Method and device for preparing a lattice structure by multi-field assisted selective laser melting

Abstract

The present invention relates to the technical field of additive manufacturing, specifically a method and device for preparing a lattice structure by multi-field assisted selective laser melting. The method includes obtaining data of the material to be processed and performance indicators of the target lattice structure; constructing a multi-field coupling physical model to predict the results of material melting, solidification, and lattice structure evolution in a multi-field environment; establishing an adaptive multi-field environment control algorithm to adjust the spatio-temporal distribution of the electric field, magnetic field, and acoustic field by real-time analyzing the laser-material interaction process; at the same time, constructing a process optimization method based on deep learning to adjust process parameters such as laser power, scanning speed, and powder layer thickness in real time through a deep learning model; outputting a combined data of the optimal material composition, laser process parameters, and multi-field environment parameters, and adjusting the selective laser melting process to prepare the target lattice structure. The present invention can significantly improve the dimensional accuracy, surface quality, and mechanical properties of the lattice structure, and expand its application fields.

Description

Technical Field

[0001] The present invention relates to the technical field of additive manufacturing, and specifically to a method and device for preparing a lattice structure by multi-field assisted selective laser melting. Background Art

[0002] Selective laser melting technology is a commonly used metal 3D printing method, which selectively melts metal powder through a high-energy laser beam and accumulatively prepares a three-dimensional solid structure layer by layer. Using selective laser melting technology to prepare a lattice structure has the advantages of flexible design and high preparation efficiency, and has broad application prospects in the fields of aerospace, biomedicine, etc.

[0003] However, traditional selective laser melting technology still faces some challenges in preparing lattice structures: the dimensional accuracy and surface quality of the lattice structure are difficult to control. Due to the complex process of selective laser melting, the dynamic behavior of the molten pool is difficult to accurately describe, resulting in the dimensional accuracy and surface quality of the lattice structure being difficult to precisely control, which limits its application range. Material defects and residual stresses affect the performance of the lattice structure. During the selective laser melting process, materials are prone to defects such as pores and cracks, and at the same time, the rapid heating and cooling cycle leads to the generation of residual stresses, affecting the mechanical properties and service life of the lattice structure. It is difficult to optimize the processing process parameters. Preparing a lattice structure by selective laser melting involves multiple process parameters such as laser power, scanning speed, and powder layer thickness. Traditional parameter optimization methods, such as the trial-and-error method and orthogonal experiments, are inefficient and costly, and it is difficult to obtain the best combination of process parameters. The multi-field coupling effect is complex and difficult to regulate. During the selective laser melting process, there are interactions between multiple physical fields such as the thermal field, flow field, and stress field, which affect the formation and evolution of the lattice structure. Traditional methods are difficult to achieve precise regulation of the multi-field coupling process, which limits the further improvement of the performance of the lattice structure.

[0004] In order to break through the above technical bottlenecks, it is urgent to develop a new method for preparing a lattice structure by selective laser melting. By introducing multi-field assisted technology and combining numerical calculation and machine learning methods, precise modeling and intelligent optimization of the entire selective laser melting process can be achieved, and then the morphology, size, and performance of the lattice structure can be precisely regulated to improve its preparation quality and efficiency. This can not only broaden the application fields of the lattice structure, but also contribute to the innovative development of additive manufacturing technology. Summary of the Invention

[0005] The purpose of the present invention is to propose a method and device for preparing a lattice structure by multi-field assisted selective laser melting in view of the deficiencies of the prior art.

[0006] To achieve the above purpose, the present invention provides the following technical solutions:

[0007] A method for preparing a lattice structure by multi-field assisted selective laser melting, including:

[0008] Obtain the composition data and powder characteristics of the material to be processed, and obtain the performance indicators of the target lattice structure;

[0009] Construct a multi-field coupling physical model, and predict the melting, solidification and lattice structure evolution results of the material to be processed under a multi-field environment through the combination of numerical calculation and machine learning model;

[0010] Establish an adaptive multi-field environment regulation algorithm, and control the lattice structure morphology and defects by adjusting the spatio-temporal distribution of the electric field, magnetic field and acoustic field through real-time analysis of the laser-material interaction process to be processed;

[0011] Construct a process optimization method based on deep learning, and adjust the process parameters of laser power, scanning speed and powder layer thickness during selective laser melting in real time through the autonomous learning and decision-making of the deep learning model;

[0012] Output the combined data of the optimal material components to be processed, laser process parameters and multi-field environment parameters for the target lattice structure, and adjust the selective laser melting process to prepare the target lattice structure.

[0013] Preferably, obtain the density, specific heat capacity and thermal conductivity data of the material to be processed; determine the porosity of the target lattice structure based on the performance requirements of the target lattice structure 、pore size distribution 、mechanical properties performance indicators;

[0014] Construct the heat conduction equation of the material to be processed to describe the heat transfer process inside the material to be processed:

[0015]

[0016] wherein, is the density of the material to be processed, is the specific heat capacity, is the thermal conductivity, is the heat source term, and the heat source term includes a laser heat source and Joule heat; is the partial derivative of the temperature T with respect to the time t, representing the rate of change of the temperature with time; represents the gradient of the temperature in space; solve the heat conduction equation using the finite element method to obtain the temperature field distribution ; represents the heat conduction term;

[0017] Construct the fluid flow equation of the material to be processed, and use the Navier-Stokes equation to describe the flow behavior of the molten metal inside the molten pool:

[0018]

[0019] Among them, is the velocity vector, is the pressure, is the dynamic viscosity, is the body force, and the body force includes the vector sum of gravity and electromagnetic force; is the gradient operator, representing the spatial derivative; represents the gradient of the pressure; represents the convection term;

[0020] The Navier-Stokes equation is solved using the finite volume method to obtain the flow field distribution and the pressure distribution of the phase change model.

[0021] Preferably, the phase change model includes introducing the liquid fraction to describe the melting and solidification processes of the material to be processed. When the temperature is lower than the solid phase temperature , it represents a completely solid phase; when the temperature is higher than the liquid phase temperature , it represents a completely liquid phase; in the solid-liquid coexistence interval, it varies linearly with temperature, and the latent heat of phase change is coupled to the heat conduction equation through the heat source term. The phase change model is specifically as follows:

[0022]

[0023] Among them, is the liquid fraction, is the solid phase temperature, is the liquid phase temperature. Combining with the temperature field distribution , calculate the melting and solidification processes of the material to be processed to obtain the solid-liquid interface position and the molten pool morphology.

[0024] Preferably, based on the support vector machine (SVM) as a machine learning model, a non-linear mapping relationship is established between the physical field distribution and the evolution of material melting, solidification, and lattice structure. The physical field distribution includes temperature field distribution, flow field distribution, pressure distribution, and liquid fraction; the Gaussian kernel function is used to map the input features to a high-dimensional space: ; among them, is the Gaussian kernel function parameter, used to control the width of the Gaussian kernel; and are feature vectors;

[0025] The objective function of the support vector machine (SVM) is as follows:

[0026]

[0027]

[0028]

[0029]

[0030] Among them, is the weight vector, is the transpose of the weight vector, is the class label of the th sample, is the total number of samples, is the penalty coefficient, and are slack variables, is the error tolerance, is the kernel function mapping, is the bias term;

[0031] By solving the optimization problem, the support vector machine SVM regression model is obtained:

[0032]

[0033] Among them, and are Lagrange multipliers, is the bias term, is the input vector.

[0034] Preferably, an adaptive multi-field environment control algorithm is established, and the electric field strength, magnetic field strength, and acoustic field strength are adjusted according to the laser-machined material interaction process analyzed in real time;

[0035] Electric field strength regulation equation:

[0036]

[0037] Among them, is the charge density, is the vacuum permittivity;

[0038] Solve the divergence equation and curl equation of the electric field to calculate the electric field strength distribution ;

[0039] Magnetic field strength regulation equation:

[0040]

[0041]

[0042] Among them, is the vacuum permeability, is the current density;

[0043] Solve the divergence equation and curl equation of the magnetic field, and calculate the magnetic field intensity distribution ;

[0044] Sound field frequency Regulation equation:

[0045]

[0046] where is the sound pressure, is the speed of sound;

[0047] Solve the acoustic wave equation and calculate the sound pressure distribution ;

[0048] Construct an adaptive multi-field environment regulation algorithm to adjust the electromagnetic acoustic field distribution.

[0049] Preferably, use a deep reinforcement learning model with an Actor-Critic architecture, including an Actor network and a Critic network. The Actor network is used to select the optimal process parameter combination according to the current state, and the Critic network is used to evaluate the pros and cons of the process parameter combination. By continuously updating the parameters of the two networks, the adaptive optimization of the process parameters for the preparation of the lattice structure is achieved.

[0050] Preferably, construct an Actor-Critic framework, and the Actor network : takes the state as the input, outputs the probability distribution of the process parameters , and updates the parameter according to the policy gradient; The Critic network : takes the state and the action as the input, outputs the action value function , and updates the parameter according to the temporal difference error;

[0051] Definition of state and action: ; ; The state includes the distribution of physical fields such as the temperature field , the flow field , the liquid fraction , the electric field strength , the magnetic field strength , the sound field frequency ; The action includes the laser power , the scanning speed , the spot diameter , the electric field strength , the magnetic field strength and the sound field frequency Process parameters;

[0052] Design the reward function of the deep reinforcement learning model with the Actor-Critic architecture The reward function measures the difference between the actual performance and the target performance through an exponential function:

[0053]

[0054]

[0055] Reward function Comprehensively consider the actual porosity Actual pore size distribution And actual mechanical properties Performance indicators, where , , Are weight coefficients, , , Are scale factors, Is the target performance index of porosity, Is the target performance index of pore size distribution, Is the target performance index of mechanical properties;

[0056] The training steps of the deep reinforcement learning model with the Actor-Critic architecture are as follows:

[0057] Use the Deep Deterministic Policy Gradient (DDPG) algorithm to train the Actor network and the Critic network;

[0058] Update the Actor network parameters through policy gradients to optimize the selection strategy of process parameters; update the Critic network parameters through the temporal difference error to optimize the estimation of the action value function, and continuously iterate the training until convergence to the optimal process parameter combination;

[0059] Output the optimal process parameter combination obtained from the training And apply it to the selective laser melting preparation process. According to the optimal parameter combination, adjust the control parameters of the selective laser melting equipment.

[0060] The device for preparing the lattice structure by multi-field assisted selective laser melting, which is used to implement the method for preparing the lattice structure by multi-field assisted selective laser melting, includes: a data acquisition module, a multi-field coupling physics module, a multi-field environment regulation module, a process optimization module, and an execution module;

[0061] The data acquisition module is used to acquire the composition data and powder characteristics of the material to be processed and the performance indicators of the target lattice structure;

[0062] The multi-field coupling physical module is used to construct a multi-field coupling physical model, and through the combination of numerical calculation and machine learning model, predict the melting, solidification and lattice structure evolution results of the material to be processed in a multi-field environment;

[0063] The multi-field environment regulation module is used to establish an adaptive multi-field environment regulation algorithm. By real-time analyzing the laser-material interaction process, adjust the spatio-temporal distribution of the electric field, magnetic field and acoustic field, and control the lattice structure morphology and defects;

[0064] The process optimization module is used to construct a process optimization method based on deep learning. Through the autonomous learning and decision-making of the deep learning model, the process parameters such as laser power, scanning speed and powder layer thickness during selective laser melting are adjusted in real time;

[0065] The execution module, based on the combined data of the optimal material composition, laser process parameters and multi-field environment parameters for the target lattice structure output, adjusts the selective laser melting process to prepare the target lattice structure.

[0066] An electronic device includes: a processor and a memory. Among them, a computer program that can be called by the processor is stored in the memory; the processor executes the method for preparing a lattice structure by multi-field assisted selective laser melting by calling the computer program stored in the memory.

[0067] A computer-readable storage medium stores instructions. When the instructions run on a computer, the computer is made to execute the method for preparing a lattice structure by multi-field assisted selective laser melting.

[0068] The beneficial effects of the present invention: The present invention realizes the precise control of the lattice structure. By obtaining the composition data and powder characteristics of the material to be processed, and combining with the performance indexes of the target lattice structure, a multi-field coupling physical model based on machine learning is constructed, which can accurately predict the evolution behavior of the lattice structure in a multi-field environment and realize the precise control of the lattice structure.

[0069] The present invention improves the preparation quality of the lattice structure. By establishing an adaptive multi-field environment regulation algorithm and real-time analyzing the laser-material interaction process, dynamically optimizing the spatio-temporal distribution of the electric field, magnetic field and acoustic field, the morphology and defects of the lattice structure can be effectively controlled, and the preparation quality of the lattice structure is improved.

[0070] Optimize the selective laser melting process parameters: Construct a process optimization method based on deep learning. Through the autonomous learning and decision-making of the deep learning model, the key process parameters such as laser power, scanning speed and powder layer thickness can be adjusted in real time, and the best process parameter combination can be found to improve the preparation efficiency and quality of the lattice structure.

[0071] The present invention expands the application scope of the lattice structure. Through the multi-field assisted selective laser melting technology, lattice structures with more excellent performance and more complex structures can be prepared.

[0072] The present invention promotes the development of the selective laser melting technology: By combining the multi-field assisted technology with the selective laser melting technology and introducing machine learning and deep learning methods, the limitations of the traditional selective laser melting technology can be broken through, promoting the innovative development of the selective laser melting technology.

[0073] The present invention provides a replicable lattice structure preparation scheme: By outputting the combined data of the optimal material components, laser process parameters and multi-field environment parameters and adjusting the selective laser melting process, a replicable solution can be provided for the preparation of lattice structures with different requirements, improving the repeatability and consistency of the lattice structure preparation. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 is a flowchart of the method for preparing a lattice structure by multi-field assisted selective laser melting according to the present invention;

[0075] Figure 2 is a structural diagram of the device for preparing a lattice structure by multi-field assisted selective laser melting according to the present invention;

[0076] Figure 3 is a schematic structural diagram of the electronic device provided by the present invention;

[0077] Figure 4 is a schematic structural diagram of the computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] To better understand the present application, more detailed descriptions will be made for various aspects of the present application with reference to the accompanying drawings. It should be understood that these detailed descriptions are only descriptions of the exemplary embodiments of the present application and do not limit the scope of the present application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0079] In the accompanying drawings, for the sake of clarity, the sizes, dimensions and shapes of the elements have been slightly adjusted. The drawings are only examples and are not drawn to an exact scale. As used herein, the terms "substantially", "about" and similar terms are used as approximate terms and not as terms of degree, and are intended to account for the inherent deviations in measured or calculated values that would be recognized by a person of ordinary skill in the art. Additionally, in the present application, the order of description of the various steps does not necessarily represent the order in which these processes occur in actual operation, unless otherwise clearly specified or derivable from the context.

[0080] It should also be understood that expressions such as "including", "comprising", "having", "containing" and / or "comprising of" are open-ended rather than closed-ended expressions in this specification, which means that the stated features, elements and / or components exist, but do not exclude the existence of one or more other features, elements, components and / or their combinations. In addition, when an expression such as "at least one of..." appears after a list of listed features, it modifies the entire list of features rather than just individual elements in the list. In addition, when describing the embodiments of the present application, the use of "may" means "one or more embodiments of the present application". And the term "exemplary" is intended to refer to an example or illustration.

[0081] Unless otherwise defined, all terms used herein (including engineering terms and technical terms) have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It should also be understood that unless there is a clear description in this application, words defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and should not be interpreted in an idealized or overly formal sense.

[0082] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the drawings and in combination with the embodiments.

[0083] Example 1

[0084] Referring to Figure 1 , the first embodiment of the present invention provides a method for preparing a lattice structure by multi-field assisted selective laser melting.

[0085] S1: Obtain the composition data and powder characteristics of the material to be processed; obtain the laser process parameters, multi-field environment parameters, and performance indicators of the target lattice structure.

[0086] Obtain the density, specific heat capacity, and thermal conductivity data of the material to be processed; based on the performance requirements of the target lattice structure, determine the porosity , pore size distribution , mechanical properties of the performance indicators.

[0087] Based on the performance requirements of the target lattice structure, determine the porosity , pore size distribution , mechanical properties of the performance indicators, the porosity is calculated by volume fraction: , where is the volume of the solid part, is the total volume; the pore size distribution Obtained through image processing and statistical analysis, calculate the average pore size and the standard deviation of the pore size ; Mechanical properties Measured through mechanical tests of tension, compression, and shear, obtain the yield strength , elastic modulus and fracture toughness .

[0088] S2: Construct a multi-field coupling physical model, and through the combination of numerical calculation and machine learning model, predict the melting, solidification, and lattice structure evolution results of the material to be processed under a multi-field environment.

[0089] Construct the heat conduction equation of the material to be processed, which is used to describe the heat transfer process inside the material to be processed:

[0090]

[0091] Among them, is the density of the material to be processed, is the specific heat capacity, is the thermal conductivity, is the heat source term, including the laser heat source and Joule heat; is the partial derivative of temperature T with respect to time t, representing the rate of change of temperature with time; represents the gradient of temperature in space, represents the heat conduction term; Solve the heat conduction equation using the finite element method to obtain the temperature field distribution , and the temperature distribution affects the morphology of the molten pool and the solidification rate.

[0092] Construct the fluid flow equation of the material to be processed, and use the Navier-Stokes equation to describe the flow behavior of the molten metal inside the molten pool:

[0093]

[0094] Among them, is the velocity vector, is the pressure, is the dynamic viscosity, is the body force, and the body force includes the vector sum of gravity and electromagnetic force; is the gradient operator, representing the spatial derivative; represents the gradient of pressure; represents the convection term.

[0095] Use the finite volume method to solve the Navier-Stokes equation to obtain the flow field distribution and the pressure distribution of the phase change model.

[0096] The phase change model includes introducing the liquid fraction to describe the melting and solidification processes of the material to be processed. When the temperature is lower than the solid phase temperature , it represents a completely solid phase; when the temperature is higher than the liquid phase temperature , it represents a completely liquid phase; in the solid-liquid coexistence range, it varies linearly with temperature, and the latent heat of phase change is coupled to the heat conduction equation through the heat source term. The phase change model is as follows:

[0097]

[0098] where, is the liquid fraction, is the solid phase temperature, is the liquid phase temperature. Combining the temperature field distribution , the melting and solidification processes of the material to be processed are calculated to obtain the solid-liquid interface position and the molten pool morphology.

[0099] Based on the support vector machine (SVM) as a machine learning model, a non-linear mapping relationship is established between the physical field distribution and the evolution of material melting, solidification, and lattice structure. The physical field distribution includes temperature field distribution, flow field distribution, pressure distribution, and liquid fraction; the Gaussian kernel function is used to map the input features to a high-dimensional space: ; where, is the Gaussian kernel function parameter, which is used to control the width of the Gaussian kernel, and are feature vectors;

[0100] The objective function of the support vector machine (SVM) is as follows:

[0101]

[0102]

[0103]

[0104]

[0105] where, is the weight vector, is the transpose of the weight vector, is the class label of the th sample, is the total number of samples, is the penalty coefficient, and are slack variables, is the error tolerance, is the kernel function mapping, is the bias term;

[0106] By solving the optimization problem, a support vector machine (SVM) regression model is obtained:

[0107]

[0108] where, and are Lagrange multipliers, is the bias term, is the input vector.

[0109] The hyperparameter optimization and model training process of the support vector machine (SVM) regression model are as follows:

[0110] The hyperparameters of the SVM model, such as the penalty coefficient , kernel function parameter , and error tolerance , are optimized using the grid search and k-fold cross-validation methods. The SVM model is trained using the training set data to obtain the optimal support vectors, Lagrange multipliers, and bias term. The model performance is evaluated on the validation set, and metrics such as the mean squared error (MSE), mean absolute error (MAE), and coefficient of determination ( ) are calculated to test the generalization ability of the model.

[0111] S3: Establish an adaptive multi-field environment regulation algorithm. By real-time analyzing the laser-material interaction process, adjust the spatio-temporal distributions of the electric field, magnetic field, and acoustic field to control the morphology and defects of the dot matrix structure.

[0112] Establish an adaptive multi-field environment regulation algorithm. According to the real-time analyzed laser-material interaction process, adjust the electric field strength, magnetic field strength, and acoustic field strength.

[0113] Regulation equation for the electric field strength :

[0114]

[0115] where, is the charge density, is the vacuum permittivity;

[0116] Solve the divergence equation and curl equation of the electric field to calculate the electric field strength distribution ; According to the temperature field and molten pool morphology information, adjust the electrode position and voltage magnitude to optimize the electric field distribution.

[0117] Regulation equation for the magnetic field strength :

[0118]

[0119]

[0120] Among them, is the vacuum permeability, is the current density;

[0121] Solve the divergence equation and curl equation of the magnetic field to calculate the magnetic field strength distribution ; According to the feedback information of the molten pool flow and solidification structure, dynamically adjust the position and current magnitude of the electromagnet to optimize the magnetic field distribution.

[0122] Sound field frequency regulation equation:

[0123]

[0124] Among them, is the sound pressure, is the speed of sound;

[0125] Solve the acoustic wave equation to calculate the sound pressure distribution ; According to the feedback information of the molten pool oscillation and solidification nucleation, dynamically adjust the transducer frequency and power to optimize the sound field distribution.

[0126] Adopt an adaptive multi-field environment regulation algorithm to adjust the electromagnetic sound field distribution. The specific algorithm steps are as follows:

[0127] According to the temperature field distribution , flow field distribution , pressure distribution and liquid fraction distribution ;

[0128] The ideal lattice structure morphology predicted by the SVM model and the current lattice structure morphology , the current electric field strength , magnetic field strength and sound pressure

[0129] The optimized electric field strength distribution , magnetic field strength distribution and sound pressure distribution ;

[0130] Calculate the error between the current lattice structure morphology and the ideal lattice structure morphology: ; Among them, is the number of feature points of the lattice structure, is the current lattice structure morphology, is the ideal lattice structure morphology.

[0131] Judge whether the error meets the requirements, that is , where is the allowable error threshold; if within the allowable error range, the current electromagnetic acoustic field distribution remains unchanged and the algorithm ends.

[0132] Otherwise, enter the adjustment step of the electromagnetic acoustic field. Calculate the morphological characteristic parameters of the molten pool according to the temperature field, flow field and liquid phase fraction information, including the molten pool size , the depth-to-width ratio of the molten pool , the curvature of the melting boundary and the maximum flow velocity .

[0133] Input the morphological characteristic parameters into the SVM model to predict the ideal electric field strength , magnetic field strength and sound pressure ; Calculate the regulation amount of the electromagnetic acoustic field distribution:

[0134] Electric field strength regulation amount: ;

[0135] Magnetic field strength regulation amount: ;

[0136] Sound pressure regulation amount: ;

[0137] Among them, are the regulation coefficients of the electric field, magnetic field and sound field respectively, controlling the regulation amplitude;

[0138] Superimpose the regulation amount and update the electromagnetic acoustic field distribution:

[0139] Electric field strength update: ;

[0140] Magnetic field strength update: ;

[0141] Sound pressure update: .

[0142] Output the updated electromagnetic acoustic field distribution to the multi-physical field model and the selective laser melting equipment, and repeat the above steps at the next sampling time point until the termination condition is met.

[0143] To improve the robustness and real-time performance of the algorithm, an adaptive regulation coefficient and a rolling optimization strategy can be introduced:

[0144] Adaptive regulation coefficient: ,

[0145] ; Among them, is the initial regulation coefficient, is the attenuation factor.

[0146] When the error is large, the regulation coefficient increases to accelerate the regulation speed; when the error decreases, the regulation coefficient decreases to avoid overshoot.

[0147] At the same time, a rolling optimization strategy is adopted: introduce a time window , and perform rolling optimization on the electromagnetic field distribution within each window, that is: ; where e represents the immediate cost function, which is a function of time , and measures the degree to which the system state deviates from the desired target at time . By solving the optimal control problem within a finite time domain, the optimal electromagnetic field distribution in the future for a period of time is predicted, improving the system's response speed and control effect.

[0148] S4: Construct a process optimization method based on deep learning, and through the autonomous learning and decision-making of the deep learning model, adjust the process parameters of laser power, scanning speed, and powder layer thickness during selective laser melting in real time.

[0149] Use a deep reinforcement learning model with an Actor-Critic architecture, including an Actor network and a Critic network. The Actor network is used to select the optimal combination of process parameters according to the current state, and the Critic network is used to evaluate the quality of the process parameter combination. By continuously updating the parameters of the two networks, the adaptive optimization of the process parameters for the preparation of the lattice structure is achieved.

[0150] Construct an Actor-Critic framework. The Actor network : takes the state as the input, outputs the probability distribution of the process parameters , and updates the parameters according to the policy gradient; the Critic network : takes the state and the action as the input, outputs the action value function , and updates the parameters .

[0151] Definitions of state and action: ; ; The state includes the distributions of physical fields such as the temperature field , the flow field , the liquid fraction , the electric field strength , the magnetic field strength , the acoustic field frequency ; The action includes the laser power , the scanning speed , spot diameter , electric field strength , magnetic field strength and acoustic field frequency of the process parameters.

[0152] Design the reward function of the deep reinforcement learning model with the Actor-Critic architecture , the reward function measures the difference between the actual performance and the target performance through an exponential function:

[0153]

[0154]

[0155] Reward function Comprehensively consider the actual porosity , actual pore size distribution and actual mechanical properties of the performance indicators, where , , are weight coefficients, , , are scale factors, is the target performance indicator of porosity, is the target performance indicator of pore size distribution, is the target performance indicator of mechanical properties. The reward function measures the difference between the actual performance and the target performance through an exponential function.

[0156] The training steps of the deep reinforcement learning model with the Actor-Critic architecture are as follows:

[0157]

[0158]

[0159]

[0160] Among them, is the Actor network, with the state as the input, and outputs the action ; represents the parameters of the Actor network; is the Critic network, with the state and the action as the input, and outputs the state-action value function ; represents the parameters of the Critic network; is the objective function of the Actor network About parameters The gradient of Indicates the distribution in state Calculate the expectation below, where It is a behavior-based strategy The state distribution of is the Q value output by the Critic network about the action The gradient of and actions conduct assessments; It is the action output by the Actor network about the parameters The gradient of conduct assessments; is the loss function of the Critic network; Indicates that in the experience replay buffer Sampling transfer tuple and calculate the expectation; is the target Q value, which is determined by the immediate reward and the estimated Q value of the next state composition; is a discount factor used to balance the importance of immediate rewards and future rewards.

[0161] It should be noted that in the DDPG algorithm, the Actor network optimizes the strategy by maximizing the Q value output by the Critic network and uses the policy gradient theorem to estimate the gradient of the objective function. The Critic network learns the state-action value function by minimizing the TD error and uses temporal difference learning to update the Q value estimate. By alternately optimizing the Actor and Critic networks, the DDPG algorithm can learn the optimal strategy in the continuous action space and achieve adaptive optimization of process parameters.

[0162] S5: Output the combined data of the best material composition to be processed, laser process parameters and multi-field environmental parameters for the target lattice structure, and adjust the laser selective melting process to prepare the target lattice structure.

[0163] Through policy gradient Update Actor Network Parameters ,Optimize the selection strategy of process parameters; Through time difference error Update Critic network parameters ,Optimize the estimation of the action value function and continuously iterate the training until it converges to the optimal process parameter combination.

[0164] Optimal parameter combination output: The optimal process parameter combination obtained through training The output is applied to the actual laser selective melting preparation process. According to the optimal parameter combination, the control parameters of the laser selective melting equipment are adjusted, including laser power, scanning speed, spot diameter, electromagnet current and transducer frequency.

[0165] Example 2

[0166] Reference Figure 2 The second embodiment of the present invention provides a device for preparing a lattice structure by multi-field assisted laser selective melting.

[0167] The device comprises: a data acquisition module, a multi-field coupling physics module, a multi-field environment control module, a process optimization module and an execution module.

[0168] The data acquisition module is used to obtain the composition data and powder characteristics of the material to be processed and to obtain the performance indicators of the target lattice structure.

[0169] The multi-field coupling physics module is used to construct a multi-field coupling physics model, and predict the melting, solidification and lattice structure evolution results of the material to be processed under a multi-field environment by combining numerical calculation and machine learning models.

[0170] The multi-field environment control module is used to establish an adaptive multi-field environment control algorithm, which adjusts the temporal and spatial distribution of the electric field, magnetic field and acoustic field by real-time analysis of the laser-processed material interaction process, and controls the lattice structure morphology and defects.

[0171] The process optimization module is used to construct a process optimization method based on deep learning, and adjusts the process parameters of laser power, scanning speed and powder layer thickness during laser selective melting in real time through autonomous learning and decision-making of the deep learning model.

[0172] The execution module adjusts the laser selective melting process to prepare the target lattice structure based on the output of the combined data of the best material composition to be processed, laser process parameters and multi-field environmental parameters for the target lattice structure.

[0173] Example 3

[0174] Figure 3 FIG. 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Figure 3 As shown, according to another aspect of the present invention, an electronic device 500 is provided. The electronic device 500 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by the one or more processors, the method for preparing a lattice structure by multi-field assisted laser selective melting as described above can be executed.

[0175] The method or system according to the embodiment of the present invention can also be used byFigure 3 implemented by the architecture of the electronic device shown below.

[0176] As Figure 3 shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc.

[0177] The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the method for preparing a lattice structure by multi-field assisted selective laser melting provided by the present invention.

[0178] The method for preparing a lattice structure by multi-field assisted selective laser melting includes: obtaining the composition data and powder characteristics of the material to be processed and obtaining the performance indexes of the target lattice structure; constructing a multi-field coupling physical model, and predicting the melting, solidification and lattice structure evolution results of the material to be processed in a multi-field environment by combining numerical calculation and a machine learning model; establishing an adaptive multi-field environment regulation algorithm, and controlling the lattice structure morphology and defects by analyzing the laser-material interaction process in real time and adjusting the spatio-temporal distribution of the electric field, magnetic field and acoustic field; constructing a process optimization method based on deep learning, and adjusting the process parameters of laser power, scanning speed and powder layer thickness during selective laser melting in real time through the autonomous learning and decision-making of the deep learning model; outputting the combined data of the optimal material components, laser process parameters and multi-field environment parameters for the target lattice structure, and adjusting the selective laser melting process to prepare the target lattice structure.

[0179] Furthermore, the electronic device 500 may further include a user interface 508. Of course, Figure 3 the architecture shown below is only exemplary, and when implementing different devices, one or more components in the electronic device shown may be omitted according to actual needs. Figure 3 shown in the electronic device.

[0180] Embodiment 4

[0181] Figure 4 is a schematic diagram of the structure of a computer-readable storage medium provided by an embodiment of the present invention.

[0182] As Figure 4 shown, it is a computer-readable storage medium 600 according to an embodiment of the present invention.

[0183] Computer-readable instructions are stored on the computer-readable storage medium 600.

[0184] When the computer-readable instructions are run by a processor, the method for preparing a lattice structure by multi-field assisted selective laser melting according to the embodiment of the present invention described with reference to the above drawings can be executed.

[0185] The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. Additionally, according to an embodiment of the present invention, the processes described above with reference to the flowcharts may be implemented as computer software programs.

[0186] For example, the present invention provides a non-transitory machine-readable storage medium storing machine-readable instructions that can be run by a processor to execute instructions corresponding to the method steps provided by the present invention, such as: obtaining composition data and powder characteristics of a material to be processed and performance indicators of a target lattice structure; constructing a multi-field coupling physical model, and predicting the melting, solidification, and lattice structure evolution results of the material to be processed in a multi-field environment through a combination of numerical calculation and machine learning models; establishing an adaptive multi-field environment regulation algorithm, and controlling the lattice structure morphology and defects by adjusting the spatio-temporal distributions of the electric field, magnetic field, and acoustic field through real-time analysis of the laser-material interaction process; constructing a process optimization method based on deep learning, and real-time adjusting process parameters such as laser power, scanning speed, and powder layer thickness during selective laser melting through autonomous learning and decision-making of the deep learning model; outputting combined data of the optimal material components, laser process parameters, and multi-field environment parameters for the target lattice structure, and adjusting the selective laser melting process to fabricate the target lattice structure. When this computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present invention are executed. The method and apparatus, device of the present invention may be implemented in many ways. For example, the method and apparatus, device of the present invention may be implemented through software, hardware, firmware, or any combination of software, hardware, and firmware.

[0187] The above order of steps for the method is only for illustration, and the steps of the method of the present invention are not limited to the specific order described above, unless otherwise specifically stated.

[0188] In addition, in some embodiments, the present invention may also be implemented as a program recorded in a recording medium, and these programs include machine-readable instructions for implementing the method according to the present invention. Thus, the present invention also covers a recording medium storing a program for executing the method according to the present invention.

[0189] Furthermore, in the above technical solutions provided in the embodiments of the present invention, parts that are consistent with the corresponding technical solutions in the prior art in terms of implementation principles are not described in detail to avoid excessive elaboration.

[0190] The specific embodiments described above further elaborate in detail the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is only for the specific embodiments of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for preparing a lattice structure by multi-field assisted laser selective melting, characterized in that: include: Obtaining the composition data and powder characteristics of the material to be processed and obtaining the performance indicators of the target lattice structure; Construct a multi-field coupling physical model, and predict the melting, solidification and lattice structure evolution of the material to be processed under a multi-field environment by combining numerical calculation and machine learning models; Establish an adaptive multi-field environment control algorithm to analyze the laser-processed material interaction process in real time, adjust the temporal and spatial distribution of the electric field, magnetic field and acoustic field, and control the lattice structure morphology and defects; Construct a process optimization method based on deep learning. Through the autonomous learning and decision-making of the deep learning model, the process parameters of laser power, scanning speed and powder layer thickness during laser selective melting can be adjusted in real time. The optimal combination data of the material composition to be processed, the laser process parameters and the multi-field environmental parameters for the target lattice structure are output, and the laser selective melting process is adjusted to prepare the target lattice structure.

2. The method for preparing a lattice structure by multi-field assisted laser selective melting according to claim 1, characterized in that: Obtain the density, specific heat capacity and thermal conductivity data of the material to be processed; Determine the porosity of the target lattice structure based on the performance requirements of the target lattice structure , pore size distribution , Mechanical properties performance indicators; Construct the heat conduction equation of the material to be processed to describe the heat transfer process inside the material to be processed: ； in, is the density of the material to be processed, is the specific heat capacity, is the thermal conductivity, is a heat source term, wherein the heat source term includes a laser heat source and Joule heat; It is the partial derivative of temperature T with respect to time t, which indicates the rate of change of temperature with time; represents the temperature gradient in space, Represents the heat conduction term; use the finite element method to solve the heat conduction equation and obtain the temperature field distribution ; Construct the fluid flow equation of the material to be processed, and use the Navier-Stokes equation to describe the flow behavior of the molten metal inside the molten pool: ； in, is the velocity vector, For pressure, is the dynamic viscosity, is a body force, wherein the body force includes the vector sum of gravity and electromagnetic force; is the gradient operator, which represents the spatial derivative; Represents the gradient of pressure; represents the convection term; The finite volume method is used to solve the Navier-Stokes equations and obtain the flow field distribution. and pressure distribution The phase change model.

3. The method for preparing a lattice structure by multi-field assisted laser selective melting according to claim 2, characterized in that: The phase change model includes introducing a liquid phase fraction to describe the melting and solidification process of the material to be processed. Below solidus temperature When the temperature is Above liquidus temperature When , it indicates a completely liquid phase; in the solid-liquid coexistence range, the temperature changes linearly, and the phase change latent heat is coupled to the heat conduction equation through the heat source term. The phase change model is as follows: ； in, is the liquid fraction, is the solid phase temperature, is the liquid temperature, combined with the temperature field distribution , calculate the melting and solidification process of the material to be processed, and obtain the position of the solid-liquid interface and the morphology of the molten pool.

4. The method for preparing a lattice structure by multi-field assisted laser selective melting according to claim 3, characterized in that: Based on support vector machine (SVM) as a machine learning model, a nonlinear mapping relationship between physical field distribution and material melting, solidification and lattice structure evolution is established, wherein the physical field distribution includes temperature field distribution, flow field distribution, pressure distribution and liquid phase fraction; The Gaussian kernel function is used to map the input features to a high-dimensional space: ;in, is the Gaussian kernel function parameter, which is used to control the width of the Gaussian kernel; and is the eigenvector; The objective function of the support vector machine SVM is as follows: ； ； ； ； in, is the weight vector, is the transpose of the weight vector, It is The class labels of samples, is the total number of samples, is the penalty coefficient, and is the slack variable, is the error tolerance, is the kernel function mapping, is the bias term; By solving the optimization problem, the support vector machine SVM regression model is obtained: ； in, and is the Lagrange multiplier, is the bias term, is the input vector.

5. The method for preparing a lattice structure by multi-field assisted laser selective melting according to claim 4, characterized in that: Establish an adaptive multi-field environment control algorithm to adjust the electric field strength, magnetic field strength and acoustic field strength according to the real-time analysis of the laser-processed material interaction process; Electric field strength The control equation is: ； in, is the charge density, is the dielectric constant of vacuum; Solve the divergence equation and curl equation of the electric field to calculate the electric field intensity distribution ; Magnetic field strength The control equation is: ； ； in, is the vacuum permeability, is the current density; Solve the divergence equation and curl equation of the magnetic field and calculate the magnetic field intensity distribution ; Sound field frequency The control equation is: ； in, is the sound pressure, is the speed of sound; Solve the acoustic wave equation and calculate the sound pressure distribution ; Construct an adaptive multi-field environmental control algorithm to adjust the distribution of the electromagnetic acoustic field.

6. The method for preparing a lattice structure by multi-field assisted laser selective melting according to claim 5, characterized in that: A deep reinforcement learning model using an Actor-Critic architecture includes an Actor network and a Critic network. The Actor network is used to select the optimal combination of process parameters according to the current state, and the Critic network is used to evaluate the pros and cons of the process parameter combination. By continuously updating the parameters of the two networks, adaptive optimization of the process parameters for lattice structure preparation is achieved.

7. The method for preparing a lattice structure by multi-field assisted laser selective melting according to claim 6, characterized in that: Building an Actor-Critic Framework, Actor Network : By status is the input and output process parameters The probability distribution of , and update the parameters according to the policy gradient ;Critic Network : By status and actions is the input and output action value function , and update the parameters according to the time difference error ; Definition of states and actions: ; ;state Including temperature field , Flow Field , liquid fraction , electric field strength , magnetic field strength Harmony Field Frequency Physical field distribution; action Including laser power , Scanning speed , Spot diameter , electric field strength , magnetic field strength And the sound field frequency Process parameters; Designing Reward Functions for Deep Reinforcement Learning Models with Actor-Critic Architecture , Award Letter The number measures the difference between the actual performance and the target performance through an exponential function: ； ； Reward Function Considering the actual porosity , actual pore size distribution and actual mechanical properties performance indicators, among which , , is the weight coefficient, , , is the scale factor, is the target performance indicator of porosity, is the target performance indicator of pore size distribution, It is the target performance index of mechanical properties; The training steps of the deep reinforcement learning model of the Actor-Critic architecture are as follows: Use the deep deterministic policy gradient DDPG algorithm to train the Actor network and the Critic network; Update the Actor network parameters through policy gradients to optimize the selection strategy of process parameters; update the Critic network parameters through time difference errors to optimize the estimation of the action value function, and continuously iterate the training until convergence to the optimal process parameter combination; The optimal process parameter combination obtained by training The output is applied to the laser selective melting preparation process, and the control parameters of the laser selective melting equipment are adjusted according to the optimal parameter combination.

8. A device for preparing a lattice structure by multi-field assisted laser selective melting, which is used to implement the method for preparing a lattice structure by multi-field assisted laser selective melting as claimed in any one of claims 1 to 7, characterized in that: include: Data acquisition module, multi-field coupling physics module, multi-field environment control module, process optimization module and execution module; The data acquisition module is used to obtain the composition data and powder characteristics of the material to be processed and to obtain the performance index of the target lattice structure; The multi-field coupling physics module is used to construct a multi-field coupling physics model, and predict the melting, solidification and lattice structure evolution results of the material to be processed under a multi-field environment by combining numerical calculation and machine learning models; The multi-field environment control module is used to establish an adaptive multi-field environment control algorithm, which adjusts the time and space distribution of the electric field, magnetic field and acoustic field by real-time analysis of the laser-to-be-processed material interaction process, and controls the lattice structure morphology and defects; The process optimization module is used to construct a process optimization method based on deep learning, and adjust the process parameters of laser power, scanning speed and powder layer thickness during laser selective melting in real time through autonomous learning and decision-making of the deep learning model; The execution module adjusts the laser selective melting process to prepare the target lattice structure based on the output of the combined data of the best material composition to be processed, laser process parameters and multi-field environmental parameters for the target lattice structure.

9. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the method for preparing a lattice structure by multi-field assisted laser selective melting as described in any one of claims 1 to 7 by calling the computer program stored in the memory.

10. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are run on a computer, the computer is caused to execute the method for preparing a lattice structure by multi-field assisted laser selective melting as claimed in any one of claims 1 to 7.

Images