Material laser ablation effect finite element model parameter inversion method based on experimental process feature matching

Abstract

The application discloses a material laser ablation effect finite element model parameter inversion method based on experimental process characteristic matching, solves the problem that existing finite element model parameter inversion technology has limited application range, cannot fully represent the characteristic quantity time evolution process, results in insufficient inversion accuracy, and cannot meet the high-precision requirement of the material laser ablation effect finite element model under the condition of multiple physical fields, and the like. The application obtains model item calculation deviations through finite element model verification experiments and Latin hypercube experimental design methods, and then provides training data for a machine learning model to obtain an optimal machine learning model, inputs expected model item calculation deviations into the optimal machine learning model, realizes material laser ablation effect finite element model parameter inversion, and has wide inversion range and high precision.

Description

Technical Field

[0001] This invention relates to the technique of parameter inversion of finite element model of laser ablation effect of materials, specifically to a method for parameter inversion of finite element model of laser ablation effect of materials based on experimental process feature matching. Background Technology

[0002] Inversion techniques for parameters in finite element model (FEM) of laser ablation effects on materials are a crucial method for improving the accuracy of FEM models. Under multiphysics conditions, the laser ablation effect of materials (especially composite materials) involves numerous complex physical processes, including but not limited to thermal diffusion, phase transition, thermal decomposition, stress response, and gas flow coupling. Two key challenges exist in constructing such FEM models: firstly, the variables involved in the governing equations of some physical processes are difficult or impossible to measure; secondly, some physical processes lack corresponding governing equations for accurate description and require approximate characterization through lumped parameters within the FEM model. Therefore, to effectively improve the reliability and computational accuracy of FEM models, it is essential to invert and optimize the model parameters in a high-dimensional parameter space.

[0003] Existing finite element model parameter inversion techniques are mainly divided into the following two types, both of which have obvious limitations.

[0004] The first approach uses neural network technology, embedding the partial differential equation (PDE) containing the parameters to be inverted into a loss function. During neural network training, the parameters are optimized and inverted. The core limitation of this method is that the parameters to be inverted must be strictly contained within the PDE equation. Furthermore, to satisfy the partial derivative form requirements of the PDE equation, the input layer of the neural network is typically limited to coordinates and time, severely restricting its applicability and making it unsuitable for parameter inversion needs corresponding to physical processes without explicit PDE equations.

[0005] The second approach involves large-scale calculations using a finite element model. This is achieved by establishing a mapping relationship between computational parameters and characteristic quantities using methods such as response surface methodology and machine learning. Then, optimization algorithms, such as genetic algorithms and particle swarm optimization, are employed to minimize the root mean square error between the calculated and measured characteristic quantities, thus obtaining a set of optimal computational parameters. The core limitation of this method is that the selected characteristic quantities are typically scalars (e.g., the highest temperature during laser heating of materials), which cannot fully characterize the temporal evolution of these characteristic quantities during parameter inversion. This results in insufficient accuracy of the inverted parameters, making it difficult to meet the high-precision requirements of finite element models for laser ablation effects under multiphysics. Summary of the Invention

[0006] To address the technical problem that existing finite element model parameter inversion techniques have limited applicability, cannot fully characterize the time evolution of characteristic quantities, resulting in insufficient inversion accuracy and difficulty in meeting the high-precision requirements of finite element models of material laser ablation effects under multi-physics conditions, this invention provides a parameter inversion method for finite element models of material laser ablation effects based on experimental process feature matching.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for parameter inversion of a finite element model of material laser ablation effect based on experimental process feature matching, characterized by the following steps:

[0009] Step 1: Divide the preset input parameters of the finite element model of material laser ablation effect into two types: working condition parameters and model parameters;

[0010] Step 2: Conduct a finite element model verification experiment based on the finite element model of the laser ablation effect of the material, and design and construct a set of working condition parameters according to the verification experiment parameters; use the Latin hypercube experimental design method to design the values ​​of each parameter in the model parameters to obtain the model parameter set;

[0011] Step 3: Use the finite element model of the laser ablation effect of the material to iteratively calculate the model parameter set and the working condition parameter set to obtain the calculation deviation of the model components; and combine the calculation deviation of the model components with the model parameter set to form a second dataset;

[0012] Step 4: Construct an initial machine learning model for parameter inversion of the finite element model of the laser ablation effect of materials, and train it using the second dataset to obtain the optimal machine learning model;

[0013] Step 5: Input the expected model's component calculation deviations into the optimal machine learning model to obtain the finite element model parameters of the material laser ablation effect finite element model, thus completing the parameter inversion of the material laser ablation effect finite element model.

[0014] Furthermore, in step 1:

[0015] The operating parameters include irradiation laser parameters, material size parameters, and environmental parameters.

[0016] Furthermore, in step 1:

[0017] The irradiated laser parameters include peak power density I peak Average power density I avg , Spot diameter d, Spot centroid coordinates (x s y s ), Irradiation time t laser ;

[0018] The material dimensional parameters include length l m Width w m and thickness h m ;

[0019] The environmental parameters include ambient temperature T. env and airflow velocity v air .

[0020] Further, in step 2, a finite element model verification experiment is conducted based on the finite element model of the laser ablation effect of the material, and a set of working condition parameters is designed and constructed according to the verification experiment parameters; specifically:

[0021] Based on the finite element model of the laser ablation effect of the material, a finite element model verification experiment was carried out. Based on the verification experiment parameters, a set of working condition parameters P was designed and constructed. exp ={p exp,kl}={I peak,k , I avg,k , d k , x s,k , y s,k , t laser,k , l m,k ,w m,k , h m,k , T env,k , v air,k}, k=1,2,...,N exp l=1,2,...,N wp k is the row number, l is the column number, and N is the column number. exp N represents the number of experiments. wp This represents the number of operating parameters used in each experiment.

[0022] Further, in step 2, the Latin hypercube experimental design method is used to design the values ​​of each parameter in the model parameters to obtain the model parameter set; specifically:

[0023] The Latin hypercube experimental design method was used to design the values ​​of each parameter in the model, resulting in the model parameter set P. model ={p model,ij}={T c,i ,t ab,i}, i=1,2,...,N mc j=1,2,...,N mp , where N mc N represents the number of sets of model parameters. mp T represents the number of model parameters in each set of model parameters. c Material removal temperature criterion, t ab This refers to the dwell time for material removal.

[0024] Further, in step 3, the finite element model of material laser ablation effect is used to iteratively calculate the model parameter set and the working condition parameter set to obtain the calculation deviation of the model components, specifically including:

[0025] Step 3.11: Obtain the model parameter set P model The values ​​of the i-th group are used to obtain the first dataset P. model,i ;

[0026] Step 3.12, from row 1 to row N exp Line-by-line traversal of the working condition parameter set P exp And set the operating parameters P for each row exp,k With the first dataset P model,i By combining these parameters, the specific input parameters of the finite element model of the laser ablation effect of the material can be obtained. Through specific input parameter C ik Get N exp Group model single calculation deviation {err t,k err s,k};

[0027] Step 3.13, based on N exp Group model single calculation deviation {err t,k err s,k The calculation model calculates the deviation E for each component. i ={E t,i E s,i};

[0028] Step 3.14: Connect i and N mc Compare;

[0029] If i<N mc If i is not found, increment i by 1, return to step 3.11, and proceed to the next iteration.

[0030] If i≥N mc Then the loop ends, and N is obtained. mc Group model component calculation deviation E i ={E t,i E s,i}

[0031] Further, in step 3, the model component calculation deviation is combined with the model parameter set to form a second dataset; specifically:

[0032] N mc Group model component calculation deviation E i ={E t,i E s,i} and model parameter set P model ={pmodel,ij A one-to-one correspondence is formed to create N. mc Group 2 dataset:

[0033] .

[0034] Furthermore, in step 3.12, the step of inputting specific parameter C... ik Get N exp Group model single calculation deviation {err t,k err s,k Specifically, it includes:

[0035] Step A: Input specific parameters C ik Input the finite element model of the laser ablation effect of the material and perform calculations to obtain the finite element model calculation results;

[0036] Step B: Extract the material ablation and perforation time t from the finite element model calculation results. m,穿孔 The perforation area s of the j´th output time step after perforation m,j´ Extract the material ablation and perforation time t from the results of the k-th verification experiment of the finite element model verification experiment. e,穿孔 The perforation area s of the j´th output time step after perforation e,j´ The calculation model's single calculation deviation {err t,k err s,k}:

[0037] ;

[0038] Where, N time This represents the total number of time steps in the output result after perforation.

[0039] Furthermore, it is characterized by:

[0040] Step 3.13 specifically involves, based on N exp Group model single calculation deviation {err t,k err s,k The calculation model calculates the deviation E for each component. i ={E t,i E s,i}:

[0041] .

[0042] Further, step 4 specifically involves constructing an initial machine learning model for parameter inversion of the finite element model of the material laser ablation effect, and training it using the second dataset D. During training, the input data is the calculated deviation E of each model component in the second dataset D. i ={E t,i Es,i The output data is the model parameter set corresponding to the second dataset D. This continues until the loss function of the machine learning model converges, thus obtaining the optimal machine learning model.

[0043] Step 5 specifically involves inputting the expected model's component calculation deviation [0, 0] into the optimal machine learning model to obtain the finite element model parameters of the material laser ablation effect finite element model, thus completing the parameter inversion of the material laser ablation effect finite element model.

[0044] The beneficial effects of this invention are:

[0045] 1. This invention obtains the calculation deviation of the model components through finite element model verification experiments and Latin hypercube experimental design method, and then provides training data for the machine learning model to obtain the optimal machine learning model. The expected calculation deviation of the model components is input into the optimal machine learning model to realize the parameter inversion of the finite element model of the laser ablation effect of materials. The inversion range is wide and the accuracy is high.

[0046] 2. This invention maps the model component calculation deviations of material ablation and perforation time, perforation area and other effect characteristic quantities to the finite element model parameters of material laser ablation effect through a data-driven approach. This solves the problem of difficult inversion of finite element model parameters in high-dimensional parameter space, especially the problem of no applicable inversion method when the model parameters contain lumped parameters without clear physical meaning.

[0047] 3. In the parameter inversion process, the present invention incorporates the perforation area calculation deviation constraint of the ablation hole evolution process, and uses machine learning methods to find the global optimum of the model parameters, which fully expresses the process matching of the material laser ablation effect characteristics. The reliability of the finite element model is improved after the model parameters are inverted. Attached Figure Description

[0048] Figure 1 This is a flowchart of an embodiment of the present invention, which is a method for parameter inversion of a finite element model of material laser ablation effect based on experimental process feature matching. Detailed Implementation

[0049] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] This invention provides a method for parameter inversion of a finite element model of material laser ablation effect based on experimental process feature matching, such as... Figure 1 As shown, it includes the following steps:

[0051] Step 1: Divide the preset input parameters of the finite element model of material laser ablation effect into two types: working condition parameters and model parameters;

[0052] The aforementioned operating parameters include irradiated laser parameters, material size parameters, and environmental parameters; the aforementioned irradiated laser parameters include peak power density I. peak Average power density I avg , Spot diameter d, Spot centroid coordinates (x s y s ), Irradiation time t laser The above material dimensional parameters include length l m Width w m and thickness h m The above environmental parameters include ambient temperature T. env and airflow velocity v air The above includes, but is not limited to, parameters with explicit physical meaning and lumped parameters without explicit physical meaning. In this embodiment, the model parameters include the ablation temperature threshold T. c and ablation removal duration t ab ;

[0053] Step 2: Based on the above finite element model of material laser ablation effect, conduct finite element model verification experiments, and design and construct a set of working condition parameters according to the verification experiment parameters; use the Latin hypercube experimental design method to design the values ​​of each parameter in the above model parameters to obtain the model parameter set;

[0054] Step 2.1: Based on the above finite element model of material laser ablation effect, conduct finite element model verification experiments, and design and construct the working condition parameter set P according to the verification experiment parameters. exp ={p exp,kl}={I peak,k , I avg,k , d k , x s,k , y s,k ,t laser,k , l m,k , w m,k , h m,k , T env,k , v air,k}, k=1,2,...,N exp l=1,2,...,N wp k is the row number, l is the column number, and N is the column number. exp N represents the number of experiments. In this embodiment, the total number of experiments is 10. wp This represents the number of operating parameters for each experiment.

[0055] Step 2.2: The Latin hypercube experimental design method is used to design the values ​​of each parameter in the above model parameters, resulting in the model parameter set P. model ={p model,ij}={T c,i ,t ab,i}, i=1,2,...,N mc j=1,2,...,N mp , where N mc N represents the number of sets of model parameters. mp T represents the number of model parameters in each set of model parameters. c Material removal temperature criterion, t ab The dwell time for material removal;

[0056] Step 3: Using the aforementioned finite element model of material laser ablation effect, iteratively calculate the model parameter set and the working condition parameter set to obtain the calculation deviations of the model components; then combine the calculation deviations of the model components with the model parameter set to form the second dataset:

[0057] Step 3.1: Using the above-mentioned finite element model of material laser ablation effect, the model parameter set P is... model and operating condition parameter set P exp To obtain the model's component calculation bias, perform a loop through the calculations using the following steps:

[0058] Step 3.11: Obtain the model parameter set P model The value of the i-th group ( (The initial value of i is 1), thus obtaining the first dataset P. model,i ;

[0059] Step 3.12, from row 1 to row N exp Line-by-line traversal of the working condition parameter set P exp And set the operating parameters P for each row exp,k With the first dataset P model,i By combining these parameters, the specific input parameters for the aforementioned finite element model of the material laser ablation effect are obtained. For each set of specific input parameters, the following calculation is performed to obtain N. exp Group model single calculation deviation {err t,k err s,k}:

[0060] Step A: Input specific parameters C ik Input the above finite element model of material laser ablation effect for calculation and obtain the finite element model calculation results;

[0061] Step B: Extract the material ablation and perforation time t from the finite element model calculation results. m,穿孔The perforation area s of the j´th output time step after perforation m,j´ Extract the material ablation and perforation time t from the results of the k-th verification experiment of the finite element model verification experiment. e,穿孔 The perforation area s of the j´th output time step after perforation e,j´ The calculation model's single calculation deviation {err t,k err s,k}:

[0062] ;

[0063] Where, N time This represents the total number of time steps in the output results after perforation.

[0064] Step 3.13, based on N exp Group model single calculation deviation {err t,k err s,k The calculation model calculates the deviation E for each component. i ={E t,i E s,i}:

[0065] ;

[0066] Step 3.14: Connect i and N mc Compare;

[0067] If i<N mc If i is not found, increment i by 1, return to step 3.11, and proceed to the next iteration.

[0068] If i≥N mc Then the loop ends, and N is obtained. mc Group model component calculation deviation E i ={E t,i E s,i};

[0069] Step 3.2, N mc Group model component calculation deviation E i ={E t,i E s,i} and model parameter set P model ={p model,ij A one-to-one correspondence is formed to create N. mc Group 2 dataset:

[0070] ;

[0071] Step 4: Construct an initial machine learning model for parameter inversion of the finite element model of the material laser ablation effect. Train the model using the second dataset D. During training, the input data is the calculated deviation E of each model component in the second dataset D. i ={E t,i E s,i The output data is the model parameter set corresponding to the second dataset D. This continues until the loss function of the machine learning model converges, thus obtaining the optimal machine learning model.

[0072] The aforementioned machine learning models include, but are not limited to, neural network models, reinforcement learning models, and support vector machines.

[0073] Step 5: Input the expected model's component calculation deviation [0, 0] into the optimal machine learning model to obtain the finite element model parameters of the material laser ablation effect finite element model, and complete the parameter inversion of the material laser ablation effect finite element model.

[0074] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present invention should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for parameter inversion of a finite element model of material laser ablation effect based on experimental process feature matching, characterized in that, Includes the following steps: Step 1: Divide the preset input parameters of the finite element model of material laser ablation effect into two types: working condition parameters and model parameters; Step 2: Conduct a finite element model verification experiment based on the finite element model of the laser ablation effect of the material, and design and construct a set of working condition parameters according to the verification experiment parameters; use the Latin hypercube experimental design method to design the values ​​of each parameter in the model parameters to obtain the model parameter set; Step 3: Use the finite element model of the laser ablation effect of the material to iteratively calculate the model parameter set and the working condition parameter set to obtain the calculation deviation of the model components; and combine the calculation deviation of the model components with the model parameter set to form a second dataset; Step 4: Construct an initial machine learning model for parameter inversion of the finite element model of the laser ablation effect of materials, and train it using the second dataset to obtain the optimal machine learning model; Step 5: Input the expected model's component calculation deviations into the optimal machine learning model to obtain the finite element model parameters of the material laser ablation effect finite element model, thus completing the parameter inversion of the material laser ablation effect finite element model.

2. The method for parameter inversion of the finite element model of material laser ablation effect based on experimental process feature matching according to claim 1, characterized in that, In step 1: The operating parameters include irradiation laser parameters, material size parameters, and environmental parameters.

3. The method for parameter inversion of the finite element model of material laser ablation effect based on experimental process feature matching according to claim 2, characterized in that, In step 1: The irradiated laser parameters include peak power density I peak Average power density I avg , Spot diameter d, Spot centroid coordinates (x s y s ), Irradiation time t laser ; The material dimensional parameters include length l m Width w m and thickness h m ; The environmental parameters include ambient temperature T. env and airflow velocity v air .

4. The method for parameter inversion of the finite element model of material laser ablation effect based on experimental process feature matching according to claim 3, characterized in that, In step 2, a finite element model verification experiment is conducted based on the finite element model of the laser ablation effect of the material, and a set of working condition parameters is designed and constructed according to the verification experiment parameters; specifically: Based on the finite element model of the laser ablation effect of the material, a finite element model verification experiment was carried out. Based on the verification experiment parameters, a set of working condition parameters P was designed and constructed. exp ={p exp,kl }={I peak,k , I avg,k , d k , x s,k , y s,k , t laser,k , l m,k , w m,k ,h m,k , T env,k , v air,k }, k=1,2,...,N exp l=1,2,...,N wp k is the row number, l is the column number, and N is the column number. exp N represents the number of experiments. wp This represents the number of operating parameters used in each experiment.

5. The method for parameter inversion of the finite element model of material laser ablation effect based on experimental process feature matching according to claim 4, characterized in that, In step 2, the Latin hypercube experimental design method is used to design the values ​​of each parameter in the model parameters to obtain the model parameter set; specifically: The Latin hypercube experimental design method was used to design the values ​​of each parameter in the model, resulting in the model parameter set P. model ={p model,ij }={T c,i ,t ab,i }, i=1,2,...,N mc j=1,2,...,N mp , where N mc N represents the number of sets of model parameters. mp T represents the number of model parameters in each set of model parameters. c Material removal temperature criterion, t ab This refers to the dwell time for material removal.

6. The method for parameter inversion of the finite element model of material laser ablation effect based on experimental process feature matching according to claim 5, characterized in that, In step 3, the finite element model of material laser ablation effect is used to iteratively calculate the model parameter set and the working condition parameter set to obtain the calculation deviation of each item in the model. Specifically, this includes: Step 3.11: Obtain the model parameter set P model The values ​​of the i-th group are used to obtain the first dataset P. model,i ; Step 3.12, from row 1 to row N exp Line-by-line traversal of the working condition parameter set P exp And set the operating parameters P for each row exp,k With the first dataset P model,i By combining these parameters, the specific input parameters of the finite element model of the laser ablation effect of the material can be obtained. Through specific input parameter C ik Get N exp Group model single calculation deviation {err t,k err s,k }; Step 3.13, based on N exp Group model single calculation deviation {err t,k err s,k The calculation model calculates the deviation E for each component. i ={E t,i E s,i }; Step 3.14: Connect i and N mc Compare; If i<N mc If i is not found, increment i by 1, return to step 3.11, and proceed to the next iteration. If i≥N mc Then the loop ends, and N is obtained. mc Group model component calculation deviation E i ={E t,i E s,i } 7. The method for parameter inversion of the finite element model of material laser ablation effect based on experimental process feature matching according to claim 6, characterized in that, In step 3, the model component calculation deviation is combined with the model parameter set to form the second dataset; specifically: N mc Group model component calculation deviation E i ={E t,i E s,i } and model parameter set P model ={p model,ij A one-to-one correspondence is formed to create N. mc Group 2 dataset: 。 8. The method for parameter inversion of the finite element model of material laser ablation effect based on experimental process feature matching according to claim 7, characterized in that, In step 3.12, the specific input parameter C is used. ik Get N exp Group model single calculation deviation {err t,k err s,k Specifically, it includes: Step A: Input specific parameters C ik Input the finite element model of the laser ablation effect of the material and perform calculations to obtain the finite element model calculation results; Step B: Extract the material ablation and perforation time t from the finite element model calculation results. m,穿孔 The perforation area s of the j´th output time step after perforation m,j´ Extract the material ablation and perforation time t from the results of the k-th verification experiment of the finite element model verification experiment. e,穿孔 The perforation area s of the j´th output time step after perforation e,j´ The calculation model's single calculation deviation {err t,k err s,k }: ； Where, N time This represents the total number of time steps in the output result after perforation.

9. The method for parameter inversion of the finite element model of material laser ablation effect based on experimental process feature matching according to claim 8, characterized in that: Step 3.13 specifically involves, based on N exp Group model single calculation deviation {err t,k err s,k The calculation model calculates the deviation E for each component. i ={E t,i E s,i }: 。 10. The method for parameter inversion of the finite element model of material laser ablation effect based on experimental process feature matching according to claim 9, characterized in that: Step 4 specifically involves constructing an initial machine learning model for parameter inversion of the finite element model of the material laser ablation effect, and training it using the second dataset D. During training, the input data consists of the calculated deviations E of each model component in the second dataset D. i ={E t,i E s,i The output data is the model parameter set corresponding to the second dataset D. This continues until the loss function of the machine learning model converges, thus obtaining the optimal machine learning model. Step 5 specifically involves inputting the expected model's component calculation deviation [0, 0] into the optimal machine learning model to obtain the finite element model parameters of the material laser ablation effect finite element model, thus completing the parameter inversion of the material laser ablation effect finite element model.

Images