Method and system for optimizing residual stress of additive manufacturing

Abstract

The invention provides an additive manufacturing residual stress optimization method and system, and the method comprises the steps: searching a global optimal solution enabling the residual stress of an additive manufacturing part to be minimum from a technological parameter space through the combination of multi-physics field numerical simulation and an intelligent optimization algorithm; the multi-physics field numerical simulation comprises the steps of sampling from a process parameter space by using an optimization algorithm to obtain process parameters, inputting the sampled process parameters into a finite element model, performing thermal analysis on the process parameters based on the finite element model, and obtaining transient thermal distribution through thermal analysis. And performing structural analysis by using transient heat distribution to obtain residual stress under the current process parameters, and transmitting the obtained residual stress to an intelligent optimization algorithm for realizing process parameter optimization. The method has the advantages of high precision, low cost and the like, and high-precision collaborative optimization of multiple types of process parameters can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of additive manufacturing, and in particular to a method and system for optimizing residual stress in additive manufacturing. Background Art

[0002] Selective Laser Melting (SLM), especially selective laser melting (SLM), as an advanced 3D printing technology, deposits metal powder layer by layer and melts it with laser, showing significant advantages such as simple process, high material utilization and shortened production cycle. These characteristics greatly enhance the freedom of design, making it a popular manufacturing method in aerospace, automotive industry, medical device manufacturing and other fields. However, SLM technology is accompanied by significant temperature gradient changes during the forming process. During the rapid melting and solidification of the material, the uneven thermal expansion and contraction of the heat-affected zone leads to the accumulation of internal residual stress. If no measures are taken to control it, these high residual stresses will cause quality problems such as deformation and cracking of parts, thereby weakening the mechanical strength and fatigue resistance of the material.

[0003] To solve the above technical problems, the existing technology has studied the influence of process parameters such as laser power, scanning speed, and hatch spacing on residual stress through a large number of experiments and physical model analysis. For example, Mugwagwa et al. found that high laser power and low scanning speed would aggravate the formation of residual stress. However, the experimental method is time-consuming and laborious, and it is difficult to cover all possible parameter combinations, which limits its application in complex parameter spaces. Li et al. and Yang et al. developed thermomechanical coupling models and transient coupling thermomechanical analysis methods, and predicted temperature evolution and residual stress distribution through numerical simulation to provide guidance for process parameter optimization. The thermoelastic model proposed by Gusarov et al. further verified preheating as an effective means to reduce residual stress. Past studies have revealed the complex relationship between process parameters and residual stress. Although the optimization of a single parameter can partially alleviate residual stress, due to the complex nonlinear interaction between parameters, a single strategy is difficult to achieve global optimization. In addition, due to time and cost constraints, extensive experimental studies are difficult to fully construct a detailed association between process parameters and residual stress evolution, making optimization methods based on numerical simulation the preferred choice. However, existing optimization methods, such as local search strategies, are often limited by model accuracy and data volume, making it difficult to effectively find the global optimal solution for multiple process parameters. Secondly, the proxy model trained based on simulation and experimental data is applied to the meta-heuristic optimization algorithm. Although it takes into account the complex interactions between multiple parameters, it is limited by the original training data and cannot guarantee the accuracy of optimization.

[0004] In summary, the main problems of the existing technical solutions are concentrated on:

[0005] 1. Limitations of experimental methods: time-consuming, costly, and difficult to cover all possible parameter combinations.

[0006] 2. Limitations of numerical simulation: Although it can provide high-precision predictions, local search strategies make it difficult to find the global optimal solution.

[0007] 3. Limitations of the agent model combined with intelligent algorithms: It has a global optimization strategy, but is limited by data accuracy and quantity. Summary of the invention

[0008] The technical problem to be solved by the present invention is as follows: In view of the above-mentioned problems in the prior art, a high-precision, low-cost optimization method and system for additive manufacturing residual stress are provided, which can realize high-precision coordinated optimization of multiple types of process parameters.

[0009] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0010] A method for optimizing residual stress in additive manufacturing includes using multi-physical field numerical simulation in combination with an intelligent optimization algorithm to find a global optimal solution that minimizes the residual stress of an additively manufactured part from a process parameter space. The multi-physical field numerical simulation includes using an optimization algorithm to sample process parameters from the process parameter space, inputting the sampled process parameters into a finite element model, performing thermal analysis on the process parameters based on the finite element model, obtaining transient thermal distribution through thermal analysis, performing structural analysis using the transient thermal distribution to obtain residual stress under current process parameters, and transmitting the obtained residual stress to the optimization algorithm for realizing process parameter optimization.

[0011] Furthermore, when the transient heat distribution is obtained by thermal analysis based on the finite element model, the functional expression of the temperature field used is as shown in the following formula:

[0012]

[0013] In the above formula, ρ is the density, C p is the specific heat capacity, T is the temperature of the powder material, t is the interaction time between the heat source and the powder bed, k is the thermal conductivity, (x, y, z) is the coordinate system coordinate to which the heat source is attached, and Q is the heat generated per volume in the component; the functional expression of the boundary conditions used is shown in the following formula:

[0014]

[0015] In the above formula, k represents the thermal conductivity, T represents the temperature of the metal powder at a certain moment, n represents the infinite tendency to infinity in the integral, and h c represents the thermal convection coefficient, T 0 represents the initial temperature, ε θrepresents the thermal radiation coefficient, σ represents the Stefan-Boltzmann constant, q represents the heat flux, (x, y, z) represents the coordinate system coordinates to which the heat source is attached, and S represents the surface subject to heat flux, convection, and radiation.

[0016] Furthermore, when the transient heat distribution is obtained by thermal analysis based on the finite element model, the heat source model used to simulate the laser energy is the Gaussian surface heat source of the following formula:

[0017]

[0018] In the above formula, q r represents the energy of the laser heat source, A represents the absorptivity of the powder material used in additive manufacturing, P represents the laser power, exp(·) represents the exponential power of e, ω is the equivalent radius of the laser spot, and r represents the radial distance from a certain point (x, y) to the center of the laser spot.

[0019] Further, the use of transient thermal distribution to perform structural analysis to obtain residual stress under current process parameters includes:

[0020] S101, constructing an additive manufacturing part model by partitioning and layering, and completing meshing of the additive manufacturing part model;

[0021] S102, defining the grid type in the thermal analysis process, inputting the thermal physical property parameters of the material in different states, loading the moving heat source model, and defining the state transition function of the material;

[0022] S103, after the initial conditions and boundary conditions are defined, a transient temperature field calculation is performed, wherein the transient temperature field calculation includes simulating the temperature change of the additive manufacturing part model during the manufacturing process until the additive manufacturing part model is cooled to room temperature, and obtaining temperature field data under the full time state;

[0023] S104, defining a mesh type, mechanical parameters related to the material, and a plastic hardening function of the material during an elastic-plastic mechanical force analysis process, loading boundary conditions and activation conditions, for evaluating the mechanical behavior and residual stress distribution of the additively manufactured part during the additive manufacturing process;

[0024] S105, according to the temperature field data, applying the unit temperature of the whole time process as a load to the mechanical model to perform coupled solution and steady-state stress calculation to obtain the residual stress under the current process parameters.

[0025] Furthermore, when the multi-physics field numerical simulation is combined with the intelligent optimization algorithm to find the global optimal solution that minimizes the residual stress of the additively manufactured part from the process parameter space, the intelligent optimization algorithm used is a particle swarm-genetic hybrid optimization algorithm, and the process parameters for additive manufacturing that minimize the residual stress from the process parameter space are found to include:

[0026] S201, extracting sample data from the process parameter space through the optimal Latin hypercube to initialize the population;

[0027] S202, transmitting the current population data to the multi-physics model to perform thermal-structural analysis to obtain residual stress;

[0028] S203, transmitting the residual stress data to an intelligent optimization algorithm as its fitness value based on the fitness value to obtain the optimal value of the population and the individual;

[0029] S204, sorting the current population in descending order based on the fitness value, and then performing crossover and mutation operations, updating the adaptive inertia weight and dynamic learning factor, and updating the speed and position of the population particles;

[0030] S205, transferring the new population data to the multi-physics model to perform thermal-structural analysis to obtain residual stress;

[0031] S206, transmitting the residual stress data to the optimization algorithm as its fitness value, and obtaining the optimal values ​​of the population and the individual based on the fitness value;

[0032] S207, determine whether the number of iterations meets the termination condition. If not, add 1 to the number of iterations and jump to step S4 to continue iterating; otherwise, output the optimal values ​​of the population and the individual.

[0033] Furthermore, the following formula is used to update the adaptive inertia weight factor and learning factor:

[0034]

[0035] In the above formula, ω 1 represents the adaptive inertia weight, ω max represents the maximum inertia weight, ω min represents the minimum inertia weight, c 1 and c 2 is the dynamic learning factor, c 1 represents the learning factor of the particle from itself, c 2 represents the learning factor of the particle learning from the global optimal particle, f represents the fitness value of the particle, and f avg represents the average fitness value of all particles at present, f min Indicates the minimum fitness value of all particles at present, ti is the current iteration number, t max is the maximum number of iterations of the particle swarm.

[0036] Furthermore, the speed and position of the population particles are updated using the following formula:

[0037]

[0038] In the above formula, represents the velocity of the k+1 generation particle, represents the velocity of the k-generation particle, represents the position of the k+1 generation particle, represents the position of the k-generation particle, represents the optimal value of individuals in generation k, represents the optimal value of the population of generation k, k represents the number of iterations, r 1 and r 2 represents a random positive number drawn from a uniform distribution [0,1], and α represents a constraint factor used to control the weight of the speed.

[0039] Furthermore, when the multi-physics field numerical simulation is combined with the intelligent optimization algorithm to find the global optimal solution that minimizes the residual stress of the additively manufactured parts from the process parameter space, the multi-physics field numerical simulation is implemented based on COMSOL simulation software, and the optimization algorithm is implemented based on Matlab. The sending of the obtained global residual stress to the optimization algorithm includes: the COMSOL simulation software establishes a data transmission interface with Matlab through the COMSOL Multiphysics Server, and the COMSOL simulation software saves the obtained global residual stress model data from the .mph file as a .m file, and then establishes a connection with the COMSOL simulation software in MATLAB through the mphstart command and imports the model data in the .m file format through the mphopen command, and then uses the mphglobal command to automatically import the model data into MATLAB, thereby realizing seamless data transmission between the COMSOL simulation software and Matlab.

[0040] A multi-parameter collaborative optimization system for additive manufacturing residual stress comprises a microprocessor and a memory connected to each other, wherein the microprocessor is programmed or configured to execute a collaborative optimization method for additive manufacturing residual stress.

[0041] A computer-readable storage medium stores a computer program / instruction, which is programmed or configured to execute a method for optimizing residual stress in additive manufacturing through a processor.

[0042] Compared with the prior art, the advantages of the present invention are:

[0043] 1. The present invention uses multi-physics field numerical simulation combined with intelligent optimization algorithm to find the global optimal solution that minimizes the residual stress of additive manufacturing parts from the process parameter space. The multi-physics field numerical simulation uses the optimization algorithm to sample the process parameters from the process parameter space, inputs the sampled process parameters into the finite element model, performs thermal analysis on the process parameters based on the finite element model, obtains transient thermal distribution through thermal analysis, and uses transient thermal distribution to perform structural analysis to obtain residual stress under the current process parameters. The present invention can more accurately predict and control the thermodynamic behavior in the additive manufacturing process, thereby more accurately calculating the distribution of residual stress. The obtained residual stress is then passed to the intelligent optimization algorithm, which can search for the global optimal solution in a larger process parameter space, greatly reducing the number of experiments and costs, and improving the optimization efficiency. Through the combination of physical field numerical simulation and intelligent optimization algorithm based on multiple iterative optimization, the process parameters can be continuously improved to ensure that the quality and performance of the final product reach the best state, significantly improve the dimensional stability and mechanical properties of the parts, reduce deformation, cracks and other problems caused by residual stress, and reduce the scrap rate, thereby improving manufacturing efficiency and economy.

[0044] 2. The present invention further adopts the optimal Latin hypercube to initialize the population, and satisfies the space filling characteristic through the maximum distance criterion, thereby ensuring the uniform distribution of each independent variable sampling point in the search space, so that the intelligent optimization algorithm can cover a larger search area at the beginning, which is more conducive to finding the global optimal solution.

[0045] 3. The present invention further introduces an adaptive inertia weight factor and a dynamic learning factor, so that the intelligent algorithm can effectively explore the entire search space in the early stage, improve the ability of global search, and avoid falling into the local optimal solution too early. As the number of iterations increases, the change strategy of the adaptive inertia weight factor and the dynamic learning factor can allow the algorithm to gradually focus on the optimal solution, accelerate the convergence process of the algorithm, and improve the solution efficiency. By dynamically adjusting the algorithm parameters, it can be ensured that the algorithm can both explore extensively and converge accurately during the search process, which helps to obtain a more accurate solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of a method for optimizing residual stress in additive manufacturing according to an embodiment of the present invention.

[0047] Figure 2 Schematic diagram of a three-dimensional finite element model and laser scanning strategy of the L-PBF process according to an embodiment of the present invention.

[0048] Figure 3 The figure is a flow chart of the adaptive particle swarm optimization method according to an embodiment of the present invention.

[0049] Figure 4 The figure is an overall flow chart of the joint operation according to an embodiment of the present invention.

[0050] Figure 5 Schematic diagram of the molten pool size simulated by the sample model of the embodiment of the present invention.

[0051] Figure 6 It is a transient residual stress cloud diagram of the top surface at different times of the embodiment of the present invention.

[0052] Figure 7 1 is a stress-temperature comparison diagram and a stress-temperature gradient comparison diagram of point P4 during the scanning process in an embodiment of the present invention.

[0053] Figure 8 It is a cloud diagram of the residual stress on the top surface of the material of the embodiment of the present invention before and after it is cooled to room temperature and the residual stress on each surface after the substrate is removed.

[0054] Fig. 9 It is a line graph of residual stress before and after the substrate is removed according to an embodiment of the present invention.

[0055] Fig.10 4 is a comparison chart of the experimental measurement results and the numerical simulation results of the embodiment of the present invention.

[0056] Fig.11 Comparison diagram of the maximum residual stress on the top surface of the iterative results of the embodiment of the present invention and sample 1, sample 2 and sample 3 (optimal process parameters) DETAILED DESCRIPTION

[0057] 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 only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] like Figure 1 As shown, a method for optimizing residual stress in additive manufacturing includes using multi-physics field numerical simulation in combination with intelligent optimization algorithm to find the global optimal solution that minimizes the residual stress of additively manufactured parts from the process parameter space. The multi-physics field numerical simulation includes using the optimization algorithm to sample process parameters from the process parameter space, inputting the sampled process parameters into a finite element model, performing thermal analysis on the process parameters based on the finite element model, obtaining transient thermal distribution through thermal analysis, performing structural analysis using transient thermal distribution to obtain residual stress under current process parameters, and passing the obtained residual stress to the intelligent optimization algorithm for realizing process parameter optimization.

[0059] In this embodiment, the residual stress of the laser powder bed fusion (L-PBF) formed parts is reduced by efficiently co-optimizing the process parameters, and the thermo-mechanical elastic-plastic finite element simulation is combined with the adaptive heuristic hybrid optimization algorithm. First, a thermo-mechanical multi-physics field finite element model is developed to accurately obtain the residual stress data of the L-PBF formed parts. Among them, the thermo-mechanical multi-physics field finite element model combines the elastic-plastic deformation of the material and adopts the thermo-mechanical indirect coupling modeling method. While ensuring the accuracy of the calculation, it will also greatly reduce the amount of calculation. At the same time, the establishment process of the thermo-mechanical multi-physics field finite element model also considers the material state transition, temperature-dependent thermal properties, enhanced anisotropic thermal conductivity and phase change latent heat, etc. In the indirect coupling thermo-mechanical modeling method, the transient heat distribution during L-PBF is initially realized by heat transfer analysis. Secondly, the transient heat distribution obtained by heat transfer analysis is used to obtain the residual stress distribution through structural analysis.

[0060] In a specific application embodiment, the finite element modeling of the laser powder bed fusion (L-PBF) process includes two processes: nonlinear transient thermal analysis and quasi-static elastoplastic mechanical analysis. Specifically, after the temperature field is obtained by nonlinear transient thermal analysis, a mechanical analysis is performed to approximate thermal stress and deformation, wherein the temperature field is used as a heat load for mechanical analysis. Thermal analysis and mechanical analysis can be performed in sequence. For uncoupled or weakly coupled analysis, the friction heat caused by elastoplastic deformation is much smaller than the thermal energy. This embodiment prefers a weakly coupled thermal-mechanical analysis method to reduce the computational cost. This model simultaneously takes into account the attenuated Gaussian heat source, the thermal dependence of physical parameters, the enhanced anisotropic thermal conductivity, the transition between powder-liquid-solid state and the latent heat of phase change, etc.

[0061] In this embodiment, when performing thermal analysis based on the finite element model to obtain transient thermal distribution, the functional expression of the temperature field used is as shown in the following formula:

[0062]

[0063] In the above formula, ρ is the density, C p is the specific heat capacity, T is the temperature of the powder material, t is the interaction time between the heat source and the powder bed, k is the thermal conductivity, (x, y, z) is the coordinate system coordinate to which the heat source is attached, and Q is the heat generated per volume in the component; the functional expression of the boundary conditions used is shown in the following formula:

[0064]

[0065] In the above formula, k represents the thermal conductivity, T represents the temperature of the metal powder at a certain moment, n represents the infinite tendency to infinity in the integral, and h c represents the thermal convection coefficient, t 0 represents the initial temperature, ε θrepresents the thermal radiation coefficient, σ represents the Stefan-Boltzmann constant, q represents the heat flux, (x, y, z) represents the coordinate system coordinates to which the heat source is attached, and S represents the surface subject to heat flux, convection, and radiation.

[0066] The heat source model has a certain influence on the calculation and analysis of the temperature field and stress of the L-PBF process. Improper selection of the heat source model will lead to a large deviation between the calculated results and the actual results. It can be understood that in specific applications, the most appropriate heat source model can be selected according to actual needs. Since the Gaussian surface heat source can well reflect the heat flux density distribution on the surface of the melt path, it is more effective when dealing with relatively shallow molten pool depths. Therefore, this embodiment selects a Gaussian surface heat source to simulate the laser energy, such as Figure 2 As shown in the figure, the size of the substrate model is 1.4mm×1.4mm×0.3mm, and the size of the A1-Mg-Sc-Zr powder bed model is 0.5mm×0.5mm×0.03mm. The powder bed is set to four tracks, and the coordinate origin of the physical model is located at the coincidence surface of the substrate and the powder bed. The X-axis is along the laser scanning direction, the Y-axis is perpendicular to the laser scanning direction, and the Z-axis is the direction of the powder layer increase. Considering the calculation accuracy and simulation efficiency, the powder bed uses a fine grid of hexahedral Solid 70 units with a grid accuracy of 0.02mm×0.02mm×0.01mm, and the substrate uses a coarser tetrahedral grid. The laser scanning strategy is a reciprocating grating mode. Three points are set on the center line to view the residual stress at different positions, and the spacing between the two points is 0.1mm. In order to be consistent with the actual forming process, a 10s powder spreading time is set after each layer of heat source is loaded. This time period is determined by the ratio of the top surface area of ​​the physical model to the actual powder feeding area in the previous experiment, and is converted into real time in the actual laser processing powder feeding process.

[0067] In this embodiment, when performing thermal analysis based on the finite element model to obtain transient heat distribution, the heat source model used to simulate the laser energy is the Gaussian surface heat source of the following formula:

[0068]

[0069] In the above formula, q r represents the energy of the laser heat source, A represents the absorptivity of the powder material used in additive manufacturing, and the reflectivity λ of the known material can be calculated, A=1-λ(the reflectivity of A1-Mg-Sc-Zr alloy is λ=0.82), P represents the laser power, exp(·) represents the exponential power of e, and ω is the equivalent radius of the laser spot, at which the energy density drops to 1 / e of the intermediate maximum value 2 , r represents the radial distance from a certain point (x, y) to the center of the laser spot.

[0070] In this embodiment, during the quasi-static elastic-plastic mechanical analysis, the control stress balance equation is:

[0071]

[0072] Where σ represents stress and the mechanical constitutive law for elastic behavior is:

[0073] σ=C·(∈ e +∈ p +∈ T ), (5)

[0074] Where C represents the fourth-order isotropic material stiffness tensor, ∈ e represents elastic strain, ∈ p represents plastic strain, ∈ T represents thermal strain, and the stress-strain components are:

[0075] σ=[σ xx σ yy σ zz σ xy σ yz σ xz ] T , (6)

[0076] ∈=[∈ xx ∈ yy ∈ zz ∈ xy ∈ yz ∈ xz ] T , (7)

[0077] The total strain ε is given by the displacement increment, σ xx σ yy σ zz σ xy σ yz σ xz Represent the six stress components at the point, ∈ xx ∈ yy ∈ zz ∈ xy ∈ yz ∈ xz Represents the six strain components at the point, U=[uvw] T Determined using the following relationship:

[0078] ∈=BU, (8)

[0079] Among them, B is the matrix related to displacement and strain, u represents the displacement component of a point in the x-axis direction, v represents the displacement component of a point in the y-axis direction, w represents the displacement component of a point in the z-axis direction, and the displacement increment U is calculated according to the following formula:

[0080] KU=F o -Fi , (9)

[0081] Where K represents the stiffness matrix, F o is the external force, F i is the Newton-Raphson restoring force resulting from the loading within the element.

[0082] When metal materials undergo plastic deformation, the yield function, flow law, and kinematic hardening law are used to study the rate-independent plasticity. The material in the plastic state conforms to the von Mises yield criterion, where the equivalent stress (σ m ) can be expressed as:

[0083]

[0084] Among them, σ 1 , σ 2 and σ 3 are the principal stresses in different directions.

[0085] If the von Mises stress is greater than the simple tensile yield stress, the material yields. The yield function f is calculated as:

[0086] T=σ m -σ Y , (11)

[0087] Among them, σ Y is the yield strength, and the initial yield stress is denoted by σ Y0 The initial yield strength used in the model is 360 MPa, and it decreases with increasing temperature.

[0088] σ Y (T) = σ Y0 (T)+σ h (∈ ep ), (12)

[0089] Among them, σ Y (T) represents the yield strength, which is a function of temperature, σ Y0 (T) represents the initial yield stress, σ h (∈ ep ) represents the strength change caused by plastic movement hardening.

[0090] Apply relevant flow laws to determine plastic strain rates

[0091]

[0092] Where f is the yield function, The hardening parameters are determined by the consistency conditions.

[0093] Thermal strain is calculated using the temperature-dependent thermal expansion coefficient α, and the thermal strain force ∈ is calculated according to the following formula T :

[0094] ∈ T =α(TT 0 ), (14)

[0095] Where α is the temperature-dependent thermal expansion coefficient, T 0 represents the reference temperature field, and T represents the current temperature.

[0096] In the SLM process simulation, the thermophysical parameters mainly include density, specific heat capacity, thermal conductivity and thermal expansion coefficient. For solid materials, their thermophysical parameters can be obtained by simulating their components using Jmatpro software, while the thermophysical parameters of powder materials are obtained by the following formulas.

[0097] The density of powder material is expressed as follows:

[0098] ρ p =φρ g +(1-φ)ρ s , (15)

[0099] In the above formula: s Represents the density of solid material; ρ p Indicates the density of the powder; ρ g represents the density of argon gas; φ represents the porosity of the powder, Here ρ s and ρ p They are selected as 1.420g / cm3 and 2.658g / cm3 respectively.

[0100] The thermal conductivity expression of powder material is as follows:

[0101] k p =k s (T)·(1-0.2φ-1.73φ 2 ), (16)

[0102] In the above formula, k s (T) represents the thermal conductivity of the solid material, k p Represents the thermal conductivity of powder material.

[0103] When the temperature exceeds the melting point of the powder, the Marangoni convection in the molten pool will accelerate the heat conduction. Marangoni refers to the flow phenomenon caused by uneven surface tension on the liquid interface. Therefore, the thermal conductivity is introduced in the simulation and the thermal conductivity of the liquid is defined as:

[0104] k xl =w x kx , k yl =w y k y , k z1 =W z k z , (17)

[0105] In the above formula, k xl , k yl , k z1 They represent the thermal conductivity of the liquid material in the x, y and z directions, k x , k y and k z represents the thermal conductivity of the solid material in the x, y and z directions respectively. In this embodiment, the thermal conductivity is set to w x =4,w y =4,w z =5.

[0106] The specific heat capacity expression of powder material is as follows:

[0107]

[0108] In the above formula, C s is the specific heat capacity of the solid material, C g is the specific heat capacity of argon, φ is the porosity of the powder, ρ s is the density of the solid material, ρ p is the density of the powder particles, ρ g is the density of nitrogen.

[0109] In summary, in this embodiment, the steps of using transient thermal distribution to perform structural analysis to obtain residual stress under current process parameters include:

[0110] Step S101, constructing an additive manufacturing part model by partitioning and layering, and completing meshing of the additive manufacturing part model;

[0111] Step S102, defining the grid type in the thermal analysis process, inputting the thermal physical property parameters of the material in different states, loading the moving heat source model, and defining the state transition function of the material;

[0112] Step S103, after defining the initial conditions and boundary conditions, performing transient temperature field calculation of the transient temperature field meter includes simulating the temperature change of the additive manufacturing part model during the manufacturing process until the additive manufacturing part model cools to room temperature, and obtaining temperature field data under the full time state;

[0113] Step S104, defining the mesh type, mechanical parameters of the material and the plastic hardening function of the material in the elastic-plastic mechanical force analysis process, loading boundary conditions and activation conditions, so as to evaluate the mechanical behavior and residual stress distribution of the additively manufactured part in the additive manufacturing process;

[0114] Step S105 , based on the temperature field data, the unit temperature of the entire time process is applied as a load to the mechanical model for coupled solution and steady-state stress calculation to obtain the residual stress under the current process parameters.

[0115] In order to simulate the actual layer-by-layer forming process, the "element birth and death" technology is used. When the first layer of powder is printed, the subsequent layers are in the "death" state of the element. In this case, the stiffness matrix of the element is temporarily multiplied by a very small factor, so the load on the element will be zero. It can be considered that the element does not participate in the calculation and has no effect on the temperature field of the current forming layer. When the next layer is reached, the element is activated and "born", and its material parameters and loads are restored to the initial values.

[0116] During the L-PBF process, the laser heat source is loaded, and the local powder absorbs energy, and its temperature rises rapidly to reach the melting point. The material changes from powder to liquid. After the heat source scans, the local powder quickly melts and solidifies into a solid. At the same time, the temperature of some powders rises, but does not reach the melting point. If so, the properties of the powder material remain unchanged. Therefore, during the simulation, different material properties of powder, liquid and solid should be specified at different times and different locations. This process is implemented through state variables and phase change modules in COMSOL. Set the state variable "state1" to display the temperature value of the unit. Set the state variable "state2" with an initial value of 0 to regulate the transformation of powder and solid material properties. The initial material properties are set to powder and "state2" to 0. When "state1" exceeds the melting point, a phase change occurs, and the material property becomes liquid. When "state1" is less than the melting point again, "state2" changes to 1, and the material property is solid. A one-way transformation occurs during the consolidation process from powder to liquid, while a two-way transformation only exists between solid and liquid.

[0117] In this embodiment, when multi-physics field numerical simulation is combined with intelligent optimization algorithm to find the global optimal solution that minimizes the residual stress of additively manufactured parts from the process parameter space, the intelligent optimization algorithm used is a particle swarm-genetic hybrid optimization algorithm. The process parameters of additive manufacturing that minimize the residual stress are found from the process parameter space. The following are included:

[0118] Step S201 , extracting sample data from the process parameter space through the optimal Latin hypercube to initialize the population.

[0119] Since the sample point sequence is a random arrangement, its space filling effect is not good. The optimal Latin hypercube sampling uses the maximum distance criterion to meet the space filling characteristics. That is, the minimum distance between each sample point and the previous sample points is the characteristic distance of the sample point, and the purpose of optimization is to maximize this characteristic distance, so that the sample points are more discrete and fill the entire space, so as to ensure the uniform distribution of each independent variable sampling point. The LHD design with the maximum minimum distance between stations can be expressed as:

[0120] Max:Mind(x i ,x j ), (19)

[0121] For n points in an m-dimensional hypercube, the maximum and minimum distance criterion can be expressed as:

[0122]

[0123] Among them, F k It is the scaling factor, d(x i ,x j ) represents particle x i ,x j The distance between.

[0124] In this embodiment, the speed and position of the population particles are updated using the following formula:

[0125]

[0126] In the above formula, represents the velocity of the k+1 generation particle, represents the velocity of the k-generation particle, represents the position of the k+1 generation particle, represents the position of the k-generation particle, represents the optimal value of individuals in generation k, represents the optimal value of the population of generation k, k represents the number of iterations, r 1 and r 2 represents a random positive number drawn from a uniform distribution [0,1], and α represents a constraint factor used to control the weight of the speed.

[0127] The inertia weight factor affects the convergence and search capabilities of the particle swarm algorithm (PSO). A larger value is conducive to global search, and a smaller value is conducive to accelerating the convergence speed. In order to balance the global search and local improvement capabilities of the PSO algorithm, this embodiment adopts an adaptive inertia weight, that is, the inertia weight changes with the change of the particle objective function value. When the particle target value is dispersed, the inertia weight is reduced; when the particle target value is consistent, the inertia weight is increased. The following formula is used to update the adaptive inertia weight factor and the learning factor to adjust the particle search behavior to balance the ability of global search and local search:

[0128]

[0129] In the above formula, ω 1 represents the adaptive inertia weight, ω max represents the maximum inertia weight, ω min represents the minimum inertia weight, f represents the fitness value of the particle, and f avg represents the average fitness value of all particles at present, f min Indicates the minimum fitness value of all particles at present. The above dynamic learning factor is used to control the learning rate or variation degree of the algorithm, and generally takes a fixed value. In the process of the particle swarm algorithm in this embodiment, the weights of the learning factors change as the algorithm proceeds. In order to prevent particles from quickly gathering around the local optimal solution in the early stage of evolution and make particles search in a large range in the global field, the learning factor c that particles learn from themselves is set 1 Taking a larger value, the learning factor c of the particle learning from the global optimal particle 2 Take a smaller value. In order to make the particles converge to the global optimal solution quickly and accurately in the later stage of the search, and to improve the convergence speed and accuracy of the algorithm, set the learning factor c that the particles learn from themselves. 1 Taking a smaller value, the learning factor c of the particle learning from the global optimal particle 2 Therefore, this embodiment constructs c 1 As a monotonically decreasing function, construct c 2 is a monotonically increasing function, and their expressions are as follows:

[0130]

[0131] In the above formula, c 1 and c 2 is the dynamic learning factor, c 1 represents the learning factor of the particle from itself, c 2 Represents the learning factor of the particle learning from the global optimal particle, t i is the current iteration number, t max is the maximum number of iterations of the particle swarm.

[0132] Step S202, transmitting the current population data to the multi-physics field model, and performing thermal-structural analysis to obtain residual stress.

[0133] Step S203, transmitting the residual stress data to the intelligent optimization algorithm as its fitness value to obtain the optimal value of the population and the individual based on the fitness value.

[0134] Step S204, sorting the current population in descending order based on the fitness value, and then performing crossover and mutation operations, updating the adaptive inertia weight and dynamic learning factor, and updating the speed and position of the population particles;

[0135] Step S205, transmitting the new population data to the multi-physics model to perform thermal-structural analysis to obtain residual stress;

[0136] Step S206, transmitting the residual stress data to the optimization algorithm as its fitness value, and obtaining the optimal values ​​of the population and the individual based on the fitness value;

[0137] Step S207, determine whether the number of iterations meets the termination condition. If not, add 1 to the number of iterations and jump to step S204 to continue iterating; otherwise, output the optimal values ​​of the population and the individual.

[0138] In this embodiment, the crossover and mutation operations of the genetic algorithm (GA) are embedded in the PSO algorithm to form a particle swarm-genetic hybrid optimization algorithm (PSO-GA) with multiple strategies. The PSO-GA hybrid optimization algorithm includes the optimal Latin hypercube initialization population, adaptive inertia weights, dynamic learning factors, and genetic operations. The optimal Latin hypercube method is used to select the initial population, which improves the uniformity of the samples and alleviates the model from falling into the local optimum. The adaptive weights and dynamic learning factors are used to learn the global overview of fitness, so as to quickly find the area where the global optimal value is located. The crossover and mutation operations of fusion genetic optimization are used to improve the population diversity and the ability of random global search, so as to improve the population diversity and the ability of random global search, and further enhance the efficiency of algorithm search.

[0139] Among them, the crossover operation is to pair individuals in pairs, randomly select a gene, use this position as the crossover point, and generate a new individual by exchanging the parts after the crossover point of the two bodies to promote information exchange between particles and increase the diversity of search. The mutation operation is to randomly decide whether to perform the mutation operation in each iteration. If the mutation operation is decided, a mutation point is randomly generated and the gene value of the point is changed. This can introduce randomness, help to jump out of the local optimal solution, and increase the global search ability of the algorithm.

[0140] like Figure 3As shown, in a specific application embodiment, the detailed steps of using the above intelligent optimization algorithm to obtain the position of the global optimal solution and the corresponding fitness value include:

[0141] Step 1, set the relevant parameters: 20 generations of particles; 40 maximum iterations; the initial inertia weight of the adaptive inertia weight factor is set to 0.8, the initial self-learning factor is set to 0.75, and the swarm learning factor is set to 2; the crossover and mutation probabilities of the genetic algorithm are set to 0.5 and 0.1 respectively; and initialize the particle swarm speed and position boundaries, the position boundaries are set to [250, 400; 900, 1400; 100, 140], and the speed boundaries are set to [-100, 100; -400, 400; -30, 30].

[0142] Step 2: The process parameter range is: laser power P 250w~400w, scanning speed V 1000mm / s~1400mm / s, scanning spacing H 100μm~140μm, which is the position boundary. The optimal Latin hypercube sampling is performed on each process parameter to be optimized to form an initial population and calculate its fitness value (i.e., the residual stress obtained by the multi-physics field model).

[0143] Step 3: For each particle, the best position and the best fitness of the population are obtained by comparing its fitness value with the fitness value of the best position it has experienced.

[0144] Step 4: Sort the fitness values ​​in descending order; for each particle, randomly decide to perform a crossover or mutation operation based on the set crossover probability and mutation probability.

[0145] Step 5: New adaptive inertia weight and dynamic learning factor, calculate new speed and position and perform boundary processing.

[0146] Step 6: Determine whether the updated particle position, i.e., process parameters, meets 60J / mm 3 <E<120J / mm 3 (E=(P*1000) / (V*H*0.03). If it is satisfied, the residual stress is calculated as the fitness value through the multi-physics field model. If it is not satisfied, 400MPa is assigned as its fitness value.

[0147] Step 7: Compare the fitness value of the current position with the individual optimal solution, and update the individual optimal solution and the corresponding value. Update the global optimal solution, compare the fitness value of the current position with the group optimal solution, and if it is better, update the group optimal solution and the corresponding value.

[0148] Step 8, check whether the convergence condition is met. If the preset maximum number of iterations (ie, 40 times) is not reached, return to step 4.

[0149] Step 9: Output the position of the global optimal solution and the corresponding fitness value as the result of the algorithm.

[0150] In summary, if Figure 4 As shown, in this embodiment, the thermo-mechanical elastic-plastic finite element simulation is combined with an adaptive heuristic hybrid optimization algorithm. First, a thermo-mechanical elastic-plastic finite element model is developed to accurately obtain the residual stress data of the L-PBF formed parts. The model takes into account the elastic-plastic deformation of the material and adopts a thermal-mechanical indirect coupling modeling method, which will greatly reduce the amount of calculation while ensuring accuracy. The model also considers material state transitions, temperature-dependent thermal properties, anisotropic thermal conductivity, and latent heat of phase change. In the indirect coupling thermo-mechanical modeling method, the transient heat distribution during L-PBF is initially realized by heat transfer analysis. Secondly, the residual stress distribution is obtained by structural analysis using the transient heat distribution obtained by heat transfer analysis.

[0151] Furthermore, a particle swarm-genetic hybrid optimization algorithm that integrates multiple strategies takes into account both optimization efficiency and accuracy. The PSO-GA hybrid optimization algorithm includes the optimal Latin hypercube to initialize the population, adaptive inertia weights, dynamic learning factors, and genetic operations. The optimal Latin hypercube method is used to select the initial population, which improves the uniformity of the samples and alleviates the model from falling into the local optimum. The adaptive weights and dynamic learning factors are used to learn the global profile of fitness, so as to quickly find the area where the global optimal value is located. The crossover and mutation operations of the fusion genetic optimization are used to improve the population diversity and the ability of random global search.

[0152] Since data cannot be directly exchanged between these two types of software, in this embodiment, when multi-physics field numerical simulation is combined with an intelligent optimization algorithm to find the global optimal solution that minimizes the residual stress of additively manufactured parts from the process parameter space, the multi-physics field numerical simulation is implemented based on COMSOL simulation software, and the optimization algorithm is implemented based on Matlab. The sending of the obtained global residual stress to the optimization algorithm includes: the COMSOL simulation software establishes a data transmission interface with Matlab through the COMSOL Multiphysics Server, and the COMSOL simulation software saves the obtained global residual stress model data from the .mph file as a .m file, and then establishes a connection with the COMSOL simulation software in MATLAB through the mphstart command and imports the model data in the .m file format through the mphopen command, and then uses the mphglobal command to automatically import the model data into MATLAB, thereby realizing seamless data transmission between the COMSOL simulation software and Matlab.

[0153] Therefore, the main advantages of the present invention include:

[0154] 1. The present invention proposes a multi-parameter collaborative optimization method for residual stress in additive manufacturing, which innovatively combines multi-physics field numerical simulation with an adaptive heuristic hybrid optimization algorithm. This optimization method can achieve high-precision collaborative optimization of multiple types of process parameters, combining the high interpretability of the model based on physical principles and consideration of the complex coupling effects between multiple parameters; it can achieve high-fidelity models and global optimal solutions, while having the high fidelity of multi-physics field finite element models and the high efficiency of global optimization of metaheuristic algorithms.

[0155] 2. The particle swarm-genetic hybrid optimization algorithm (PSO-GA) designed by the present invention integrates multiple strategies. The particle swarm optimization algorithm is improved by optimal Latin hypercube initialization, adaptive weights, dynamic learning factors and fusion genetic algorithms to improve the efficiency of the optimization algorithm and the global optimization effect. The optimal Latin hypercube method is used to select the initial population, which improves the uniformity of the sample and alleviates the model from falling into the local optimum. The global profile of fitness is learned using adaptive weights and dynamic learning factors, so as to quickly find the area where the global optimal value is located. The crossover and mutation operations of fusion genetic optimization are used to improve the population diversity and the ability of random global search. The optimization parameters include laser power, scanning speed, hatching spacing and substrate preheating temperature, and the optimization goal is to minimize the global residual stress.

[0156] 3. The present invention constructs a high-fidelity thermomechanical multi-physics field coupling model. The multi-physics field model adopts a method that can couple thermal analysis and mechanical analysis in sequence, which ensures high accuracy while greatly reducing the amount of calculation. In addition, this model considers the decaying Gaussian heat source, thermal dependence of physical parameters, enhanced anisotropic thermal conductivity, transition between powder-liquid-solid state and latent heat of phase change in the thermal analysis stage, and considers elastic-plastic stress and isotropic hardening strain in the force analysis stage, which is more suitable for the additive manufacturing process, ensuring the high fidelity of the model while taking into account the computational efficiency.

[0157] The effectiveness and accuracy of the additive manufacturing residual stress optimization method proposed in the present invention are verified by comparing experimental data. The specific verification contents are as follows:

[0158] In the L-PBF process, as the laser heat source scans successively, the powder absorbs energy and melts into a molten pool; as the laser moves away, the molten pools overlap and solidify. Accurate simulation of the molten pool size is a necessary condition for simulating residual stress. Verify the accuracy of the model simulation of the molten pool size, such as Figure 5 As shown, the solid ellipse line in the numerical simulation diagram of the molten pool size and the metallographic diagram represents the melting point of the material, and the inside is the molten pool area. Figure 5 (a) is sample 1, the laser power is set to 200W, the scanning speed is 1000mm / s, and the scanning interval is 140μm. Figure 5(b) is sample 2, with the laser power set to 300W, the scanning speed to 1200mm / s, and the scanning interval to 120μm. The simulated molten pool width of sample 1 is 94mm and the depth is 20.5mm. The simulated molten pool width of sample 2 is 126mm and the depth is 35mm. Figure 5 As shown in (b) and (c), the molten pool sizes of sample 1 and sample 2 printed under the same process parameters were measured. The measured molten pool width of sample 1 is about 91mm and the depth is about 21mm. The measured molten pool width of sample 2 is about 120mm and the depth is about 36mm. Overall, the simulated molten pool width is slightly larger than the experimental data, and the simulated molten pool depth is slightly smaller than the experimental data. Comparing the real data with the simulated data, the maximum width error is less than 6.8%, and the maximum depth error is less than 8.9%. This shows that the model has high accuracy and can be used for residual stress simulation.

[0159] like Figure 6 As shown, the von Mises stress cloud diagrams of the top surface of the material at (a) 4E-4s, (b) 8E-4s, (c) 1.3E-3s and (d) 1.8E-3s during the scanning process are compared. When the powder bed is heated to a temperature much higher than the bottom layer, the new material layer will expand uniformly. This expansion will be restricted by the low-temperature substrate, resulting in the generation of compressive stress in the printed layer and tensile stress in the substrate. When the heat source is removed, the printed layer will cool rapidly, shrinking faster than the cooling part below can accommodate, and the shrinkage of the top layer will be constrained by the surrounding material, resulting in tensile stress in the printed layer and compressive stress in the bottom material part. The predicted tensile stress of the deposited layer and the accumulated compressive stress of the substrate are consistent with the temperature gradient mechanism of residual stress formation. In addition, the residual stress at the beginning of the scanning track corner is larger, such as Figure 6 As shown in (b), (c) and (d), the elliptical wireframe marks are shown.

[0160] Extract the (a) stress and temperature curves and (b) stress and temperature gradient curves of the center point P2 in the first 3e-3s of the printing process. Figure 6 As shown. Figure 7 and Figure 6It can be seen that the stress at point P2 begins to accumulate even before the laser coincides with it. When the laser scans to point P2, its temperature is higher than the melting point of the material. Due to the phase change and temperature dependence of the material properties, the von Mose stress reaches a minimum at this time. Because when it is in a molten state, the stiffness and yield stress are very low. As the laser moves away from point P2 and begins to cool and solidify, the high-temperature area behind the molten pool shows obvious stretching, and its stress rises to a higher value. During the laser scanning process, point P2 undergoes multiple heating and cooling cycles, resulting in several cycles of stress relaxation and stress accumulation. As the molten track continues to move away from point P2, these reheating cycles become smaller and smaller, and the cyclic behavior of stress continues to decrease. The complex combination of stress relaxation and stress accumulation phenomena leads to the final stress distribution inside the components manufactured by L-PBF. In addition, through Figure 6 (b) It can be concluded that stress increases with the increase of temperature gradient, and the stress peak appears along with the temperature gradient peak.

[0161] like Figure 8 As shown in Figure 1, the residual stress distribution on the top surface is obtained when the laser power is 200W, the scanning speed is 1000mm / s, and the scanning spacing is 140μm. Figure 8 As shown in (a), the maximum stress is located at the interface between the molten powder layer and the substrate. The possible reason for the maximum stress value around the interface area of ​​the powder bed and the substrate is the constrained nature of this location. As the temperature tends to decrease, the material tends to shrink, but since the material is surrounded by multiple sides, its shrinkage is restricted, resulting in a higher stress value at this location. In addition, compared with the manufactured parts, the size and conductivity of the substrate are significantly larger, which will lead to a larger temperature gradient and cooling rate at the interface between the molten powder layer and the substrate, and will also lead to higher residual stress at this location. Since the four corners of the molten powder layer can deform freely without constraints, the stress at the four corners of the molten powder layer is lower. Due to the higher temperature gradient and the constraints of the adjacent solidified material, there is a larger stress in the middle of each edge. The von Mises stress results before and after removing the substrate constraints are compared as shown in the figure. Figure 8 (b) and Figure 8 As shown in (c), after the fixed boundary is removed, part of the stress is released. After the base plate is removed, the constraint nature of the connection between the base plate and the part changes.

[0162] Extract the residual stress values ​​at the three curve positions before and after the substrate is removed, such as Fig. 9 As shown in Figure 2, since the scanning direction is parallel to the x-axis, Fig. 9 (d) and Fig. 9The symmetric behavior of stress in the X-axis direction can be easily observed in (e). The residual stress is maximum near the edge of the print, and the residual stress is uniform near the center area. The unconstrained nature of the free edge leads to the lowest residual stress values ​​at these points. However, due to the combined effect of the constrained nature and the higher cooling rate, the highest stress values ​​are observed along the X-axis in the area close to the free edge. There are three peaks and bottoms in the stress distribution along the Y-axis perpendicular to the scanning direction. The adjacent areas of the tracks where these three peaks are located are not connected because the energy density is not high enough. Therefore, the temperature gradient in the adjacent areas of the tracks is small, resulting in lower residual stress. Larger stress values ​​are observed near the edge where the scan stops. The residual stress first decreases slightly and then increases with the deposition height (i.e., along the build direction). The deformation after the substrate is removed releases part of the previously generated residual stress, which is the reason for the significant reduction in residual stress after the substrate is removed.

[0163] The values ​​of equivalent residual stresses in the finite element model (FEM) were extracted at the same locations as the experimental measurements. The numerical measurements extracted from the FEM model were compared with the experimental investigation values, as shown in Fig.10 As shown. Among them, Fig.10 (a) The process parameters are laser power 200 W, scanning spacing 140 μm and scanning speed 1000 mm / s. Fig.10 (b) The process parameters are laser power 300W, scanning pitch 120μm, and scanning speed 1200mm / s. It is observed that all simulation values ​​are comparable to the measured values, with the difference within 16%. The reasons for the difference between the simulation results and the experimental results may be that the influence of multi-layer scanning is not considered, the model size is simplified, and the measurement error. In addition, Fig.10 The error ratio of the three points in (b) Fig.10 (a) is larger. This may be due to the simplified size of the model, which reduces the influence of heat accumulation and molten pool overlap on residual stress. The residual stress value at point P3 is higher than that at point P2, which indicates that the stress value at the starting direction of the adjacent track corner is higher than the stress value at the ending direction.

[0164] This study will conduct numerical tests on the joint simulation operation to evaluate its effectiveness. The joint operation is terminated when the number of iterations of multi-process parameter optimization reaches 40. The iteration results are as follows: Fig.11 The number of iterations stabilized after 20 times, and the optimal residual stress value was 197MPa. The optimal process parameters were laser power 283W, scanning speed 1209mm / s, and scanning spacing 123um. The maximum residual stress on the top surface of sample 1, sample 2, and sample 3 (optimal process parameters) was shown as follows: Fig.11(b) is shown. The maximum residual stresses of Sample 1, Sample 2 and Sample 3 are 316MPa, 238MPa and 197MPa respectively. The maximum residual stress under the optimal parameters is 37.6% lower than that of Sample 1 and 17.2% lower than that of Sample 2, and the maximum residual stress is significantly reduced. Therefore, the joint optimization method combining numerical simulation with adaptive optimization design can provide an important technical reference for the collaborative optimization design of L-PBF multi-parameters.

[0165] This embodiment also provides an optimization system for additive manufacturing residual stress, including a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the same optimization method for additive manufacturing residual stress.

[0166] The embodiment further provides a computer-readable storage medium having a computer program / instruction stored therein, wherein the computer program / instruction is programmed or configured to execute the method for optimizing residual stress in additive manufacturing through a processor.

[0167] This embodiment also provides a computer program product, including a computer program / instruction, which is programmed or configured to execute a method for optimizing additive manufacturing residual stress through a processor.

[0168] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the functions in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0169] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for optimizing residual stress in additive manufacturing, characterized in that: It includes the use of multi-physics field numerical simulation combined with intelligent optimization algorithm to find the global optimal solution that minimizes the residual stress of additively manufactured parts from the process parameter space. The multi-physics field numerical simulation includes using the optimization algorithm to sample process parameters from the process parameter space, inputting the sampled process parameters into a finite element model, performing thermal analysis on the process parameters based on the finite element model, obtaining transient thermal distribution through thermal analysis, performing structural analysis using transient thermal distribution to obtain residual stress under current process parameters, and passing the obtained residual stress to the intelligent optimization algorithm for realizing process parameter optimization.

2. The method for optimizing residual stress in additive manufacturing according to claim 1, characterized in that: When performing thermal analysis based on the finite element model to obtain transient thermal distribution, the functional expression of the temperature field used is as follows: In the above formula, ρ is the density, C p is the specific heat capacity, T is the temperature of the powder material, t is the interaction time between the heat source and the powder bed, k is the thermal conductivity, (x, y, z) is the coordinate system coordinate to which the heat source is attached, and Q is the heat generated per volume in the component; the functional expression of the boundary conditions used is shown in the following formula: In the above formula, k represents the thermal conductivity, T represents the temperature of the metal powder at a certain moment, n represents the infinite tendency to infinity in the integral, and h c represents the thermal convection coefficient, T0 represents the initial temperature, ε θ represents the thermal radiation coefficient, σ represents the Stefan-Boltzmann constant, q represents the heat flux, (x, y, z) represents the coordinate system coordinates to which the heat source is attached, and S represents the surface subject to heat flux, convection, and radiation.

3. The method for optimizing residual stress in additive manufacturing according to claim 1, characterized in that: When the transient heat distribution is obtained by thermal analysis based on the finite element model, the heat source model used to simulate the laser energy is the Gaussian surface heat source of the following formula: In the above formula, q r represents the energy of the laser heat source, A represents the absorptivity of the powder material used in additive manufacturing, P represents the laser power, exp(·) represents the exponential power of e, ω is the equivalent radius of the laser spot, and r represents the radial distance from a certain point (x, y) to the center of the laser spot.

4. The method for optimizing residual stress in additive manufacturing according to claim 1, characterized in that: The method of using transient thermal distribution to perform structural analysis to obtain residual stress under current process parameters includes: S101, constructing an additive manufacturing part model by partitioning and layering, and completing meshing of the additive manufacturing part model; S102, defining the grid type in the thermal analysis process, inputting the thermal physical property parameters of the material in different states, loading the moving heat source model, and defining the state transition function of the material; S103, after defining the initial conditions and boundary conditions, performing transient temperature field calculation, wherein the transient temperature field calculation includes simulating the temperature change of the additive manufacturing part model during the manufacturing process until the additive manufacturing part model cools to room temperature, and obtaining temperature field data under the full time state; S104, defining a mesh type, mechanical parameters related to the material, and a plastic hardening function of the material in an elastic-plastic mechanical force analysis process, loading boundary conditions and activation conditions, for evaluating the mechanical behavior and residual stress distribution of the additively manufactured part during the additive manufacturing process; S105, according to the temperature field data, applying the unit temperature of the whole time process as a load to the mechanical model to perform coupled solution and steady-state stress calculation to obtain the residual stress under the current process parameters.

5. The method for optimizing residual stress in additive manufacturing according to claim 1, characterized in that: When the multi-physics field numerical simulation is combined with the intelligent optimization algorithm to find the global optimal solution that minimizes the residual stress of the additively manufactured part from the process parameter space, the intelligent optimization algorithm used is a particle swarm-genetic hybrid optimization algorithm. The process parameters of additive manufacturing that minimize the residual stress from the process parameter space are found to include: S201, extracting sample data from the process parameter space through the optimal Latin hypercube to initialize the population; S202, transmitting the current population data to the multi-physics model to perform thermal-structural analysis to obtain residual stress; S203, transmitting the residual stress data to an intelligent optimization algorithm as its fitness value based on the fitness value to obtain the optimal value of the population and the individual; S204, sorting the current population in descending order based on the fitness value, and then performing crossover and mutation operations, updating the adaptive inertia weight and dynamic learning factor, and updating the speed and position of the population particles; S205, transferring the new population data to the multi-physics model to perform thermal-structural analysis to obtain residual stress; S206, transmitting the residual stress data to the optimization algorithm as its fitness value, and obtaining the optimal values ​​of the population and the individual based on the fitness value; S207, determine whether the number of iterations meets the termination condition. If not, add 1 to the number of iterations and jump to step S204 to continue iterating; otherwise, output the optimal values ​​of the population and the individual.

6. The method for optimizing residual stress in additive manufacturing according to claim 5, characterized in that: The function expression for updating the adaptive inertia weight and the dynamic learning factor in step S204 is: In the above formula, ω1 represents the adaptive inertia weight, ω max represents the maximum inertia weight, ω min represents the minimum inertia weight, c1 and c2 are dynamic learning factors, c1 represents the learning factor that the particle learns from itself, c2 represents the learning factor that the particle learns from the global optimal particle, f represents the fitness value of the particle, and f avg represents the average fitness value of all particles at present, f min Indicates the minimum fitness value of all particles at present, t i is the current iteration number, t max is the maximum number of iterations of the particle swarm.

7. The method for optimizing residual stress in additive manufacturing according to claim 6, characterized in that: The function expression for updating the speed and position of the population particles in step S204 is: In the above formula, represents the velocity of the k+1 generation particle, represents the velocity of the k-generation particle, represents the position of the k+1 generation particle, represents the position of the k-generation particle, represents the optimal value of individuals in generation k, represents the optimal value of the population of generation k, k represents the number of iterations, r1 and r2 represent random positive numbers drawn from a uniform distribution [0, 1], and α represents the constraint factor, which is used to control the weight of the speed.

8. The method for optimizing residual stress in additive manufacturing according to claim 1, characterized in that: When the multi-physics field numerical simulation is combined with the intelligent optimization algorithm to find the global optimal solution that minimizes the residual stress of the additively manufactured part from the process parameter space, the multi-physics field numerical simulation is implemented based on the COMSOL simulation software, and the optimization algorithm is implemented based on Matlab. The sending of the obtained global residual stress to the optimization algorithm includes: the COMSOL simulation software establishes a data transmission interface with Matlab through the COMSOL Multiphysics Server, and the COMSOL simulation software saves the obtained global residual stress model data from the .mph file as a .m file, and then establishes a connection with the COMSOL simulation software in MATLAB through the mphstart command and imports the model data in the .m file format through the mphopen command, and then uses the mphglobal command to automatically import the model data into MATLAB, thereby realizing seamless data transmission between the COMSOL simulation software and Matlab.

9. An optimization system for additive manufacturing residual stress, comprising a microprocessor and a memory connected to each other, characterized in that: The microprocessor is programmed or configured to execute the additive manufacturing residual stress optimization method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program / instruction stored therein, characterized in that: The computer program / instructions are programmed or configured to execute the method for optimizing additive manufacturing residual stress as claimed in any one of claims 1 to 8 through a processor.

Images