Welding method for precise heat dissipation tooth-shaped plate

Abstract

The invention relates to the technical field of metal material welding, and discloses a welding method for a precise heat dissipation tooth-shaped plate. The method comprises the following steps: carrying out primary welding on a precise heat dissipation tooth-shaped plate, collecting multi-source original sensing data of a welding joint, and carrying out multiple preprocessing and three-dimensional structure reconstruction and image enhancement on the welding joint; performing defect partition segmentation on the obtained three-dimensional defect fusion data, extracting various defect characteristic parameters from a welding abnormal region, performing comprehensive calculation and defect type judgment, performing heat conduction path calculation and thermal resistance value evaluation based on the obtained fine welding defect characteristic type and spatial position information, and predicting the heat dissipation performance. Generating a primary welding quality report; and according to the welding quality report and the associated database, welding process control parameters and welding optimization control are adjusted, and the heat dissipation tooth-shaped plate with the target welding quality is obtained. Accurate detection of welding defects of the heat dissipation tooth-shaped plate and closed-loop optimization of technological parameters are achieved, and the heat dissipation performance and the welding reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of metal material welding, and particularly relates to a welding method for a precision heat dissipation tooth-shaped plate. Background Art

[0002] In the field of heat dissipation of electronic devices, the precision heat dissipation tooth-shaped plate is a key component for the thermal management of high-performance electronic devices, and is widely used in 5G base stations, high-performance servers, and new energy vehicle electronic control systems. The welding quality of the precision heat dissipation tooth-shaped plate directly affects its heat dissipation performance and equipment reliability. Therefore, a high-quality welding method plays a decisive role in ensuring the stable operation of electronic devices and extending their service life.

[0003] Currently, the welding process of precision heat dissipation tooth-shaped plates mostly adopts a single process control mode based on empirical parameter settings, or attempts to use a single detection method for welding quality evaluation, and even begins to apply artificial intelligence technology to welding defect identification to improve the controllability of welding quality. However, these methods still face significant challenges in integrating multi-source detection data, processing micro-welding defect features, and adapting to complex tooth-shaped structures. Traditional welding methods for heat dissipation tooth-shaped plates often ignore key factors such as the influence mechanism of welding defects on the heat conduction path, the fusion strategy of multi-modal detection information, and the defect-thermal performance correlation, which have a decisive impact on welding quality evaluation and process optimization. That is, the existing welding methods for precision heat dissipation tooth-shaped plates lack accurate identification of micro-welding defects and quantitative evaluation of the influence on thermal performance, making it difficult to achieve closed-loop optimization of the welding process and efficient quality control, resulting in serious limitations in the heat dissipation efficiency and reliability of the heat dissipation tooth-shaped plate in the actual working environment, especially in thin-walled structures and high-density tooth sheet areas. Summary of the Invention

[0004] The main purpose of the present invention is to solve the problem that the existing welding methods for precision heat dissipation tooth-shaped plates lack accurate identification of micro-welding defects and quantitative evaluation of the influence on thermal performance, making it difficult to achieve closed-loop optimization of the welding process and efficient quality control, resulting in serious limitations in the heat dissipation efficiency and reliability of the heat dissipation tooth-shaped plate in the actual working environment, especially in thin-walled structures and high-density tooth sheet areas.

[0005] The first aspect of the present invention provides a welding method for a precision heat dissipation tooth-shaped plate, and the welding method for the precision heat dissipation tooth-shaped plate includes: Based on the preset welding process parameters, the precision heat dissipation tooth-shaped plate is initially welded, and multi-source original sensing data of the tooth-shaped plate welding joint in the precision heat dissipation tooth-shaped plate after the initial welding is collected, and the multi-source original sensing data is subjected to multiple preprocessings to obtain preprocessed multi-modal sensing data; the preprocessed multi-modal sensing data is used for the reconstruction of the three-dimensional structure of the welding joint and three-dimensional image enhancement to obtain three-dimensional defect fusion data, and the three-dimensional defect fusion data is subjected to various defect zone segmentations to obtain multiple types of welding abnormal regions; multiple preset welding defect characteristic parameters are extracted from the multiple types of welding abnormal regions to generate a multi-dimensional welding defect characteristic space, and the multi-dimensional welding defect characteristic space is subjected to defect characteristic comprehensive calculation and defect type discrimination to obtain the fine welding defect characteristic type and defect space position information; based on the fine welding defect characteristic type and the defect space position information, the heat conduction path of the precision heat dissipation tooth-shaped plate is calculated and the thermal resistance value is evaluated to obtain a heat dissipation performance prediction result, and based on the fine welding defect characteristic type, the defect space position information and the heat dissipation performance prediction result, a primary welding quality report of the precision heat dissipation tooth-shaped plate after the primary welding is generated; based on the primary welding quality report and the preset welding process parameter-defect-thermal performance correlation database, the welding process control parameters of the precision heat dissipation tooth-shaped plate are adjusted, and based on the welding process control parameters, the precision heat dissipation tooth-shaped plate is subjected to welding optimization control to obtain a precision heat dissipation tooth-shaped plate with the target welding quality.

[0006] Optionally, in the first implementation manner of the first aspect of the present invention, the multi-source original sensing data includes a ray projection image sequence, original ultrasonic scan signal data, and an original temperature field image sequence. Collecting the multi-source original sensing data of the tooth-shaped plate welding joint in the precision heat dissipation tooth-shaped plate after the initial welding includes: performing multi-angle X-ray scanning on the welding joint of the precision heat dissipation tooth-shaped plate to obtain a ray projection image sequence, performing high-frequency ultrasonic scanning on the welding joint of the precision heat dissipation tooth-shaped plate to obtain original ultrasonic scan signal data, and performing infrared thermal imaging on the surface of the precision heat dissipation tooth-shaped plate to obtain an original temperature field image sequence.

[0007] Optionally, in the second implementation manner of the first aspect of the present invention, the multiple pre-processings of the multi-source original sensing data to obtain pre-processed multi-modal sensing data include: filtering out noises of multiple frequency components and correcting circular artifacts in the ray projection image sequence to obtain enhanced ray projection data, performing Hilbert transform and gain compensation for signal propagation depth on the original ultrasonic scan signal data to obtain balanced ultrasonic echo signals, and performing pixel response correction and image dead pixel repair on the original temperature field image sequence to obtain calibrated temperature field data; synchronizing and integrating the enhanced ray projection data, the balanced ultrasonic echo signals, and the calibrated temperature field data in terms of time to obtain pre-processed multi-modal sensing data.

[0008] Optionally, in the third implementation manner of the first aspect of the present invention, the reconstruction of the three-dimensional structure of the welded joint and the enhancement of the three-dimensional image for the pre-processed multi-modal sensing data to obtain three-dimensional defect fusion data include: enhancing the image contrast of the ray projection data in the pre-processed multi-modal sensing data and arranging the spatial positions corresponding to the projection angles to obtain a three-dimensional projection data matrix, and performing voxel back-projection reconstruction calculation and metal material hardening artifact correction on the three-dimensional projection data matrix to obtain corrected three-dimensional structure data; performing phase correction and multi-point delay superposition on the ultrasonic echo signals in the pre-processed multi-modal sensing data to obtain enhanced ultrasonic signals, and arranging and integrating the enhanced ultrasonic signals based on a preset scanning position to obtain a high-resolution ultrasonic scan image; calculating the pixel displacement vector field between adjacent frames and compensating and adjusting the positions of each frame image for the temperature field data in the pre-processed multi-modal sensing data to obtain a temperature sequence with the influence of micro-vibrations eliminated, and performing multi-frame super-resolution integration on the temperature sequence with the influence of micro-vibrations eliminated to obtain a high-precision thermal field distribution map; performing spatial registration calculation on the corrected three-dimensional structure data, the high-resolution ultrasonic scan image, and the high-precision thermal field distribution map to obtain multi-modal data in a unified coordinate system, and performing multi-channel data fusion and feature complementary enhancement on the multi-modal data in the unified coordinate system to obtain three-dimensional defect fusion data.

[0009] Optionally, in the fourth implementation manner of the first aspect of the present invention, the multi-type welding abnormal area includes a micro-porosity type defect area, a micro-crack planar defect area, and a heat conduction abnormal area. The multi-type welding abnormal area is obtained by performing multi-defect partition segmentation on the three-dimensional defect fusion data, including: based on the structural area characteristics corresponding to the precision heat dissipation tooth-shaped plate, performing adaptive spatial area division on the three-dimensional defect fusion data and calculating the statistical characteristics of each spatial sub-area to obtain area adaptive segmentation parameters, and based on the area adaptive segmentation parameters, performing multi-threshold segmentation on each spatial sub-area and calculating the smooth transition of the area boundary to obtain a volume segmentation result with consistent spatial segmentation; performing three-dimensional connected domain analysis and marking on the volume segmentation result to obtain a set of candidate volume areas, and performing three-dimensional morphological operations and size filtering on the set of candidate volume areas to obtain a micro-porosity type defect area; calculating the interface reflection signal gradient of the corresponding ultrasonic feature component in the three-dimensional defect fusion data to obtain a planar feature edge map, and calculating the time temperature change rate of the corresponding thermal field feature component in the three-dimensional defect fusion data to obtain a temperature anomaly index map, and determining the edge connection and area closure of the planar feature edge map to obtain a micro-crack planar defect area, and performing region growing and boundary extraction on the temperature anomaly index map to obtain a heat conduction abnormal area; calculating the spatial coincidence degree of the micro-porosity type defect area, the micro-crack planar defect area, and the heat conduction abnormal area to obtain a multi-type welding abnormal area.

[0010] Optionally, in the fifth implementation manner of the first aspect of the present invention, the extraction of various preset welding defect characteristic parameters from the multi-type welding abnormal area to generate a multi-dimensional welding defect characteristic space includes: Calculate various pore geometric feature parameters for the micro-pore type defect area to obtain a pore geometric feature set, and perform three-dimensional principal axis calculation and moment of inertia calculation on the micro-pore type defect area to obtain a pore morphology feature set, and measure the positions of the corresponding weld center line and heat dissipation channel for the micro-pore type defect area to obtain a pore position feature set; calculate various crack geometric feature parameters for the micro-crack planar defect area to obtain a crack geometric feature set, and perform edge curvature and irregularity calculations on the micro-crack planar defect area to obtain a crack morphology feature set, and extract ultrasonic echo amplitude and phase features for the micro-crack planar defect area to obtain a crack acoustic feature set; perform numerical calculations of various thermal feature parameters on the heat conduction abnormal area to obtain a thermal feature parameter set, and calculate the heating rate and cooling time constant for the heat conduction abnormal area to obtain a thermal dynamic feature set, and perform a difference calculation on the heat conduction abnormal area based on a preset welding thermal field model to obtain a thermal resistance index feature set; calculate the corresponding position weight coefficients for the micro-pore type defect area, the micro-crack planar defect area, and the heat conduction abnormal area based on the device function partition corresponding to the precision heat dissipation tooth-shaped plate to obtain a region weight set, and perform weighted integration on the pore geometric feature set, the pore morphology feature set, the pore position feature set, the crack geometric feature set, the crack morphology feature set, the crack acoustic feature set, the thermal feature parameter set, the thermal dynamic feature set, and the thermal resistance index feature set based on the region weight set to construct a multi-dimensional welding defect feature space.

[0011] Optionally, in the sixth implementation manner of the first aspect of the present invention, the comprehensive calculation of defect features and discrimination of defect types for the multi-dimensional welding defect feature space to obtain the subtle welding defect feature types and defect space position information includes: normalizing the data of the multi-dimensional welding defect feature space to obtain a standardized feature vector, and calculating the signal-to-noise ratio and information entropy corresponding to the three modal data in the standardized feature vector to obtain a modal data reliability index; calculating the importance weights of each feature dimension in the standardized feature vector based on the modal data reliability index and the corresponding structural features of the precision heat dissipation tooth-shaped plate to obtain a feature weight coefficient, and performing a weighted calculation on the standardized feature vector based on the feature weight coefficient to obtain a weighted feature vector; calculating the inter-feature correlation degree of the weighted feature vector to obtain a feature correlation matrix, and extracting a feature subset greater than a preset complementarity degree and lower than a preset redundancy degree based on the feature correlation matrix to obtain an optimized feature vector; calculating the similarity of the optimized feature vector based on a preset defect typical feature template to obtain a defect type recognition result, and performing a matching combination on the defect type recognition result and the multi-dimensional welding defect feature space to obtain the subtle welding defect feature types and defect space position information.

[0012] Optionally, in the seventh implementation manner of the first aspect of the present invention, the calculation of the heat conduction path and the evaluation of the thermal resistance value for the precision heat dissipation tooth-shaped plate based on the subtle welding defect feature types and the defect space position information to obtain a heat dissipation performance prediction result includes: calculating the heat flow obstruction effect value for the pore-type defects in the subtle welding defect feature types to obtain a pore thermal resistance increment, and calculating the heat flow bypass effect value for the micro-crack-type defects in the subtle welding defect feature types to obtain a crack thermal resistance increment, and calculating the position weight of the defect space position information based on the tooth-shaped plate functional area distribution corresponding to the precision heat dissipation tooth-shaped plate to obtain a defect position weighting coefficient; performing a weighted calculation on the pore thermal resistance increment and the crack thermal resistance increment based on the defect position weighting coefficient to obtain a defect thermal resistance value, and performing a heat flow path perturbation calculation on the precision heat dissipation tooth-shaped plate based on the tooth-shaped plate heat flow channel distribution corresponding to the precision heat dissipation tooth-shaped plate and the defect thermal resistance value to obtain an actual heat flow channel distribution; performing a series-parallel combination calculation on the actual heat flow channel distribution to obtain the overall equivalent thermal resistance increase value of the precision heat dissipation tooth-shaped plate, and performing a simulation calculation of the working temperature field on the overall equivalent thermal resistance increase value to obtain a heat dissipation performance prediction result.

[0013] Optionally, in the eighth implementation manner of the first aspect of the present invention, generating a primary welding quality report for the precision heat dissipation tooth-shaped plate after primary welding based on the subtle welding defect feature type, the defect spatial position information, and the heat dissipation performance prediction result, including: performing three-dimensional visualization and color coding of the defect type on the subtle welding defect feature type and the defect spatial position information to obtain a defect visualization diagram, performing density statistical calculation on the defect spatial position information to obtain a defect density heat map, and performing a thermal field distribution mapping on the heat dissipation performance prediction result to obtain a heat performance influence distribution diagram; performing quantitative statistical calculation on the defect features in the defect visualization diagram to generate a defect characteristic score and a heat performance score, and performing weighted calculation on the defect characteristic score and the heat performance score to obtain the comprehensive quality score and quality grade of the precision heat dissipation tooth-shaped plate; integrating the defect visualization diagram, the defect density heat map, the heat performance influence distribution diagram, the comprehensive quality score, and the quality grade to obtain the primary welding quality report for the precision heat dissipation tooth-shaped plate after primary welding.

[0014] Optionally, in the ninth implementation manner of the first aspect of the present invention, adjusting the welding process control parameters of the precision heat dissipation tooth-shaped plate based on the primary welding quality report and a preset welding process parameter-defect-heat performance association database, including: based on the preset welding process parameter-defect-heat performance association database, performing an associated comparison of defect parameters on the primary welding quality report to obtain a set of similar defect cases, and performing parameter sensitivity calculation on the set of similar defect cases to obtain the process parameter adjustment direction for the corresponding defect type; based on the process parameter adjustment direction, performing parameter quantification calculation on the corresponding pore-type defects and micro-crack planar defects in the subtle welding defect feature type to obtain a defect suppression parameter set, and performing fusion parameter calculation on the heat conduction abnormal region based on the process parameter adjustment direction and the heat dissipation performance prediction result to obtain a heat performance optimization parameter set; performing a balance calculation of multi-objective control parameters and a constraint check and parameter adjustment of the welding equipment process window on the defect suppression parameter set and the heat performance optimization parameter set to obtain the welding process control parameters of the precision heat dissipation tooth-shaped plate.

[0015] The above-mentioned welding method for a precision heat dissipation tooth-shaped plate. In the embodiment of the present invention, the precision heat dissipation tooth-shaped plate is first subjected to primary welding with preset parameters, and multi-source sensing data of the welded joint is collected. These data are subjected to multiple pre-treatments to obtain multi-modal sensing data, and then three-dimensional structure reconstruction and image enhancement are performed to form three-dimensional defect fusion data. Subsequently, defect partitioning and segmentation are carried out to obtain multiple types of welding abnormal regions. Then, preset defect feature parameters are extracted from these abnormal regions to construct a multi-dimensional feature space. Through comprehensive calculation and type discrimination, the types and spatial position information of subtle welding defect features are obtained. Furthermore, based on the defect information, heat conduction path calculation and thermal resistance evaluation are carried out to predict the heat dissipation performance and generate a primary welding quality report. Finally, using this report in combination with the process parameter-defect-thermal performance correlation database, the welding process control parameters are adjusted and optimized welding is executed to obtain a precision heat dissipation tooth-shaped plate with the target quality. Through multi-level data processing and feature analysis, the accurate identification of micro-welding defects is realized. Especially in multi-modal data fusion and thermal performance evaluation, the influence mechanism of defects on heat dissipation performance is fully considered. And a closed-loop process optimization strategy is adopted, which not only realizes the identification of micro-defects but also establishes the correlation evaluation between defects and heat dissipation performance. In addition, through intelligent adjustment of process parameters, the welding quality is accurately controlled, thus overall realizing high-quality welding of precision heat dissipation tooth-shaped plates.

[0016] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the specification, claims, and drawings.

[0017] To make the above objectives, features, and advantages of the present invention more obvious and understandable, the following specifically enumerates preferred embodiments and, in conjunction with the accompanying drawings, provides a detailed description as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the first embodiment of the welding method for a precision heat dissipation tooth-shaped plate in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] As used in the embodiments of the present invention, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally further include other unlisted steps or units, or may optionally further include other steps or units inherent to these processes, methods, products, or devices.

[0021] For ease of understanding of this embodiment, the following describes the specific process of the embodiments of the present invention. Please refer to Figure 1 , the first embodiment of the welding method of the precision heat dissipation tooth-shaped plate in the embodiments of the present invention includes: 101. Based on preset welding process parameters, perform primary welding on the precision heat dissipation tooth-shaped plate, collect multi-source original sensing data of the tooth-shaped plate welding joint in the precision heat dissipation tooth-shaped plate after primary welding, and perform multiple preprocessings on the multi-source original sensing data to obtain preprocessed multi-modal sensing data; The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is a theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results.

[0022] Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. Artificial intelligence software technologies mainly include several major directions such as computer vision technology, robotics, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0023] In this embodiment, the multi-source original sensing data includes a ray projection image sequence, original ultrasonic scan signal data, and an original temperature field image sequence. An X-ray scan of the welded joint of the precision heat dissipation tooth-shaped plate is performed at multiple angles to obtain a ray projection image sequence, and a high-frequency ultrasonic scan of the welded joint of the precision heat dissipation tooth-shaped plate is performed to obtain original ultrasonic scan signal data, and an infrared thermal image of the surface of the precision heat dissipation tooth-shaped plate is collected to obtain an original temperature field image sequence; noise filtering of multiple frequency components and correction of circular artifacts are performed on the ray projection image sequence to obtain enhanced ray projection data, and Hilbert transform and gain compensation for signal propagation depth are performed on the original ultrasonic scan signal data to obtain balanced ultrasonic echo signals, and pixel response correction and image dead pixel repair are performed on the original temperature field image sequence to obtain calibrated temperature field data; time synchronization integration is performed on the enhanced ray projection data, balanced ultrasonic echo signals, and calibrated temperature field data to obtain preprocessed multi-modal sensing data.

[0024] In practical applications, during the initial welding stage, according to the material properties, structural complexity, and heat dissipation function requirements of the precision heat dissipation tooth-shaped plate, appropriate welding process parameters are preset in advance, including key parameters such as laser power, welding speed, focal position, and shielding gas ratio. And during the initial welding of the precision heat dissipation tooth-shaped plate, it is necessary to pay attention to the control of energy input at the thin-wall structure and high-density tooth slices to avoid functional damage caused by excessive melting or thermal deformation, so as to complete the initial welding stage of the precision heat dissipation tooth-shaped plate. Then, after welding is completed, the internal and external characteristics of the welded joint are comprehensively captured through three detection methods with different physical mechanisms. That is, first, a CT system composed of a micro-focus X-ray source and a high-resolution detector is deployed to perform a 360° rotational scan of the welded joint at an angular step of 0.5°. Multiple frames of images are collected at each angular position to improve the signal-to-noise ratio and effectively cope with the signal occlusion problem that may be caused by high-density areas with tooth pitch ≤ 2 mm, thereby obtaining a complete sequence of X-ray projection images. And a high-frequency focused phased array ultrasonic probe with a center frequency of 50 MHz is configured, the pulse repetition frequency is set to 2 kHz, the sampling rate reaches 500 MSa / s, and the scanning step is controlled within 50 μm to ensure that tiny weld defects that may exist in the thin-wall structure with a thickness of less than 0.5 mm can be accurately detected, and high-quality original ultrasonic scan signal data is obtained. And an infrared thermal imager with a thermal sensitivity better than 0.05 °C is used to collect the temperature field distribution on the surface of the heat dissipation tooth-shaped plate at a high frame rate of 60 fps under standard temperature difference conditions (ΔT = 30 °C), record the temperature change characteristics of the abnormal heat conduction area, and form a sequence of original temperature field images. Then, targeted preprocessing is performed on the collected original data. Among them, for the X-ray projection image sequence, wavelet decomposition technology is used to decompose the image into different frequency components, and high-frequency quantum noise and low-frequency circular artifacts are processed separately, and the enhanced image is restored through wavelet reconstruction, significantly improving the signal-to-noise ratio from the initial 15 dB to 25 dB, making defects such as tiny pores and lack of fusion more clearly visible. For the original ultrasonic scan signal data, the Hilbert transform is applied to extract the signal envelope, and combined with the attenuation characteristics of the sound wave propagation depth in the material, an adaptive time gain compensation algorithm is designed to balance the echo signal intensities at different depths and ensure the effective detection of defects such as deep microcracks. For the original temperature field image sequence, pixel response non-uniformity correction is performed to eliminate the inherent response differences of the sensor, and at the same time, bad pixels and noise in the image are identified and repaired, and the infrared radiation intensity is accurately converted into the actual temperature value through the temperature calibration curve. Then, based on the time-stamp synchronization trigger device of the PTP protocol, the time error of the three detection devices during data collection is controlled within 1 ms, and the data in different coordinate systems is mapped to a unified reference system through a spatial coordinate conversion algorithm to achieve perfect integration of the data and obtain a multi-modal sensing data set.

[0025] 102. Reconstruct the three-dimensional structure of the welded joint and enhance the three-dimensional image for the preprocessed multi-modal sensing data to obtain three-dimensional defect fusion data, and perform multi-defect partition segmentation on the three-dimensional defect fusion data to obtain multiple types of welding abnormal regions; In this embodiment, image contrast enhancement and spatial arrangement of the corresponding projection angles and positions are performed on the ray projection data in the preprocessed multi-modal sensing data to obtain a three-dimensional projection data matrix, and voxel back-projection reconstruction calculation and metal material hardening artifact correction are performed on the three-dimensional projection data matrix to obtain corrected three-dimensional structure data; phase correction and multi-point delay superposition are performed on the ultrasonic echo signals in the preprocessed multi-modal sensing data to obtain enhanced ultrasonic signals, and based on the preset scanning positions, the enhanced ultrasonic signals are arranged and integrated to obtain a high-resolution ultrasonic scan image; pixel displacement vector field calculation between adjacent frames and position compensation adjustment of each frame image are performed on the temperature field data in the preprocessed multi-modal sensing data to obtain a temperature sequence with the influence of micro-vibrations eliminated, and multi-frame super-resolution integration is performed on the temperature sequence with the influence of micro-vibrations eliminated to obtain a high-precision thermal field distribution map; spatial registration calculation is performed on the corrected three-dimensional structure data, high-resolution ultrasonic scan image, and high-precision thermal field distribution map to obtain multi-modal data in a unified coordinate system, and multi-channel data fusion and feature complementary enhancement are performed on the multi-modal data in the unified coordinate system to obtain three-dimensional defect fusion data; based on the structural region characteristics corresponding to the precision heat dissipation tooth-shaped plate, adaptive spatial region division is performed on the three-dimensional defect fusion data and statistical characteristics of each spatial sub-region are calculated to obtain region adaptive segmentation parameters, and based on the region adaptive segmentation parameters, multi-threshold segmentation and smooth transition calculation of the region boundaries are performed on each spatial sub-region to obtain a volume segmentation result with consistent spatial segmentation; three-dimensional connected component analysis and labeling are performed on the volume segmentation result to obtain a set of candidate volume regions, and three-dimensional morphological operations and size filtering are performed on the set of candidate volume regions to obtain a micro-porosity type defect region; interface reflection signal gradient calculation is performed on the corresponding ultrasonic feature components in the three-dimensional defect fusion data to obtain a planar feature edge map, and time-temperature change rate calculation is performed on the corresponding thermal field feature components in the three-dimensional defect fusion data to obtain a temperature anomaly index map, and edge connection and region closure determination are performed on the planar feature edge map to obtain a micro-crack planar defect region, and region growing and boundary extraction are performed on the temperature anomaly index map to obtain a heat conduction anomaly region; spatial overlap calculation is performed on the micro-porosity type defect region, micro-crack planar defect region, and heat conduction anomaly region to obtain multiple types of welding abnormal regions.

[0026] In practical applications, first, frequency domain enhancement technology is adopted. The ray projection data in the preprocessed multi-modal sensing data is converted to the frequency domain through Fourier transform. The intermediate frequency components (usually containing edge and detail information) are selectively enhanced, while high-frequency noise and low-frequency background variations are suppressed, effectively enhancing the image details. And they are precisely arranged and organized according to the angle and position information during their acquisition to form a complete three-dimensional projection data matrix (this matrix contains all the information of the welded joint observed from different angles). Then, using the improved Feldkamp-Davis-Kress (FDK) algorithm, the two-dimensional projection data is reconstructed into the initial three-dimensional volume data through voxel back-projection calculation technology. During the reconstruction process, special attention needs to be paid to solving the hardening artifact problem commonly found in the X-ray imaging of the heat dissipation tooth-shaped plate metal material. Through iterative multi-energy spectrum modeling and physical correction calculation, the energy spectrum change when the X-ray beam passes through metal materials with different thicknesses is accurately simulated, thereby eliminating the stripes and dark areas caused by hardening artifacts, and finally obtaining the corrected high-quality three-dimensional structure data (wherein, this data can clearly present the internal microstructure of the welded joint, including pores as small as 20 microns in diameter and fine unfused areas). And for the ultrasonic echo signals, by calculating the cross-correlation function between each signal, the phase lag amount is accurately determined, and the original signals are corrected to make all signals phase-consistent. Then, multi-point delay superposition processing is carried out. The echo signals of the same reflector collected by probes at different positions are aligned and superimposed in the time domain, significantly improving the signal-to-noise ratio and spatial resolution to obtain enhanced ultrasonic signals (even tiny crack reflections can be clearly captured by these signals). Then, according to the accurate position information during the ultrasonic probe scanning, the enhanced one-dimensional A-scan signals are arranged and integrated according to their spatial acquisition coordinates, and the signal values between sampling points are filled through two-dimensional and three-dimensional interpolation algorithms, and finally high-resolution B-scan and C-scan ultrasonic images (that is, high-resolution ultrasonic scanning images) are generated (these images can visually display the planar defects inside the welded joint, such as unfused, interlayer separation, and micro-cracks, and the detection ability reaches 50 microns). And for the temperature field data, the optical flow method is used to calculate the pixel displacement vector field between adjacent frames. By solving the corresponding brightness constancy equation, the motion vector of each pixel point is obtained. Then, precise position compensation and adjustment are carried out for the subsequent frames to eliminate the vibration influence and ensure the spatial consistency of the temperature field data. And the temperature sequence after position compensation is then input into the multi-frame super-resolution reconstruction algorithm. Using the complementary information contained in the multi-frame low-resolution images, through sub-pixel-level image registration and iterative back-projection, a high-precision thermal field distribution map with a resolution increased by 1 - 2 times is generated.Furthermore, feature points or landmark structures in each modality data (i.e., the corrected three-dimensional structure data, high-resolution ultrasonic scan images, and high-precision thermal field distribution maps) are identified, such as weld edges, heat dissipation teeth of specific shapes, etc. Then, a non-rigid registration method based on maximizing mutual information is used to calculate the spatial mapping function between different modalities. This function takes into account possible non-linear deformations and is described using a B-spline transformation model. The registration accuracy is controlled within 10 microns to ensure that the three modality data can accurately correspond in space. Furthermore, an adaptive weighting strategy is adopted to generate three-dimensional defect fusion data by fusing the three modality data into a multi-channel three-dimensional data volume, where each spatial position contains X-ray density values, ultrasonic echo intensities, and temperature field information, forming complementary feature descriptions. The fusion value calculation formula for each voxel point is:; (wherein, ); Wherein, represents the fusion value at position , is the normalized data of the i-th modality at position , is the weight coefficient of the i-th modality (a weight coefficient adaptively calculated based on signal-to-noise ratio and feature significance) (for example, in the micro-porosity area of a precision heat dissipation tooth-shaped plate, the weight of X-ray CT data is usually set to 0.6 - 0.7, while at the weld interface, the weight of ultrasonic data can be increased to 0.5 - 0.6 to ensure the best detection effect for various defects); and a feature-level fusion method is adopted. Through techniques such as principal component analysis (PCA) and independent component analysis (ICA), the key features of each modality data are extracted and weighted combined to enhance the defect features with common indication and suppress the noise interference of a single modality, finally generating three-dimensional defect fusion data with high signal-to-noise ratio and rich features.

[0027] Secondly, according to the structural design drawings and actual scanned morphology of the heat dissipation tooth-shaped plate, the three-dimensional defect fusion data is adaptively divided into spatial regions, and the entire data volume is divided into multiple sub-regions, including tooth slice regions, welded joint regions, substrate regions, etc. Each sub-region has relatively consistent material properties and structural characteristics, and the gray statistical characteristics (such as mean, variance, skewness, kurtosis) and gradient distribution characteristics (such as gradient amplitude histogram, gradient direction distribution) of each sub-region are calculated. These characteristics reflect the uniformity and boundary characteristics of the materials within the region, and then the adaptive segmentation parameters optimized for each sub-region are obtained, such as the optimal threshold, region growing seed points, etc. Based on these adaptive parameters, different segmentation strategies are adopted for each sub-region. For example, the multi-threshold OTSU algorithm is used for thin-walled regions, the region growing method is used for the tooth root transition region, and the morphological watershed algorithm is used for the weld center region, etc. The differences in regional characteristics are fully considered to improve the accuracy of segmentation. In addition, to solve the possible discontinuity problems in the sub-region segmentation results, a spatial weighted smoothing transition algorithm is adopted at the regional boundaries. The transition weights are designed through a distance function to smoothly fuse the segmentation results of adjacent regions to obtain a volume segmentation result with consistent spatial segmentation. Then, a three-dimensional labeling algorithm is used to assign a unique identifier to each spatially connected region in the volume segmentation result to generate a set of candidate volume regions, and three-dimensional morphological operations are performed on these candidate regions, that is, an opening operation (erosion first and then dilation) is performed using a spherical structural element with a radius of 2 voxels to effectively remove noise points and small-sized false positive regions. Then, a closing operation (dilation first and then erosion) is performed using a spherical structural element with a radius of 3 voxels to fill the small holes that may exist inside the defect and make the defect morphology more complete. Thus, according to the typical size characteristics of the welding defects of the heat dissipation tooth-shaped plate, multi-dimensional filtering conditions such as volume, surface area, sphericity, etc. are set, and only the regions that meet the pore characteristics are retained, and finally, an accurate segmentation result of the micro-pore type defect region is obtained. Furthermore, for the ultrasonic feature components in the three-dimensional defect fusion data, the spatial gradient of the reflected signal is calculated. Since planar defects (such as cracks, lack of fusion) will produce significant acoustic impedance mutations in the ultrasonic propagation direction, they appear as high-value regions in the gradient map, forming a planar feature edge map. And an edge connection algorithm based on tensor voting is adopted to utilize the consistency of the local structure tensor to complement the missing edge parts, and through the discrimination of regional closure, the edge groups forming closed regions are identified, and finally, the micro-crack planar defect region is determined. And for the thermal field feature components, a time-temperature change rate analysis is performed, and the derivative of the temperature of each voxel point with respect to time is calculated to generate a temperature anomaly index map. Since the heat conduction efficiency of the defect region is lower than that of the normal region, it will show a different temperature change pattern from the surrounding regions during the thermal field change process. Therefore, a region growing algorithm is adopted, starting from the points with abnormally high temperature values, gradually expanding the regions containing similar temperature change characteristics, and the exact boundary of the heat conduction anomaly region is determined through a boundary extraction algorithm;Furthermore, for the already identified small pore-type defect regions, micro-crack planar defect regions, and heat conduction anomaly regions, by calculating the ratios of the spatial intersections and unions of different types of defect regions, high-confidence defect regions jointly indicated by multiple detection methods are identified, and at the same time, the reliability of single-modal detection results is evaluated. For regions with highly overlapping spatial positions, a higher defect confidence level is assigned; for regions detected only by a single modality, additional feature analysis is performed to confirm their authenticity. Thus, through this multi-modal cross-validation mechanism, multi-type welding anomaly regions are finally obtained.

[0028] 103. Extract multiple preset welding defect characteristic parameters for the multi-type welding anomaly regions to generate a multi-dimensional welding defect characteristic space, and perform comprehensive calculation of defect characteristics and discrimination of defect types on the multi-dimensional welding defect characteristic space to obtain the types of fine welding defect characteristics and defect spatial position information; In this embodiment, various pore geometric feature parameters are calculated for the micro-pore type defect area to obtain a pore geometric feature set, and the three-dimensional major axis calculation and moment of inertia calculation are performed on the micro-pore type defect area to obtain a pore morphology feature set, and the position measurement of the corresponding weld center line and heat dissipation channel is performed on the micro-pore type defect area to obtain a pore position feature set; various crack geometric feature parameters are calculated for the micro-crack planar defect area to obtain a crack geometric feature set, and the edge curvature and irregularity are calculated for the micro-crack planar defect area to obtain a crack morphology feature set, and the ultrasonic echo amplitude and phase characteristics are extracted for the micro-crack planar defect area to obtain a crack acoustic feature set; numerical calculations of various thermal feature parameters are performed on the thermal conduction abnormal area to obtain a thermal feature parameter set, and the heating rate and cooling time constant are calculated for the thermal conduction abnormal area to obtain a thermal dynamic feature set, and based on a preset welding thermal field model, the difference degree is calculated for the thermal conduction abnormal area to obtain a thermal resistance index feature set; based on the device function partition corresponding to the precision heat dissipation tooth-shaped plate, the position weight coefficients corresponding to the micro-pore type defect area, the micro-crack planar defect area and the thermal conduction abnormal area are calculated to obtain a regional weight set, and based on the regional weight set, the pore geometric feature set, the pore morphology feature set, the pore position feature set, the crack geometric feature set, the crack morphology feature set, the crack acoustic feature set, the thermal feature parameter set, the thermal dynamic feature set and the thermal resistance index feature set are weighted and integrated to construct a multi-dimensional welding defect feature space; data normalization is performed on the multi-dimensional welding defect feature space to obtain a standardized feature vector, and the signal-to-noise ratio and information entropy corresponding to the three modal data in the standardized feature vector are calculated to obtain a modal data reliability index; based on the modal data reliability index and the structural characteristics corresponding to the precision heat dissipation tooth-shaped plate, the importance weights of each feature dimension in the standardized feature vector are calculated to obtain a feature weight coefficient, and based on the feature weight coefficient, the standardized feature vector is weighted and calculated to obtain a weighted feature vector; the feature correlation degree is calculated for the weighted feature vector to obtain a feature correlation matrix, and based on the feature correlation matrix, a feature subset greater than the preset complementarity degree and lower than the preset redundancy degree is extracted to obtain an optimized feature vector; based on a preset defect typical feature template, the similarity is calculated for the optimized feature vector to obtain a defect type recognition result, and the defect type recognition result and the multi-dimensional welding defect feature space are matched and combined to obtain the fine welding defect feature type and defect space position information.

[0029] In practical applications, first, for the micro-porosity defect area, a series of pore geometric feature parameters are calculated using 3D image processing techniques, including volume (obtained by multiplying the voxel count by the voxel volume), surface area (calculated by the mosaic surface algorithm or triangular mesh approximation method), equivalent diameter (the diameter of an equal-volume sphere calculated from the volume), maximum inscribed sphere diameter, and minimum circumscribed sphere diameter, etc. These parameters comprehensively describe the spatial size characteristics of the pores, constituting the pore geometric feature set. And the main extension direction of the pores in space is determined by calculating the 3D principal axis, which is based on the eigen-decomposition of the covariance matrix of the pore area, obtaining three orthogonal principal directions and their corresponding eigenvalues (these eigenvalues reflect the extension degree of the pores in the three principal axis directions), and calculating the 3D moment of inertia of the pores. Shape factors such as sphericity (the ratio of the maximum eigenvalue to the minimum eigenvalue), flatness, and elongation are derived from these parameters. These complex morphological features can effectively distinguish spherical pores, ellipsoidal pores, and inclusions with irregular shapes, forming the pore morphological feature set. And the welding defects are directly associated with the product function. The distance and depth of the pores relative to the weld centerline are measured, and the Euclidean distance from the pores to the nearest heat dissipation channel is calculated to evaluate the potential impact of the pores on the heat dissipation function. At the same time, the relative position of the pores in the weld cross-section (such as the root, middle, or crown) is analyzed, and these position features constitute the pore position feature set. Furthermore, for the micro-crack planar defect, the geometric feature parameters of the crack are calculated, including area (obtained by the projected area or 3D surface integral), perimeter, major axis length, minor axis length, major-to-minor axis ratio (reflecting the slenderness of the crack), and average thickness, etc. These parameters constitute the crack geometric feature set. And uniform sampling is carried out along the crack edge to calculate the local curvature change and generate curvature statistical features such as curvature mean, standard deviation, maximum value, and minimum value. These features can distinguish linear cracks and cracks with irregular shapes, and the irregularity index of the crack is calculated, including shape complexity (the ratio of the actual perimeter to the perimeter of an equal-area circle), fractal dimension (reflecting the complexity of the boundary), and concavity-convexity (the ratio of the crack area to the area of its convex hull). These morphological features constitute the crack morphological feature set. And for the ultrasonic signal features corresponding to the crack area, the amplitude features (such as maximum amplitude, average amplitude, integral intensity) and phase features (such as phase jump, phase consistency) of the ultrasonic echo are extracted. These acoustic features can reflect the acoustic impedance characteristics and internal structure of the crack, constituting the crack acoustic feature set;Furthermore, for the regions with abnormal heat conduction, a thermodynamic analysis method is used to extract various thermal characteristic parameters, including the maximum temperature of the region, the average temperature, the standard deviation of temperature, the temperature difference between the region and the surrounding normal region, and the temperature gradient (calculating the spatial derivative of temperature by the central difference method). These static thermal characteristics constitute the thermal characteristic parameter set. And the dynamic performance of heat conduction is analyzed by using time-series temperature data, calculating the heating rate (the maximum value of the time derivative of temperature rise during the heating stage) and the cooling time constant (the time required for the temperature to drop to 63.2% of the initial value during the cooling stage) of each voxel point. These dynamic parameters reflect the heat capacity and thermal conductivity characteristics of the region and constitute the thermal dynamic characteristic set. And based on the pre-established ideal welding thermal field model (a temperature distribution model under defect-free conditions constructed by the finite element method), the root mean square error, correlation coefficient, and structural similarity index between the measured temperature field and the theoretically predicted temperature field are calculated. These difference degree indicators directly quantify the severity of abnormal heat conduction and constitute the thermal resistance index characteristic set. Furthermore, based on the device functional partition design of the precision heat dissipation tooth-shaped plate (such as the chip direct contact area, the main heat dissipation channel, the secondary heat dissipation channel, and the non-critical area), the functional importance weights of the positions where various defects are located are calculated. That is, the highest weight (such as 0.9 - 1.0) is given to the defects in the region directly in contact with the heat source, followed by the main heat dissipation channel region (such as 0.7 - 0.9), then the secondary heat dissipation channel region (such as 0.5 - 0.7), and the lowest for the non-critical area (such as 0.3 - 0.5). This position weight evaluation based on function constitutes the regional weight set. Subsequently, based on these weight values, all the characteristic sets extracted in the previous steps are weighted and integrated. The pore geometric characteristic set, pore morphology characteristic set, pore position characteristic set, crack geometric characteristic set, crack morphology characteristic set, crack acoustic characteristic set, thermal characteristic parameter set, thermal dynamic characteristic set, and thermal resistance index characteristic set are systematically organized. Each characteristic value is multiplied by the corresponding position weight to form a high-dimensional vector space, that is, the multi-dimensional welding defect characteristic space (where this space comprehensively describes all aspects of welding defects and their potential impacts).;

[0030] Secondly, the Z-score normalization method is used to normalize the data in the multi-dimensional welding defect feature space, so as to convert the feature variables with different physical dimensions and numerical ranges to a unified scale (usually the interval of [0, 1] or [-1, 1]), obtaining the standardized feature vector. By calculating the signal-to-noise ratio (the ratio of signal intensity to background noise) and information entropy (the amount of information contained in the data) of the X-ray, ultrasonic and infrared thermography data respectively, the clarity and information richness of each modality data are reflected, constituting the reliability index of the modality data. Furthermore, based on the above reliability index and the unique structural characteristics of the heat dissipation tooth-shaped plate (such as thin-wall structure, high-density teeth, etc.), considering the discrimination ability (quantified by Fisher discriminant ratio or information gain), stability (evaluated by cross-validation) and relevance to product function (set based on domain knowledge) of each feature, the importance weights of each feature dimension in the standardized feature vector are calculated. Finally, the feature weight coefficients are generated, and these weight coefficients are used to perform weighted calculation on the standardized feature vector to obtain the weighted feature vector (this vector can better reflect the key characteristics of welding defects and their potential impact on product function). Then, the correlation degree between features of the weighted feature vector is calculated. By calculating the Pearson correlation coefficient, Spearman rank correlation coefficient or mutual information between each pair of features, a feature correlation matrix is constructed. Based on this correlation matrix, a feature subset with high complementarity (correlation coefficient lower than a preset threshold, such as 0.3) and low redundancy (not highly correlated with the selected features, such as correlation coefficient lower than 0.7) is selected to form an optimized feature vector (this vector contains an information-rich and mutually independent feature combination, greatly reducing the computational complexity and improving the accuracy of subsequent classification). Then, based on the pre-established defect typical feature template library (a set of feature patterns of various typical defects constructed through a large amount of historical data and expert knowledge), the optimized feature vector is subjected to similarity matching, that is, calculating the Euclidean distance, cosine similarity or Mahalanobis distance between the current feature vector and various typical defect templates, identifying the most matching defect type, and estimating the credibility of the matching, so as to obtain the defect type recognition result. This defect type recognition result is then matched and combined with the position information in the multi-dimensional welding defect feature space to obtain a complete description including the defect type, position, size, shape and severity, that is, the subtle welding defect feature type and the defect space position information.

[0031] 104. Based on the subtle welding defect feature type and the defect space position information, calculate the heat conduction path and evaluate the thermal resistance value of the precision heat dissipation tooth-shaped plate to obtain the heat dissipation performance prediction result. Based on the subtle welding defect feature type, the defect space position information and the heat dissipation performance prediction result, generate the initial welding quality report of the precision heat dissipation tooth-shaped plate after the initial welding. In this embodiment, the heat flow obstruction effect value of the porosity-type defects among the fine welding defect feature types is calculated to obtain the thermal resistance increment of the pores, and the heat flow bypass effect value of the micro-crack-type defects among the fine welding defect feature types is calculated to obtain the crack thermal resistance increment. Based on the distribution of the tooth-shaped plate functional areas corresponding to the precision heat dissipation tooth-shaped plate, the position weight calculation of the defect spatial position information is carried out to obtain the defect position weighting coefficient; based on the defect position weighting coefficient, the thermal resistance increment of the pores and the crack thermal resistance increment are weighted and calculated to obtain the defect thermal resistance value. Based on the heat flow channel distribution of the precision heat dissipation tooth-shaped plate corresponding to the defect thermal resistance value, the heat flow path perturbation calculation of the precision heat dissipation tooth-shaped plate is carried out to obtain the actual heat flow channel distribution; the series-parallel combination calculation of the actual heat flow channel distribution is carried out to obtain the overall equivalent thermal resistance increment of the precision heat dissipation tooth-shaped plate, and the simulation calculation of the working temperature field of the overall equivalent thermal resistance increment is carried out to obtain the heat dissipation performance prediction result; the three-dimensional visualization and color coding of the defect type of the fine welding defect feature type and the defect spatial position information are carried out to obtain the defect visualization diagram, and the density statistical calculation of the defect spatial position information is carried out to obtain the defect density heat map, and the heat field distribution mapping of the heat dissipation performance prediction result is carried out to obtain the heat performance influence distribution map; the quantitative statistical calculation of the defect features in the defect visualization diagram is carried out to generate the defect characteristic score and the heat performance score, and the defect characteristic score and the heat performance score are weighted and calculated to obtain the comprehensive quality score and quality grade of the precision heat dissipation tooth-shaped plate; the defect visualization diagram, the defect density heat map, the heat performance influence distribution map, the comprehensive quality score and the quality grade are reported and integrated to obtain the primary welding quality report of the precision heat dissipation tooth-shaped plate after the primary welding.

[0032] In practical applications, first, for the porosity-type defects among the fine welding defect feature types, the quantitative calculation of the heat flow obstruction effect is carried out by using the heat conduction theory. The pores are regarded as cavity obstacles on the heat flow path, and their obstruction effect on the heat flow is analyzed through the heat conduction equation and the equivalent medium model. Therefore, the geometric characteristics of the pores (such as volume, shape, and distribution) and the thermal conductivity of the material where they are located need to be considered in the calculation process, that is, it is calculated based on the modified Maxwell-Eucken equation. The formula for the thermal resistance increment of the pores is as follows: ; Where is the thermal resistance increment caused by the pores, V is the pore volume, ρ is the pore density, is the ideal thermal conductivity, is the equivalent thermal conductivity, is a calibrated coefficient, and finally the thermal resistance increment caused by pores, i.e., the pore thermal resistance increment (this parameter directly quantifies the negative impact of pores on the heat dissipation performance); and the precise calculation of the heat flow bypass effect for microcrack defects among the types of fine welding defect characteristics, based on the geometric characteristics of the crack (such as area, length, width, thickness, and direction), a thermal resistance calculation model is established, and its crack thermal resistance increment formula is: ; where L is the crack length, W is the crack width, D is the thickness in the heat flow direction, θ is the angle between the crack and the main heat flow direction, is an empirical coefficient. The blocking effect is the largest when the crack is perpendicular to the heat flow direction and the smallest when it is parallel to the heat flow direction, to calculate the thermal resistance increment caused by the crack, i.e., the crack thermal resistance increment; and based on the functional area distribution of the precision heat dissipation tooth-shaped plate, a weight evaluation of the defect position is carried out, that is, according to the heat flow channel distribution and functional area division designed by the tooth-shaped plate (such as the chip contact area, the main heat dissipation channel, the secondary heat dissipation channel, and the non-critical area), a position weight function w(x, y, z) is established, and this function maps the three-dimensional space to a weight value. For example, the weight of the area directly in contact with the heat source is the highest (0.9 - 1.0), followed by the main heat dissipation channel (0.7 - 0.9), then the secondary heat dissipation channel (0.5 - 0.7), and the non-critical area is the lowest (0.3 - 0.5). Applying this weight function to the defect spatial position information, the position weighting coefficient of each defect is obtained; furthermore, based on the above position weighting coefficient, a weighted calculation of the pore thermal resistance increment and the crack thermal resistance increment is carried out, and the formula is: , where is the increment of thermal resistance (which can be the increment of pore thermal resistance or crack thermal resistance), so as to obtain the defect thermal resistance considering the influence of position, making the defects on the critical heat flow path receive higher attention, while the influence weight of the defects in the non-critical area is appropriately reduced, thus realizing the differential evaluation of the heat dissipation function. Then, these weighted defect thermal resistances are combined with the ideal heat flow channel distribution model of the precision heat dissipation tooth-shaped plate, and through the heat flow network analysis technology, the redistribution of heat flow under the condition of the existence of defects is calculated, the changes of the temperature of each node and the heat flow path are calculated, and finally the actual heat flow channel distribution considering the influence of defects is obtained (showing how the heat flow bypasses or passes through the defect area and the influence degree of the defects on the overall heat conduction efficiency); furthermore, the thermal resistance network analysis of the entire heat dissipation tooth-shaped plate is carried out, that is, for multiple defects distributed in series on the heat flow path, their thermal resistances are directly added; for defects distributed in parallel, the composite thermal resistance is calculated by adding the reciprocals and then taking the reciprocal; for a complex network with both series and parallel connections, the node analysis method or mesh analysis method of network theory is used to solve, and finally the overall equivalent thermal resistance increment of the precision heat dissipation tooth-shaped plate is calculated, which represents the degree of decline in heat dissipation performance caused by the combined action of all defects. Based on the equivalent thermal resistance increment and combined with the working environmental conditions of the tooth-shaped plate (such as ambient temperature, heat source power, etc.), the computational fluid dynamics technology is applied to accurately simulate the working temperature field and predict the influence of defects on the temperature distribution in the actual working state, including key indicators such as the maximum temperature increase, the change of temperature uniformity and the transfer of hot spot position. These simulation results constitute the heat dissipation performance prediction results.

[0033] Secondly, volume rendering and surface rendering techniques are used to precisely reconstruct the defects in three-dimensional space, and different color coding schemes are applied according to the defect types. For example, red represents pores, blue represents cracks, yellow represents lack of fusion, etc. At the same time, the severity of the defects or the reliability of the detection is represented by the change in transparency, and finally an intuitive defect visualization map is generated; and density statistical analysis is carried out on the defect spatial position information. The kernel density estimation method is used to calculate the distribution density of the defects in three-dimensional space, and the result is visualized as a defect density heat map. This map shows the defect concentration area in a color gradient manner, with red representing the high-density area and blue representing the low-density area, providing an intuitive reference for the improvement of the welding process; and the thermal field distribution mapping of the heat dissipation performance prediction results is carried out to generate a heat performance influence distribution map, which intuitively shows the degree of influence of the defects on the temperature field, including the temperature rise area, the heat flow obstruction area and the hot spot transfer situation; furthermore, quantitative statistical calculations are carried out on the defect characteristics in the defect visualization map, including indicators such as the total number of defects, the maximum defect size, the defect density, and the defect distribution concentration degree, and compared with the preset quality standards to generate a defect characteristic score (0-100 points). At the same time, based on the heat performance prediction results, indicators such as the increase ratio of thermal resistance, the maximum temperature rise increase value, and the change in temperature uniformity are calculated to generate a heat performance score (0-100 points), and according to the product application requirements, the defect characteristic score and the heat performance score are weighted and averaged to obtain the comprehensive quality score of the precision heat dissipation tooth-shaped plate. Based on this score, according to the preset grade division standard (such as 90-100 points for grade A, 75-89 points for grade B, 60-74 points for grade C, and below 60 points for grade D), the quality grade of the product is determined; furthermore, all analysis results and evaluation information are integrated to generate a comprehensive initial welding quality report, which includes product basic information, detection parameter settings, defect visualization maps, defect density heat maps, heat performance influence distribution maps, defect statistical data, heat performance prediction results, comprehensive quality scores and quality grades, etc., and provides a detailed description of quality problems and possible cause analysis to achieve accurate evaluation and effective control of the welding quality of the precision heat dissipation tooth-shaped plate, and finally ensures the heat dissipation performance and reliability of the product in actual applications.

[0034] 105. Based on the initial welding quality report and the preset welding process parameter-defect-thermal performance correlation database, adjust the welding process control parameters of the precision heat dissipation tooth-shaped plate, and based on the welding process control parameters, carry out welding optimization control on the precision heat dissipation tooth-shaped plate to obtain a precision heat dissipation tooth-shaped plate with the target welding quality.

[0035] In this embodiment, based on a preset welding process parameter - defect - thermal performance correlation database, the defect parameters in the initial welding quality report are correlated and compared to obtain a set of similar defect cases. Then, the parameter sensitivity of the set of similar defect cases is calculated to obtain the adjustment direction of the process parameters for the corresponding defect type. Based on the adjustment direction of the process parameters, the parameter quantification calculation is carried out for the corresponding pore - type defects and micro - crack planar defects in the fine welding defect feature types to obtain a set of defect suppression parameters. And based on the adjustment direction of the process parameters and the heat dissipation performance prediction result, the fusion parameter calculation is carried out for the abnormal heat conduction region to obtain a set of thermal performance optimization parameters. The balance calculation of the multi - objective control parameters and the constraint check and parameter adjustment of the welding equipment process window are carried out for the set of defect suppression parameters and the set of thermal performance optimization parameters to obtain the welding process control parameters of the precision heat dissipation tooth - shaped plate.

[0036] In practical applications, based on a pre-set welding process parameter - defect - thermal performance correlation database (which is a knowledge base constructed through a large amount of historical production data, experimental research results, and expert experience, and contains the mapping relationships between different combinations of welding process parameters (such as laser power, welding speed, focus position, shielding gas ratio, etc.) and the formation of various defects (such as pores, cracks, lack of fusion, etc.) and their corresponding thermal performance impacts), the defect feature data (including defect type, size, distribution, and severity) in the initial welding quality report is multi-dimensionally matched and compared with historical cases in the database. The nearest neighbor search algorithm or similarity calculation method is used to identify a set of historical cases with similar defect patterns (these cases contain process parameter settings and optimization experiences under similar defect conditions), obtaining a set of similar defect cases. Then, through multivariate regression analysis or partial derivative calculation methods on the selected set of similar defect cases, the influence degree and direction of each process parameter (such as laser power, welding speed, etc.) on the formation of specific defects are evaluated to reveal the causal relationship between parameter changes and defect suppression (for example, the analysis may find that increasing the laser power by 10% can reduce the porosity by 30%, or decreasing the welding speed by 5% can significantly reduce the incidence of lack of fusion defects), so as to determine the direction of process parameter adjustment for the specific defects of the current heat dissipation tooth-shaped plate; furthermore, based on the determined direction of process parameter adjustment, precise parameter quantification calculations are carried out for different types of defects. For pore-type defects, the system calculates the optimal laser power adjustment amount, welding speed correction value, and focus position offset amount according to the volume, distribution, and morphological characteristics of the pores, combined with the physical mechanisms of pore formation (such as metal evaporation, shielding gas entrainment, or gas release within the material, etc.). These parameter adjustments aim to optimize the melt pool dynamics behavior, reduce gas retention and entrainment, thereby suppressing pore formation. And for micro-crack planar defects, the system analyzes the direction, length, and position of the cracks, combined with the thermal stress theory, to calculate the reasonable range of welding energy density, preheating temperature requirements, and cooling rate control values. The goal of these parameter adjustments is to reduce the temperature gradient and stress concentration in the welding thermal cycle, thereby preventing the initiation and propagation of cracks. Thus, a set of defect suppression parameters is formed based on the parameter calculation results to guide the precise adjustment of the welding process; and based on the direction of process parameter adjustment and the heat dissipation performance prediction results, the optimal weld width, penetration ratio, and metal mixing ratio are calculated, and the corresponding shielding gas ratio (such as the mixing ratio of argon and helium) and pulse parameters (such as frequency, duty cycle) are determined (the optimization of these parameters aims to improve the melt pool fluidity, reduce the formation of inclusions, and increase the density of the weld metal, thereby enhancing the heat conduction efficiency of the welding area). Thus, a set of heat performance optimization parameters is generated from the calculation results;Furthermore, a multi-objective optimization algorithm (such as the weighted summation method or the Pareto front method) is adopted to balance the parameters, finding the best compromise point among multiple optimization objectives. At the same time, the process window constraints of the welding equipment are also considered to ensure that the calculated parameters are within the operating range of the equipment (such as the upper limit of laser power, speed control accuracy, etc.). Parameter correction is carried out if necessary to ensure operability. Finally, the welding process control parameters of the precision heat dissipation tooth-shaped plate are obtained. These parameters can not only effectively suppress welding defects but also ensure that the heat dissipation performance of the product meets the design requirements. Furthermore, based on the optimized welding process control parameters, the precision heat dissipation tooth-shaped plate is subjected to optimized welding operations, and real-time monitoring and fine-tuning are also carried out during the welding process to ensure the precise execution of the process parameters. Finally, the precision heat dissipation tooth-shaped plate with the target welding quality is obtained. These products have a lower defect rate, more uniform weld quality, and more excellent heat dissipation performance, significantly improving the reliability and service life of the products. Through this data-driven closed-loop optimization control method, the continuous improvement and precise control of the welding quality of the precision heat dissipation tooth-shaped plate are realized.

[0037] In the embodiment of the present invention, through the initial welding of the precision heat dissipation tooth-shaped plate with preset parameters and the acquisition of multi-source sensing data of the welded joint, these data are subjected to multiple preprocessings to obtain multi-modal sensing data, and then three-dimensional structure reconstruction and image enhancement are carried out to form three-dimensional defect fusion data, and then defect zoning and segmentation are carried out to obtain multiple types of welding abnormal regions; then preset defect feature parameters are extracted from these abnormal regions to construct a multi-dimensional feature space, and through comprehensive calculation and type discrimination, the types and spatial position information of subtle welding defect features are obtained; then based on the defect information, the heat conduction path is calculated and the thermal resistance is evaluated, the heat dissipation performance is predicted and the initial welding quality report is generated; finally, using this report in combination with the process parameter-defect-thermal performance correlation database, the welding process control parameters are adjusted and the optimized welding is executed to obtain the precision heat dissipation tooth-shaped plate with the target quality. Through multi-level data processing and feature analysis, the precise identification of micro-welding defects is realized. Especially in the aspect of multi-modal data fusion and thermal performance evaluation, the influence mechanism of defects on the heat dissipation performance is fully considered; and a closed-loop process optimization strategy is adopted, which not only realizes the identification of micro-defects but also establishes the correlation evaluation between defects and heat dissipation performance; in addition, through the intelligent adjustment of process parameters, the welding quality is accurately controlled, thus realizing the high-quality welding of the precision heat dissipation tooth-shaped plate as a whole.

[0038] This application can be used in numerous general-purpose or special-purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This application can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0039] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A welding method for a precision heat dissipation toothed plate, characterized in that: The welding method of the precision heat dissipation toothed plate comprises: Based on preset welding process parameters, the precision heat dissipation toothed plate is initially welded and multi-source original sensor data of the toothed plate welding joint in the precision heat dissipation toothed plate after the initial welding is collected, and the multi-source original sensor data is preprocessed multiple times to obtain preprocessed multi-modal sensor data; Reconstructing the three-dimensional structure of the welded joint and enhancing the three-dimensional image on the preprocessed multimodal sensing data to obtain three-dimensional defect fusion data, and performing multiple defect partitioning on the three-dimensional defect fusion data to obtain multiple types of welding abnormality areas; Extracting multiple preset welding defect characteristic parameters from the multiple types of welding abnormal areas to generate a multi-dimensional welding defect characteristic space, and performing defect characteristic comprehensive calculation and defect type discrimination on the multi-dimensional welding defect characteristic space to obtain subtle welding defect characteristic types and defect space position information; Based on the characteristic type of the minute welding defect and the spatial position information of the defect, the heat conduction path of the precision heat dissipation toothed plate is calculated and the thermal resistance value is evaluated to obtain a heat dissipation performance prediction result, and based on the characteristic type of the minute welding defect, the spatial position information of the defect and the heat dissipation performance prediction result, an initial welding quality report of the precision heat dissipation toothed plate after the initial welding is generated; Based on the initial welding quality report and the preset welding process parameter-defect-thermal performance association database, the welding process control parameters of the precision heat dissipation toothed plate are adjusted, and based on the welding process control parameters, the welding optimization control of the precision heat dissipation toothed plate is performed to obtain a precision heat dissipation toothed plate with target welding quality.

2. The welding method of the precision heat dissipation toothed plate according to claim 1, characterized in that: The multi-source original sensor data includes a ray projection image sequence, original ultrasonic scanning signal data and an original temperature field image sequence, and collects multi-source original sensor data of the toothed plate welding joint in the precision heat dissipation toothed plate after the initial welding, including: The welding joints of the precision heat dissipation toothed plates are subjected to multi-angle X-ray scanning to obtain a ray projection image sequence, and the welding joints of the precision heat dissipation toothed plates are subjected to high-frequency ultrasonic scanning to obtain original ultrasonic scanning signal data, and the surface of the precision heat dissipation toothed plates is subjected to infrared thermal imaging to obtain an original temperature field image sequence.

3. The welding method of the precision heat dissipation toothed plate according to claim 2 is characterized in that: The performing multiple preprocessing on the multi-source original sensor data to obtain preprocessed multi-modal sensor data includes: Performing noise filtering and ring artifact correction on multiple frequency components of the ray projection image sequence to obtain enhanced ray projection data, performing Hilbert transform and gain compensation of signal propagation depth on the original ultrasonic scanning signal data to obtain balanced ultrasonic echo signals, and performing pixel response correction and image bad pixel repair on the original temperature field image sequence to obtain calibrated temperature field data; The enhanced ray projection data, the balanced ultrasonic echo signal and the calibrated temperature field data are time-synchronously integrated to obtain pre-processed multi-modal sensing data.

4. The welding method of the precision heat dissipation toothed plate according to claim 3 is characterized in that: The reconstructing the three-dimensional structure of the weld joint and enhancing the three-dimensional image of the pre-processed multi-modal sensing data to obtain three-dimensional defect fusion data includes: Performing image contrast enhancement and spatial arrangement of corresponding projection angles and positions on the ray projection data in the preprocessed multimodal sensing data to obtain a three-dimensional projection data matrix, and performing voxel back-projection reconstruction calculation and metal material hardening artifact correction on the three-dimensional projection data matrix to obtain corrected three-dimensional structure data; Performing phase correction and multi-point delay superposition on the ultrasonic echo signals in the preprocessed multimodal sensing data to obtain enhanced ultrasonic signals, and arranging and integrating the enhanced ultrasonic signals based on a preset scanning position to obtain a high-resolution ultrasonic scanning image; The temperature field data in the pre-processed multimodal sensing data are subjected to pixel displacement vector field calculation between adjacent frames and position compensation adjustment of each frame image to obtain a temperature sequence that eliminates the influence of micro-vibration, and multi-frame super-resolution integration is performed on the temperature sequence that eliminates the influence of micro-vibration to obtain a high-precision thermal field distribution map; The corrected three-dimensional structural data, the high-resolution ultrasonic scanning image and the high-precision thermal field distribution map are spatially registered to obtain multimodal data of a unified coordinate system, and multi-channel data fusion and feature complementary enhancement are performed on the multimodal data of the unified coordinate system to obtain three-dimensional defect fusion data.

5. The welding method of the precision heat dissipation toothed plate according to claim 1, characterized in that: The multi-type welding abnormality regions include micro-pore defect regions, micro-crack surface defect regions and heat conduction abnormality regions. The three-dimensional defect fusion data is subjected to multiple defect partitioning and segmentation to obtain the multi-type welding abnormality regions, including: Based on the structural region characteristics corresponding to the precision heat dissipation toothed plate, the three-dimensional defect fusion data is adaptively divided into spatial regions and the statistical characteristics of each spatial sub-region are calculated to obtain regional adaptive segmentation parameters, and based on the regional adaptive segmentation parameters, each spatial sub-region is subjected to multi-threshold segmentation and smooth transition calculation of the regional boundary to obtain a volume segmentation result with consistent spatial segmentation; Performing three-dimensional connected domain analysis and marking on the volume segmentation result to obtain a candidate volume region set, and performing three-dimensional morphological operation and size filtering on the candidate volume region set to obtain a micro-pore type defect region; Performing interface reflection signal gradient calculation on the corresponding ultrasonic feature components in the three-dimensional defect fusion data to obtain a planar feature edge map, performing time-temperature change rate calculation on the corresponding thermal field feature components in the three-dimensional defect fusion data to obtain a temperature anomaly index map, performing edge connection and region closure determination on the planar feature edge map to obtain a microcrack planar defect region, and performing region growth and boundary extraction on the temperature anomaly index map to obtain a heat conduction anomaly region; The spatial overlap of the micropore-type defect region, the microcrack surface defect region and the heat conduction abnormal region is calculated to obtain multiple types of welding abnormal regions.

6. The welding method of the precision heat dissipation toothed plate according to claim 5, characterized in that: The extracting of multiple preset welding defect characteristic parameters from the multiple types of welding abnormal areas to generate a multi-dimensional welding defect characteristic space includes: Calculating multiple pore geometric characteristic parameters of the tiny pore defect area to obtain a pore geometric characteristic set, performing three-dimensional principal axis calculation and moment of inertia calculation on the tiny pore defect area to obtain a pore morphological characteristic set, and measuring the position of the corresponding weld center line and heat dissipation channel of the tiny pore defect area to obtain a pore position characteristic set; Calculating multiple crack geometric characteristic parameters of the microcrack surface defect region to obtain a crack geometric characteristic set, calculating edge curvature and irregularity of the microcrack surface defect region to obtain a crack morphological characteristic set, and extracting ultrasonic echo amplitude and phase characteristics of the microcrack surface defect region to obtain a crack acoustic characteristic set; Numerical calculations are performed on a plurality of thermal characteristic parameters of the abnormal heat conduction region to obtain a thermal characteristic parameter set, and a heating rate and a cooling time constant are calculated on the abnormal heat conduction region to obtain a thermal dynamic characteristic set, and a difference calculation is performed on the abnormal heat conduction region based on a preset welding thermal field model to obtain a thermal resistance index characteristic set; Based on the device functional zoning corresponding to the precision heat dissipation toothed plate, the position weight coefficients corresponding to the tiny pore-type defect area, the microcrack surface defect area and the heat conduction abnormality area are calculated to obtain a regional weight set, and based on the regional weight set, the pore geometric feature set, the pore morphology feature set, the pore position feature set, the crack geometric feature set, the crack morphology feature set, the crack acoustic feature set, the thermal feature parameter set, the thermal dynamic feature set and the thermal resistance index feature set are weighted and integrated to construct a multi-dimensional welding defect feature space.

7. The welding method of the precision heat dissipation toothed plate according to claim 1, characterized in that: The defect feature comprehensive calculation and defect type discrimination are performed on the multi-dimensional welding defect feature space to obtain the subtle welding defect feature type and defect space position information, including: Normalizing the multidimensional welding defect feature space to obtain a standardized feature vector, and calculating the signal-to-noise ratio and information entropy corresponding to the three modal data in the standardized feature vector to obtain a modal data reliability index; Based on the modal data reliability index and the structural features corresponding to the corresponding precision heat dissipation toothed plate, the importance weight of each feature dimension in the standardized feature vector is calculated to obtain a feature weight coefficient, and based on the feature weight coefficient, the standardized feature vector is weighted to obtain a weighted feature vector; Calculating the inter-feature correlation degree of the weighted feature vector to obtain a feature correlation matrix, and based on the feature correlation matrix, extracting a feature subset with a degree greater than a preset complementarity and less than a preset redundancy to obtain an optimized feature vector; Based on a preset defect typical feature template, the optimized feature vector is similarity calculated to obtain a defect type recognition result, and the defect type recognition result and the multi-dimensional welding defect feature space are matched and combined to obtain subtle welding defect feature types and defect space position information.

8. The welding method of the precision heat dissipation toothed plate according to claim 1, characterized in that: Based on the characteristic type of the minute welding defect and the spatial position information of the defect, the heat conduction path calculation and thermal resistance value evaluation of the precision heat dissipation toothed plate are performed to obtain the heat dissipation performance prediction result, including: The heat flow obstruction effect value of the pore defects in the characteristic type of the minor welding defects is calculated to obtain the pore thermal resistance increment, and the heat flow detour effect value of the microcrack defects in the characteristic type of the minor welding defects is calculated to obtain the crack thermal resistance increment, and based on the distribution of the toothed plate functional areas corresponding to the precision heat dissipation toothed plate, the position weight of the defect spatial position information is calculated to obtain the defect position weighting coefficient; Based on the defect position weighting coefficient, the pore thermal resistance increment and the crack thermal resistance increment are weightedly calculated to obtain the defect thermal resistance value, and based on the toothed plate heat flow channel distribution corresponding to the precision heat dissipation toothed plate and the defect thermal resistance value, the heat flow path disturbance calculation is performed on the precision heat dissipation toothed plate to obtain the actual heat flow channel distribution; The actual heat flow channel distribution is calculated in series and parallel combination to obtain the overall equivalent thermal resistance increase of the precision heat dissipation toothed plate, and the working temperature field simulation calculation is performed on the overall equivalent thermal resistance increase to obtain the heat dissipation performance prediction result.

9. The welding method of the precision heat dissipation toothed plate according to claim 1, characterized in that: The method of generating an initial welding quality report of the precision heat dissipation toothed plate after initial welding based on the characteristic type of the minor welding defect, the spatial position information of the defect and the heat dissipation performance prediction result includes: Performing three-dimensional visualization and color coding of the defect types on the characteristic types of the minor welding defects and the spatial position information of the defects to obtain a defect visualization map, performing density statistical calculation on the spatial position information of the defects to obtain a defect density thermodynamic map, and performing thermal field distribution mapping on the heat dissipation performance prediction results to obtain a thermal performance impact distribution map; Performing quantitative statistical calculation on the defect features in the defect visualization diagram to generate a defect characteristic score and a thermal performance score, and performing weighted calculation on the defect characteristic score and the thermal performance score to obtain a comprehensive quality score and quality grade of the precision heat dissipation toothed plate; The defect visualization map, the defect density thermogram, the thermal performance impact distribution map, the comprehensive quality score and the quality grade are reported and integrated to obtain an initial welding quality report of the precision heat dissipation toothed plate after initial welding.

10. The welding method of the precision heat dissipation toothed plate according to claim 5, characterized in that: The step of adjusting the welding process control parameters of the precision heat dissipation toothed plate based on the initial welding quality report and a preset welding process parameter-defect-thermal performance association database includes: Based on a preset welding process parameter-defect-thermal performance association database, the defect parameters of the initial welding quality report are compared to obtain a similar defect case set, and parameter sensitivity calculation is performed on the similar defect case set to obtain a process parameter adjustment direction corresponding to the defect type; Based on the process parameter adjustment direction, the corresponding pore-type defects and microcrack surface defects in the micro welding defect characteristic type are quantitatively calculated to obtain a defect suppression parameter set, and based on the process parameter adjustment direction and the heat dissipation performance prediction result, the fusion parameter of the heat conduction abnormal area is calculated to obtain a thermal performance optimization parameter set; The defect suppression parameter set and the thermal performance optimization parameter set are subjected to a balance calculation of multi-objective control parameters and a constraint check and parameter adjustment of the welding equipment process window to obtain welding process control parameters of the precision heat dissipation toothed plate.

Images