Additive manufacturing temperature field rapid simulation method based on mixing precision

Abstract

The invention discloses an additive manufacturing temperature field rapid simulation method based on mixing precision, which fully considers the problem of low calculation efficiency caused by a large number of time steps in a metal additive manufacturing process, divides the time steps into a plurality of sub-intervals, and solves each interval in different kernels in parallel, so that the calculation efficiency is improved. And the initial temperature value of each interval is corrected based on the result obtained on each core, the error of the temperature solution is iteratively reduced, the process is standard and convenient to realize in a programmed manner, and if computing resources are sufficient, the parallel solution efficiency of the method is higher than that of serial solution.

Description

Technical Field

[0001] The present invention belongs to the technical field of additive manufacturing, and in particular relates to a rapid simulation method for an additive manufacturing temperature field based on mixed precision. Background Art

[0002] Additive Manufacturing (AM) is a manufacturing technique that creates three-dimensional solid parts by incrementally adding material. Metal AM is a high-value-added metal part manufacturing technology, with the most representative processes including Direct Energy Deposition (DED) and Selective Laser Melting (SLM). In recent years, metal AM has experienced rapid development and has been successfully applied in fields such as aviation, aerospace, mold manufacturing, and medical treatment.

[0003] The technical advantage of additive manufacturing technology is rapid, mold-free, free-form forming, enabling the rapid creation of complex, irregular spatial structures starting from a single metal substrate. However, the additive manufacturing process involves intense powder melting and cooling, and the large temperature gradients lead to very large transient thermal stresses. This can easily cause significant deformation and even cracking during the additive manufacturing process, leading to manufacturing failure. Additive manufacturing often involves single-piece, small-batch production, where the first piece is the final product. Therefore, the first-time yield of additive manufacturing has become a critical factor in commercial success. Traditional additive manufacturing design or optimization, driven by the experience of process engineers, can no longer meet the demands of fierce market competition.

[0004] Therefore, data-driven modeling and simulation, or digital twin technology, has emerged in additive manufacturing allowance design, placement design and optimization, support design, residual stress prediction, deformation control, and other aspects.

[0005] Modeling and simulating macroscopic thermal stress fields for additive manufacturing still faces numerous challenges, the most significant of which is slow simulation speed, often requiring weeks or even months to simulate a single set of results. This is no longer sufficient for the rapid delivery of additive manufacturing technology, and rapid modeling and computation has become a crucial research direction. Professor Stuck's team in the United States, focusing on symmetric boundary conditions for additive manufacturing and triangular decomposition of sparse matrices, developed a fast solution algorithm for additive manufacturing and founded 3DSim, which was recently acquired by Ansys and became the Ansys Additive Manufacturing module. A team at Pennsylvania State University developed Cubic, a software that utilizes adaptive meshing to improve computational efficiency by reducing the number of meshes. This software has been integrated into NetFabb, a simulation module within the software. MSC, using the inherent strain method as its mathematical foundation, uses temperature-induced inherent strain as the data source for mechanical analysis. Layer-by-layer inherent strain is applied to enable rapid simulation of selected laser melting processes. While computationally efficient, this method requires experimentally measured inherent strain data, resulting in relatively poor fidelity in solving complex coupled thermal problems. Dassault Systèmes' Abaqus simulation software has launched a control word for additive manufacturing, allowing users to simulate the additive manufacturing process by adjusting material properties. Compared with the traditional Model Change method, the calculation time is 17% of the traditional method.

[0006] Generally speaking, current simulation techniques can be divided into three categories: the full thermoelastic method, the inherent strain method, and the inherent shrinkage method. The full thermoelastic method offers the highest computational accuracy and the most mature algorithmic foundation, but also carries the greatest computational time and development complexity. The inherent strain method offers the fastest computational time, but requires the measurement of inherent strain data and exhibits poor simulation fidelity for repeated thermal cycles. The inherent shrinkage method falls somewhere in between, aiming to leverage the accuracy of the full thermoelastic method while also aiming to rapidly apply heat source conditions layer by layer. However, this method has not achieved the desired results, and with limited research results in academia, it has been largely dismissed.

[0007] The metal additive manufacturing process adds materials point by point, line by line, surface by body, which inevitably leads to the "life and death" control of units in the modeling and simulation process. Traditional methods generally use two methods: 1. Simulate additive manufacturing by continuously adding units, that is, changing the geometric configuration of the model; 2. The geometric configuration (units and nodes) is completely constructed at one time, and the additive manufacturing process is simulated by adjusting the material properties of the units. The second method is faster than the first method, but there is a problem that, for full-size parts, if the entire model is completely discretized in time and space at one time, the number of discrete time and space grids is still large (even up to 1*10 9Since the temperature field calculation of existing technology and software is a serial solution without relevant parallel technology, the large number of time steps for temperature field solution in the modeling and simulation of real parts will lead to low efficiency of finite element solution. Summary of the Invention

[0008] In order to overcome the shortcomings of the existing technology, the present invention provides a rapid simulation method for the temperature field of additive manufacturing based on mixed precision. It fully considers the problem of low computational efficiency caused by the large number of time steps in the metal additive manufacturing process. The time step is divided into several sub-intervals, and each interval is solved in parallel in different cores. The initial temperature value in each interval is corrected based on the result obtained on each core, and the error of the temperature solution is iteratively reduced. The process is standardized and easy to implement in a program. If the computing resources are sufficient, the parallel solution efficiency of this method is higher than that of the serial solution.

[0009] The technical solutions adopted by the present invention to solve the technical problems are as follows:

[0010] Step 1: Create the geometry, divide the mesh, and set the heat source function, material parameters, and boundary conditions.

[0011] Step 2: Discretize the total time domain [0, T] into several large steps, denoted as ΔT, with a total of N;

[0012] Step 3: Within each large step, the time ΔT is discretized into several small steps Δt, satisfying ΔT = JΔt, where Δt meets the accuracy requirements of the solution;

[0013] Step 4: Set the number of cores to be N, and each core is simply referred to as process 1, 2, ...N.

[0014] Step 5: Solve N times with large step size single precision starting from time 0 to obtain the initial temperature guess of each large step size node θ c represents the solution obtained by large step size calculation;

[0015] Step 6: Guess the temperature from the previous step on each core As the initial value, with Δt as the step size, solve J times in parallel with double precision to obtain θ f represents the solution obtained by small step calculation;

[0016] Step 7: Each core will Send to process 1, and the serial correction part will be solved uniformly in process 1;

[0017] Step 8: Calculate the correction amount of the large step node solution in process 1 And calculate the correction amount of the large step size coarse grid node solution: in Indicates is the initial value, with ΔT as the step size, and the temperature solution obtained with single precision, where

[0018] Step 9: Correct the large step node temperature in process 1: n plus 1 and repeat steps 8 and 9 until n + 1 = N;

[0019] Step 10: Solve the corrected temperature from process 1 Broadcast to all processes;

[0020] Step 11: Repeat steps 6 to 10 until the temperature solution meets the iteration tolerance.

[0021] A computer program enables a computer to execute the above-mentioned method for rapid simulation of temperature field in additive manufacturing.

[0022] An electronic device comprises: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device performs the above-mentioned additive manufacturing temperature field rapid simulation method.

[0023] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned method for rapid simulation of temperature field in additive manufacturing.

[0024] A chip includes: a processor for calling and running a computer program from a memory, so that a device equipped with the chip executes the above-mentioned method for rapid simulation of the additive manufacturing temperature field.

[0025] A computer program product includes a computer storage medium storing a computer program, wherein the computer program includes instructions executable by at least one processor, and when the instructions are executed by the at least one processor, the method for rapid simulation of the additive manufacturing temperature field is implemented.

[0026] The beneficial effects of the present invention are as follows:

[0027] The present invention fully considers the characteristics of solving the temperature field of real additive manufacturing: that is, the temperature gradient in the cladding area is large, the simulation time step is small, and the number of time steps is large, so that the temperature field can be solved in multi-core parallel to improve the computational efficiency. The present invention adopts a method of single-precision serial guessing with a large step size, double-precision accurate solution with a small step size and iterative correction to simulate the evolution process of the temperature field of additive manufacturing. This method maintains the working accuracy unchanged during the entire simulation process, can solve the temperature field distribution of additive manufacturing in parallel with multiple cores, and adopts low-precision solution in the serial correction part to reduce the amount of calculation, and has high computational efficiency. In addition, this method has a good convergence speed, and the temperature solution meets the solution accuracy requirements after a few iterations. Through numerical experimental comparison, the computational efficiency of this method is nearly 40 times higher than that of the traditional serial process, and it has good practical engineering value. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is the grid model diagram of the present invention.

[0029] Figure 2 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0030] The present invention will be further described below with reference to the accompanying drawings and examples.

[0031] To address the problem of slow calculation speed in the modeling and simulation of the macroscopic thermal stress field in the additive manufacturing process, the present invention proposes an accelerated solution technology based on mixed precision, which can solve the finite element model of the additive manufacturing temperature field in parallel on multiple cores and continuously iterate to reduce the error, thereby improving the calculation efficiency.

[0032] This invention fully considers the problem of low computational efficiency caused by the large number of time steps in the metal additive manufacturing process. It divides the time step into several sub-intervals and solves each interval in parallel in different cores. Based on the results obtained in each core, the initial temperature value in each interval is corrected, and the error of the temperature solution is iteratively reduced. This process is standardized and easy to implement in a program. If there are sufficient computing resources, the parallel solution efficiency of this method is higher than that of the serial solution. The specific steps are as follows:

[0033] The first step is to establish the geometric structure, divide the mesh, set the heat source function, material parameters and boundary conditions, etc.

[0034] The second step is to discretize the total time domain [0, T] into several large steps, denoted as ΔT, with a total of N;

[0035] The third step is to discretize the time ΔT into several small steps Δt within each large step, where ΔT = JΔt, where Δt meets the accuracy requirements of the solution;

[0036] Step 4: Set the number of cores to N (equal to the number of large steps), and each core is simply referred to as process 1, 2...N.

[0037] Step 5: Start from time 0 and quickly solve N times with large step size single precision to obtain the initial temperature guess of each large step size node (θ c represents the solution obtained by large step size calculation);

[0038] Step 6: Guess the temperature of each core based on the previous step. As the initial value, with Δt as the step size, solve J times in parallel with double precision to obtain (θ f represents the solution obtained by small step calculation);

[0039] In the seventh step, each core will Send to process 1, and the serial correction part will be solved uniformly in process 1;

[0040] Step 8: Calculate the correction amount of the large step node solution in process 1 And calculate the correction amount of the large step size coarse grid node solution: in Indicates is the initial value, with ΔT as the step size, and the temperature solution obtained with single precision, where

[0041] Step 9: Correct the temperature of the large-step node in process 1: n=n+1 Repeat steps 8 and 9 until n+1=N;

[0042] Step 10: Solve the corrected temperature from process 1 Broadcast to all processes;

[0043] In step 11, repeat steps 6 to 10 until the temperature solution meets the iteration tolerance.

[0044] Example:

[0045] The present invention provides a method for parallelizing the temperature field in additive manufacturing. The process is standardized, easy to implement through programming, and has a series of advantages, such as fast convergence and high solution accuracy. The following is intended to explain rather than limit the present invention. Its implementation process includes the following steps:

[0046] Step 1: First, establish a two-dimensional finite element model and set the boundary conditions, including a square with a length of 2 and a width of 2, a cladding pass width of 0.02, a grid size of 0.25 (80 grids in the x and y directions), and the grid model is as follows: Figure 1The material parameters are IN625, whose thermal conductivity and specific heat capacity vary with temperature, the density is 8440, the initial temperature and the four boundary temperatures are 25 degrees Celsius, the heat source is a Gaussian heat source, and the scanning method is a circular scan.

[0047] Step 2: The scanning speed is 1, and the body heat flux load is applied to the units on the cladding path one by one to simulate the movement of the heat source. The total scanning time is 36.

[0048] Step 3: Discretize the total time domain [0, 36] into several large steps, denoted as ΔT, with a total of N (equal to the number of cores, set to 12, 24, 48, and 96 respectively). Within each large step, discretize the time ΔT into several small steps Δt = 1 / 36864, where ΔT = JΔt, where the number of cores J is different;

[0049] Step 4: Set the initial temperature to 25 and the boundary temperature to 25, and quickly solve N times (Picard iterations several times) with a large step size ΔT and single precision starting from time 0 to obtain the initial temperature guess for each large step size node. (θ c represents the solution obtained by large step size calculation);

[0050] Step 5: Guess the temperature from the previous step on each core As the initial value, with Δt as the step size, solve J times in parallel with double precision to obtain (θ f represents the solution obtained by small step calculation);

[0051] Step 6: Guess the temperature from the previous step on each core is the initial value, 25 is the boundary temperature, Δt is the step size, and the double precision (Picard iteration is solved once) is solved J times in parallel to obtain (θ f represents the solution obtained by small step calculation), each core will Send to process 1;

[0052] Step 7: Calculate the correction amount of the large step node solution in process 1 And calculate the correction amount of the large step size coarse grid node solution (Picard iteration several times): in Indicates is the initial value, 0 is the boundary condition, ΔT is the step size, heat source is not included, and the temperature solution is obtained with single precision.

[0053] Step 8: Correct the large step node temperature in process 1: n=n+1 Repeat steps 6 and 7 until n+1=N, and solve the corrected temperature from process 1. Broadcast to all processes;

[0054] Step 9: Repeat steps 6 to 8 until the temperature solution meets the iteration tolerance (the number of iterations is 4);

[0055] IN625 material parameters are shown in Table 1:

[0056] Table 1

[0057]

[0058] The comparison of convergence error (maximum absolute error) is shown in Table 2:

[0059] Table 2

[0060]

[0061] Computational efficiency comparison is shown in Table 3:

[0062] After completing the temperature field solution example, the data was processed and the speedup ratios of the unified precision and mixed precision parallel algorithms were compared (the speedup ratio is calculated by dividing the serial solution time by the parallel solution time):

[0063] Table 3

[0064]

[0065] For the above model, the parallel solution of the temperature field solution time is more efficient than the serial solution. The more cores there are, the greater the improvement. After applying mixed precision, the amount of calculation of the serial correction part is reduced, and the parallel solution efficiency is improved compared with the unified double precision.

[0066] The present invention fully considers the characteristics of solving the temperature field of real additive manufacturing: that is, the temperature gradient in the cladding zone is large, the time step is small, and the number is large, so that the temperature field can be solved in parallel by multiple cores to improve the calculation efficiency. The present invention adopts a method of single-precision serial guessing with a large step size, double-precision parallel solution with a small step size and iterative correction to simulate the evolution process of the temperature field of additive manufacturing. This method maintains the working accuracy unchanged during the entire simulation process, can solve the temperature field distribution of additive manufacturing in parallel by multiple cores, and uses low-precision solution in the serial correction part to reduce the amount of calculation, with high calculation efficiency. In addition, this method has a good convergence speed, and the temperature solution meets the solution accuracy requirements after a few iterations, which can realize the full automation of the process, thereby realizing fully automated simulation calculation without human intervention.

Claims

1. A rapid simulation method for temperature field in additive manufacturing based on mixed precision, characterized in that: The steps include: Step 1: Create the geometry, divide the mesh, and set the heat source function, material parameters, and boundary conditions. Step 2: Discretize the total time domain [0, T] into several large steps, denoted as ΔT, with a total of N; Step 3: Within each large step, the time ΔT is discretized into several small steps Δt, satisfying ΔT = JΔt, where Δt meets the accuracy requirements of the solution; Step 4: Set the number of cores to be N, and each core is simply referred to as process 1, 2, ...N. Step 5: Solve N times with large step size single precision starting from time 0 to obtain the initial temperature guess of each large step size node θ c represents the solution obtained by large step size calculation; Step 6: Guess the temperature from the previous step on each core As the initial value, with Δt as the step size, solve J times in parallel with double precision to obtain θ f represents the solution obtained by small step calculation; Step 7: Each core will Send to process 1, and the serial correction part will be solved uniformly in process 1; Step 8: Calculate the correction amount of the large step node solution in process 1 And calculate the correction amount of the large step size coarse grid node solution: in Indicates is the initial value, with ΔT as the step size, and the temperature solution obtained with single precision, where Step 9: Correct the large step node temperature in process 1: n plus 1 and repeat steps 8 and 9 until n + 1 = N; Step 10: Solve the corrected temperature from process 1 Broadcast to all processes; Step 11: Repeat steps 6 to 10 until the temperature solution meets the iteration tolerance.

2. A computer program, characterized in that The computer program enables a computer to execute the method according to claim 1.

3. An electronic device, characterized in that: include: processor and memory; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device performs the method as claimed in claim 1.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to claim 1 is implemented.

5. A chip, characterized in that: include: A processor, configured to call and run a computer program from a memory, so that a device equipped with the chip executes the method as claimed in claim 1.

6. A computer program product, characterized in that The computer program product comprises a computer storage medium storing a computer program, wherein the computer program comprises instructions executable by at least one processor, and when the instructions are executed by the at least one processor, the method according to claim 1 is implemented.

Images