Semiconductor equipment framework laser precision welding deformation control method

Abstract

The invention discloses a semiconductor equipment framework laser precision welding deformation control method, which relates to the technical field of laser welding, and comprises the following steps: S1, identifying thermophysical parameters of materials on two sides of a welding seam, and executing differential preheating to enable heterogeneous materials to reach a preset initial temperature on a welding initial line; s2, welding is conducted based on the initial temperature, a molten pool and thermal stress distribution are monitored in real time, and a laser mode and energy input are dynamically adjusted; and S3, synchronously starting two-stage cooling according to a monitoring result. According to the laser precision welding deformation control method for the semiconductor equipment framework, a cooperative control system covering the whole welding process is constructed, welding deformation of the semiconductor equipment framework is effectively restrained through systematic regulation and control from the source to the terminal, personalized temperature control is implemented according to material differences in the preheating link, and the welding deformation of the semiconductor equipment framework is effectively restrained. Foundation deformation caused by different thermophysical properties is avoided, intelligent adaptation to complex welding seams is achieved, and thermal stress concentration is reduced.

Description

Technical Field

[0001] This invention relates to the field of laser welding technology, specifically to a method for controlling deformation during precision laser welding of semiconductor equipment skeletons. Background Technology

[0002] As the core structure of key semiconductor manufacturing equipment such as etching machines and thin-film deposition equipment, the framework of this semiconductor equipment is often constructed from dissimilar thin-walled materials such as stainless steel and aluminum alloys. Its complex structure involves numerous weld seams and often exhibits unique characteristics such as intersections and variable cross-sections. Laser welding, with its advantages of concentrated energy and small heat-affected zone, has become the mainstream technology for welding such precision structures. However, in practical applications, due to differences in the thermophysical properties of dissimilar materials and uneven energy distribution requirements of complex weld seams, thermal deformation remains a key challenge affecting welding accuracy, significantly impacting the operational stability and lifespan of semiconductor equipment.

[0003] Existing laser welding deformation control technologies still have many areas for improvement in practical applications. These technologies often focus on optimizing a single stage, suppressing deformation by adjusting welding parameters, adding mechanical constraints, or relying on post-weld straightening, lacking effective coordination and linkage between stages, and exhibiting a passive response overall. In the preheating stage, a uniform heating method is often used, failing to fully consider the differences in thermal conductivity and thermal expansion coefficients of dissimilar materials, leading to a mismatch in the initial thermal state of the weld and creating potential hazards for subsequent deformation. During the welding process, the laser mode, scanning path, and energy output are often kept fixed or only slightly adjusted, making it difficult to dynamically adapt to the energy distribution requirements of complex welds, easily causing localized thermal stress concentration. The cooling stage is often disconnected from the welding process, using a single medium and a fixed cooling rate, which is mismatched with the welding heat input rhythm, easily causing secondary stress and deformation; while post-weld contact straightening methods may also introduce the risk of surface damage or internal microcracks for thin-walled, precision semiconductor device skeletons. To address these issues, we propose a laser precision welding deformation control method for semiconductor device skeletons. Summary of the Invention

[0004] To address the aforementioned technical problems, a method for controlling deformation during precision laser welding of semiconductor equipment skeletons is provided. This technical solution solves the problems of: lack of coordinated linkage among the various control links, resulting in mostly passive responses; failure to consider the differences in thermophysical properties of dissimilar materials during preheating, leading to mismatches in initial thermal states; lack of dynamic adaptability in welding process control; mismatch between cooling and welding heat input rhythms; and the potential for damage to precision structures due to post-weld contact straightening.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for controlling deformation during precision laser welding of semiconductor device skeletons includes the following steps: S1. Identify the thermophysical parameters of the materials on both sides of the weld and perform differentiated preheating so that the dissimilar materials reach the preset initial temperature at the welding start line; S2. Welding is performed based on the initial temperature, and the molten pool and thermal stress distribution are monitored in real time, and the laser mode and energy input are dynamically adjusted. S3. Based on the monitoring results, the two-stage cooling is started simultaneously, and the cooling rate is dynamically adjusted to match the current heat input; S4. After welding, the overall deformation distribution is measured to obtain deformation data, and non-contact electromagnetic pulses are applied for straightening based on the deformation data; S5 integrates preheating, welding, cooling, and straightening data, and optimizes process parameters through closed-loop feedback.

[0006] Preferably, in step S1, identifying the thermophysical parameters of the materials on both sides of the weld specifically involves: Based on the workpiece 3D model and weld path planning, material identification information of the areas to be welded on both sides of the weld is extracted; Based on the material identification information, a preset material database is retrieved, and the standard thermophysical parameter set of the corresponding material is obtained. The thermophysical parameter set includes: thermal conductivity, specific heat capacity and coefficient of linear expansion. When a material identifier is missing or there is no match in the database, the online thermal property identification process is initiated as follows: A localized stepped heat flow was applied to a safe area next to the weld, and the temperature rise response curve was simultaneously acquired using an infrared thermal imager. The actual thermal conductivity and specific heat capacity of the area were obtained by fitting the inversion algorithm. By combining a spectral composition analyzer to detect the micro-area composition of the region, the equivalent thermophysical parameters suitable for the current working conditions are corrected according to the material composition and output. Associate the identified or measured thermophysical parameter sets with the corresponding weld sections.

[0007] Preferably, the differentiated preheating is specifically performed as follows: By comparing the thermophysical parameters of the identified materials on both sides of the weld, the weld area is divided into a high thermal conductivity side and a low thermal conductivity side based on their relative thermal conductivity. For the side with high thermal conductivity, a first preheating strategy is adopted: a first target preheating temperature and a first heating rate are set, and heating is carried out through a first heating device; For the test of low thermal conductivity, a second preheating strategy is adopted: a second target preheating temperature higher than the first target preheating temperature and a second heating rate higher than the first heating rate are set, and heating is carried out by a second heating device; The temperature on both sides of the weld is monitored in real time, and the output of the heating devices on both sides is dynamically adjusted through closed-loop control so that the temperature on both sides of the weld at the welding start line reaches and stabilizes within the preset initial temperature range simultaneously.

[0008] Preferably, in step S2, real-time monitoring of the molten pool and thermal stress distribution specifically involves: During the welding process, visual sensors and infrared thermal imagers are used simultaneously to collect real-time images of the geometric shape of the molten pool and the temperature field distribution around the weld. Based on the collected temperature field distribution, the real-time thermal stress distribution field is calculated using a thermo-mechanical coupling inversion model.

[0009] Preferably, in step S2, dynamically adjusting the laser mode and energy input specifically involves: When the weld cross-section is found to be uniform and the heat distribution requirement is stable, the laser welding head is controlled to output the fundamental mode beam for welding. When weld intersections, variable cross sections, or areas of concentrated thermal stress are detected, the laser output is dynamically switched to a high-order vortex laser mode, and the scanning trajectory of the laser beam is simultaneously adjusted to a preset oscillating path and spiral path, while the pulse frequency and duty cycle of the laser are modulated. For long straight welds, the long weld is divided into multiple sections, and welding is carried out in a non-continuous order, with cooling intervals set between adjacent welded sections.

[0010] Preferably, in step S3, the simultaneous initiation of the two-stage cooling process specifically involves: The two-stage cooling includes a first stage of rapid cooling and shaping immediately following the laser weld spot, and a second stage of slow cooling and stress relief following the first stage. The first stage of rapid cooling and shaping uses an inert gas nozzle located at a first preset distance behind the laser weld spot to implement forced convection cooling of the molten pool and the adjacent high-temperature zone; The second stage of slow cooling stress relief uses a controllable rate cooling device located at a second preset distance behind the first stage cooling area to uniformly cool the welding heat-affected zone at a preset rate.

[0011] Preferably, the dynamic adjustment of the cooling rate to match the current heat input specifically involves: Establish a linkage control model between instantaneous laser power and cooling device control parameters; Based on the heat input intensity monitored in real time in S2, the flow rate and temperature of the cooling medium in the controllable rate cooling device during the second stage of slow cooling stress release are dynamically adjusted through the linkage control model.

[0012] Preferably, in step S4, measuring the overall deformation distribution specifically involves: After the welding and cooling processes are completed, a 3D optical scanner is used to perform surface scanning measurements on the entire workpiece. Based on the point cloud data obtained from the scan, the measured model is fitted and compared with the theoretical 3D model to identify the location, direction and magnitude of the deformation.

[0013] Preferably, the straightening process based on applying non-contact electromagnetic pulses according to deformation data specifically involves: Based on the deformation data, an electromagnetic pulse straightening scheme is planned, and the coil arrangement position, application direction, amplitude, frequency and timing of the pulse current of the electromagnetic pulse generator are determined. The workpiece is placed in the induction field of the coil, and an electromagnetic pulse with set parameters is applied by program control. The generated non-contact Lorentz force induces micro-plastic deformation in the deformation area of ​​the workpiece.

[0014] Preferably, in step S5, optimizing process parameters through closed-loop feedback specifically involves: The information on molten pool instability and abnormal thermal stress detected in S2 is fed back to S1 to adjust the preheating parameters of the corresponding area. The stress distribution assessment results after cooling in S3 and the residual deformation after straightening in S4 are fed back to S2 to adjust the laser mode, scanning path and energy input strategy for subsequent welding. The straightening effect data of S4 is fed back to S2 and S3 to adjust the welding energy parameters and cooling strategy.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The proposed method for controlling deformation during laser precision welding of semiconductor equipment skeletons effectively suppresses welding deformation by constructing a collaborative control system covering the entire welding process, from source to end. In the preheating stage, personalized temperature control is implemented based on material differences, avoiding basic deformation caused by varying thermal properties. During welding, energy output is dynamically adjusted based on real-time monitoring, achieving intelligent adaptation to complex welds and significantly reducing thermal stress concentration. The cooling stage is linked to welding for step-by-step temperature control, ensuring weld formation while smoothly releasing stress and eliminating the risk of secondary deformation. Post-weld, non-contact electromagnetic micro-straightening technology is used to achieve high-precision, non-destructive correction of residual deformation. The entire process forms a closed-loop feedback loop, and parameters at each stage can be self-optimized and iterated based on real-time data and historical experience, thereby improving the overall stability, repeatability, and structural reliability of the welding quality. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1As shown, a method for controlling deformation during precision laser welding of a semiconductor device skeleton includes the following steps: S1. Identify the thermophysical parameters of the materials on both sides of the weld and perform differentiated preheating so that the dissimilar materials reach the preset initial temperature at the welding start line; In step S1, the thermophysical parameters of the materials on both sides of the weld are specifically identified as follows: Based on the workpiece 3D model and weld path planning, material identification information of the areas to be welded on both sides of the weld is extracted; according to the material identification information, a preset material database is retrieved, and the standard thermophysical parameter set of the corresponding material is obtained. The thermophysical parameter set includes: thermal conductivity, specific heat capacity and coefficient of linear expansion. When the material identification is missing or there is no match in the database, the online thermal property identification process is initiated: a micro-focus laser is used as the heat flow application device to apply local stepped heat flow in a safe area at least 5mm away from the weld edge on the side of the weld. The heat flow power is increased in a gradient of 50W, 100W, 150W and 200W, and each power level is maintained stably for 30 seconds.

[0019] The temperature rise response curve is acquired synchronously using an infrared thermal imager. The sampling frequency of the infrared thermal imager is set to 50Hz, and the spatial resolution is 320×240 pixels. The application of heat flow and image acquisition are linked through a synchronous trigger module, with a trigger delay of no more than 1ms.

[0020] The temperature rise response curve was fitted using the least squares inversion algorithm, with the objective function being minimizing the curve fitting error. Iterative optimization yielded the actual thermal conductivity and specific heat capacity of the region. Micro-area composition analysis was then performed on this region using a spectrometer. Three evenly distributed points were selected around the center and periphery of the heat flux application area, with each micro-area having a diameter of 50 μm. Based on the volume fractions of each component obtained, the thermophysical parameters were corrected using a mixing principle, expressed as follows: in, The equivalent thermal conductivity is For equivalent specific heat capacity, For the first Volume fraction of the components For the first The thermal conductivity of the components, For the first The specific heat capacity of each component is determined, and the final output is the equivalent thermophysical parameters applicable to the current working conditions. The thermophysical parameter sets obtained by identification or measured and fitted are associated with the corresponding weld sections to establish a segmented parameter mapping table.

[0021] The specific implementation of differentiated preheating is as follows: By comparing the thermophysical parameters of the identified materials on both sides of the weld, the weld area is divided into a high thermal conductivity side and a low thermal conductivity side based on their relative thermal conductivity. For the high thermal conductivity side, a first preheating strategy is adopted: an induction coil is selected as the first heating device, wound around the area to be preheated, maintaining an 8mm distance from the weld edge to avoid interference. The first target preheating temperature is set at 40-80℃, and the first heating rate is 0.5-1℃ / s, using electromagnetic induction heating through the induction coil. For the low thermal conductivity side, a second preheating strategy is adopted: an infrared heating lamp is selected as the second heating device, installed 15mm directly above the area to be preheated. The second target preheating temperature is set at 60-130℃, which is 20-50℃ higher than the first target preheating temperature. The second heating rate is set... The heating rate is 1-2℃ / s, which is 0.5-1℃ / s higher than the first heating rate. Radiation heating is achieved through infrared heating lamps. Two temperature monitoring points are arranged 2mm on each side of the weld start line. Thermocouples are used to collect temperature data in real time at a frequency of 10Hz. A closed-loop control system is constructed based on the proportional-integral-derivative control algorithm. The output power of the heating devices on both sides is dynamically adjusted according to the deviation between the measured temperature and the preset initial temperature, so that the temperature on both sides of the weld at the weld start line reaches and stabilizes within the preset initial temperature range synchronously. The temperature fluctuation range is controlled within ±5℃, and the stable maintenance time is not less than 3 seconds.

[0022] S2. Welding is performed based on the initial temperature, and the molten pool and thermal stress distribution are monitored in real time, and the laser mode and energy input are dynamically adjusted. In step S2, real-time monitoring of the molten pool and thermal stress distribution specifically includes: During the welding process, a high-speed CMOS vision sensor and an infrared thermal imager are used to work synchronously. The high-speed CMOS vision sensor is installed at a 45° position on the side of the laser welding head, 300mm away from the molten pool area. It is equipped with a narrow-band filter with a center wavelength of 808nm to block the interference of strong welding light. It acquires the geometric shape image of the molten pool in real time at a frame rate of 1000fps and an image resolution of 2048×1088 pixels. The infrared thermal imager is installed 600mm directly above the welding head, with a field of view covering the weld and a 50mm surrounding area. It works synchronously with the vision sensor via a trigger signal, with a synchronization error of ≤2ms, and collects the temperature field distribution around the weld in real time. Based on the collected temperature field distribution, a thermo-mechanical coupling inversion model is used to calculate the real-time thermal stress distribution field. This model is a simplified finite element model. The input parameters include temperature field data, material thermophysical parameters and workpiece geometry. The boundary condition is set as natural convection heat dissipation, and the convection heat transfer coefficient is taken as 15W / (m²·K). The mesh is refined with the molten pool as the center. The mesh size in the molten pool region is 0.1mm, and the mesh size in the surrounding region is gradually expanded to 1mm. The thermal stress distribution field is obtained through iterative calculation.

[0023] In S2, the dynamic adjustment of the laser mode and energy input specifically involves: when the weld cross-section is found to be uniform and the heat distribution requirement is stable, the laser welding head is controlled to output a fundamental mode beam for welding. The diameter of the focused spot of the fundamental mode beam is 0.3 mm, and the laser power is stabilized at 800-1200 W. When weld intersections, variable cross-sections, or areas of concentrated thermal stress are detected, the laser output is dynamically switched to a higher-order vortex laser mode through an electrically controlled variable phase plate. The laser beam scanning trajectory is adjusted synchronously, the oscillation path adopts a sinusoidal oscillation, the oscillation amplitude is 0.5-2 mm, the oscillation frequency is 10-50 Hz, the spiral path radius is 0.3-1 mm, the number of spiral lines is 2-3 turns, and the laser pulse frequency is modulated to 5-20 kHz with a duty cycle of 30%-70%. The parameters are linearly adjusted according to the degree of stress concentration. For long straight welds exceeding 500mm in length, divide them into several 50-100mm sections and weld them in a non-continuous sequence of 1, 3, 5, 2, 4. Set a cooling interval between adjacent welded sections. The cooling interval time is determined according to the plate thickness. For plates with a thickness of 2-5mm, the interval is 10-20 seconds to ensure that the temperature between sections drops below 150℃ before welding the next section.

[0024] S3. Based on the monitoring results, the two-stage cooling is started simultaneously, and the cooling rate is dynamically adjusted to match the current heat input; In S3, the simultaneous initiation of the two-stage cooling process specifically involves: The two-stage cooling includes a first stage of rapid cooling and shaping immediately following the laser weld spot, and a second stage of slow cooling and stress relief following the first stage. The first preset distance for rapid cooling and shaping in the first stage is 3-5mm behind the laser weld spot. This distance is dynamically adjusted according to the welding speed. It is 3mm when the welding speed is 5mm / s and 5mm when the welding speed is 2mm / s. Forced convection cooling is implemented through an inert gas nozzle installed at this position. The nozzle adopts a flat structure with an outlet width of 2mm. The airflow direction is at a 30° angle with the weld. Argon gas with a purity of 99.99% is selected as the cooling gas, and the gas flow rate range is 10-25L / min. The second preset distance for the second stage of slow cooling stress relief is 8-12mm behind the cooling area of ​​the first stage. A temperature-controlled cold air knife is used as a controllable rate cooling device to uniformly cool the weld heat-affected zone. The cooling medium is dry air, the air temperature is controllable within the range of 20-80℃, the airflow velocity is adjustable within the range of 5-15m / s, and the cold air knife outlet width is adapted to the weld width to ensure that the cooling covers the entire heat-affected zone.

[0025] The dynamic adjustment of the cooling rate to match the current heat input specifically involves: establishing a linkage control model between the instantaneous laser power and the control parameters of the cooling device, and constructing the correlation using empirical formulas, the formulas being: in, The flow rate of the cooling medium in the second stage (unit: m / s). This is the proportionality constant, with a value of 0.02 m / (s·W). Instantaneous laser power (unit: W).

[0026] Based on the real-time heat input intensity monitored in S2, the cooling medium flow rate and temperature are dynamically adjusted through this linkage model. When the laser power increases from 800W to 1200W, the cooling medium flow rate is simultaneously adjusted from 16m / s to 24m / s. At the same time, if the temperature of the heat-affected zone exceeds 300℃, the cooling medium temperature is reduced from 25℃ to 20℃ to ensure that the cooling rate matches the heat input.

[0027] S4. After welding, the overall deformation distribution is measured to obtain deformation data, and non-contact electromagnetic pulses are applied for straightening based on the deformation data; In S4, measuring the overall deformation distribution specifically involves: after the welding and cooling processes are completed, a three-dimensional optical scanner is used to perform a surface scan measurement on the entire workpiece. The scanner accuracy is ±0.01mm, the point cloud acquisition point distance is 0.05mm, and the scanning range covers the entire workpiece and the reference surface.

[0028] Based on the point cloud data obtained from the scan, the iterative nearest point algorithm is used to fit and compare the measured model with the theoretical 3D model. The algorithm flow is as follows: First, feature points are extracted from the point cloud and theoretical model. The feature points are selected from the workpiece corners, holes, and other reference features. Then, coarse registration and fine registration are performed. The coarse registration uses a random sampling consistency algorithm to remove outliers. The fine registration iterates no more than 50 times, and the registration error is controlled within ≤0.02mm. Finally, a deviation field cloud map is generated, and the area with an absolute deviation value exceeding 0.03mm is automatically identified as a deformation area. The three-dimensional coordinates of the deformation location, the deformation direction vector, and the deformation amplitude data are calculated to distinguish deformation types such as bending and warping, and a deformation quantification report is output.

[0029] The straightening process based on applying non-contact electromagnetic pulses using deformation data specifically involves: Based on the deformation data, an electromagnetic pulse straightening scheme was planned. A capacitor discharge electromagnetic pulse generator was selected, with an output pulse energy range of 1-10kJ. The coil adopted a rectangular structure with dimensions of 50mm×30mm, a coil thickness of 5mm, and 20 turns. The coil arrangement was adjusted according to the deformation position and direction. The angle between the coil and the workpiece surface was 30°-90°. When the deformation direction was perpendicular to the workpiece surface, the angle was 90°, and when the deformation direction was inclined, the angle was adjusted accordingly.

[0030] The pulse current parameters are determined based on the deformation amount, material thickness, and material composition, with an amplitude range of 10-50kA, a frequency range of 1-10Hz, and 1-5 applications. For a 2mm thick stainless steel skeleton, when the deformation amplitude is 0.1mm, the pulse current amplitude is set to 20kA, the frequency to 5Hz, and the application is performed twice. The workpiece is placed within the coil's induction field, with the distance between the coil and the workpiece surface controlled at 5-10mm. Electromagnetic pulses with set parameters are applied via program control, inducing micro-plastic deformation in the workpiece's deformation area using the generated non-contact Lorentz force. By adjusting the coil position and current parameters, the Lorentz force is concentrated on the deformation area. The micro-plastic deformation is verified offline using a 3D optical scanner, and a second scan is performed after straightening to ensure that the residual deformation is ≤0.02mm.

[0031] S5 integrates preheating, welding, cooling, and straightening data, and optimizes process parameters through closed-loop feedback.

[0032] In step S5, optimizing process parameters through closed-loop feedback specifically involves: A central process database and machine learning optimization model are constructed. The central process database stores preheating parameters, laser welding parameters, cooling parameters, straightening parameters and monitoring data of each stage. The data is associated with the corresponding weld section and workpiece batch. The machine learning model employs the random forest algorithm, aiming to minimize residual deformation and achieve uniform thermal stress distribution, and constructs parameter mapping relationships. The training dataset for this model is built from historical production data collected from the welding process of similar semiconductor equipment skeletons, covering common heterogeneous material combinations such as stainless steel-aluminum alloy and stainless steel-copper. Simultaneously, process verification experiments were conducted to supplement the dataset with measured data under different weld structures, plate thicknesses, laser power, and other variables. The dataset input dimensions include three core parameters: preheating parameters, welding parameters, and cooling parameters; the output dimensions include two target parameters: residual deformation and peak thermal stress.

[0033] In the feature engineering stage, key influencing features were selected, redundant features with a correlation of less than 0.3 with the target parameters were removed, continuous features were normalized to eliminate dimensional differences, and discrete features were one-hot encoded. The Laida criterion was used to remove outliers other than 3σ to ensure the validity of the dataset. Residual deformation was classified into three levels: ≤0.02mm, 0.02-0.05mm, and >0.05mm, and peak thermal stress was classified into three levels: ≤200MPa, 200-300MPa, and >300MPa, achieving the classification optimization objective. During model training, the dataset was divided into training, validation, and test sets in a 7:2:1 ratio. The number of decision trees was set to 100, and the maximum depth of the decision trees was set to 10 layers. Five-fold cross-validation was used to optimize the model hyperparameters. On the test set, the model achieved a prediction accuracy of ≥92% for residual deformation levels and ≥90% for peak thermal stress levels, meeting the accuracy requirements for process optimization. The information flow architecture is a multi-stage bidirectional feedback system. The information on molten pool instability and abnormal thermal stress detected in S2 is fed back to S1. If the molten pool fluctuates and the thermal stress exceeds 200MPa, the preheating temperature of the high thermal conductivity side is reduced by 5-10℃, while the heating rate of the low thermal conductivity side is increased by 0.2℃ / s. The stress distribution assessment results after cooling in S3 and the residual deformation after straightening in S4 are fed back to S2. If the local stress concentration after cooling exceeds 300MPa and the residual deformation after straightening is 0.03mm, the laser mode in the corresponding area is adjusted to a higher-order vortex mode, the pulse frequency is increased by 5kHz, the duty cycle is reduced by 10%, and the length of the long straight weld section is shortened to 50mm. The straightening effect data of S4 is fed back to S2 and S3. If the number of straightening times exceeds 3, it indicates that the welding heat input is too large. The laser power is reduced by 100-150W, and the flow rate of the cooling medium in the second stage is increased by 3m / s.

[0034] This closed-loop system can achieve dual-mode optimization. It can be used to adjust the process parameters of subsequent welding sections of the same workpiece online in real time, and it can also be used to optimize the initial process parameters of the next workpiece of the same type offline. For example, after the first workpiece is optimized, the optimized parameters can be directly called for subsequent workpieces in the same batch, and then fine-tuned according to real-time monitoring data to gradually reduce the residual deformation and the number of straightening times.

[0035] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for controlling deformation during precision laser welding of a semiconductor device skeleton, characterized in that, Includes the following steps: S1. Identify the thermophysical parameters of the materials on both sides of the weld and perform differentiated preheating so that the dissimilar materials reach the preset initial temperature at the welding start line; S2. Welding is performed based on the initial temperature, and the molten pool and thermal stress distribution are monitored in real time, and the laser mode and energy input are dynamically adjusted. S3. Based on the monitoring results, the two-stage cooling is started simultaneously, and the cooling rate is dynamically adjusted to match the current heat input; S4. After welding, the overall deformation distribution is measured to obtain deformation data, and non-contact electromagnetic pulses are applied for straightening based on the deformation data; S5 integrates preheating, welding, cooling, and straightening data, and optimizes process parameters through closed-loop feedback.

2. The method for controlling deformation during laser precision welding of a semiconductor device skeleton according to claim 1, characterized in that, In step S1, the thermophysical parameters of the materials on both sides of the weld are specifically identified as follows: Based on the workpiece 3D model and weld path planning, material identification information of the areas to be welded on both sides of the weld is extracted; Based on the material identification information, a preset material database is retrieved, and the standard thermophysical parameter set of the corresponding material is obtained. The thermophysical parameter set includes: thermal conductivity, specific heat capacity and coefficient of linear expansion. When a material identifier is missing or there is no match in the database, the online thermal property identification process is initiated as follows: A localized stepped heat flow was applied to a safe area next to the weld, and the temperature rise response curve was simultaneously acquired using an infrared thermal imager. The actual thermal conductivity and specific heat capacity of the area were obtained by fitting the inversion algorithm. By combining a spectral composition analyzer to detect the micro-area composition of the region, the equivalent thermophysical parameters suitable for the current working conditions are corrected according to the material composition and output. Associate the identified or measured thermophysical parameter sets with the corresponding weld sections.

3. The method for controlling deformation during laser precision welding of a semiconductor device skeleton according to claim 2, characterized in that, The specific implementation of differentiated preheating is as follows: By comparing the thermophysical parameters of the identified materials on both sides of the weld, the weld area is divided into a high thermal conductivity side and a low thermal conductivity side based on their relative thermal conductivity. For the side with high thermal conductivity, a first preheating strategy is adopted: a first target preheating temperature and a first heating rate are set, and heating is carried out through a first heating device; For the test of low thermal conductivity, a second preheating strategy is adopted: a second target preheating temperature higher than the first target preheating temperature and a second heating rate higher than the first heating rate are set, and heating is carried out by a second heating device; The temperature on both sides of the weld is monitored in real time, and the output of the heating devices on both sides is dynamically adjusted through closed-loop control so that the temperature on both sides of the weld at the welding start line reaches and stabilizes within the preset initial temperature range simultaneously.

4. The method for controlling deformation during laser precision welding of a semiconductor device skeleton according to claim 3, characterized in that, In step S2, real-time monitoring of the molten pool and thermal stress distribution specifically includes: During the welding process, visual sensors and infrared thermal imagers are used simultaneously to collect real-time images of the geometric shape of the molten pool and the temperature field distribution around the weld. Based on the collected temperature field distribution, the real-time thermal stress distribution field is calculated using a thermo-mechanical coupling inversion model.

5. The method for controlling deformation during laser precision welding of a semiconductor device skeleton according to claim 4, characterized in that, In S2, the dynamic adjustment of the laser mode and energy input specifically involves: When the weld cross-section is found to be uniform and the heat distribution requirement is stable, the laser welding head is controlled to output the fundamental mode beam for welding. When weld intersections, variable cross sections, or areas of concentrated thermal stress are detected, the laser output is dynamically switched to a high-order vortex laser mode, and the scanning trajectory of the laser beam is simultaneously adjusted to a preset oscillating path and spiral path, while the pulse frequency and duty cycle of the laser are modulated. For long straight welds, the long weld is divided into multiple sections, and welding is carried out in a non-continuous order, with cooling intervals set between adjacent welded sections.

6. The method for controlling deformation during laser precision welding of a semiconductor device skeleton according to claim 5, characterized in that, In S3, the simultaneous initiation of the two-stage cooling process specifically involves: The two-stage cooling includes a first stage of rapid cooling and shaping immediately following the laser weld spot, and a second stage of slow cooling and stress relief following the first stage. The first stage of rapid cooling and shaping uses an inert gas nozzle located at a first preset distance behind the laser weld spot to implement forced convection cooling of the molten pool and the adjacent high-temperature zone; The second stage of slow cooling stress relief uses a controllable rate cooling device located at a second preset distance behind the first stage cooling area to uniformly cool the welding heat-affected zone at a preset rate.

7. The method for controlling deformation during laser precision welding of a semiconductor device skeleton according to claim 6, characterized in that, The dynamic adjustment of the cooling rate to match the current heat input specifically refers to: Establish a linkage control model between instantaneous laser power and cooling device control parameters; Based on the heat input intensity monitored in real time in S2, the flow rate and temperature of the cooling medium in the controllable rate cooling device during the second stage of slow cooling stress release are dynamically adjusted through the linkage control model.

8. The method for controlling deformation during laser precision welding of a semiconductor device skeleton according to claim 7, characterized in that, In step S4, measuring the overall deformation distribution specifically involves: After the welding and cooling processes are completed, a 3D optical scanner is used to perform surface scanning measurements on the entire workpiece. Based on the point cloud data obtained from the scan, the measured model is fitted and compared with the theoretical 3D model to identify the location, direction and magnitude of the deformation.

9. The method for controlling deformation during laser precision welding of a semiconductor device skeleton according to claim 8, characterized in that, The straightening process based on applying non-contact electromagnetic pulses using deformation data specifically involves: Based on the deformation data, an electromagnetic pulse straightening scheme is planned, and the coil arrangement position, application direction, amplitude, frequency and timing of the pulse current of the electromagnetic pulse generator are determined. The workpiece is placed in the induction field of the coil, and an electromagnetic pulse with set parameters is applied by program control. The generated non-contact Lorentz force induces micro-plastic deformation in the deformation area of ​​the workpiece.

10. The method for controlling deformation during laser precision welding of a semiconductor device skeleton according to claim 9, characterized in that, In step S5, optimizing process parameters through closed-loop feedback specifically involves: The information on molten pool instability and abnormal thermal stress detected in S2 is fed back to S1 to adjust the preheating parameters of the corresponding area. The stress distribution assessment results after cooling in S3 and the residual deformation after straightening in S4 are fed back to S2 to adjust the laser mode, scanning path and energy input strategy for subsequent welding. The straightening effect data of S4 is fed back to S2 and S3 to adjust the welding energy parameters and cooling strategy.

Images