Femtosecond laser thin film precision welding system and method

Abstract

The invention relates to the technical field of laser welding, and discloses a femtosecond laser thin film precision welding system and method. According to the system and method, through dynamic energy density adjustment, self-adaptive light spot track planning, multi-material welding parameter collaborative optimization and real-time quality monitoring feedback, the problems that in existing femtosecond film welding, energy control is not accurate, light spot track adaptability is poor, and multi-material welding quality is not stable are solved. And through an innovative algorithm and system design, the welding precision, efficiency and material adaptability are remarkably improved, and an efficient solution is provided for high-precision thin film welding.

Description

Technical Field

[0001] This invention relates to the field of laser welding technology, and discloses a femtosecond laser thin film precision welding system and method. Background Technology

[0002] In the field of femtosecond laser thin film welding, there are many specific technical problems in practical applications that restrict the improvement of welding quality and efficiency:

[0003] Firstly, energy control is inaccurate. Existing femtosecond welding systems mostly use a fixed energy output mode, which cannot adjust the energy density in real time according to changes in film thickness and material. When there are slight thickness differences in the welding area, overmelting or incomplete welding can easily occur, resulting in welding strength fluctuations exceeding 15%, which seriously affects product consistency.

[0004] Secondly, the adaptability of the laser spot trajectory is poor. Traditional welding trajectories are mostly preset straight lines or arcs, which are difficult to adapt to the micro-undulations of the thin film surface (such as unevenness at the ±5μm level). During the welding process, the focal plane offset rate between the laser spot and the thin film surface exceeds 20%, which leads to a decrease in energy coupling efficiency and the formation of local welding defects.

[0005] Third, the quality of multi-material welding is unstable. When welding dissimilar materials (such as PET and aluminum foil, PI and copper film), the existing system lacks a targeted parameter adjustment mechanism. The welding crack rate caused by the mismatch of the interface thermal expansion coefficients exceeds 30%, and it cannot be completely avoided through post-inspection. Summary of the Invention

[0006] This invention aims to solve the problems of inaccurate energy control, poor adaptability of light spot trajectory, and unstable welding quality of multiple materials in existing femtosecond film welding.

[0007] To address the above problems, this invention provides a femtosecond laser thin-film precision welding system and method, comprising the following four inventive points:

[0008] Dynamic energy density adjustment algorithm: Based on real-time film thickness detection data, precise energy density control is achieved through coordinated adjustment of pulse width and repetition frequency. The algorithm formula is as follows:

[0009] E(x,y)=k1×d(x,y) 2 +k2×f -1 +k3×τ

[0010] Where E(x,y) is the energy density of the welding point (x,y); d(x,y) is the film thickness at that point; f is the laser repetition frequency; τ is the pulse width; k1, k2, and k3 are material property coefficients, satisfying k1+k2+k3=1.

[0011] Adaptive spot trajectory planning method: Combining surface contour scanning data, a Bézier curve is used to fit the dynamic trajectory. The algorithm formula is as follows:

[0012] P(t=Σ n i＝0 B i,n (t)×P i (0≤t≤1)

[0013] Where P(t) is a point on the trajectory; B i,n (t) is an nth-degree Bessel basis function; P i denoted as contour feature points; n represents the number of feature points (n≥3).

[0014] Multi-material welding parameter collaborative optimization module: Establish a material combination-parameter mapping database, predict the optimal welding parameter combination through a neural network model, and realize the control of interface matching between dissimilar materials.

[0015] Real-time quality monitoring and feedback system: Integrating coherent optical interferometry and infrared thermal imaging modules, it monitors the weld pool morphology at the millisecond level and corrects welding parameters through PID closed-loop control.

[0016] Beneficial effects: This solution addresses the problems of inaccurate energy control, poor spot trajectory adaptability, and unstable welding quality across multiple materials in existing femtosecond film welding processes. Through innovative algorithms and system design, it significantly improves welding accuracy, efficiency, and material adaptability, providing a highly efficient solution for high-precision thin film welding. Attached Figure Description

[0017] Figure 1 Adaptive trajectory planning flowchart;

[0018] Figure 2 Multi-material welding control flowchart. Detailed Implementation

[0019] Example 1: Implementation of Dynamic Energy Density Regulation

[0020] Step 1: Scan the welding area of ​​the PET film (thickness range 5-20μm) using a laser thickness gauge to obtain thickness distribution data d(x,y) and transmit it to the control system.

[0021] Step 2: The system calls the dynamic energy density adjustment algorithm to calculate the required energy density E(x,y) at each point based on d(x,y) and generate an adjustment sequence of pulse width τ and repetition frequency f.

[0022] Step 3: The femtosecond laser outputs laser according to the adjustment sequence and responds to thickness changes in real time during the welding process. When a sudden change in thickness is detected (such as from 10μm to 15μm), the energy parameter is adjusted within 100μs.

[0023] Compared with existing fixed energy modes, this embodiment reduces the welding strength fluctuation range from ±15% to ±3% and reduces the overmelting defect rate by more than 90%.

[0024] Example 2: Implementation of Adaptive Spot Trajectory Planning

[0025] Step 1: Use a white light interferometer to scan the welding path of the PI thin film (surface undulation ±8μm) and extract 200 contour feature points P. i .

[0026] Step 2: The system uses an adaptive trajectory planning algorithm to substitute feature points into the Bézier curve formula to generate a smooth trajectory, ensuring that the focal plane of the light spot fits the thin film surface ≥95%.

[0027] Step 3: The galvanometer system moves along the planned trajectory. When the welding speed is maintained at 100 mm / s, the trajectory tracking error is ≤ ±0.5 μm.

[0028] Compared to traditional preset trajectories, this embodiment improves energy coupling efficiency by 25% and reduces the local defect rate from 20% to below 3%.

[0029] Example 3: Implementation of Collaborative Control of Multi-Material Welding Parameters

[0030] Step 1: Before welding aluminum foil (5μm) and PET (12μm) dissimilar materials, the system retrieves the initial parameters of the material combination (laser wavelength 1030nm, initial power 5W) from the database.

[0031] Step 2: The neural network model corrects the parameters based on the ambient temperature (25℃) and humidity (45%), and outputs the optimal pulse width of 50fs and repetition frequency of 200kHz.

[0032] Step 3: During the welding process, the real-time monitoring system acquires an image of the molten pool every 5ms. When a microcrack precursor is detected, the energy density is finely adjusted by ±2% through PID control.

[0033] Example 4: Thin Film Precision Welding Method and System

[0034] Flexible electronic packaging ultrathin polyimide (PI) film welding scenario (multi-dimensional adaptive energy control algorithm)

[0035] Application Background

[0036] In flexible electronic packaging, ultrathin polyimide (PI) films (below 10 μm) are the core packaging material, and their welding quality directly determines the bending durability and signal stability of electronic devices. These films are extremely thin and thermally sensitive. Traditional welding techniques, lacking targeted energy control, often encounter two extreme problems: first, excessive energy leads to film ablation, which not only affects the packaging seal but may also damage internal electronic circuitry; second, insufficient energy results in low weld strength, making interlayer delamination prone to occur when the device is bent. Furthermore, unavoidable minor thickness fluctuations during film production (such as the difference in thickness between the edge and center) further exacerbate energy mismatch, leading to significant differences in weld quality within the same batch, making it difficult to meet the precision requirements of large-scale flexible electronics production.

[0037] Multidimensional adaptive energy regulation algorithm principle

[0038] The innovative "multi-dimensional adaptive energy regulation algorithm" developed in this embodiment breaks away from the traditional crude control logic of "fixed energy parameters" and constructs an energy management closed loop of "real-time perception - dynamic prediction - precise adaptation", allowing laser energy to be "distributed on demand" according to the actual state of the thin film.

[0039] The core logic of the algorithm is divided into three stages: The first stage is multi-dimensional state perception, which comprehensively captures key information of the welding scene through high-precision detection equipment. This includes not only the film thickness distribution in the welding area (such as the difference between thicker areas at the center and thinner areas at the edges), but also factors affecting energy absorption such as ambient temperature and humidity and film surface flatness, ensuring a comprehensive understanding of the welding scene. The second stage is dynamic thermal field prediction, which simulates the temperature change trend of the film under different energy inputs based on the perceived film thickness and material thermophysical properties (such as the thermal conductivity and melting point of the PI film). This identifies areas where heat accumulation (easily leading to ablation) or insufficient energy (easily leading to insufficient strength) may occur in advance, providing a predictive basis for energy adjustment. The third stage is dynamic energy allocation, which generates differentiated energy schemes for different areas based on the thermal field prediction results. For areas with larger thickness, the energy is appropriately increased to ensure welding strength, while for areas with thinner thickness, the energy is reduced to avoid ablation. Gradient energy is used to smoothly adjust the edge transition area to prevent film cracking due to sudden energy changes. At the same time, the energy is corrected in real time in combination with environmental parameters (such as appropriately compensating for energy in low temperature and high humidity environments to offset the weakening effect of the environment on energy absorption).

[0040] The innovation of this algorithm lies in upgrading energy control from "passive adaptation" to "active prediction". Through multi-dimensional perception and thermal field simulation, it achieves precise matching between energy and thin film state and environmental conditions, which not only solves the ablation problem of ultrathin films, but also ensures the consistency of welding strength, while significantly improving the system's adaptability to thin film thickness fluctuations.

[0041] Modeling and Solving Process

[0042] The modeling process revolves around three core stages: "data accumulation, model building, and strategy optimization," forming a practical algorithm system.

[0043] The first step is to construct a multi-dimensional database to provide data support for the algorithm. A high-precision laser thickness gauge is used to scan the entire area of ​​thousands of ultrathin PI films from different batches, recording the thickness distribution characteristics of each film (such as average thickness, location of extreme regions, and thickness transition range). Simultaneously, welding experiments are conducted under different environmental conditions (temperature 15-35℃, humidity 30%-70%), recording the welding effects corresponding to different energy parameters (such as ablation area, peel strength, and heat-affected zone size). This ultimately forms a multi-dimensional database containing hundreds of thousands of samples, covering complete correlation data of "film state - environmental conditions - energy parameters - welding effect".

[0044] The second step is to build a thermal field prediction model to simulate the effects of energy in advance. Using the finite element method, the film thickness distribution and material thermophysical properties (such as the thermal conductivity, specific heat capacity, and melting point of PI) are used as input parameters, while laser energy and repetition frequency are used as control variables. The film temperature field distribution and the range of the heat-affected zone are used as output results to build a thermal field simulation model. By repeatedly comparing the simulation results with the actual welding effect (e.g., controlling the heat-affected zone size error within 5%), the model parameters are continuously optimized. Ultimately, the model can accurately predict the film temperature changes under different energy inputs, with a prediction accuracy of over 98%, providing a scientific basis for energy adjustment.

[0045] The third step is the generation of energy regulation strategies, transforming model predictions into executable energy schemes. Fuzzy control logic is used to design regulation rules, employing "thickness deviation (the difference between actual and target thickness), predicted thermal field temperature, and environmental correction coefficient" as judgment criteria. Over two hundred regulation rules are established, such as "if the thickness deviation is positive and the thermal field temperature is below the melting point, then energy is increased by X%; if the thickness deviation is negative and the thermal field temperature is close to the melting point, then energy is decreased by Y%." These rules are then transformed into specific energy adjustment instructions through fuzzy reasoning. A smooth transition mechanism is also incorporated to avoid welding defects caused by sudden changes in energy parameters, ensuring the accuracy and stability of energy adjustment.

[0046] Implementation steps

[0047] Thin film status and environmental perception: The 8μm ultrathin PI film to be welded was fixed on a vacuum adsorption stage. The welding area was scanned by a laser thickness gauge. It was found that the thickness of the film was uniform in the central area (8.0±0.2μm), and the thickness of the edge area was slightly thinner (7.5±0.3μm). There was also a local thin spot of 7.0μm. At the same time, the environmental sensor detected that the current temperature was 25℃ and the humidity was 50%, with no special interference factors.

[0048] Thermal field prediction and energy scheme formulation: Thickness data and environmental parameters are input into the thermal field prediction model. Model simulation shows that if a fixed energy is used, the temperature in the thinner edge region will exceed the melting point of PI (420℃), posing a risk of ablation, while the temperature in the central region will be too low, potentially leading to insufficient strength. Based on this, the algorithm generates a dynamic energy scheme: the energy density in the central region is moderately increased to ensure strength, the energy gradient in the edge region is reduced, and the energy in the thinner region is minimized. At the same time, a 2mm wide energy transition zone is set to avoid abrupt parameter changes.

[0049] Collaborative Welding and Real-Time Adjustment: The trajectory planning module generates a smooth Bézier curve welding path. The energy adjustment module works in conjunction with the trajectory module—as the galvanometer system moves along the path, the energy parameters are switched in real time according to the dynamic scheme, with the response delay controlled within 80μs, ensuring precise matching of energy and position. During the welding process, the quality monitoring module monitors the temperature in real time through infrared thermal imaging. When the temperature in the thin edge area is close to the melting point, the energy is immediately fine-tuned by reducing it by 5%; when the temperature in the central area is too low, the energy is appropriately increased by 3%, dynamically correcting the energy deviation.

[0050] Quality Inspection and Algorithm Optimization: After welding, tensile testing revealed that the peel strength of each region was above 15 N / cm, with a strength difference of only 0.7 N / cm, far superior to the traditional fixed energy method (strength difference ≥ 4 N / cm). Optical microscopy showed no ablation marks, and the width of the heat-affected zone was controlled within 10 μm. The welding data (thickness distribution, energy scheme, and quality results) were fed back into the algorithm model to further optimize the thermal field prediction parameters and control rules, thereby improving the accuracy of subsequent welding.

[0051] Enhancement Explanation

[0052] The application of this algorithm achieves three core efficiency improvements: First, the welding quality is significantly improved, with the thin film ablation rate reduced from the traditional 15% to below 1%, the heat-affected zone width reduced by 60%, and the welding strength uniformity improved by 85%, fully meeting the high-precision requirements of flexible electronic packaging; Second, material adaptability is enhanced, with a significant increase in tolerance to thin film thickness fluctuations (±0.5μm), and the thin film utilization rate increased from 70% to 98%, reducing raw material screening costs; Third, production efficiency is improved, with parameter debugging time shortened from 30 minutes to 5 minutes, single-piece welding time reduced by 25%, and rework rate reduced from 18% to 1.2%, providing a guarantee for large-scale production.

[0053] Welding scenario of multilayer heterogeneous films (PET / aluminum / PI) (interface fusion enhanced trajectory planning algorithm)

[0054] Application Background

[0055] Multilayer heterogeneous films (such as 15μm PET / 10μm aluminum / 20μm PI) are widely used in new energy battery pouch packaging, flexible display panels, and other fields. Their welding requires solving three major challenges simultaneously: weak interlayer bonding, significant differences in material properties, and difficulty in edge alignment. The PET layer has weak heat resistance (melting point 250℃) and is easily ablated due to excessive energy; the aluminum layer has high thermal conductivity (237W / (m·K)), and energy is easily lost rapidly, leading to insufficient interlayer fusion; the PI layer has strong heat resistance (melting point 420℃) and requires sufficient energy to achieve effective welding. Traditional straight or simple curved trajectories cannot accommodate these differences, often resulting in uneven interlayer strength (difference ≥8N / cm) and excessive edge alignment deviations, seriously affecting packaging sealing and device reliability.

[0056] Interface-integrated enhanced trajectory planning algorithm principle

[0057] The innovative “interface fusion enhanced trajectory planning algorithm” proposed in this embodiment breaks away from the limitations of the traditional “trajectory is only responsible for path guidance” and deeply binds trajectory planning with interlayer fusion requirements. Through “trajectory morphology optimization”, it promotes uniform energy distribution and precise alignment between layers, and realizes high-quality welding of multilayer heterogeneous thin films.

[0058] The core logic of the algorithm consists of three stages: The first stage is the accurate identification of interface features. By using a white light interferometer in conjunction with a multispectral sensor, it can not only scan the surface contour of the thin film, but also locate the interlayer interface positions (such as the interface coordinates of PET-aluminum and aluminum-PI), identify interface gaps (which can easily lead to poor fusion), material boundaries (which can easily cause energy mismatch), and surface defects (such as wrinkles and scratches), and construct an "interface feature map" of the multilayer thin film; The second stage is the coordinated distribution of interlayer energy. Based on the interface features and the thermophysical properties of each layer, it determines the energy requirements of different interfaces—the aluminum layer interface needs to increase energy. To compensate for heat loss, the PET layer interface needs to reduce energy to avoid ablation, and the gap area needs to be superimposed with pulse energy to enhance fusion, forming an "interlayer differentiated energy scheme"; the third stage is adaptive trajectory generation, which designs the trajectory shape according to the energy coordination scheme: the interface flat area adopts a smooth trajectory to ensure energy uniformity, the gap area adopts a reciprocating scanning trajectory to enhance energy accumulation, and the material boundary area adopts a gradual transition trajectory to avoid energy abrupt change. At the same time, the trajectory fine-tuning achieves precise alignment of the multilayer film edges (error ≤ 0.5μm), so that the trajectory can play both the role of "path guidance" and assist in "energy optimization distribution".

[0059] The innovation of this algorithm lies in upgrading trajectory planning from "single path design" to "fusion auxiliary tool". By coordinating trajectory shape and energy demand, it solves the problem of uneven interlayer fusion caused by the differences in the properties of multilayer heterogeneous materials, while taking into account edge alignment accuracy, thus greatly improving the stability of welding quality.

[0060] Modeling and Solving Process

[0061] The modeling process revolves around "interface feature analysis - energy co-modeling - trajectory strategy optimization", forming an algorithm system adapted to multilayer heterogeneous thin films.

[0062] The first step is to construct an interface feature database to provide foundational data for the algorithm. Using a white light interferometer (scanning accuracy ±0.1μm) and a multispectral sensor, hundreds of PET / aluminum / PI multilayer films were scanned across their entire surface, recording interlayer interface coordinates, gap distribution, material boundary locations, and surface defect information. Simultaneously, ultrasonic scanning was used to detect the initial interlayer bonding state and mark weak areas. This process ultimately resulted in a database containing "interface features - material parameters - initial bonding state," providing a basis for subsequent modeling.

[0063] The second step is to build an interlayer energy coordination model to determine the energy distribution rules for each layer. Using a multiphysics coupling analysis method, layer thickness, thermal conductivity, and melting point are used as input parameters, while interlayer bonding strength and damage rate of each layer are used as output parameters to simulate the interlayer heat conduction and fusion process under different energy distribution ratios. Through extensive simulations and experimental comparisons, it was found that the energy ratio at the aluminum layer interface needs to reach 60%-70% of the total energy to compensate for heat loss, while the energy ratio at the PET layer interface needs to be controlled within 10%-15% to avoid ablation. Based on this, the energy distribution range for each layer was determined, and the model's prediction accuracy for interlayer bonding strength exceeded 97%.

[0064] The third step is to optimize the adaptive trajectory generation strategy, transforming energy requirements into trajectory shape. Based on an energy-coordinated approach, an improved Bézier curve algorithm is used to design the trajectory: a 3rd-order Bézier curve is used to ensure smoothness in flat areas, a 5th-order Bézier curve with 2-3 reciprocating scans is used in gap areas, and a gradient-order curve is used to achieve a smooth transition in material boundary areas. Simultaneously, edge alignment correction logic is added, dynamically adjusting the trajectory coordinates based on the alignment deviation detected in real-time by the visual positioning system to ensure that the edge alignment error is ≤0.5μm. Through extensive experimental verification, the distribution of trajectory control points is optimized, keeping the trajectory tracking error within ±0.3μm.

[0065] Implementation steps

[0066] Interface characteristics and material state perception: The PET / aluminum / PI multilayer film was fixed on a high-precision worktable. The precise coordinates of the PET-aluminum interface (Z-axis 15μm) and the aluminum-PI interface (Z-axis 25μm) were obtained by scanning with a white light interferometer and a multispectral sensor. Three gaps (maximum 0.8μm) were found at the aluminum-PI interface, and wrinkles with a height of 1μm were found at the edge of the PET layer. The initial weak areas between the layers were marked by ultrasonic scanning.

[0067] Interlayer energy coordination scheme formulation: Interface features are input into the energy coordination model, and the model outputs an energy distribution scheme: total energy density 1.5 J / cm³. 2 The PET-aluminum interface accounts for 15% (low energy to prevent ablation), the aluminum-PI interface accounts for 70% (high energy to compensate for heat loss), and the bottom of the PI layer accounts for 15%. An additional 20% pulse energy is added to the gap area to enhance fusion, and the energy in the wrinkled area is reduced by 10% to avoid stress concentration ablation.

[0068] Adaptive trajectory generation and alignment correction: The trajectory planning module generates differentiated trajectories based on the energy scheme: 3rd order Bézier curves are used in flat areas, 5th order curves with 2 reciprocating scans are used in gap areas, and gradually changing order curves are used in wrinkled areas; the visual positioning system detects an edge alignment deviation of 0.7μm, and the trajectory module automatically corrects the coordinates to reduce the deviation to 0.4μm.

[0069] Collaborative welding and real-time monitoring: During welding, the galvanometer moves along an adaptive trajectory, and the energy module synchronously distributes energy—switching to low energy when passing the PET-aluminum interface and switching to high energy when passing the aluminum-PI interface, with pulsed energy superimposed in the gap area; infrared thermal imaging monitors the temperature in real time, with the PET layer reaching a maximum of 230℃ (below the melting point), the aluminum layer at 350℃, and the PI layer at 380℃, all within the safe fusion range; coherent light interferometry monitoring shows that the gap area is well fused.

[0070] Quality Inspection and Algorithm Optimization: Post-weld tensile testing showed that the peel strength between each layer was above 16.5 N / cm, with a difference of only 0.7 N / cm (compared to ≥8 N / cm for traditional trajectories); no voids were observed in the cross-section, and the heat-affected zone was ≤10% of the layer thickness; the edge alignment error was 0.4 μm. The data was fed back to the model to optimize the energy distribution ratio and trajectory control points, thereby improving the subsequent welding quality.

[0071] Micro-sensor packaging ultra-thin ceramic / metal composite film welding scenario (multi-physics coupling adaptation algorithm)

[0072] Application Background

[0073] In miniaturized sensor packaging, the welding of 5-8μm ultrathin ceramic / metal composite films (such as alumina ceramic / copper foil composite films) must simultaneously meet three major requirements: electrical conductivity, mechanical strength, and high temperature resistance. The ceramic layer is insulating but brittle (low fracture toughness), while the metal layer is conductive but has a significantly different coefficient of thermal expansion. Traditional welding methods are prone to two problems: first, the ceramic layer cracks due to energy concentration; second, the metal-ceramic interface peels off due to thermal expansion mismatch. Furthermore, the miniaturization requirements of sensor packaging necessitate solder joint diameters ≤50μm. Traditional welding methods struggle to balance miniaturization with high reliability, severely impacting the signal transmission stability and lifespan of the sensor.

[0074] Multiphysics Coupling Adaptation Algorithm Principle

[0075] The innovative "multi-physics coupling adaptation algorithm" developed in this embodiment breaks away from the traditional single control logic of "only focusing on the thermal field". By comprehensively regulating the three major physical fields of "thermal field, stress field and electric field", it achieves low-damage and high-reliability welding of ceramic / metal composite films.

[0076] The core logic of the algorithm is divided into three stages: The first stage is multi-physics state perception, which uses high-precision detection equipment to simultaneously capture data on the thermal field (temperature distribution of each layer), stress field (stress changes between the ceramic layer and the metal layer), and electric field (conductivity of the metal layer) during the welding process, while identifying the interface bonding state of the composite film (such as initial gap and oxide layer thickness) to construct a multi-physics "state map"; The second stage is multi-field coupling effect prediction, which, based on the perceived data and material properties (fracture toughness of ceramics, thermal expansion coefficient of metals, and interface bonding energy), simulates the influence of the thermal field on the stress distribution under different welding parameters. The first stage involves assessing the risk of ceramic cracking from the stress field and the effect of the electric field on conductivity, predicting potential problems such as "ceramic cracking," "interface peeling," and "poor conductivity." The second stage involves multi-parameter coordinated control, simultaneously adjusting parameters such as laser energy, pulse frequency, and welding speed based on the prediction results. This involves controlling the thermal field through gradient energy to avoid localized overheating and stress concentration in the ceramic layer; optimizing stress release through pulse intervals to reduce thermal expansion mismatch stress at the metal-ceramic interface; and ensuring the conductivity of the metal layer through energy density adaptation, ultimately achieving a multi-objective balance of "uniform thermal field, controllable stress, and stable electric field."

[0077] The innovation of this algorithm lies in upgrading the welding control from "single thermal field regulation" to "multi-physics field collaborative adaptation". By comprehensively considering the coupling effects of heat, force and electricity, it solves three major problems at the same time: ceramic cracking, interface peeling and poor conductivity, thus meeting the stringent requirements of micro sensor packaging.

[0078] Modeling and Solving Process

[0079] The modeling process revolves around "multi-field data accumulation - coupling effect modeling - regulation strategy optimization", forming an algorithm system adapted to composite thin films.

[0080] The first step is to construct a multiphysics database to provide data support for the algorithm. Using infrared thermal imaging (temperature difference ±1℃), miniature stress sensors (measurement accuracy ±1MPa), and a conductivity detector, thousands of sets of thermal, stress, and electric field data are simultaneously collected during the ceramic / metal composite film welding process. At the same time, welding parameters (energy, frequency, speed) are changed, and the corresponding ceramic cracking rate, interface peeling rate, and conductivity resistance values ​​are recorded, forming a database of hundreds of thousands of data points containing "multiphysics data, parameters, and quality indicators," covering samples from different interface states and material batches.

[0081] The second step is to build a multi-field coupling effect model to predict problems in advance. A multi-physics coupling simulation method is used, taking the material parameters of the composite thin film (ceramic fracture toughness, metal thermal expansion coefficient, interface bonding energy) and welding parameters as inputs, and the thermal field distribution, stress distribution, and conduction resistance as outputs to build a coupled simulation model. Simulations revealed that the risk of cracking increases sharply when the temperature gradient of the ceramic layer exceeds 50℃ / mm, delamination is easy when the stress at the metal-ceramic interface exceeds 20MPa, and the energy density of the metal layer is below 0.6J / cm². 2 As the on-resistance increases, multi-field control thresholds are determined based on this, and the model's prediction accuracy for quality problems reaches over 96%.

[0082] The third step is the generation of a multi-parameter collaborative control strategy, transforming the predicted results into a control scheme. A multi-objective optimization logic is employed, with the objectives of "ceramic cracking rate ≤ 0.5%, interface peeling rate ≤ 0.3%, and conduction resistance ≤ 5mΩ" as the starting points. Parameter control rules are established as follows: when the predicted risk of ceramic cracking is high, the energy gradient is reduced and the pulse interval is extended; when the predicted risk of interface peeling is high, the welding speed is optimized to reduce heat retention; when the predicted poor conductivity is high, the energy density is appropriately increased. Through repeated experiments to optimize the parameter weights, a balance between multiple objectives is ensured, and the execution accuracy of the control scheme reaches over 95%.

[0083] Implementation steps

[0084] Multi-physics state sensing: A 5μm alumina ceramic / 7μm copper foil composite film was fixed on a miniature worktable, and initial state data was collected simultaneously by infrared thermal imaging, stress sensor and continuity detector: there is an initial gap of 0.3μm at the interface between the ceramic layer and the metal layer, and there is a thin oxide layer on the surface of the metal layer; the ambient temperature is 23℃ and there are no interference factors.

[0085] Multi-field coupling effect prediction: Sensing data is input into the coupling model. Model simulation shows that if a fixed energy is used, the temperature gradient of the ceramic layer reaches 60℃ / mm (exceeding the threshold), resulting in a high risk of cracking; the stress at the metal-ceramic interface reaches 25MPa (exceeding the threshold), resulting in a high risk of delamination; and insufficient energy in the metal layer may cause the on-resistance to exceed 10mΩ. Based on this, the algorithm predicts that the thermal field gradient and stress distribution need to be carefully controlled.

[0086] Multi-parameter coordinated control: Based on the prediction results, the algorithm generates a control scheme: energy is distributed in a gradient (low energy on the ceramic side and high energy on the metal side) to reduce the temperature gradient to 40℃ / mm; the pulse interval is extended from 100ns to 150ns to allow time for stress release; the welding speed is reduced from 5mm / s to 3mm / s to reduce interface heat retention; at the same time, the conductivity of the metal layer is ensured through energy density adaptation.

[0087] Micro-welding and real-time monitoring: During the welding process, the status of multiple physical fields is monitored simultaneously: the thermal field shows that the highest temperature of the ceramic layer is 280℃, with a temperature gradient of 40℃ / mm; the stress field shows that the interface stress is 18MPa; and the electric field shows that the on-resistance of the metal layer is 4mΩ, all of which meet the threshold requirements. When the local stress of the ceramic layer is found to rise to 19MPa, the pulse interval is immediately finely adjusted to 160ns, and the stress drops to 17MPa.

[0088] Quality Inspection and Algorithm Optimization: Post-weld inspection showed that the ceramic layer was crack-free (crack rate 0%), the interface peeling rate was 0.2%, the on-resistance was 3.8mΩ, and the solder joint diameter was 48μm (deviation ±2μm), fully meeting the requirements for micro-sensor packaging. The data was fed back to the model to optimize multi-field coupling parameters and control rules, improving the reliability of subsequent welding.

Claims

1. A femtosecond laser precision welding method for thin films, characterized in that, Includes the following steps: (1) Obtain the film thickness distribution d(x,y) in the welding area by thickness detection, and substitute it into the dynamic energy density algorithm to calculate the energy parameters. The algorithm formula is: E (x,y) = k1×d (x,y)² + k2×f⁻¹ + k3×τ Where E(x,y) is the energy density, f is the repetition frequency, τ is the pulse width, and k1, k2, and k3 are material property coefficients and their sum is 1; (2) Feature points are extracted from the surface contour of the thin film, and a welding path is generated using an adaptive trajectory algorithm. The algorithm formula is as follows: P (t) = Σⁿᵢ₌0 Bᵢ, n (t)×Pᵢ (0≤t≤1) Where P(t) is the trajectory point, Bᵢ, n (t) represents the Bessel basis function, and Pᵢ represents the characteristic point; (3) Based on the material combination, call the parameter database and optimize the welding parameters through a neural network model; (4) Monitor welding quality in real time and provide feedback for corrective parameters.

2. The method according to claim 1, characterized in that, In step (1), the thickness is measured using a laser thickness gauge with a scanning accuracy of ≤ ±0.1 μm.

3. The method according to claim 1, characterized in that, In step (2), the contour scanning uses a white light interferometer, and the feature point extraction density is ≥10 points / mm.

4. The method according to claim 1, characterized in that, The input parameters of the neural network model in step (3) include ambient temperature, humidity and material thickness ratio.

5. The method according to claim 1, characterized in that, In step (4), quality monitoring uses coherent optical interferometry combined with infrared thermal imaging, with a sampling frequency ≥200Hz.

6. A femtosecond laser thin-film precision welding system, characterized in that, include: (1) Energy regulation module: used to receive thickness data and output parameters through a dynamic energy algorithm, the algorithm formula is: E (x,y) = k1×d (x,y)² + k2×f⁻¹ + k3×τ Where E(x,y) is the energy density, f is the repetition frequency, τ is the pulse width, and k1, k2, and k3 are material property coefficients and their sum is 1; (2) Trajectory planning module: used to process contour data and generate paths through an adaptive algorithm, the algorithm formula is: P (t) = Σⁿᵢ₌0 Bᵢ, n (t)×Pᵢ (0≤t≤1) Where P(t) is the trajectory point, Bᵢ, n (t) represents the Bessel basis function, and Pᵢ represents the characteristic point; (3) Parameter optimization module: used to store the material parameter database and output optimized parameters through a neural network model; (4) Quality monitoring module: used to collect images of the welding area and correct parameters through feedback control.

7. The system according to claim 6, characterized in that, The energy regulation module is connected to a femtosecond laser with a response delay of ≤100μs.

8. The system according to claim 6, characterized in that, The trajectory planning module is linked with the galvanometer system, and the trajectory tracking error is ≤ ±0.5μm.

9. The system according to claim 6, characterized in that, The parameter optimization module contains more than 500 sets of benchmark parameters for material combinations.

10. The system according to claim 6, characterized in that, The spatial resolution of the quality monitoring module is ≥1μm, and the temperature measurement accuracy is ≤±2℃.

Images