A parameter optimization method for powder laser 3D printing molding process

Abstract

The present invention relates to the field of data processing technology, and more specifically, to a parameter optimization method for a powder laser 3D printing molding process, the method comprising: determining a solution space of a particle swarm algorithm, performing finite element analysis on each process parameter, determining an objective function of the particle swarm algorithm, obtaining simulated particles and measured particles in the solution space, and calculating the credibility of the simulated particles at any training time during iterative training of a pre-constructed support vector machine using measured particles, when the product of the credibility and the normalized simulated fitness value is greater than a preset value, and the measured fitness value of the simulated particles is greater than a conversion threshold, the simulated particles are used as measured particles in subsequent training processes, the training process is repeated, and the target process parameters are obtained after the training is completed. The present invention can improve the accuracy of the determined target process parameters while reducing the amount of calculation and time cost.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology. More specifically, the present invention relates to a parameter optimization method for a powder laser 3D printing process. Background Art

[0002] Powder laser 3D printing is an additive manufacturing technology mainly used for the processing of metal and alloy materials. This process uses a high-energy laser beam to melt metal powder layer by layer and solidify it into a predetermined shape to build parts. The printing parameters directly determine the quality and performance of the final product. In order to obtain high-quality finished products, it is crucial to optimize the printing parameters.

[0003] In the related technology, such as the patent application document with publication number CN108197384A, a parameter optimization method for powder laser 3D printing molding process is disclosed. The method includes: using a computer finite element program to analyze and calculate the temperature, stress and strain field of metal powder when it is molded under different process parameters, tracking the stress and strain values ​​during the part printing process, and predicting the deformation and defects of the part after molding under different process parameters, and then based on the analysis results, quickly optimizing the existing process parameters to quickly obtain the optimal process parameters for powder molding.

[0004] However, the above scheme only uses the finite element analysis model, selects discrete process parameters to input into the finite element analysis model, and compares the finite element analysis results to obtain the optimal discrete process parameters. The calculation amount is huge and the calculation efficiency is low. In addition, the obtained optimal discrete process parameters are greatly affected by the accuracy of discrete data selection, which may lead to low accuracy of the determined optimal discrete process parameters to a certain extent. Summary of the invention

[0005] In order to solve the problem of low efficiency and accuracy in determining optimal process parameters, the present invention proposes a parameter optimization method for a powder laser 3D printing process.

[0006] According to one aspect of an embodiment of the present invention, a method for optimizing parameters of a powder laser 3D printing process is provided, comprising:

[0007] The range of multiple process parameters of the workpiece to be printed is used as the solution space of the particle swarm algorithm, and finite element analysis is performed on each process parameter to determine the objective function of the particle swarm algorithm;

[0008] Get the simulated particles and measured particles in the preset solution space, and use the measured particles to iteratively train the pre-built support vector machine, and calculate the credibility of any simulated particle at any training time , is the number of particles measured during historical training, For the The number of particles measured during training, For the During the next training session, The simulated fitness value obtained by inputting the simulated particles into the support vector machine is is the first The measured fitness value of the process parameter at the location of the measured particle, , are the simulated particle and the measured particle, respectively. process parameters, is the number of process parameters, is the normalization function, is the absolute value symbol;

[0009] When the product of the credibility and the normalized simulation fitness value is greater than the preset value, and the measured fitness value of any simulated particle is greater than the conversion threshold, any simulated particle will be used as the measured particle in the subsequent training process and the training process will be repeated. After the training is completed, the target process parameters are obtained.

[0010] The present invention sets simulated particles and uses a trained support vector machine to determine the simulated measured fitness value of the simulated particles, which can reduce the amount of calculation for finite element analysis of each particle, and uses the characteristic of the support vector machine that can complete the regression task, improves the optimization process using the particle swarm algorithm, and can reduce the time cost when using the particle swarm algorithm to determine the optimal parameters; and in the training process of the support vector machine, when the credibility of any simulated particle meets certain conditions and the measured fitness value of the simulated particle is greater than a conversion threshold, the simulated particle can be converted into a measured particle to participate in the subsequent training of the support vector machine, thereby improving the search accuracy when determining the optimal solution of the process parameters and ensuring the accuracy of the determined target process parameters.

[0011] In an exemplary embodiment of the present invention, a method for obtaining a solution space of a particle swarm algorithm includes:

[0012] The three-dimensional CAD model of the printed part is discretized to obtain a finite element simulation model, and the ranges of multiple process parameters for simulation using the finite element simulation model are set, and the ranges of each process parameter are used as a solution space.

[0013] The present invention sets a plurality of ranges of process parameters in combination with a finite element simulation model corresponding to the workpiece to be printed, thereby ensuring that an optimal solution of the process parameters exists in the solution space.

[0014] In an exemplary embodiment of the present invention, finite element analysis is performed on each process parameter to determine the objective function of the particle swarm algorithm, including:

[0015] Using the finite element simulation model, the temperature field, stress field and strain field under various process parameters are simulated to obtain the residual stress function, deformation function, density function and defect function;

[0016] Using preset weights, the residual stress function, deformation function, density function and defect function are weighted summed to obtain the objective function of the particle swarm algorithm.

[0017] In an exemplary embodiment of the present invention, a method for obtaining a measured fitness value includes:

[0018] Input the process parameters of any particle location into the finite element analysis model, simulate the temperature field, stress field and strain field under the corresponding process parameters, and determine the corresponding residual stress, deformation, density and defects;

[0019] Substituting the residual stress, deformation, density and defects into the objective function, the measured fitness values ​​of the corresponding particles are obtained.

[0020] In an exemplary embodiment of the present invention, a method for obtaining simulated particles and measured particles in a solution space includes:

[0021] A number of particles are randomly selected from the solution space, and the particles are randomly divided according to a preset ratio to obtain simulated particles and measured particles.

[0022] The present invention can significantly reduce the number of finite element analyses by setting up simulated particles, thereby reducing the amount of calculation.

[0023] In an exemplary embodiment of the present invention, iterative training of a pre-built support vector machine using measured particles includes:

[0024] Take a number of particles as initial training samples, take the measured fitness values ​​corresponding to each particle in the initial training samples as labels, and perform the first training on the support vector machine;

[0025] The measured particles obtained by random division according to a preset ratio are used as basic measured particles, the initial training samples after the current training times minus one are used as basic samples, and the sample set consisting of basic samples, basic measured particles and simulated particles converted into measured particles determined in the historical training process is used as the target training samples for the current training. The measured fitness value corresponding to each particle in the target training sample is used as a label, and the support vector machine is iteratively trained.

[0026] The present invention trains a support vector machine based on initial training samples, thereby ensuring a sufficient number of samples, thereby ensuring the accuracy of the output results of the trained support vector machine.

[0027] In an exemplary embodiment of the present invention, a method for obtaining a conversion threshold includes:

[0028] Obtain the maximum value of the data sequence composed of the simulated fitness values ​​of all historical simulated particles and the measured fitness values ​​of all historical measured particles;

[0029] The preset conversion experience coefficient is obtained, and the product of the maximum value and the conversion experience coefficient is calculated to obtain the conversion threshold.

[0030] The present invention sets a conversion threshold based on the fitness value, so that the difference between the measured fitness value corresponding to the simulated particle that can be converted into the measured particle and the measured fitness value of the historical measured particle is small, thereby further ensuring the accuracy of the simulated particle that can be converted into the measured particle.

[0031] In an exemplary embodiment of the present invention, the process of obtaining the target process parameters includes:

[0032] During the training of the support vector machine, if the positions of all simulated particles and all measured particles tend to converge, the training is stopped;

[0033] When the training is stopped, the particle corresponding to the maximum value in the data sequence composed of the simulated fitness values ​​of all current simulated particles and the measured fitness values ​​of all current measured particles is taken as the target particle, and the process parameter corresponding to the position of the target particle is taken as the target process parameter.

[0034] The present invention has the following effects:

[0035] The present invention can comprehensively determine the credibility of simulated particles during training, and determine a condition based on the credibility of the simulated particles and a preset value, set a condition based on the measured fitness value of the simulated particles and a conversion threshold, and combine the two conditions to determine whether to convert the simulated particles into measured particles, which can ensure the accuracy of the simulated particles converted into measured particles, and converting the simulated particles into measured particles can increase the search accuracy of the measured particles, thereby improving the accuracy of the determined target process parameters; and the present invention uses the trained support vector machine to determine the simulated fitness value of the simulated particles, which can reduce the number of particles subjected to finite element analysis and the time cost of determining the simulated fitness value of the simulated particles. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0037] Figure 1It is a schematic flow chart of the steps of a parameter optimization method for a powder laser 3D printing molding process according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0039] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0040] Reference Figure 1 , a parameter optimization method for a powder laser 3D printing molding process, comprising steps S1 to S3, specifically as follows:

[0041] S1: Taking the range of multiple process parameters of the workpiece to be printed as the solution space of the particle swarm algorithm, performing finite element analysis on each process parameter, and determining the objective function of the particle swarm algorithm.

[0042] In an exemplary embodiment of the present invention, process parameters refer to parameters that can affect the quality and performance of the printed part when printing the printed part using a 3D printing molding process. The process parameters selected in this embodiment are powder thickness, laser power, scanning width, scanning speed, spot radius and overlap rate.

[0043] Particle Swarm Optimization (PSO) is a swarm intelligence optimization algorithm. The core idea of ​​the algorithm is to simulate the movement of a group of particles in the search space to find the optimal solution. The solution space refers to the search space when the particle swarm algorithm is used to determine the optimal solution of the process parameters; the objective function refers to the function that determines the fitness value of the particles using the particle swarm algorithm. The fitness value is a professional term in the particle swarm algorithm, which will not be described in detail in this embodiment.

[0044] In an exemplary embodiment of the present invention, the determination of the solution space of the particle swarm algorithm can be achieved by the following steps:

[0045] The three-dimensional CAD model of the printed part is discretized to obtain a finite element simulation model, and the ranges of multiple process parameters for simulation using the finite element simulation model are set, and the ranges of each process parameter are used as a solution space.

[0046] The 3D CAD model of the part to be printed refers to a 3D image of the part to be printed, which is used as a basic model to construct a finite element simulation model of the part to be printed. Specifically, the 3D CAD model of the part to be printed can be discretized in the finite element analysis software, that is, the 3D CAD model can be meshed using the software to obtain a finite number of small units, thereby obtaining a finite element simulation model of the part to be printed.

[0047] The finite element simulation model refers to a model used to simulate the temperature field, stress field and strain field of each process parameter under different values. It should be noted that before using the particle swarm algorithm to determine the optimal solution of the process parameters, it is necessary to simulate the temperature field, stress field and strain field of each process parameter under different values, so that the optimal solution of the process parameters can be determined by comparing the differences between the temperature field, stress field and strain field under different process parameters.

[0048] Optionally, the three-dimensional CAD model of the printed part may be discretized using finite element analysis software, such as ANSYS software or ABAQUS software. This embodiment does not specifically limit the type of finite element analysis software selected.

[0049] Optionally, the range of each process parameter simulated by the finite element simulation model can be directly set in the finite element analysis software, so that the set range of each process parameter is used as the solution space of the particle swarm algorithm.

[0050] In an exemplary embodiment of the present invention, the objective function of the particle swarm algorithm can be determined by the following steps:

[0051] Using the finite element simulation model, the temperature field, stress field and strain field under various process parameters are simulated to obtain the residual stress function, deformation function, density function and defect function;

[0052] Using preset weights, the residual stress function, deformation function, density function and defect function are weighted summed to obtain the objective function of the particle swarm algorithm.

[0053] Optionally, based on empirical values, the weight of the residual stress function can be set to -1, the weight of the deformation function can be set to -1, the weight of the density function can be set to 1, and the weight of the defect function can be set to -1, so that the corresponding weights can be used to perform weighted summation on the residual stress function, deformation function, density function and defect function to obtain the objective function of the particle swarm algorithm. This embodiment does not specifically limit the weights of each function.

[0054] Optionally, when the particle swarm algorithm is used to adjust the positions of particles, the maximum value of the objective function can be used as the optimization direction of each particle.

[0055] S2: obtaining simulated particles and measured particles in a preset solution space, and calculating the credibility of any simulated particle at any training time when iteratively training the pre-built support vector machine using the measured particles.

[0056] In an exemplary embodiment of the present invention, simulated particles refer to particles whose corresponding fitness values ​​are determined by support vector machines; measured particles refer to particles whose corresponding fitness values ​​are determined by finite element simulation models and objective functions; credibility refers to data used to measure the possibility of the existence of an optimal solution in the process parameters corresponding to the position of the corresponding simulated particle and the surrounding positions; support vector machines (SVM) are a supervised learning algorithm that can be used to complete regression tasks.

[0057] It should be noted that when using the traditional particle swarm algorithm to determine the optimal solution for the process parameters, all particles set in the solution space need to undergo a finite element analysis, that is, the process parameters at the location of each particle need to be analyzed by finite element analysis, so as to determine the fitness value corresponding to each particle to adjust the position and speed of the particle until the optimal solution is determined. The amount of calculation is huge and the time cost is high, resulting in low efficiency in determining the optimal solution. The support vector machine can complete the regression task, can replace the optimization process of the particle swarm algorithm, and can output the fitness value of the particle by training the support vector machine, so that when determining the fitness value of some particles, it is not necessary to perform finite element analysis on the particle, but to determine it through the trained support vector machine, thereby improving the efficiency of using the particle swarm algorithm to determine the optimal solution for the process parameters.

[0058] In an exemplary embodiment of the present invention, the determination of simulated particles and measured particles in the solution space can be achieved by the following steps:

[0059] A number of particles are randomly selected from the solution space, and the particles are randomly divided according to a preset ratio to obtain simulated particles and measured particles.

[0060] Optionally, 300 particles can be randomly selected in the solution space, and then based on empirical values, the 300 particles are randomly divided into simulated particles and measured particles in a ratio of 8:2, that is, the number of simulated particles is 240, and the number of measured particles is 60. This embodiment does not specifically limit the number of particles selected from the solution space and the division ratio.

[0061] Furthermore, after the simulated particles and measured particles in the solution space are determined, all the measured particles can be used to construct and train a support vector machine. It should be noted that due to the small number of measured particles, if only all the measured particles obtained by division are used to construct and train the support vector machine, the reference degree is low. Therefore, when training the support vector machine, it is necessary to increase the number of particles in the training sample. The process of constructing a support vector machine is a prior art and will not be described in detail in this embodiment.

[0062] In an exemplary embodiment of the present invention, the training of a support vector machine can be implemented by the following steps:

[0063] (1) Take a number of particles as initial training samples, use the measured fitness values ​​corresponding to each particle in the initial training samples as labels, and perform the first training of the support vector machine;

[0064] The initial training sample refers to a set of all particles randomly selected from the solution space, and the measured fitness value refers to the value obtained by simulating and calculating using a finite element simulation model and then substituting the calculated result into the objective function.

[0065] In an exemplary embodiment of the present invention, the determination of the measured fitness value may be achieved by the following steps:

[0066] The process parameters of any particle location are input into the finite element analysis model, the temperature field, stress field and strain field under the corresponding process parameters are simulated, and the corresponding residual stress, deformation, density and defects are determined; the residual stress, deformation, density and defects are substituted into the objective function to obtain the measured fitness value of the corresponding particle.

[0067] It should be noted that, whether calculating the measured fitness value of a particle randomly selected from the solution space, or calculating the measured fitness value of a simulated particle or measured particle obtained by division, the method for determining the measured fitness value in this embodiment is adopted.

[0068] Furthermore, after determining the measured fitness values ​​corresponding to all particles in the initial training sample, the initial training sample can be used as the input of the support vector machine, and the measured fitness values ​​corresponding to all particles in the initial training sample can be used as labels, so as to perform the first training on the support vector machine, wherein the training process of the support vector machine using the training sample set is a prior art and is not described in detail in this embodiment.

[0069] Optionally, after the support vector machine is trained for the first time, the position of each particle in the initial training sample set in the solution space will change, and the position of each particle in the solution space will move closer to the position of the particle with the largest measured fitness value.

[0070] (2) In each subsequent training, the measured particles obtained by dividing according to the preset ratio are used as the basic measured particles, the initial training samples after the current training times minus one are used as the basic samples, and the sample set consisting of the basic samples, the basic measured particles, and the simulated particles converted into measured particles determined in the historical training process are used as the target training samples for the current training, and the measured fitness values ​​corresponding to each particle in the target training samples are used as labels to iteratively train the support vector machine.

[0071] Optionally, when 300 particles are randomly selected from the solution space and divided into simulated particles and measured particles in a ratio of 8:2, the number of particles in the initial training sample is 300, and the support vector machine is trained for the first time. The number of particles in the target training sample during training ,in, For the The number of simulated particles converted to measured particles during training, To perform the support vector machine training, is the summation symbol, 60 is the number of measured particles divided, that is, the number of basic measured particles; it should be noted that since there may be particles with overlapping positions during the movement of particles, the number of particles in the target training sample is often slightly smaller than the calculated number.

[0072] Furthermore, when training the support training set, the credibility of the simulated particles at any training time can be calculated. Specifically, the credibility of the simulated particles at any training time satisfies the following relationship:

[0073] ;

[0074] In the formula, For the The simulated particles are Credibility during training; is the number of particles measured during historical training, For the The number of particles measured during training, For the During the next training session, The simulated fitness value obtained by inputting the simulated particles into the support vector machine is is the first The measured fitness value of the process parameter at the location of the measured particle, , are the simulated particle and the measured particle, respectively. process parameters, is the number of process parameters, is the normalization function, is the absolute value symbol.

[0075] in, This reflects the During the first training, The simulated fitness value obtained by inputting the first simulated particle into the support vector machine is The difference between the measured fitness values ​​corresponding to the measured particles is calculated. The larger the value, the lower the possibility that there is an optimal solution in the process parameters corresponding to the position of the simulated particle and the surrounding positions, and the lower the credibility of the corresponding simulated particle.

[0076] This reflects the During the first training session, The process parameters corresponding to the position of the simulated particle are The larger the value is, the lower the possibility that there is an optimal solution in the process parameters corresponding to the position of the simulated particle and the surrounding positions, and the lower the credibility of the corresponding simulated particle.

[0077] Furthermore, after determining the credibility of each simulated particle at any training time, the number of simulated particles that can be converted into measured particles can be determined based on the credibility of each simulated particle at that training time and the set judgment conditions, so that the target training samples for the next training time can be determined to continue training the support vector machine.

[0078] S3: When the product of the credibility and the normalized simulation fitness value is greater than the preset value, and the measured fitness value of any simulated particle is greater than the conversion threshold, any simulated particle will be used as the measured particle in the subsequent training process and the training process will be repeated. After the training is completed, the target process parameters are obtained.

[0079] The product of the credibility of any simulated particle at any training time and the simulated fitness value of the simulated particle determined at the training time can reflect the possibility of performing a fine search at the location of the simulated particle. The larger the product, the greater the possibility that the location of the simulated particle requires a fine search.

[0080] Optionally, the preset value can be set to 0.7 based on experience. When the product of the credibility of any simulated particle in any training and the simulated fitness value determined for the simulated particle in the training is greater than 0.7, a detailed search can be performed at the location of the simulated particle.

[0081] The conversion threshold refers to the standard for determining whether to convert a simulated particle into a measured particle when it is determined that a detailed search is required for the location of any simulated particle. It should be noted that when using measured particles to train the support vector machine, the label is the measured fitness value corresponding to the measured particle, and the measured fitness value corresponding to the measured particle has a greater impact on the training results. Therefore, when determining whether to convert a simulated particle into a measured particle, the measured fitness value corresponding to the simulated particle can be calculated, and based on the measured fitness value corresponding to the simulated particle and the conversion threshold, it is determined whether to convert the simulated particle into a measured particle, thereby ensuring the accuracy of the simulated particles that can be converted into measured particles.

[0082] The target process parameters refer to the optimal solution of the process parameters determined by using the support vector machine and the particle swarm algorithm, which are used as the process parameters of powder laser 3D printing, so that the 3D printer with optimized process parameters can be used to print the parts to be printed, so as to improve the quality and performance of the printed parts.

[0083] In an exemplary embodiment of the present invention, the determination of the conversion threshold may be achieved by the following steps:

[0084] The maximum value of the data sequence composed of the simulated fitness values ​​of all historical simulated particles and the measured fitness values ​​of all historical measured particles is obtained; the preset conversion experience coefficient is obtained, and the product of the maximum value and the conversion experience coefficient is calculated to obtain the conversion threshold.

[0085] Optionally, the conversion empirical coefficient can be set to 0.8 according to the empirical value, and the maximum value of the data sequence composed of the simulated fitness values ​​of all historical simulated particles and the measured fitness values ​​of all historical measured particles is recorded as , when the support vector machine is During the first training, if The simulated particles are The credibility of the simulated particle in the first training When the product of the simulated fitness values ​​determined during the training is greater than 0.7, the measured fitness value of the simulated particle is calculated. If the measured fitness value is greater than 0.8 , then in the support vector machine For training of the next number of times, the simulated particles are converted into measured particles to participate in the training of the support vector machine; when the measured fitness value is less than or equal to 0.8 , the simulated particle is not converted.

[0086] In an exemplary embodiment of the present invention, the determination of the target process parameters can be achieved by the following steps:

[0087] During the training of the support vector machine, if the positions of all simulated particles and all measured particles tend to converge, the training is stopped; after the training is stopped, the particle corresponding to the maximum value in the data sequence composed of the simulated fitness values ​​of all current simulated particles and the measured fitness values ​​of all current measured particles is used as the target particle, and the process parameter corresponding to the position of the target particle is used as the target process parameter.

[0088] Among them, all current simulated particles refer to all simulated particles whose positions converge in the solution space after training is completed; all current measured particles refer to all measured particles whose positions converge in the solution space after training is completed. It should be noted that the maximum value in the data sequence composed of the simulated fitness value of all current simulated particles and the measured fitness value of all current measured particles is the optimal solution of the fitness value, and the process parameter corresponding to the position of the particle corresponding to this value is the target process parameter.

[0089] Furthermore, after the target process parameters are determined, the target process parameters can be used as process parameters for powder laser 3D printing, so that the 3D printer after adjusting the process parameters can print the workpiece.

[0090] Optionally, in the process of determining the target process parameters, setting simulated particles can effectively reduce the amount of calculation of finite element analysis, and as the iterative training process of the support vector machine proceeds, the simulated particles that meet the conversion conditions are converted into measured particles, which can improve the search accuracy. Therefore, the present invention can effectively search for the optimal solution of the process parameters while reducing the time cost.

[0091] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0092] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A parameter optimization method for a powder laser 3D printing process, characterized in that: include: Taking the range of multiple process parameters of the to-be-printed workpiece as the solution space of the particle swarm algorithm, performing finite element analysis on each of the process parameters, and determining the objective function of the particle swarm algorithm; The simulated particles and measured particles in the preset solution space are obtained, and when the pre-constructed support vector machine is iteratively trained using the measured particles, the credibility of any simulated particle in any training is calculated. , is the number of particles measured during historical training, For the The number of particles measured during training, For the During the next training session, The simulated fitness value obtained by inputting the simulated particles into the support vector machine is is the first The measured fitness value of the process parameter at the location of the measured particle, , are the simulated particle and the measured particle, respectively. process parameters, is the number of process parameters, is the normalization function, is the absolute value symbol; When the product of the credibility and the normalized simulation fitness value is greater than a preset value, and the measured fitness value of any simulated particle is greater than a conversion threshold, any simulated particle is used as a measured particle in the subsequent training process and the training process is repeated, and the target process parameters are obtained after the training is completed; Simulated particles refer to particles whose corresponding fitness values ​​are determined by support vector machines; measured particles refer to particles whose corresponding fitness values ​​are determined by finite element simulation models and objective functions; conversion threshold refers to the standard for converting simulated particles into measured particles when it is determined that a detailed search is required for the location of any simulated particle.

2. The parameter optimization method of a powder laser 3D printing molding process according to claim 1, characterized in that: The method for obtaining the solution space of the particle swarm algorithm comprises: The three-dimensional CAD model of the workpiece to be printed is discretized to obtain a finite element simulation model, and the ranges of multiple process parameters for simulation using the finite element simulation model are set, and the ranges of each process parameter are used as the solution space.

3. The parameter optimization method of a powder laser 3D printing process according to claim 2, characterized in that: The performing of finite element analysis on each of the process parameters to determine the objective function of the particle swarm algorithm includes: Using the finite element simulation model, the temperature field, stress field and strain field under each of the process parameters are simulated to obtain a residual stress function, a deformation function, a density function and a defect function; The residual stress function, the deformation function, the density function and the defect function are weightedly summed using preset weights to obtain the objective function of the particle swarm algorithm.

4. The parameter optimization method for a powder laser 3D printing process according to claim 1, characterized in that: The method for obtaining the measured fitness value includes: Inputting the process parameters of any particle location into the finite element analysis model, simulating the temperature field, stress field and strain field under the corresponding process parameters, and determining the corresponding residual stress, deformation, density and defects; The residual stress, the deformation, the density and the defect are substituted into the objective function to obtain the measured fitness value of the corresponding particle.

5. The parameter optimization method for a powder laser 3D printing process according to claim 1, characterized in that: The method for obtaining simulated particles and measured particles in the solution space includes: A number of particles are randomly selected from the solution space, and the particles are randomly divided according to a preset ratio to obtain simulated particles and measured particles.

6. The parameter optimization method for a powder laser 3D printing process according to claim 5, characterized in that: The iterative training of the pre-built support vector machine includes: Using the plurality of particles as initial training samples, using the measured fitness value corresponding to each particle in the initial training samples as a label, and performing a first training on the support vector machine; In each subsequent training, the measured particles obtained by dividing according to the preset ratio are used as basic measured particles, the value of initial training samples after the current training times is reduced by one is used as basic samples, and the sample set consisting of the basic samples, the basic measured particles and the simulated particles converted into measured particles determined in the historical training process is used as the target training sample for the current training, and the measured fitness value corresponding to each particle in the target training sample is used as a label to iteratively train the support vector machine.

7. The parameter optimization method for a powder laser 3D printing process according to claim 1, characterized in that: The method for obtaining the conversion threshold comprises: Obtain the maximum value of the data sequence composed of the simulated fitness values ​​of all historical simulated particles and the measured fitness values ​​of all historical measured particles; A preset conversion empirical coefficient is obtained, and the product of the maximum value and the conversion empirical coefficient is calculated to obtain the conversion threshold.

8. The parameter optimization method for a powder laser 3D printing process according to claim 7, characterized in that: The process of obtaining the target process parameters includes: During the training of the support vector machine, if the positions of all simulated particles and all measured particles tend to converge, the training is stopped; When the training is stopped, the particle corresponding to the maximum value in the data sequence composed of the simulated fitness values ​​of all the current simulated particles and the measured fitness values ​​of all the current measured particles is taken as the target particle, and the process parameter corresponding to the position of the target particle is taken as the target process parameter.

Images