Temperature control method and device for laser annealing

Abstract

The invention relates to the technical field of laser material processing, and discloses a laser annealing temperature control method and device.The method comprises the steps that the specific heat capacity, the heat conductivity coefficient, the heat diffusivity, the heat radiation rate, the heat reflectivity and the heat loss coefficient are obtained, and a material thermal characteristic model is established; performing grid division according to the material thermal characteristic model to obtain a laser energy density parameter; extracting a thermal strain coefficient according to the laser energy density parameter, and establishing a temperature-thermal strain mapping relation prediction model; dynamically adjusting a laser power proportion and a pulse width time sequence according to the temperature-thermal strain mapping relation prediction model to obtain temperature field distribution data; according to the temperature field distribution data, carrying out shortest path analysis to obtain a temperature regulation and control decision; and according to the temperature regulation and control decision, inputting a preset control regulation model, and combining reinforcement learning iteration optimization parameters to obtain a temperature control scheme. According to the method, large-range temperature crossing and fine adjustment in an extremely short time can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of laser material processing, and particularly to a temperature control method and device for laser annealing. Background Art

[0002] Currently, laser annealing technology is widely used in the fields of semiconductor wafer doping activation, metal material surface modification, and flexible electronic device manufacturing. However, in the processing of large-size and complex-shaped materials, due to the significant spatial differences in the thermal properties of the materials and the complex coupling effect between the laser energy distribution and the scanning path, it is a severe challenge to control the temperature field uniformity. Traditional methods usually adopt a temperature control method of matching a fixed laser power ratio with a uniform linear scanning path. For example, the laser power density and scanning speed are preset to match the average thermal property parameters of the material, and closed-loop regulation is achieved by using infrared temperature measurement feedback.

[0003] In an existing technology, a laser annealing control method based on a static thermal property model and a fixed scanning path determines the average thermal conductivity and specific heat capacity parameters of the material in advance, sets a unified laser power density and a uniform scanning speed, and performs proportional-integral-derivative regulation by combining the temperature data fed back by an infrared thermometer. For example, in the semiconductor wafer doping process, this method matches the thermal conductivity characteristics of the center region of the wafer by fixing the laser power density. However, when the laser scans to the edge region, due to the significant reduction in the thermal impedance caused by the thinning of the material thickness and the change in curvature, the static parameters cannot adjust the energy input in real time, resulting in temperature overshoot in the edge region. At the same time, the PID control has a lag in responding to the sudden change in heat in the laser scanning turning or overlapping regions, resulting in the inability to effectively suppress the temperature fluctuation within milliseconds. Such methods lack a dynamic compensation mechanism for the emissivity and reflectivity, and when processing large-area special-shaped workpieces, the mismatch of energy distribution and the thermal lag effect are superimposed, resulting in limited temperature crossing rate and insufficient local temperature control accuracy. Therefore, there is a problem in the prior art that it is difficult to achieve a large-range temperature crossing and fine adjustment in a very short time. Summary of the Invention

[0004] The present invention provides a temperature control method and device for laser annealing to achieve a large-range temperature crossing and fine adjustment in a very short time.

[0005] In a first aspect, to solve the above technical problems, the present invention provides a temperature control method for laser annealing, including:

[0006] Obtain the specific heat capacity, thermal conductivity, thermal diffusivity, emissivity, reflectivity, and heat loss coefficient of the material surface, and establish a material thermal property model integrating multi-dimensional thermal property parameters;

[0007] According to the material thermal characteristic model, the material is meshed by using the finite element analysis method, the thermal impedance values of each region are calculated, and combined with the gradient descent method and linear interpolation optimization, the laser energy density parameters of each region are obtained;

[0008] According to the laser energy density parameters, the dynamic simulation of the surface temperature field of the material is carried out by using a preset laser scanning path, the regions with significant thermal hysteresis effect are identified and the thermal strain coefficients are extracted, and a temperature-thermal strain mapping relationship prediction model is established;

[0009] According to the temperature-thermal strain mapping relationship prediction model, an adaptive algorithm is used to dynamically adjust the laser power ratio and pulse width timing sequence to obtain the temperature field distribution data;

[0010] According to the temperature field distribution data, combined with the multi-point temperature data of the material surface collected in real time, the graph theory algorithm is used to calculate the node centrality and the clustering coefficient of the connected domain, and the shortest path analysis of the graph structure is carried out to obtain the temperature control decision;

[0011] According to the temperature control decision, the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data is calculated, and the temperature control decision and the temperature deviation value are input into a preset control and adjustment model to calculate a multi-parameter dynamic adjustment vector including laser power, pulse width and energy density, and the parameters are iteratively optimized by combining reinforcement learning to obtain a temperature control scheme.

[0012] In an optional implementation manner, the obtaining of the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity and thermal loss coefficient of the material surface and the establishment of a material thermal characteristic model integrating multi-dimensional thermal characteristic parameters include:

[0013] Obtain the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity and thermal loss coefficient of the material surface;

[0014] According to the specific heat capacity, the thermal conductivity and the thermal diffusivity, the support vector machine algorithm is used for multi-variable collaborative analysis to construct a basic material thermal characteristic model;

[0015] According to the basic material thermal characteristic model, thermal emissivity, thermal reflectivity and thermal loss coefficient, a data fusion algorithm is used for parameter correlation matching to obtain a fused material thermal characteristic model;

[0016] According to the fused material thermal characteristic model, the linear regression algorithm is used for model parameter optimization to obtain a material thermal characteristic model integrating multi-dimensional thermal characteristic parameters.

[0017] In an alternative embodiment, according to the material thermal property model, the finite element analysis method is used to divide the mesh of the material, calculate the thermal impedance values of each region, and combine gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region:

[0018] According to the material thermal property model, the finite element analysis method is used to perform polygon topological structure mesh division on the material surface to obtain the thermal property parameters of each mesh;

[0019] According to the thermal property parameters of each mesh, combined with the heat conduction equation, thermal impedance calculation is performed to obtain the thermal impedance values of different regions;

[0020] According to the thermal impedance values, thermodynamic derivation is performed to obtain the initial laser energy density parameters;

[0021] According to the initial laser energy density parameters and the thermal impedance values, when the thermal impedance value is greater than the preset first thermal impedance threshold, the gradient descent method is used for optimization to obtain the laser energy density parameters. When the thermal impedance value is less than the preset second thermal impedance threshold, the linear interpolation method is used for optimization to obtain the laser energy density parameters. When the thermal impedance value is less than the preset first thermal impedance threshold and greater than the preset second thermal impedance threshold, the initial laser energy density parameters are determined as the laser energy density parameters.

[0022] In an alternative embodiment, according to the laser energy density parameters, a dynamic simulation of the temperature field on the material surface is performed using a preset laser scanning path, the regions with significant thermal hysteresis effects are identified and the thermal strain coefficients are extracted, and a temperature-thermal strain mapping relationship prediction model is established, including:

[0023] According to the laser energy density parameters, a dynamic simulation of the temperature field is performed through a preset grid scanning path to generate a dynamic temperature distribution map containing temperature gradient distribution and heat conduction characteristics;

[0024] According to the dynamic temperature distribution map, the relationship between the temperature change rate of adjacent regions and the preset thermal response threshold is analyzed to obtain the regions with significant thermal hysteresis effects;

[0025] According to the regions with significant thermal hysteresis effects, the thermal strain coefficient values at the corresponding positions are extracted and a data set is generated to obtain a thermal strain coefficient data set;

[0026] According to the dynamic temperature distribution map and the thermal strain coefficient data set, a mapping relationship between the surface temperature and thermal strain of the material is constructed. When non-linear characteristics appear in the mapping relationship, the polynomial regression algorithm is used to fit the change law of the thermal strain coefficient to obtain a linear mapping relationship;

[0027] Classify the thermal strain response patterns of the material surface according to the mapping relationship, the dynamic temperature distribution map, and the thermal strain coefficient data set to obtain a thermal strain response pattern classification map;

[0028] Construct a temperature-thermal strain mapping relationship prediction model according to the thermal strain response pattern classification map and in combination with the boundary condition constraints of the heat conduction equation.

[0029] In an optional implementation manner, according to the temperature-thermal strain mapping relationship prediction model, an adaptive algorithm is used to dynamically adjust the laser power ratio and the pulse width time sequence to obtain temperature field distribution data, including:

[0030] Analyze the dynamic characteristics of the temperature rise and fall of the regional material according to the temperature-thermal strain mapping relationship prediction model to obtain the thermal response curve of each regional material;

[0031] According to the thermal response curve and in combination with the preset thermal stress constraint conditions, perform calculations to obtain the required heating rate corresponding to each region and the required cooling rate corresponding to each region;

[0032] According to the required heating rate and the required cooling rate, use an adaptive control algorithm to analyze the dynamic deviation between the real-time temperature monitoring data and the target temperature, and match the heat capacity characteristics and the heat conduction delay response characteristics of the regional material to obtain the laser power ratio parameter;

[0033] According to the required heating rate and the required cooling rate, dynamically adjust the laser duty cycle through a preset pulse width time sequence control strategy. When in the heating stage, determine the laser duty cycle in the high-frequency pulse mode as the pulse width time sequence control parameter. When in the cooling stage, determine the laser duty cycle in the low-frequency pulse mode as the pulse width time sequence control parameter;

[0034] According to the laser power ratio parameter and the pulse width time sequence control parameter, perform dynamic simulation in combination with the three-dimensional heat conduction equation to obtain temperature field distribution data.

[0035] In an optional implementation manner, according to the temperature field distribution data and in combination with the multi-point temperature data of the material surface collected in real time, use a graph theory algorithm to calculate the node centrality and the clustering coefficient of the connected domain, and perform the shortest path analysis of the graph structure to obtain a temperature control decision, including:

[0036] According to the temperature field distribution data and in combination with the multi-point temperature data of the material surface collected in real time, perform topological analysis. Take each temperature point as a node in the graph structure, calculate the temperature gradient value as the edge weight, and generate a temperature distribution map;

[0037] According to the temperature distribution map, calculate the degree centrality and eigenvector centrality of each node. When the node degree is greater than the preset degree threshold, it is determined as a key temperature control node;

[0038] According to the temperature distribution map and the key temperature control nodes, analyze the distribution characteristics of the connected domains, and determine the independent connected domain as the temperature isolation area;

[0039] According to the temperature distribution map and the temperature isolation area, calculate the clustering coefficient of each connected domain, perform abnormal temperature identification, and obtain the temperature abnormal area;

[0040] According to the temperature distribution map, the key temperature control nodes and the temperature abnormal area, use the shortest path algorithm to calculate the shortest heat conduction path between the key temperature control nodes, and obtain the energy transfer efficiency of the temperature field;

[0041] According to the energy transfer efficiency, the temperature distribution map, the key temperature control nodes and the temperature abnormal area, conduct a multi - element collaborative analysis by integrating the node spatial distribution characteristics, the regional heat gradient characteristics and the shortest heat conduction path, and obtain the temperature control decision.

[0042] In an alternative embodiment, according to the temperature control decision, calculate the temperature deviation value between the real - time collected temperature distribution data and the preset temperature distribution data. Input the temperature control decision and the temperature deviation value into a preset control and adjustment model to calculate a multi - parameter dynamic adjustment vector including laser power, pulse width and energy density, and combine reinforcement learning to iteratively optimize the parameters to obtain a temperature control scheme, including:

[0043] Perform spatial matching and calculation on the real - time collected temperature distribution data and the preset temperature distribution data to obtain the temperature deviation value;

[0044] Use the historical temperature control decision and the historical temperature deviation value as inputs, and the historical multi - parameter dynamic adjustment vector as the output to construct and train a control and adjustment model. When the number of training times is greater than or equal to the preset number of training times, it is determined that the training is completed, and a trained control and adjustment model is obtained;

[0045] Input the temperature control decision and the temperature deviation value into the trained control and adjustment model to obtain a multi - parameter dynamic adjustment vector including laser power, pulse width and energy density;

[0046] According to the multi - parameter dynamic adjustment vector, use the deep deterministic policy gradient algorithm, and verify the physical feasibility of the adjustment parameters in combination with the boundary conditions of the three - dimensional heat conduction equation, and perform iterative optimization to obtain a temperature control scheme.

[0047] In a second aspect, the present invention provides a temperature control device for laser annealing, including:

[0048] A data acquisition module, configured to acquire the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and thermal loss coefficient of the material surface, and establish a material thermal characteristic model integrating multi-dimensional thermal characteristic parameters;

[0049] A grid calculation module, configured to perform grid division on the material according to the material thermal characteristic model by using the finite element analysis method, calculate the thermal impedance values of each region, and combine gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region;

[0050] A simulation mapping module, configured to perform dynamic simulation on the material surface temperature field according to the laser energy density parameters by using a preset laser scanning path, identify regions with significant thermal hysteresis effects and extract thermal strain coefficients, and establish a temperature-thermal strain mapping relationship prediction model;

[0051] A dynamic adjustment module, configured to dynamically adjust the laser power ratio and pulse width timing according to the temperature-thermal strain mapping relationship prediction model by using an adaptive algorithm to obtain temperature field distribution data;

[0052] A graph theory analysis module, configured to calculate the node centrality and clustering coefficient of the connected domain by using graph theory algorithms according to the temperature field distribution data and in combination with the multi-point temperature data of the material surface collected in real time, perform the shortest path analysis of the graph structure, and obtain a temperature control decision;

[0053] A result output module, configured to calculate the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data according to the temperature control decision, input the temperature control decision and the temperature deviation value into a preset control and adjustment model to calculate a multi-parameter dynamic adjustment vector including laser power, pulse width, and energy density, and combine reinforcement learning to iteratively optimize the parameters to obtain a temperature control scheme.

[0054] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the temperature control method of laser annealing described in any one of the above is implemented.

[0055] In a fourth aspect, the present invention further provides a computer-readable storage medium, where the computer-readable storage medium includes a stored computer program. When the computer program runs, the device where the computer-readable storage medium is located is controlled to execute the temperature control method of laser annealing described in any one of the above.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The present invention discloses a temperature control method for laser annealing, which includes obtaining the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity and thermal loss coefficient of the material surface, and establishing a material thermal characteristic model integrating multi-dimensional thermal characteristic parameters; according to the material thermal characteristic model, using the finite element analysis method to perform mesh division on the material, calculating the thermal impedance values of each region, and combining gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region; according to the laser energy density parameters, using a preset laser scanning path to dynamically simulate the temperature field on the material surface, identifying the regions with significant thermal hysteresis effect and extracting the thermal strain coefficient, and establishing a temperature-thermal strain mapping relationship prediction model; according to the temperature-thermal strain mapping relationship prediction model, using an adaptive algorithm to dynamically adjust the laser power ratio and pulse width timing to obtain the temperature field distribution data; according to the temperature field distribution data, combining the multi-point temperature data of the material surface collected in real time, using the graph theory algorithm, calculating the node centrality and the clustering coefficient of the connected domain, and performing the shortest path analysis of the graph structure to obtain the temperature control decision; according to the temperature control decision, calculating the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data, inputting the temperature control decision and the temperature deviation value into a preset control and adjustment model to calculate a multi-parameter dynamic adjustment vector including laser power, pulse width and energy density, and combining reinforcement learning to iteratively optimize the parameters to obtain a temperature control scheme.

[0058] The present invention establishes a material thermal property model integrating multi-dimensional thermal property parameters. First, the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and thermal loss coefficient of the material are obtained, and a basic material thermal property model is constructed based on these thermal property parameters. Further, a data fusion algorithm and a support vector machine algorithm are used to optimize the model to obtain a material thermal property model integrating multi-dimensional thermal property parameters. Subsequently, the surface of the material is meshed by the finite element analysis method, the thermal impedance values of each region are calculated, and the laser energy density parameters of each region are obtained by combining the gradient descent method and the linear interpolation method. According to the laser energy density parameters, the surface temperature field of the material is dynamically simulated through a preset laser scanning path, the regions with significant thermal hysteresis effects are identified, and the thermal strain coefficient is extracted. Finally, a temperature-thermal strain mapping relationship prediction model is established. Then, in combination with the temperature-thermal strain mapping relationship prediction model, the laser power ratio and pulse width timing are dynamically adjusted through an adaptive algorithm to obtain temperature field distribution data. On this basis, in combination with the multi-point temperature data of the material surface collected in real time, the node centrality and connected domain clustering coefficient are calculated using the graph theory algorithm, and the temperature control decision is obtained through the shortest path analysis of the graph structure. Finally, the temperature control decision and the temperature deviation value are input into a preset control adjustment model, a multi-parameter dynamic adjustment vector including laser power, pulse width, and energy density is calculated, and iterative optimization is performed in combination with reinforcement learning to obtain a temperature control scheme. Therefore, the present invention can accurately control the temperature change during the laser annealing process, provide fast response and ensure temperature stability, and thus achieve a large range of temperature crossing and fine adjustment in an extremely short time. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a schematic flowchart of the temperature control method for laser annealing provided by the first embodiment of the present invention;

[0060] Figure 2 is a schematic structural diagram of the temperature control device for laser annealing provided by the second embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. 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.

[0062] Referring to Figure 1 , the first embodiment of the present invention provides a temperature control method for laser annealing, including the following steps:

[0063] S11. Obtain the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and thermal loss coefficient of the material surface, and establish a material thermal property model integrating multi-dimensional thermal property parameters;

[0064] S12. According to the material thermal property model, use the finite element analysis method to divide the mesh of the material, calculate the thermal impedance values of each region, and combine gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region;

[0065] S13. According to the laser energy density parameters, use a preset laser scanning path to dynamically simulate the temperature field of the material surface, identify the regions with significant thermal hysteresis effect and extract the thermal strain coefficient, and establish a temperature-thermal strain mapping relationship prediction model;

[0066] S14. According to the temperature-thermal strain mapping relationship prediction model, use an adaptive algorithm to dynamically adjust the laser power ratio and pulse width timing to obtain the temperature field distribution data;

[0067] S15. According to the temperature field distribution data, combined with the multi-point temperature data of the material surface collected in real time, use the graph theory algorithm to calculate the node centrality and the clustering coefficient of the connected domain, and perform the shortest path analysis of the graph structure to obtain the temperature control decision;

[0068] S16. According to the temperature control decision, calculate the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data, input the temperature control decision and the temperature deviation value into the preset control adjustment model to calculate the multi-parameter dynamic adjustment vector including laser power, pulse width, and energy density, and combine reinforcement learning to iteratively optimize the parameters to obtain the temperature control scheme.

[0069] In step S11, it is necessary to obtain the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and thermal loss coefficient of the material surface, and establish a material thermal property model integrating multi-dimensional thermal property parameters.

[0070] In one implementation, obtaining the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and thermal loss coefficient of the material surface, and establishing a material thermal property model integrating multi-dimensional thermal property parameters includes:

[0071] Obtain the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and heat loss coefficient of the material surface; according to the specific heat capacity, the thermal conductivity, and the thermal diffusivity, use the support vector machine algorithm to perform multivariate collaborative analysis and construct a basic material thermal property model; according to the basic material thermal property model, the thermal emissivity, the thermal reflectivity, and the heat loss coefficient, use the data fusion algorithm to perform parameter correlation matching to obtain a fused material thermal property model; according to the fused material thermal property model, use the linear regression algorithm to optimize the model parameters to obtain a material thermal property model with fused multi-dimensional thermal property parameters.

[0072] It should be noted that the material thermal property model is a mathematical model constructed based on the thermophysical properties of the material, aiming to describe the response and change behavior of the material to heat during the laser annealing process. First, obtain the specific heat capacity, thermal conductivity, and thermal diffusivity of the material surface, and use the support vector machine algorithm to perform multivariate collaborative analysis to construct a basic material thermal property model. The basic material thermal property model is further combined with the thermal emissivity, the thermal reflectivity, and the heat loss coefficient, and the data fusion algorithm is used to perform parameter correlation matching to obtain a fused material thermal property model. Finally, the parameters of the fused model are optimized by the linear regression algorithm to obtain the final material thermal property model. The material thermal property model will be used for temperature control and optimization in the subsequent laser annealing process, can more accurately describe the heat conduction characteristics of the material under different conditions, and provide a basis for the adjustment of laser power and pulse width.

[0073] In step S12, it is necessary to use the finite element analysis method to perform mesh division on the material according to the material thermal property model, calculate the thermal impedance values of each region, and combine gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region.

[0074] In one implementation, according to the material thermal property model, using the finite element analysis method to perform mesh division on the material, calculate the thermal impedance values of each region, and combine gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region, including:

[0075] According to the material thermal characteristic model, using the finite element analysis method, polygon topology structure meshing is carried out on the material surface to obtain the thermal characteristic parameters of each mesh; according to the thermal characteristic parameters of each mesh, combined with the heat conduction equation, heat impedance calculation is carried out to obtain the heat impedance values of different regions; according to the heat impedance values, thermodynamic derivation is carried out to obtain the initial laser energy density parameters; according to the initial laser energy density parameters and the heat impedance values, when the heat impedance value is greater than the preset first heat impedance threshold, the gradient descent method is used for optimization to obtain the laser energy density parameters, when the heat impedance value is less than the preset second heat impedance threshold, the linear interpolation method is used for optimization to obtain the laser energy density parameters, and when the heat impedance value is less than the preset first heat impedance threshold and greater than the preset second heat impedance threshold, the initial laser energy density parameters are determined as the laser energy density parameters.

[0076] It should be noted that the thermal characteristic parameters of each mesh refer to the thermophysical parameters corresponding to each mesh unit after the material surface is meshed in the finite element analysis method, including specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and heat loss coefficient. These parameters reflect the thermal response characteristics of the material in the local area. By dividing the material surface into multiple small mesh units and assigning corresponding thermal characteristic parameters to each mesh unit according to the thermophysical properties of the material, the heat conduction process on the material surface can be accurately simulated. The laser energy density parameter is obtained during the calculation process and represents the distribution density of laser energy on the material surface in different regions. The laser energy density parameter is derived from the heat impedance value of the material and the initial thermal characteristic model and is optimized by the gradient descent method and the linear interpolation method. In practical applications, the laser energy density parameter is used to control the energy output of the laser, ensuring appropriate energy input in different regions during the annealing process, so as to achieve a uniform heating effect and avoid material surface quality problems caused by too high or too low energy density.

[0077] In step S13, according to the laser energy density parameter, a dynamic simulation of the material surface temperature field is carried out using a preset laser scanning path to identify the regions with significant thermal hysteresis effect and extract the thermal strain coefficient, and a temperature-thermal strain mapping relationship prediction model is established.

[0078] In one implementation, according to the laser energy density parameter, a dynamic simulation of the material surface temperature field is carried out using a preset laser scanning path to identify the regions with significant thermal hysteresis effect and extract the thermal strain coefficient, and a temperature-thermal strain mapping relationship prediction model is established, including:

[0079] According to the laser energy density parameter, perform dynamic simulation of the temperature field through a preset grid scanning path to generate a dynamic temperature distribution map containing the temperature gradient distribution and heat conduction characteristics; according to the dynamic temperature distribution map, analyze the relationship between the temperature change rate of adjacent regions and the preset thermal response threshold to obtain the regions with significant thermal hysteresis effect; according to the regions with significant thermal hysteresis effect, extract the thermal strain coefficient values at the corresponding positions and generate a data set to obtain the thermal strain coefficient data set; according to the dynamic temperature distribution map and the thermal strain coefficient data set, construct the mapping relationship between the surface temperature and thermal strain of the material. When non-linear characteristics appear in the mapping relationship, use the polynomial regression algorithm to fit the variation law of the thermal strain coefficient to obtain a linear mapping relationship; according to the mapping relationship, the dynamic temperature distribution map and the thermal strain coefficient data set, classify the thermal strain response mode on the material surface to obtain the thermal strain response mode classification map; according to the thermal strain response mode classification map, combined with the boundary condition constraints of the heat conduction equation, construct a temperature-thermal strain mapping relationship prediction model.

[0080] It should be noted that the dynamic temperature distribution map is generated by performing dynamic simulation of the temperature field according to the laser energy density parameter and the preset grid scanning path. The dynamic temperature distribution map reflects the temperature gradient distribution and heat conduction characteristics at different positions on the material surface. Through the dynamic temperature distribution map, the temperature distribution in different regions can be observed. Especially during the laser scanning process, the change law of temperature and the thermal hysteresis effect can be accurately captured. The temperature-thermal strain mapping relationship prediction model is constructed by analyzing the relationship between the dynamic temperature distribution map and the thermal strain coefficient data set, which is a mapping model between the surface temperature and thermal strain of the material. The temperature-thermal strain mapping relationship prediction model is used to predict the thermal strain that occurs in the material at a specific temperature. When non-linear characteristics appear in the mapping relationship during the temperature change process, the polynomial regression algorithm can be used for fitting to obtain a more accurate linear mapping relationship. This temperature-thermal strain mapping relationship prediction model can not only provide a prediction basis for the thermal strain behavior of the material, but also provide scientific support for temperature regulation during the laser annealing process, helping to achieve more precise temperature control and avoid excessive thermal strain from affecting the material performance.

[0081] In step S14, it is necessary to dynamically adjust the laser power ratio and pulse width timing according to the temperature-thermal strain mapping relationship prediction model to obtain the temperature field distribution data.

[0082] In one implementation, dynamically adjusting the laser power ratio and pulse width timing according to the temperature-thermal strain mapping relationship prediction model to obtain the temperature field distribution data includes:

[0083] According to the temperature-thermal strain mapping relationship prediction model, analyze the dynamic characteristics of the temperature rise and fall of the regional materials to obtain the thermal response curves of the materials in each region; according to the thermal response curves, combined with the preset thermal stress constraint conditions, perform calculations to obtain the required heating rate corresponding to each region and the required cooling rate corresponding to each region; according to the required heating rate and the required cooling rate, adopt an adaptive control algorithm to analyze the dynamic deviation between the real-time temperature monitoring data and the target temperature, match the heat capacity characteristics and thermal conduction delay response characteristics of the regional materials, and obtain the laser power ratio parameter; according to the required heating rate and the required cooling rate, dynamically adjust the laser duty cycle through a preset pulse width timing control strategy. When in the heating stage, determine the laser duty cycle in the high-frequency pulse mode as the pulse width timing control parameter. When in the cooling stage, determine the laser duty cycle in the low-frequency pulse mode as the pulse width timing control parameter; according to the laser power ratio parameter and the pulse width timing control parameter, combined with the three-dimensional heat conduction equation, perform dynamic simulation to obtain the temperature field distribution data.

[0084] It should be noted that the laser power ratio parameter is a parameter dynamically adjusted by adopting an adaptive control algorithm after analyzing the combination of the thermal response curves of the regional materials and the real-time temperature monitoring data. The laser power ratio parameter reflects the laser power distribution ratio required for different regions, aiming to accurately control the temperature change during the heating and cooling stages according to the heat capacity characteristics and thermal conduction delay response characteristics of the materials. The temperature field distribution data is the temperature distribution map obtained through dynamic simulation, calculated according to the three-dimensional heat conduction equation by combining the laser power ratio parameter and the pulse width timing control parameter. By analyzing the temperature distribution in each region, the temperature change situation of each point on the entire material surface during the laser scanning process can be obtained. The temperature field distribution data not only helps to monitor and control the temperature of the materials, but also provides a basis for further temperature regulation decisions to ensure that the temperature distribution can reach the best control effect during the laser annealing process.

[0085] In step S15, it is necessary to calculate the node centrality and the clustering coefficient of the connected domain according to the temperature field distribution data, combined with the multi-point temperature data of the material surface collected in real time, and perform the shortest path analysis of the graph structure to obtain the temperature regulation decision.

[0086] In one implementation, according to the temperature field distribution data, combined with the multi-point temperature data of the material surface collected in real time, adopt a graph theory algorithm to calculate the node centrality and the clustering coefficient of the connected domain, and perform the shortest path analysis of the graph structure to obtain the temperature regulation decision, including:

[0087] Based on the temperature field distribution data and combined with the multi-point temperature data of the material surface collected in real time, topological analysis is carried out. Each temperature point is used as a node in the graph structure, and the temperature gradient value is calculated as the edge weight to generate a temperature distribution map; according to the temperature distribution map, the degree centrality and eigenvector centrality of each node are calculated. When the node degree is greater than the preset degree threshold, it is determined as a key temperature control node; according to the temperature distribution map and the key temperature control node, the distribution characteristics of the connected domain are analyzed, and the independent connected domain is determined as the temperature isolation area; according to the temperature distribution map and the temperature isolation area, the clustering coefficient of each connected domain is calculated to identify abnormal temperatures and obtain the temperature abnormal area; according to the temperature distribution map, the key temperature control node and the temperature abnormal area, the shortest heat conduction path between key temperature control nodes is calculated using the shortest path algorithm to obtain the energy transfer efficiency of the temperature field; according to the energy transfer efficiency, the temperature distribution map, the key temperature control node and the temperature abnormal area, a multi-factor collaborative analysis is carried out by integrating the node space distribution characteristics, regional heat gradient characteristics and the shortest heat conduction path to obtain a temperature control decision.

[0088] It should be noted that the temperature distribution map is a graph structure generated by topological analysis by combining the temperature field distribution data and the multi-point temperature data of the material surface collected in real time. Each temperature point is regarded as a node in the graph structure, and the temperature gradient value between nodes is used as the weight of the edge. Through the temperature distribution map, the temperature distribution characteristics of the material surface can be intuitively displayed to identify key areas. The temperature control decision is a decision obtained through the shortest path analysis of graph theory algorithms based on the analysis results of the temperature distribution map. The temperature control decision comprehensively considers multiple factors such as the temperature abnormal area, key temperature control nodes and temperature isolation areas, and identifies the path with the optimal energy transfer efficiency through the shortest path analysis, so as to effectively control the temperature of the material to ensure that the temperature distribution during the laser annealing process can accurately reach the predetermined target. The core of the temperature control decision is to use multi-factor collaborative analysis to comprehensively consider the regional heat gradient, node space distribution and the efficiency of the heat conduction path, and finally achieve precise control of the temperature.

[0089] In step S16, according to the temperature control decision, the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data needs to be calculated. The temperature control decision and the temperature deviation value are input into a preset control adjustment model to calculate a multi-parameter dynamic adjustment vector including laser power, pulse width and energy density, and the parameters are iteratively optimized by combining reinforcement learning to obtain a temperature control scheme.

[0090] In one implementation, according to the temperature regulation decision, calculate the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data, input the temperature regulation decision and the temperature deviation value into a preset control and adjustment model to calculate a multi-parameter dynamic adjustment vector including laser power, pulse width, and energy density, and combine reinforcement learning to iteratively optimize the parameters to obtain a temperature control scheme, including:

[0091] Perform spatial matching and calculation on the temperature distribution data collected in real time and the preset temperature distribution data to obtain the temperature deviation value; use the historical temperature regulation decision and the historical temperature deviation value as inputs, and the historical multi-parameter dynamic adjustment vector as the output to construct and train a control and adjustment model. When the number of training times is greater than or equal to the preset number of training times, it is determined that the training is completed, and a trained control and adjustment model is obtained; input the temperature regulation decision and the temperature deviation value into the trained control and adjustment model to obtain a multi-parameter dynamic adjustment vector including laser power, pulse width, and energy density; according to the multi-parameter dynamic adjustment vector, use the deep deterministic policy gradient algorithm, combined with the boundary conditions of the three-dimensional heat conduction equation to verify the physical feasibility of the adjustment parameters, and perform iterative optimization to obtain a temperature control scheme.

[0092] It should be noted that the deep deterministic policy gradient algorithm is an algorithm based on reinforcement learning, which is used to solve the decision optimization problem in a continuous action space. In the present invention, the deep deterministic policy gradient algorithm is used to optimize the multi-parameter dynamic adjustment vector for temperature control, including laser power, pulse width, and energy density. The deep deterministic policy gradient algorithm simulates the process of temperature regulation decision by training a neural network model and performs iterative adjustment according to the environmental feedback. In each iteration, the deep deterministic policy gradient algorithm will verify the physical feasibility of the adjustment parameters based on the current temperature deviation value and control decision, and use the boundary conditions of the three-dimensional heat conduction equation to gradually adjust the parameter values, and finally obtain an optimized temperature control scheme. The temperature control scheme is a specific adjustment strategy for laser power, pulse width, and energy density given according to the optimization result of the deep deterministic policy gradient algorithm, which is used to accurately control the temperature distribution on the material surface during the laser annealing process to ensure that the preset temperature control target is achieved.

[0093] For the convenience of understanding the present invention, some preferred embodiments of the present invention will be further described below.

[0094] The working process of the present invention will be described below with a relatively common scenario as an example. Please refer to Figure 2 , which is Figure 1 a schematic diagram of the working scenario of the method.

[0095] In the laser annealing process for doping activation of semiconductor wafers, the present invention realizes precise regulation of the temperature field on the surface of large-area wafers through the collaborative optimization of multi-physics field modeling and intelligent control algorithms. Taking the manufacturing of 12-inch silicon-based wafers as an example, the surface thereof needs to complete the uniform activation of boron and phosphorus impurities through laser annealing. However, due to the curvature change at the wafer edge and the thickness difference of the thin film layer, the traditional method is prone to the problem of uneven activation efficiency between the center and the edge. The temperature control method of the present invention realizes large-range temperature crossing and fine adjustment in an extremely short time through the following process:

[0096] First, obtain the basic parameters such as the specific heat capacity, thermal conductivity, and thermal diffusivity of the thin film layer on the wafer surface. At the same time, use a radiation spectrometer to measure its thermal emissivity and thermal reflectivity, and combine the thermal loss test data in the vacuum chamber to construct a material thermal property model including multi-dimensional thermal property parameters. In this process, use the support vector machine algorithm to extract the characteristics of the non-linear relationship between the thermal conductivity and the thermal diffusivity, use the data fusion algorithm to correlate and match the thermal emissivity with the basic material thermal property model, and finally obtain a material thermal property model that accurately characterizes the thermal response of the wafer surface through linear regression optimization.

[0097] Based on the material thermal property model, use the finite element analysis method to divide the wafer surface into hexagonal meshes to obtain the thermal property parameters of each mesh. Calculate the thermal impedance value of each mesh unit through the heat conduction equation. For the edge bending area with a thermal impedance higher than the threshold, use the gradient descent method to gradually reduce the input of the laser energy density to avoid temperature overshoot caused by heat accumulation; for the central flat area with a thermal impedance lower than the threshold, generate a smooth energy density transition band between adjacent meshes through linear interpolation to ensure continuous energy distribution. The laser energy density parameters formed in this process accurately match the geometric shape and thermal property distribution of the wafer surface.

[0098] Subsequently, perform dynamic simulation on the temperature field of the wafer surface with a preset spiral progressive laser scanning path. During the simulation process, monitor the temperature change rate of each mesh unit in real time. When the temperature difference change between adjacent areas exceeds the set threshold, mark it as a region with significant thermal hysteresis effect. By extracting the thermal strain coefficient of the region with significant thermal hysteresis effect, establish a linear mapping relationship between temperature and thermal strain, and use the polynomial regression algorithm to perform piecewise linear fitting on the thermal strain response in the non-linear interval, and then classify the thermal strain response mode of the material surface to generate a thermal strain response mode classification map that can predict the thermal strain behavior under different temperature gradients. According to the thermal strain response mode classification map, combined with the boundary condition constraints of the heat conduction equation, construct a temperature-thermal strain mapping relationship prediction model. The temperature-thermal strain mapping relationship prediction model can accurately reflect the deformation trend of the wafer during the rapid heating and cooling process, providing a theoretical basis for subsequent parameter adjustment.

[0099] Enter the actual processing stage. According to the temperature-thermal strain mapping relationship prediction model, analyze the dynamic characteristics of the temperature rise and fall of the regional materials, obtain the thermal response curves of the materials in each region, and combine the preset thermal stress constraint conditions to perform calculations to obtain the required heating rate corresponding to each region and the required cooling rate corresponding to each region. Then, use an adaptive control algorithm to analyze the dynamic deviation between the real-time temperature monitoring data and the target temperature, and dynamically adjust the laser parameters. In the heating stage, adopt a high-frequency pulse mode for the high-thermal conductivity region to improve the thermal input efficiency, and achieve rapid heating by increasing the laser duty cycle; in the cooling stage, switch to a low-frequency pulse mode, use the pulse interval to promote heat diffusion, and perform dynamic simulation in combination with the three-dimensional heat conduction equation to obtain the temperature field distribution data.

[0100] To optimize the coordination of the large-scale temperature field, construct a graph theory network from the multi-point temperature data of the material surface collected in real time. Take each temperature point as a node in the graph structure, calculate the temperature gradient value as the edge weight, generate a temperature distribution map, calculate the degree centrality of each node to identify the key temperature control nodes that have a significant impact on the global temperature field. Analyze the distribution characteristics of the connected domains, determine the independent connected domains as temperature isolation regions, and use the Dijkstra algorithm to analyze the shortest heat conduction paths between the key temperature control nodes. Conduct a multi-element collaborative analysis by integrating the node spatial distribution characteristics, regional thermal gradient characteristics, and the shortest heat conduction paths to obtain a temperature control decision.

[0101] Finally, perform spatial matching and calculation on the temperature distribution data collected in real time and the preset temperature distribution data to obtain the temperature deviation value. Input the temperature control decision and the temperature deviation value into a pre-trained control and adjustment model, and output a multi-parameter dynamic adjustment vector including laser power, pulse width, and energy density. The reinforcement learning framework uses the temperature uniformity index as the reward function and continuously optimizes the adjustment parameter combination through the policy gradient algorithm. During the iteration process, the physical constraints of the three-dimensional heat conduction equation ensure that the parameter adjustment conforms to the actual heat conduction law, avoiding adjustment schemes that are theoretically feasible but physically impossible. The final control scheme generated after multiple rounds of optimization enables the surface of a 12-inch wafer to complete the transition from room temperature to the target temperature within 3 seconds, and the temperature difference across the entire region is controlled within the threshold range required by the process, successfully achieving uniform doping of the sub-micron activation layer.

[0102] In summary, the present invention discloses a temperature control method for laser annealing, which includes obtaining the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity and thermal loss coefficient of the material surface, and establishing a material thermal property model integrating multi-dimensional thermal property parameters; according to the material thermal property model, using the finite element analysis method to perform mesh division on the material, calculating the thermal impedance values of each region, and combining gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region; according to the laser energy density parameters, using a preset laser scanning path to dynamically simulate the temperature field on the material surface, identifying the regions with significant thermal hysteresis effect and extracting the thermal strain coefficient, and establishing a temperature-thermal strain mapping relationship prediction model; according to the temperature-thermal strain mapping relationship prediction model, using an adaptive algorithm to dynamically adjust the laser power ratio and pulse width timing to obtain temperature field distribution data; according to the temperature field distribution data, combining the multi-point temperature data of the material surface collected in real time, using the graph theory algorithm, calculating the node centrality and the clustering coefficient of the connected domain, and performing the shortest path analysis of the graph structure to obtain a temperature control decision; according to the temperature control decision, calculating the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data, inputting the temperature control decision and the temperature deviation value into a preset control and adjustment model to calculate a multi-parameter dynamic adjustment vector including laser power, pulse width and energy density, and combining reinforcement learning to iteratively optimize the parameters to obtain a temperature control scheme.

[0103] The present invention establishes a material thermal property model integrating multi-dimensional thermal property parameters. First, the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and thermal loss coefficient of the material are obtained, and a basic material thermal property model is constructed based on these thermal property parameters. Further, a data fusion algorithm and a support vector machine algorithm are used to optimize the model, and a material thermal property model integrating multi-dimensional thermal property parameters is obtained. Subsequently, the surface of the material is meshed by the finite element analysis method, the thermal impedance values of each region are calculated, and the laser energy density parameters of each region are obtained by combining the gradient descent method and the linear interpolation method. According to the laser energy density parameters, the surface temperature field of the material is dynamically simulated through a preset laser scanning path, the regions with significant thermal hysteresis effects are identified, and the thermal strain coefficient is extracted. Finally, a temperature-thermal strain mapping relationship prediction model is established. Then, in combination with the temperature-thermal strain mapping relationship prediction model, the laser power ratio and pulse width time sequence are dynamically adjusted by an adaptive algorithm to obtain temperature field distribution data. On this basis, in combination with the multi-point temperature data of the material surface collected in real time, the node centrality and connected domain clustering coefficient are calculated by using the graph theory algorithm, and the temperature control decision is obtained through the shortest path analysis of the graph structure. Finally, the temperature control decision and the temperature deviation value are input into a preset control and adjustment model, a multi-parameter dynamic adjustment vector including laser power, pulse width, and energy density is calculated, and iterative optimization is carried out in combination with reinforcement learning to obtain a temperature control scheme. Therefore, the present invention can accurately control the temperature change during the laser annealing process, provide fast response and ensure temperature stability, and thus achieve a large range of temperature crossing and fine adjustment in an extremely short time.

[0104] Referring to Figure 2 , the second embodiment of the present invention provides a temperature control device for laser annealing, including:

[0105] A data acquisition module, configured to acquire the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and thermal loss coefficient of the material surface, and establish a material thermal property model integrating multi-dimensional thermal property parameters;

[0106] A grid calculation module, configured to mesh the material by the finite element analysis method according to the material thermal property model, calculate the thermal impedance values of each region, and obtain the laser energy density parameters of each region by combining the gradient descent method and linear interpolation optimization;

[0107] A simulation mapping module, configured to dynamically simulate the surface temperature field of the material through a preset laser scanning path according to the laser energy density parameters, identify the regions with significant thermal hysteresis effects and extract the thermal strain coefficient, and establish a temperature-thermal strain mapping relationship prediction model;

[0108] A dynamic adjustment module, configured to dynamically adjust the laser power ratio and pulse width time sequence by an adaptive algorithm according to the temperature-thermal strain mapping relationship prediction model to obtain temperature field distribution data;

[0109] A graph theory analysis module, which is configured to calculate the node centrality and the clustering coefficient of the connected domain by using a graph theory algorithm according to the temperature field distribution data and in combination with the multi-point temperature data of the material surface collected in real time, perform the shortest path analysis of the graph structure, and obtain a temperature regulation decision.

[0110] A result output module, which is configured to calculate the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data according to the temperature regulation decision, input the temperature regulation decision and the temperature deviation value into a preset control and adjustment model to calculate a multi-parameter dynamic adjustment vector including laser power, pulse width and energy density, and obtain a temperature control scheme by combining reinforcement learning to iteratively optimize the parameters.

[0111] It should be noted that a temperature control device for laser annealing provided by an embodiment of the present invention is used to execute all the process steps of a temperature control method for laser annealing in the above embodiment, and the working principles and beneficial effects of the two correspond one by one, so details will not be described herein again.

[0112] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a temperature control program for laser annealing. When the processor executes the computer program, the steps in the above embodiments of the temperature control method for laser annealing are implemented, such as Figure 1 the step S11 shown. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above device embodiments are implemented, such as the graph theory analysis module.

[0113] Exemplarily, the computer program can be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0114] The electronic device can be a computing device such as a desktop computer, a notebook, a palm computer, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device, and do not constitute a limitation to the electronic device. The electronic device may include more or fewer components than the above, or combine some components, or different components. For example, the electronic device may further include input / output devices, network access devices, a bus, etc.

[0115] The so-called processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device, and connects various parts of the entire electronic device through various interfaces and circuits.

[0116] The memory can be used to store the computer program and / or modules. The processor realizes various functions of the electronic device by running or executing the computer program and / or modules stored in the memory, and by calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0117] Among them, if the modules / units integrated in the electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice within the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0118] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0119] The specific embodiments described above further elaborate on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. It is particularly pointed out that for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A temperature control method for laser annealing, characterized in that Including: Obtain the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity and heat loss coefficient of the material surface, and establish a material thermal property model integrating multi-dimensional thermal property parameters; According to the material thermal property model, use the finite element analysis method to mesh the material, calculate the thermal impedance values of each region, and combine gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region; According to the laser energy density parameters, use a preset laser scanning path to dynamically simulate the temperature field on the material surface, identify the regions with significant thermal hysteresis effect and extract the thermal strain coefficient, and establish a temperature-thermal strain mapping relationship prediction model; According to the temperature-thermal strain mapping relationship prediction model, use an adaptive algorithm to dynamically adjust the laser power ratio and pulse width timing to obtain the temperature field distribution data; According to the temperature field distribution data, combined with the multi-point temperature data of the material surface collected in real time, use the graph theory algorithm to calculate the node centrality and the clustering coefficient of the connected domain, and perform the shortest path analysis of the graph structure to obtain the temperature control decision; According to the temperature control decision, calculate the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data, and input the temperature control decision and the temperature deviation value into a preset control adjustment model to calculate the multi-parameter dynamic adjustment vector including laser power, pulse width and energy density, and combine reinforcement learning to iteratively optimize the parameters to obtain the temperature control scheme.

2. The temperature control method for laser annealing according to claim 1, wherein, The obtaining the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity and heat loss coefficient of the material surface, and establishing a material thermal property model integrating multi-dimensional thermal property parameters includes: Obtain the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity and heat loss coefficient of the material surface; According to the specific heat capacity, the thermal conductivity and the thermal diffusivity, use the support vector machine algorithm to perform multi-source collaborative analysis and construct a basic material thermal property model; According to the basic material thermal property model, thermal emissivity, thermal reflectivity and heat loss coefficient, use the data fusion algorithm to perform parameter correlation matching to obtain a fused material thermal property model; According to the fused material thermal property model, use the linear regression algorithm to optimize the model parameters to obtain a material thermal property model integrating multi-dimensional thermal property parameters.

3. The temperature control method for laser annealing according to claim 1, wherein, The according to the material thermal property model, using the finite element analysis method to mesh the material, calculate the thermal impedance values of each region, and combining gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region includes: According to the material thermal property model, use the finite element analysis method to perform polygon topological structure meshing on the material surface to obtain the thermal property parameters of each mesh; According to the thermal property parameters of each mesh, combined with the heat conduction equation, perform thermal impedance calculation to obtain the thermal impedance values of different regions; According to the thermal impedance values, perform thermodynamic derivation to obtain the initial laser energy density parameters; According to the initial laser energy density parameter and the thermal impedance value, when the thermal impedance value is greater than a preset first thermal impedance threshold, the gradient descent method is used for optimization to obtain the laser energy density parameter; when the thermal impedance value is less than a preset second thermal impedance threshold, the linear interpolation method is used for optimization to obtain the laser energy density parameter; when the thermal impedance value is less than the preset first thermal impedance threshold and greater than the preset second thermal impedance threshold, the initial laser energy density parameter is determined as the laser energy density parameter.

4. The temperature control method for laser annealing according to claim 1, characterized in that, Based on the laser energy density parameter, a dynamic simulation of the surface temperature field of the material is carried out using a preset laser scanning path, the regions with significant thermal hysteresis effects are identified and the thermal strain coefficients are extracted, and the establishment of the temperature-thermal strain mapping relationship prediction model includes: According to the laser energy density parameter, a dynamic simulation of the temperature field is carried out through a preset grid scanning path to generate a dynamic temperature distribution map including the temperature gradient distribution and heat conduction characteristics; According to the dynamic temperature distribution map, the relationship between the temperature change rate of adjacent regions and a preset thermal response threshold is analyzed to obtain the regions with significant thermal hysteresis effects; According to the regions with significant thermal hysteresis effects, the thermal strain coefficient values at the corresponding positions are extracted and a data set is generated to obtain the thermal strain coefficient data set; According to the dynamic temperature distribution map and the thermal strain coefficient data set, the mapping relationship between the surface temperature and thermal strain of the material is constructed. When non-linear characteristics appear in the mapping relationship, the polynomial regression algorithm is used to fit the variation law of the thermal strain coefficient to obtain a linear mapping relationship; According to the mapping relationship, the dynamic temperature distribution map and the thermal strain coefficient data set, the thermal strain response modes of the material surface are classified to obtain the thermal strain response mode classification map; According to the thermal strain response mode classification map and combined with the boundary condition constraints of the heat conduction equation, the temperature-thermal strain mapping relationship prediction model is constructed.

5. The temperature control method for laser annealing according to claim 1, characterized in that, Based on the temperature-thermal strain mapping relationship prediction model, an adaptive algorithm is used to dynamically adjust the laser power ratio and pulse width timing sequence to obtain the temperature field distribution data, including: According to the temperature-thermal strain mapping relationship prediction model, the dynamic characteristics of the temperature rise and fall of the regional materials are analyzed to obtain the thermal response curves of the materials in each region; According to the thermal response curves and combined with the preset thermal stress constraint conditions, calculations are carried out to obtain the required temperature rise rate corresponding to each region and the required temperature fall rate corresponding to each region; According to the required temperature rise rate and the required temperature fall rate, an adaptive control algorithm is used to analyze the dynamic deviation between the real-time temperature monitoring data and the target temperature, and the heat capacity characteristics and thermal conduction delay response characteristics of the regional materials are matched to obtain the laser power ratio parameter; According to the required temperature rise rate and the required temperature fall rate, the laser duty cycle is dynamically adjusted through a preset pulse width timing control strategy. When in the heating stage, the laser duty cycle in the high-frequency pulse mode is determined as the pulse width timing control parameter, and when in the cooling stage, the laser duty cycle in the low-frequency pulse mode is determined as the pulse width timing control parameter; According to the laser power ratio parameter and the pulse width timing control parameter, dynamic simulation is carried out in combination with the three-dimensional heat conduction equation to obtain temperature field distribution data.

6. The temperature control method for laser annealing according to claim 1, wherein Based on the temperature field distribution data and combined with the multi-point temperature data of the material surface collected in real time, using graph theory algorithms, calculate the node centrality and the clustering coefficient of the connected domain, and perform the shortest path analysis of the graph structure to obtain the temperature regulation decision, including: Based on the temperature field distribution data and combined with the multi-point temperature data of the material surface collected in real time, perform topological analysis. Taking each temperature point as a node in the graph structure, calculate the temperature gradient value as the edge weight to generate a temperature distribution map; According to the temperature distribution map, calculate the degree centrality and eigenvector centrality of each node. When the node degree is greater than the preset degree threshold, it is determined as a key temperature regulation node; Based on the temperature distribution map and the key temperature regulation nodes, analyze the distribution characteristics of the connected domain, and determine the independent connected domain as the temperature isolation area; Based on the temperature distribution map and the temperature isolation area, calculate the clustering coefficient of each connected domain, perform abnormal temperature identification, and obtain the temperature abnormal area; Based on the temperature distribution map, the key temperature regulation nodes, and the temperature abnormal area, use the shortest path algorithm to calculate the shortest heat conduction path between the key temperature regulation nodes to obtain the energy transfer efficiency of the temperature field; Based on the energy transfer efficiency, the temperature distribution map, the key temperature regulation nodes, and the temperature abnormal area, perform multivariate collaborative analysis by integrating the node spatial distribution characteristics, regional heat gradient characteristics, and the shortest heat conduction path to obtain the temperature regulation decision.

7. The temperature control method for laser annealing according to claim 1, wherein, Based on the temperature regulation decision, calculate the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data. Input the temperature regulation decision and the temperature deviation value into the preset control adjustment model to calculate the multi-parameter dynamic adjustment vector including laser power, pulse width, and energy density, and combine reinforcement learning to iteratively optimize the parameters to obtain the temperature control scheme, including: Perform spatial matching and calculation on the temperature distribution data collected in real time and the preset temperature distribution data to obtain the temperature deviation value; Use the historical temperature regulation decision and the historical temperature deviation value as inputs, and the historical multi-parameter dynamic adjustment vector as the output to construct and train a control adjustment model. When the number of training times is greater than or equal to the preset number of training times, it is determined that the training is completed, and the trained control adjustment model is obtained; Input the temperature regulation decision and the temperature deviation value into the trained control adjustment model to obtain a multi-parameter dynamic adjustment vector including laser power, pulse width, and energy density; According to the multi-parameter dynamic adjustment vector, use the deep deterministic policy gradient algorithm, and combine the boundary conditions of the three-dimensional heat conduction equation to verify the physical feasibility of the adjustment parameters, and perform iterative optimization to obtain the temperature control scheme.

8. A temperature control device for laser annealing, characterized in that, Including: A data acquisition module for acquiring the specific heat capacity, thermal conductivity, thermal diffusivity, thermal emissivity, thermal reflectivity, and heat loss coefficient of the material surface, and establishing a material thermal characteristic model integrating multi-dimensional thermal characteristic parameters; A grid computing module, which is used to divide the grid of the material by using the finite element analysis method according to the material thermal characteristic model, calculate the thermal impedance values of each region, and combine gradient descent method and linear interpolation optimization to obtain the laser energy density parameters of each region; A simulation mapping module, which is used to dynamically simulate the surface temperature field of the material according to the laser energy density parameters by using a preset laser scanning path, identify the regions with significant thermal hysteresis effect and extract the thermal strain coefficients, and establish a temperature-thermal strain mapping relationship prediction model; A dynamic adjustment module, which is used to dynamically adjust the laser power ratio and pulse width timing by using an adaptive algorithm according to the temperature-thermal strain mapping relationship prediction model to obtain the temperature field distribution data; A graph theory analysis module, which is used to calculate the node centrality and the clustering coefficient of the connected domain by using graph theory algorithms according to the temperature field distribution data and in combination with the multi-point temperature data of the material surface collected in real time, and perform the shortest path analysis of the graph structure to obtain the temperature control decision; A result output module, which is used to calculate the temperature deviation value between the temperature distribution data collected in real time and the preset temperature distribution data according to the temperature control decision, input the temperature control decision and the temperature deviation value into a preset control adjustment model to calculate a multi-parameter dynamic adjustment vector including laser power, pulse width and energy density, and combine reinforcement learning to iteratively optimize the parameters to obtain a temperature control scheme.

9. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the temperature control method for laser annealing according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the temperature control method for laser annealing according to any one of claims 1 to 7.

Images