrared temperature measurement are complemented.

[0070] In one embodiment, the equivalent area of ​​the heat-affected zone on the metal side is equal to the sum of the membership degrees of all pixels relative to the cluster center located on the metal side; the monitoring results of the welding process are obtained by determining the thermal stress asymmetry index based on the difference in the equivalent area of ​​the heat-affected zones on different metal sides, and outputting a warning signal to indicate the imbalance of welding heat input when the thermal stress asymmetry index exceeds a preset safety threshold.

[0071] The composite plate to be monitored can be, for example, a titanium-steel composite plate, which can be obtained by combining titanium plates and steel plates. The heat-affected zones on different metal sides include the heat-affected zone on the steel side and the heat-affected zone on the titanium side.

[0072] Taking the composite plate to be monitored as a titanium-steel composite plate as an example, the image can be divided into the side where the steel is located and the other side where the titanium is located by using the pre-calibrated metal interface line.

[0073] The thermal stress asymmetry index is determined based on the difference in equivalent area of ​​the heat-affected zone on different metal sides. This can include: determining the difference in equivalent area of ​​the heat-affected zone on different metal sides, determining the sum of equivalent area of ​​the heat-affected zone on different metal sides, and using the ratio of the difference to the sum as the thermal stress asymmetry index.

[0074] The equivalent area of ​​the heat-affected zone on the metal side is equal to the sum of the membership degrees of all pixels relative to the cluster centers located on the metal side; that is, the equivalent area of ​​the heat-affected zone on the titanium side is equal to the sum of the membership degrees of all pixels relative to the cluster centers of the titanium heat-affected zone; the equivalent area of ​​the heat-affected zone on the steel side is equal to the sum of the membership degrees of all pixels relative to the cluster centers of the steel heat-affected zone.

[0075] For example, the thermal stress asymmetry index can be equal to... , as well as The equivalent areas of the heat-affected zones on different metal sides are shown in sequence. The preset safety threshold can be, for example, equal to 0.15. If the thermal stress asymmetry index is greater than the preset safety threshold, an alarm can be triggered, indicating that the heat input is seriously biased to one side.

[0076] The difficulty in welding titanium-steel composite plates lies in the significant difference in their thermophysical properties. For example, the thermal conductivity of steel is much greater than that of titanium. The welding process should adjust the heat source offset or waveform to achieve a relatively balanced heating on both sides in order to reduce residual stress.

[0077] If the ratio of the equivalent area of ​​the heat-affected zone on one side to the equivalent area of ​​the heat-affected zone on the other side exceeds a predetermined ratio, it indicates that heat is excessively accumulated on one side, which can easily lead to coarse grains or the formation of brittle phases in the obtained metal composite plate, affecting the performance of the welded metal composite plate.

[0078] Through iterative clustering of infrared thermal images, the obtained cluster centers are more accurate, and the membership degree of different pixels relative to different cluster centers can better characterize whether a pixel belongs to the category to which the cluster center belongs. Therefore, the sum of the membership degrees of all pixels relative to the cluster centers located on the metal side is used as the equivalent area of ​​the heat-affected zone on the metal side. The obtained equivalent area can be better used to compare the differences in the affected area of ​​different metal sides.

[0079] The thermal stress asymmetry index obtained in this application embodiment is constructed by using the difference ratio and sum of values, which can achieve normalization of the thermal equivalent area, eliminate the overall area scaling effect caused by the fluctuation of the total welding power, and make the thermal stress asymmetry index focus on the symmetry of thermal stress.

[0080] This application embodiment converts the information of thermal infrared images into a quantifiable thermal stress asymmetry index, which can intuitively and in real time reflect the internal thermal stress state of the welded metal composite plate. Operators can adjust the welding torch angle or current waveform according to the thermal stress asymmetry index, thereby effectively suppressing welding cracks and incomplete fusion defects.

[0081] Figure 2 This is a schematic diagram showing the clustering results of different locations of the composite plate in the clustering space, such as... Figure 2 As shown, the temperature assessment values ​​corresponding to different molten pools, titanium-affected zones, steel-affected zones, and base materials are different, which makes the categories of different clusters mainly distributed in different intervals of temperature assessment values, and the local spatial variance of different data points is different.

[0082] Figure 3 This is a schematic diagram comparing the thermal stress asymmetry index determined by the clustering results of this application embodiment with the thermal stress asymmetry index determined by the clustering results of fuzzy C-means clustering, as shown in the diagram. Figure 3 As shown, when the membership degree determined by the fuzzy C-means clustering algorithm is used directly to determine the equivalent area of ​​the heat-affected zone before optimization, the equivalent areas of different metal sides are 61171 and 53291, respectively, and the obtained thermal stress asymmetry index is equal to (61171-53291) / (61171+53291)=0.069.

[0083] When determining the equivalent area using the membership degree obtained from the clustering process in this application embodiment, a clearer division of the membership degree of different pixels in the welding area is achieved. This avoids determining the membership degree of pixels that actually belong to the molten pool or the base material as the titanium side influence area or the steel side influence area, making the sum of the membership degrees of the corresponding titanium side influence area or the steel side influence area smaller overall. The equivalent areas of the different metal sides determined are 48462 and 36741, respectively. The obtained thermal stress asymmetry index is equal to (48462-36741) / (48462+36741)=0.138, which is larger than the value of the thermal stress asymmetry index obtained by using fuzzy C-means clustering, and can improve the monitoring sensitivity of thermal stress asymmetry in the welding process.

[0084] Figure 4 This is a schematic diagram illustrating the structure of a composite plate welding process monitoring system 1000 according to an exemplary embodiment. (Refer to...) Figure 4 The composite plate welding process monitoring system 1000 includes a processor 1100 and a memory 1200. The memory 1200 stores computer program instructions, which, when executed by the processor 1100, implement all or part of the steps of the composite plate welding process monitoring method in this application.

[0085] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only.

[0086] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for monitoring the welding process of composite plates, characterized in that, include: Acquire thermal infrared images of the welding area of ​​the composite plate to be monitored, and determine the temperature assessment value corresponding to each pixel. Use the temperature assessment values ​​to determine multiple initial cluster centers from the thermal infrared images, including: sorting all pixels in the thermal infrared images according to the temperature assessment value from largest to smallest, and using the pixels corresponding to the multiple preset rankings as the initial multiple cluster centers corresponding to different welding influence areas; different welding influence areas include the molten pool area, the base material area, and the heat-affected zones on different metal sides; The first membership degree of a pixel relative to the cluster center is determined based on the difference in temperature evaluation values ​​of the pixel relative to the cluster center and the initial membership degree determined by fuzzy C-means clustering. The first membership degree is adjusted to obtain the second membership degree by utilizing the distribution of the first membership degree of other pixels in the neighborhood window of the pixel relative to the cluster center. The temperature assessment value of the cluster center is updated according to the second membership degree, and the cluster center of the thermal infrared image is iterated using the second membership degree of different pixels relative to different cluster centers to obtain the clustering result that makes the iteration process converge. The pre-labeled metal interface line in the thermal infrared image is obtained. The membership degree of the pixel points in the clustering results relative to the cluster centers of different metal sides of the metal interface line is used to determine the equivalent area of ​​the heat-affected zone on different metal sides, so as to obtain the monitoring results of the welding process. The first membership degree of a pixel relative to the cluster center is determined in the following way: ,in, For the first For the pixel point The first membership degree of each cluster center, where exp is the natural exponential function. For the first Temperature evaluation value of each pixel For the first Temperature assessment values ​​for each cluster center To determine the first clustering step using the fuzzy C-means clustering algorithm The pixel and the The initial membership degree between the cluster centers; The second membership degree is obtained by adjusting the first membership degree, including: ,in, For the first For the pixel point The second membership degree of each cluster center This indicates normalization processing. For the first The number of other pixels within the spatial neighborhood window of a pixel, excluding itself. For the first The first pixel For the nth neighboring pixel point The first membership degree of each cluster center. For the first The variance of the temperature evaluation values ​​of all pixels within the spatial neighborhood window of a pixel; The temperature assessment values ​​of the cluster centers are updated based on the second membership degree, including: ,in, For the first After the nth iteration Temperature assessment values ​​for each cluster center This represents the total number of pixels in the thermal infrared image. is the fuzzy coefficient.

2. The method for monitoring the welding process of composite plates according to claim 1, characterized in that, The temperature assessment value corresponding to a pixel is determined in the following way: For a target pixel in a thermal infrared image, the sum of the pixel values ​​of the target pixel in the red, green, and blue channels of the thermal infrared image is determined. The ratio of the pixel value of the target pixel in the red channel to the sum of the pixel values ​​is used as the temperature evaluation value of the target pixel.

3. The method for monitoring the welding process of composite plates according to claim 1, characterized in that, Whether the iterative process has converged is determined by the following method: The change in the temperature assessment value of the cluster center during the iteration process relative to the temperature assessment value of the cluster center in the previous iteration process is determined. If the change in the temperature assessment value of the cluster center in two consecutive iterations is less than the preset convergence threshold, the iteration process is considered to have converged.

4. The method for monitoring the welding process of composite plates according to claim 1, characterized in that, Metal interface lines in thermal infrared images are pre-defined in the following way: A visible light camera is used to acquire a visible light image of the weld bevel of the composite plate to be monitored before welding. The position of the metal interface line in the spatial coordinate system of the workpiece is determined by image calibration. The position of the metal interface line in the workpiece's spatial coordinate system is mapped to the pixel coordinate system of the thermal infrared image to obtain the position coordinates of the metal interface line in the thermal infrared image, thereby obtaining the pre-calibrated metal interface line in the thermal infrared image.

5. The method for monitoring the welding process of composite plates according to claim 1, characterized in that, The equivalent area of ​​the heat-affected zone on the metal side is equal to the sum of the membership degrees of all pixels relative to the cluster centers located on the metal side; the monitoring results of the welding process are obtained in the following way: Based on the difference in equivalent area of ​​the heat-affected zone on different metal sides, a thermal stress asymmetry index is determined. When the thermal stress asymmetry index exceeds a preset safety threshold, a warning signal is output to indicate an imbalance in welding heat input.

6. A composite plate welding process monitoring system, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the composite plate welding process monitoring method according to any one of claims 1-5.

Images