SLM (selective laser melting) forming part process parameter optimization method based on evolutionary mixing and melting method

Abstract

The invention discloses an SLM forming part process parameter optimization method based on an evolutionary mixing and melting method, and belongs to the field of metal additive manufacturing. The method comprises the following steps: firstly, selecting key process parameters such as laser power, scanning speed, scanning spacing and powder laying thickness, performing numerical calculation on a deformation problem in an SLM forming process by taking deformation of an SLM forming part as a response index to obtain deformation data, and constructing a mathematical relationship model between the process parameters and the deformation through a response surface method, so as to obtain the deformation data of the SLM forming part; therefore, the deformation can be predicted; then, an evolutionary mixing and melting method is adopted for optimizing technological parameters, the evolutionary mixing and melting method is used for conducting evolutionary operation in predation and migration behaviors of the black wing plinostratums, candidate solutions are generated through two evolutionary strategies, evolution vectors are generated through weighted mixing, and mixed probability factors are introduced to generate test individuals; and whether a new solution is accepted or not is determined by comparing the fitness of the test individual and the fitness of the original solution, so that the technological parameters are optimized, and the deformation of the SLM forming part is reduced. The method can accurately predict and optimize the process parameters, reduce the deformation of the formed part, simplify the parameter adjustment process, reduce the production cost and improve the production efficiency.

Description

Technical Field

[0001] The present invention belongs to the field of metal additive manufacturing, and in particular relates to a method for optimizing process parameters of SLM molded parts based on an evolutionary blending method. Background Art

[0002] Selective Laser Melting (SLM), as a key technology in the field of metal additive manufacturing, has shown great potential in many high-end manufacturing industries such as aerospace and medical devices with its unique processing method of melting metal powder layer by layer and solidifying it. However, in the SLM molding process, the reasonable selection of process parameters plays a vital role. Many process parameters such as laser power, scanning speed, layer thickness and ambient atmosphere directly affect the quality of the molded parts. If the process parameters are not selected properly, it may lead to unreasonable accumulation and release of thermal stress, which in turn causes deformation of parts. This deformation not only affects the geometric dimensional accuracy of the molded parts, but also may cause uneven microstructure inside the parts, thereby greatly affecting their mechanical properties.

[0003] At present, although methods such as preheating the substrate, optimizing the scanning strategy, introducing support structures and adjusting process parameters have been widely adopted to reduce the deformation of SLM molded parts, these traditional methods are often difficult to fully play their role in actual operation due to equipment limitations, cost control and operational complexity considerations, resulting in the deformation problem has not been fundamentally solved. What is particularly critical is that the optimization process of process parameters has limitations and fails to accurately balance the influence of the interaction between various parameters on the quality of molded parts, thereby limiting the improvement of the dimensional accuracy of the molded parts and the optimization of the overall performance. Therefore, it is urgent to develop new control strategies and technical means to achieve effective optimization of process parameters, so as to further reduce the deformation of SLM molded parts and improve their dimensional accuracy and overall performance. This has also become one of the key research directions in the field of metal additive manufacturing. To this end, a new method for optimizing process parameters for SLM molded parts is proposed to improve the dimensional accuracy and overall performance of SLM molded parts. Summary of the invention

[0004] In view of this, the object of the present invention is to provide a method for optimizing process parameters of SLM molded parts based on the evolutionary blending method, so as to improve the quality of molded parts by controlling and optimizing process parameters.

[0005] The technical solution to achieve the purpose of the present invention is: a method for optimizing process parameters of SLM molded parts based on evolutionary blending method, comprising the following steps:

[0006] 1. A method for optimizing process parameters of SLM molded parts based on evolutionary blending method, characterized by comprising the following steps:

[0007] Step 1: Select key process parameters such as laser power, scanning speed, scanning spacing, and powder thickness, take the deformation of SLM molded parts as the response index, use Simufact Additive software to perform numerical calculations on the deformation problem of 316L stainless steel powder in the SLM molding process, and obtain the deformation data of SLM molded parts under different process parameters; use the collected experimental data to construct a mathematical relationship model between process parameters and deformation of molded parts through the response surface method to predict the deformation;

[0008] Step 2, the evolutionary hybrid method is used to optimize the process parameters in the model; the evolutionary hybrid method is to perform evolutionary operations in the predation and migration behaviors of black kites, use two different types of evolutionary strategies to generate candidate solutions, and apply weighted mixing to generate the final evolution vector; then, the hybrid probability factor is introduced to generate experimental individuals by combining the evolutionary vector and some features of the original solution; and on this basis, the fitness of the experimental individuals and the original solution is compared to decide whether to accept the new solution, increase the global search capability, and thus optimize the process parameters in the model, and find the optimal combination of process parameters to reduce the deformation of SLM molded parts.

[0009] 2. According to claim 1, a method for optimizing process parameters of SLM molded parts based on evolutionary blending method is characterized in that the evolutionary blending method described in step 2 is to perform evolutionary operations in the predation and migration behaviors of black kites, generate candidate solutions using two different types of evolutionary strategies, and apply weighted blending to generate the final evolution vector; then, introduce a hybrid probability factor, and generate a test individual by combining the evolutionary vector and some features of the original solution; and on this basis, compare the fitness of the test individual and the original solution to decide whether to accept the new solution, specifically:

[0010] The evolution operation formula is:

[0011]

[0012] Among them, v i,j is the value of the jth characteristic of the i-th individual in the new solution of evolution strategy 1, x i,j is the value of the jth characteristic of the ith individual in the original solution, γ is the scaling factor of evolution strategy 1, exp(fi-τ) is the natural exponential function, and f i is the fitness of the ith individual, τ is the threshold parameter, C(0,1) is a random number drawn from the Cauchy distribution with location and scale parameters of 0 and 1 respectively, and u i,j is the value of the jth characteristic of the i-th individual in the new solution of evolution strategy 2, x r1,j 、x r2,j 、x r3,j Represents the value of the jth feature of three randomly selected individuals, F dynamicis the scaling factor of evolution strategy 2, F0 is the initial scaling factor, β is the parameter that controls the rate of change of the scaling factor, and f min and f max are the minimum and maximum fitness in the current population, respectively, and α is the mixing coefficient;

[0013] In the evolutionary hybrid method, the evolution operation first takes the eigenvalue x of the original solution as i,j As a basis, add an adjustment term to get the new solution v i,j The adjustment term combines the scaling factor γ, the random number C(0,1) drawn from the Cauchy distribution, and the natural exponential function exp(fi-τ). The setting of each coefficient strategically controls the amplitude and direction of the adjustment. The purpose is to jump out of the local solution, increase the diversity of the search and the possibility of global optimization. Secondly, based on the eigenvalues ​​x of three different individuals randomly selected r1,j 、x r2,j 、x r3,j Calculate and then multiply by the scaling factor F dynamic , and get a new solution u i,j ; F dynamic According to the minimum fitness f in the current population min , maximum fitness f max , parameter β and initial scaling factor F0 are dynamically adjusted; the two new solutions v i,j 、u i,j As candidate solutions, the mixing coefficient α is introduced to combine these candidate solutions through weighted mixing to generate the final evolution vector; Since the evolutionary blending method improves the global search ability of the algorithm and the quality of the solution through evolutionary operations, the following will specifically explain how it guides the solution update process through the mixing probability factor R to further enhance the algorithm's exploration and utilization capabilities;

[0014] The formula for introducing the mixed probability factor R to generate and compare the test individuals is:

[0015]

[0016] Among them, y i,j is the value of the jth characteristic of the ith individual in the test, w i,j is the value of the jth feature of the i-th individual in the evolution vector, j r is a randomly selected cross-reference dimension, rand j is the random number associated with dimension j, R is the mixing probability factor, x i is the i-th individual of the original solution, y i is the i-th individual of the test individual, f(y i ) is the experimental individual y i The fitness of f(x i) is the original solution x i Adaptability;

[0017] Based on the above formula, firstly, according to the given mixing probability factor R, the evolution vector w i,j With the current individual x i Mix to generate the experimental individuals y i,j Specifically, for each dimension j, if the randomly generated number is less than R or j is equal to a randomly selected dimension index, then the value y of the test individual in that dimension is i,j is equal to the value w of the evolution vector in this dimension i,j Otherwise, y i,j Equal to the value x of the current individual in this dimension i,j ; Then, by comparing the experimental individuals y i and the current individual x i The fitness of the test individual y determines whether to update the current individual; if the test individual y i The fitness of y is better, then the current individual is updated to y i , hoping to gradually find better solutions during the evolution process.

[0018] Beneficial effects of the present invention:

[0019] 1) By applying the present invention, the process parameters in the SLM forming process can be accurately predicted and optimized, and the deformation of the molded parts can be significantly reduced.

[0020] 2) Traditional parameter adjustment often relies on experience and a lot of trial and error, while the optimization method provided by the present invention can quickly and efficiently find the optimal process parameters, simplifying the parameter adjustment process.

[0021] 3) The optimized process parameters help reduce the workload of correcting molded parts in subsequent processing steps, reduce production costs and improve production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of the SLM molding process parameter optimization method based on the evolutionary blending method;

[0023] Figure 2 To optimize the minimum deformation in the previous experimental specimen;

[0024] Figure 3 The minimum deformation for optimizing the process parameters of SLM molding parts based on the evolutionary hybrid method. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The following examples are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the scope of protection of the present invention.

[0026] Combination Figure 1 The present invention provides a method for optimizing process parameters of SLM molded parts based on an evolutionary blending method, which is characterized by comprising the following steps:

[0027] Step 1: Select key process parameters such as laser power, scanning speed, scanning spacing, and powder thickness, take the deformation of SLM molded parts as the response index, use Simufact Additive software to perform numerical calculations on the deformation problem of 316L stainless steel powder in the SLM molding process, and obtain the deformation data of SLM molded parts under different process parameters; use the collected experimental data to construct a mathematical relationship model between process parameters and deformation of molded parts through the response surface method to predict the deformation;

[0028] Step 2, the evolutionary hybrid method is used to optimize the process parameters in the model; the evolutionary hybrid method is to perform evolutionary operations in the predation and migration behaviors of black kites, use two different types of evolutionary strategies to generate candidate solutions, and apply weighted mixing to generate the final evolution vector; then, the hybrid probability factor is introduced to generate experimental individuals by combining the evolutionary vector and some features of the original solution; and on this basis, the fitness of the experimental individuals and the original solution is compared to decide whether to accept the new solution, which increases the global search capability, thereby optimizing the process parameters in the model and finding the optimal combination of process parameters to reduce the deformation of SLM molded parts.

[0029] Example

[0030] In the specific implementation process, the specific implementation steps of the present invention are as follows:

[0031] The present invention sets up an experiment with four variables as factors, namely, laser power (A), scanning speed (B), scanning interval (C) and powder thickness (D). Since the numerical differences between the elements are too large, the experimental data are normalized and all parameters are mapped to the interval [0,2]. Then, the actual scanning interval is C / 10 and the actual powder thickness is D / 10. The actual unit of laser power is kW, the actual unit of scanning speed is m / s, the actual unit of scanning interval is mm, the actual unit of powder thickness is mm, and the unit of deformation is mm.

[0032] The BBD (Box-Behnken Design) design method in the response surface experiment is a three-level design method composed of factor design and incomplete pool design. It requires the minimum, maximum and central values ​​of the factors, so it is necessary to specify the value range of the variables, as shown in Table 1.

[0033] Table 1 Variable value range

[0034]

[0035] Taking deformation (Δ) as the response index and four variables including laser power (A), scanning speed (B), scanning interval (C) and powder thickness (D) as factors, a simulation experiment with 4 factors, 3 levels and 1 response was designed by using the BBD design method in the response surface experiment. Then, the deformation problem in the SLM forming process of 316L stainless steel powder was numerically calculated using SimufactAdditive software, and the deformation data of SLM molded parts under different process parameters were obtained.

[0036] When the factor arrangement is close to the optimal point, the model is used to approximate the response, so a binary model is used to express the relationship between the response and various factors. According to the experimental data, a multivariate regression fitting is performed on the response, and the fitting result of the deformation is as follows:

[0037] Δ=0.0970+0.0043A-0.0020B+0.0008C+0.0092D

[0038] +0.0013AB+0.0025AC+0.0045AD-0.0013BC+0.0010BD

[0039] +0.0022CD+0.0016A 2 +0.0012B 2 -0.0038C 2 -0.0004D 2

[0040] After obtaining the mathematical model between process parameters and deformation, the evolutionary hybrid method is used to optimize the process parameters in the model.

[0041] The evolutionary hybrid method is divided into the following steps:

[0042] (1) Population initialization

[0043] Evenly distribute the positions of each black-winged kite:

[0044] X i =BK lb +rand(BK ub -BK lb )

[0045] Where i is an integer between 1 and N, BK lb and BK ub are the lower and upper bounds of the j-th dimension black-winged kite, and rand is a random number between [0,1].

[0046] Since the main purpose of this invention is to find the minimum value of deformation, the deformation of the molded part is used as the fitness value, and the constructed fitness function is shown in the following formula:

[0047] Y1=|Δ|

[0048] (2) Aggressive behavior

[0049] This behavior is used to perform a detailed search of the area near the optimal solution of the current process parameters to find a better solution.

[0050]

[0051] Among them, y i,j (t) and y i,j (t+1) represents the position of the i-th black-winged kite in the j-th dimension and the (t+1)-th iteration step respectively; r is a random number between 0 and 1, p is a constant of 0.9; T is the total number of iterations, and t is the number of iterations completed so far.

[0052] (3) Migration behavior

[0053] Through dynamic leader selection and Cauchy mutation operator, the black kite population can achieve global search in the search space. During migration, the black kite adjusts the migration direction and step size according to the fitness difference between the current position and the optimal solution (leader) and the random position, thereby jumping out of the local optimum and enhancing the global optimization ability.

[0054]

[0055] m=2×sin(r+π / 2)

[0056] Among them, L j (t) represents the leader of the j-th black-winged kite in the t-th iteration so far; y i,j (t) and y i,j (t+1) represents the position of the i-th black-winged kite in the j-th dimension and the (t+1) iteration step respectively; F i represents the j-th dimension current position obtained by any black-winged kite in the t-th iteration; F ri represents the fitness value of any black kite at a random position in the jth dimension in the tth iteration; C(0,1) represents the Cauchy mutation.

[0057] (4) Evolutionary Operation

[0058] The evolution operation formula is:

[0059]

[0060] Among them, v i,j is the value of the jth characteristic of the i-th individual in the new solution of evolution strategy 1, x i,j is the value of the jth characteristic of the ith individual in the original solution, γ is the scaling factor of evolution strategy 1, exp(fi-τ) is the natural exponential function, and f i is the fitness of the ith individual, τ is the threshold parameter, C(0,1) is a random number drawn from the Cauchy distribution with location and scale parameters of 0 and 1 respectively, and u i,j is the value of the jth characteristic of the i-th individual in the new solution of evolution strategy 2, x r1,j 、x r2,j 、x r3,j Represents the value of the jth feature of three randomly selected individuals, F dynamic is the scaling factor of evolution strategy 2, F0 is the initial scaling factor, β is the parameter that controls the rate of change of the scaling factor, and f min and f max are the minimum and maximum fitness in the current population, respectively, and α is the mixing coefficient;

[0061] In the evolutionary hybrid method, the evolution operation first takes the eigenvalue x of the original solution as i,j As a basis, add an adjustment term to get the new solution v i,j The adjustment term combines the scaling factor γ, the random number C(0,1) drawn from the Cauchy distribution, and the natural exponential function exp(fi-τ). The setting of each coefficient strategically controls the amplitude and direction of the adjustment. The purpose is to jump out of the local solution, increase the diversity of the search and the possibility of global optimization. Secondly, based on the eigenvalues ​​x of three different individuals randomly selected r1,j 、x r2,j 、x r3,j Calculate and then multiply by the scaling factor F dynamic , and get a new solution u i,j ; F dynamic According to the minimum fitness f in the current population min , maximum fitness f max , parameter β and initial scaling factor F0 are dynamically adjusted; the two new solutions v i,j 、u i,jAs candidate solutions, the mixing coefficient α is introduced to combine these candidate solutions through weighted mixing to generate the final evolution vector; Since the evolutionary blending method improves the global search ability of the algorithm and the quality of the solution through evolutionary operations, the following will specifically explain how it guides the solution update process through the mixing probability factor R to further enhance the algorithm's exploration and utilization capabilities;

[0062] The formula for introducing the mixed probability factor R to generate and compare the test individuals is:

[0063]

[0064]

[0065] Among them, y i,j is the value of the jth characteristic of the ith individual in the test, w i,j is the value of the jth feature of the i-th individual in the evolution vector, j r is a randomly selected cross-reference dimension, rand j is the random number associated with dimension j, R is the mixing probability factor, x i is the i-th individual of the original solution, y i is the i-th individual of the test individual, f(y i ) is the experimental individual y i The fitness of f(x i ) is the original solution x i Adaptability;

[0066] Based on the above formula, firstly, according to the given mixing probability factor R, the evolution vector w i,j With the current individual x i Mix to generate the experimental individuals y i,j Specifically, for each dimension j, if the randomly generated number is less than R or j is equal to a randomly selected dimension index, then the value y of the test individual in that dimension is i,j is equal to the value w of the evolution vector in this dimension i,j Otherwise, y i,j Equal to the value x of the current individual in this dimension i,j ; Then, by comparing the experimental individuals y i and the current individual x i The fitness of the test individual y determines whether to update the current individual; if the test individual y i The fitness of y is better, then the current individual is updated to y i , hoping to gradually find better solutions during the evolution process.

[0067] After iteration, check whether the preset number of iterations T is reached or other stopping conditions are met. If reached, the iteration ends; otherwise, return to the attack behavior for the next round of iteration until the conditions are met. In this invention, N = 30, T = 100, dim = 4, F0 = 0.9, γ = 0.01, R = 0.9, α = 0.1, τ = 0.1, β = 0.1, the defined variables are used as input, the deformation is used as output, and the optimal solution is obtained through the evolutionary hybrid method.

[0068] The deformation obtained after optimization is compared with the minimum deformation under all parameter combinations in the previous experiment to verify the optimization effect of the proposed evolutionary blending method. The parameter combination with the minimum deformation in the data before optimization is laser power of 150W, scanning speed of 1100mm / s, scanning interval of 0.1mm, powder thickness of 0.02mm, and deformation of 0.058mm. Figure 2 The parameter combination obtained by the process parameter optimization method based on the evolutionary blending method is a laser power of 106.54W, a scanning speed of 1525.6mm / s, a scanning interval of 0.08mm, and a powder thickness of 0.02mm. The optimal parameter combination is brought into the Simufact Additive simulation software for verification, and the final deformation is 0.055mm, as shown in Figure 3 Compared with the minimum deformation of 0.058mm in the experimental sample, the deformation is reduced by 0.003mm, which is optimized by 5.17%, and the optimization effect is obvious.

[0069] The above embodiments are merely examples of the present invention and do not limit the protection scope of the present invention. All designs that are the same or similar to the present invention fall within the protection scope of the present invention.

Claims

1. A method for optimizing process parameters of SLM molded parts based on evolutionary blending method, characterized in that: The following steps are involved: Step 1: Select key process parameters such as laser power, scanning speed, scanning spacing, and powder thickness, take the deformation of SLM molded parts as the response index, use SimufactAdditive software to perform numerical calculations on the deformation problem of 316L stainless steel powder in the SLM molding process, and obtain the deformation data of SLM molded parts under different process parameters; use the collected experimental data to construct a mathematical relationship model between process parameters and deformation of molded parts through the response surface method to predict the deformation; Step 2, the evolutionary hybrid method is used to optimize the process parameters in the model; the evolutionary hybrid method is to perform evolutionary operations in the predation and migration behaviors of black kites, use two different types of evolutionary strategies to generate candidate solutions, and apply weighted mixing to generate the final evolution vector; then, the hybrid probability factor is introduced to generate experimental individuals by combining the evolutionary vector and some features of the original solution; and on this basis, the fitness of the experimental individuals and the original solution is compared to decide whether to accept the new solution, increase the global search capability, and thus optimize the process parameters in the model, and find the optimal combination of process parameters to reduce the deformation of SLM molded parts.

2. The method for optimizing process parameters of SLM molded parts based on evolutionary blending method according to claim 1 is characterized in that: The evolutionary hybrid method described in step 2 is to perform evolutionary operations in the predation and migration behaviors of black-winged kites, generate candidate solutions using two different types of evolutionary strategies, and apply weighted hybridization to generate the final evolution vector; then, introduce a hybrid probability factor to generate test individuals by combining the evolution vector and some features of the original solution; and on this basis, compare the fitness of the test individuals and the original solution to decide whether to accept the new solution, specifically: The evolution operation formula is: Among them, v i,j is the value of the jth characteristic of the i-th individual in the new solution of evolution strategy 1, x i,j is the value of the jth characteristic of the ith individual in the original solution, γ is the scaling factor of evolution strategy 1, exp(fi-τ) is the natural exponential function, and f i is the fitness of the ith individual, τ is the threshold parameter, C(0,1) is a random number drawn from the Cauchy distribution with location and scale parameters of 0 and 1 respectively, and u i,j is the value of the jth characteristic of the i-th individual in the new solution of evolution strategy 2, x r1,j 、x r2,j 、x r3,j Represents the value of the jth feature of three randomly selected individuals, F dynamic is the scaling factor of evolution strategy 2, F0 is the initial scaling factor, β is the parameter that controls the rate of change of the scaling factor, and f min and f max are the minimum and maximum fitness in the current population, respectively, and α is the mixing coefficient; In the evolutionary hybrid method, the evolution operation first takes the eigenvalue x of the original solution as i,j As a basis, add an adjustment term to get the new solution v i,j The adjustment term combines the scaling factor γ, the random number C(0,1) drawn from the Cauchy distribution, and the natural exponential function exp(fi-τ). The setting of each coefficient strategically controls the amplitude and direction of the adjustment. The purpose is to jump out of the local solution, increase the diversity of the search and the possibility of global optimization. Secondly, based on the eigenvalues ​​x of three different individuals randomly selected r1,j 、x r2,j 、x r3,j Calculate and then multiply by the scaling factor F dynamic , and get a new solution u i,j ; F dynamic According to the minimum fitness f in the current population min , maximum fitness f max , parameter β and initial scaling factor F0 are dynamically adjusted; the two new solutions v i,j 、u i,j As candidate solutions, the mixing coefficient α is introduced to combine these candidate solutions through weighted mixing to generate the final evolution vector; Since the evolutionary blending method improves the global search ability of the algorithm and the quality of the solution through evolutionary operations, the following will specifically explain how it guides the solution update process through the mixing probability factor R to further enhance the algorithm's exploration and utilization capabilities; The formula for introducing the mixed probability factor R to generate and compare the test individuals is: Among them, y i,j is the value of the jth characteristic of the ith individual in the test, w i,j is the value of the jth feature of the i-th individual in the evolution vector, j r is a randomly selected cross-reference dimension, rand j is the random number associated with dimension j, R is the mixing probability factor, x i is the i-th individual of the original solution, y i is the i-th individual of the test individual, f(y i ) is the experimental individual y i The fitness of f(x i ) is the original solution x i Adaptability; Based on the above formula, firstly, according to the given mixing probability factor R, the evolution vector w i,j With the current individual x i Mix to generate the experimental individuals y i,j Specifically, for each dimension j, if the randomly generated number is less than R or j is equal to a randomly selected dimension index, then the value y of the test individual in that dimension is i,j is equal to the value w of the evolution vector in this dimension i,j Otherwise, y i,j Equal to the value x of the current individual in this dimension i,j ; Then, by comparing the experimental individuals y i and the current individual x i The fitness of the test individual y determines whether to update the current individual; if the test individual y i The fitness of y is better, then the current individual is updated to y i , hoping to gradually find better solutions during the evolution process.

Images