Hybrid Optimization Method for Layout of Heat Pipe Constrained Components

By establishing populations on the x-axis and y-axis and combining genetic algorithms and firework algorithms, the high-dimensional and multi-constraint problems in the layout optimization of heat pipe constraint components are solved, and more efficient search and accuracy are achieved, and the existing technology is overcome. Premature maturity agglomeration and local optimal problems.

CN116776552BActive Publication Date: 2024-12-03SUZHOU UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310555440.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-05-17
Publication Date
2024-12-03
Estimated Expiration
2043-05-17

AI Technical Summary

Technical Problem

The prior art is difficult to effectively deal with the problem of component layout optimization that relies on heat pipe heat dissipation, especially in the case of high-dimensionality and multi-constraints, and it is difficult to obtain satisfactory layout solutions, and the existing methods cannot effectively deal with the special problems caused by heat pipes. .

Method used

A hybrid optimization method for heat pipe constrained component layout is adopted. By establishing populations on the x-axis and y-axis, combining genetic algorithms, firework algorithms and championship selection operators, the combination of global search and local search is achieved, and the search ability and Solve the accuracy.

Benefits of technology

The layout optimization problem of heat pipe constraint components is effectively dealt with, the solution accuracy and speed are improved, and better layout solutions can be found under high-dimensional and multi-constraint conditions, avoiding premature convergence and local optimal traps.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116776552B_ABST
    Figure CN116776552B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of component layout optimization, and discloses a hybrid optimization method for the layout of a heat pipe constraint component. An x-axis and a y-axis are respectively established along two directions perpendicular and parallel to the direction of the heat pipe, and a population X is established on the x-axis and a population Y is established on the y-axis according to the structure of the heat pipe constraint component. Population X and population Y respectively represent a group of layout solutions of the heat pipe constraint component in the x-axis and y-axis directions. A genetic algorithm, a fireworks algorithm and a tournament selection operator are combined to find the optimal layout solution of population X on the x-axis, and the optimal layout solution of population Y on the y-axis is found according to the optimal layout solution of population X on the x-axis. The optimal feasible solution of the layout of the heat pipe constraint component is obtained by combining the two groups of optimal layout solutions. The present invention can effectively handle the HCLO problem, obtain a better layout solution, improve the search capability, and improve the solution accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of component layout optimization, and in particular to a hybrid optimization method for heat pipe constraint component layout. Background Art

[0002] As the heat generation power of electronic components increases, the heat dissipation requirements of cutting-edge electronic devices are getting higher and higher. In order to ensure the normal operation of the equipment, the heat dissipation performance of the equipment is improved by optimizing the component layout, which has been widely used in engineering fields such as printed circuit boards (PCBs), satellites, electric vehicles, and mechanical systems. In order to meet the engineering application needs of component layout optimization (CLO), researchers proposed a formal problem model with the goal of optimizing thermal performance and considering component layout constraints. The purpose is to solve the component layout solution that optimizes the heat dissipation performance of electronic equipment under relevant constraints such as component non-overlap and layout stability.

[0003] Among various types of heat dissipation materials, heat pipe heat dissipation has been increasingly used in the engineering field due to its advantages such as high thermal conductivity and high heat flux density. With the promotion of heat pipe heat dissipation technology, some studies have proposed the component layout optimization (HCLO) problem considering heat pipe constraints. This problem is mainly aimed at application scenarios that rely on heat pipes for component heat dissipation. Unlike other CLO problem models, the HCLO problem not only considers the constraints of conventional CLO problems such as component non-overlap and static stability, but also considers specific layout goals and constraints caused by heat pipes, such as heat pipe heat dissipation capacity constraints and component heat pipe overlap constraints. The above goals and constraints make the HCLO problem have the characteristics of high dimensionality, multiple constraints, and sparse feasible domain, making it more difficult to solve.

[0004] In solving component layout optimization (CLO) problems, metaheuristic algorithms, such as genetic algorithms (GA), particle swarm optimization (PSO) and differential evolution (DE), have been widely used in recent years to solve CLO problems due to their advantages of high efficiency and strong scalability. Component layout optimization (CLO) problems are usually formalized as constrained optimization problems. CLO varies in terms of model objectives and model constraints according to different application scenarios. In view of the differences in CLO problems, current research has incorporated problem-related solution strategies based on different metaheuristic algorithm frameworks and designed various solution strategies. However, HCLO is a new type of CLO problem. Due to the need to specifically consider the specific layout objectives and constraints caused by heat pipes, the HCLO problem has the characteristics of high dimensionality, multiple constraints and sparse feasible domain, which is more difficult than other CLO problems. In order to solve the HCLO problem, Yao proposed an improved genetic algorithm that implements simulated crossover (SBX), polynomial mutation (PLM) and tournament selection. However, this method still suffers from the problem of premature convergence (Yao et al, 2022) and cannot obtain satisfactory solutions for high-dimensional HCLO problems. In order to cope with the challenges brought by HCLO, Ye proposed a dual-population tribe-based genetic algorithm (BCGA), which consists of a dual-population strategy, a framework based on and a redesigned genetic operator (Ye et al, 2023). However, the BCGA algorithm is prone to fall into local optimality and has poor local search ability.

[0005] In summary, there are few existing technologies that target application scenarios that rely on heat pipes for component heat dissipation. The specific solution strategies designed cannot be directly applied to the HCLO problem, and because the particularity of the problem caused by heat pipes is not considered, the existing methods cannot effectively deal with the HCLO problem. Although some studies have designed specific solution strategies for the HCLO problem, the success rate of solving complex HCLO problems is low and a satisfactory layout solution cannot be given. Summary of the invention

[0006] To this end, the technical problem to be solved by the present invention is to overcome the deficiencies in the prior art and provide a hybrid optimization method for the layout of heat pipe constraint components, which can effectively handle the HCLO problem, obtain a better layout solution, enhance the search capability, and improve the solution accuracy.

[0007] In order to solve the above technical problems, the present invention provides a hybrid optimization method for the layout of heat pipe constraint components, comprising:

[0008] S1: Establish an x-axis and a y-axis in two directions perpendicular and parallel to the direction of the heat pipe, respectively, and establish a population X on the x-axis and a population Y on the y-axis according to the structure of the heat pipe constraint assembly, wherein the population X represents a set of layout solutions of the heat pipe constraint assembly in the x-axis direction, and the population Y represents a set of layout solutions of the heat pipe constraint assembly in the y-axis direction;

[0009] S2: Combine the genetic algorithm, the fireworks algorithm and the tournament selection operator to find the optimal layout solution of population X on the x-axis, find the optimal layout solution of population Y on the y-axis based on the optimal layout solution of population X on the x-axis, and combine the optimal layout solution of population X on the x-axis and the optimal layout solution of population Y on the y-axis to obtain the optimal feasible solution for the layout of the heat pipe constraint assembly.

[0010] In one embodiment of the present invention, the combination of genetic algorithm, fireworks algorithm and tournament selection operator is used to find the optimal layout solution of population X on the x-axis, and the optimal layout solution of population Y on the y-axis is found according to the optimal layout solution of population X on the x-axis. The optimal feasible solution of the layout of the heat pipe constraint assembly is obtained by combining the optimal layout solution of population X on the x-axis and the optimal layout solution of population Y on the y-axis, which is specifically:

[0011] S2-1: Set the maximum number of function evaluations and clan size;

[0012] S2-2: Evaluate population X. If the number of evaluations does not reach the maximum number of function evaluations, execute S2-3; if the number of evaluations reaches the maximum number of function evaluations, execute S2-10;

[0013] S2-3: Combine the genetic algorithm, the fireworks algorithm and the tournament selection operator to update the optimal layout solution of population X on the x-axis;

[0014] S2-4: If the layout solution of population X on the x-axis satisfies the component-heat pipe overlap constraint, then activate population Y and execute S2-5; otherwise, execute S2-2;

[0015] S2-5: Determine whether it is the first time to activate population X or whether a better layout solution of population X on the x-axis is obtained. If so, execute S2-6; otherwise, execute S2-7;

[0016] S2-6: Use the layout solution of population X on the x-axis at this time to evaluate population Y and execute S2-7;

[0017] S2-7: Determine whether the optimal feasible solution for the complete heat pipe constraint assembly layout has been obtained based on the layout solution of the current population X on the x-axis, if not, execute S2-8, if yes, execute S2-2;

[0018] S2-8: Combine the genetic algorithm and the tournament selection operator to update the optimal layout solution of population Y on the y-axis, and execute S2-9;

[0019] S2-9: Determine whether the current population Y has obtained a feasible solution on the y-axis. If so, store the feasible solution, and the population Y goes into hibernation and executes S2-2; if not, directly execute S2-2;

[0020] S2-10: Output the solution with the smallest target value from the stored feasible solution set, thereby obtaining the optimal feasible solution for the layout of the heat pipe constraint component.

[0021] In one embodiment of the present invention, the combination of the genetic algorithm, the fireworks algorithm and the tournament selection operator to update the optimal layout solution of the population X on the x-axis is specifically:

[0022] S2-3-1: Get the size of population X at this time, construct all clans in population X, and obtain the offspring of all clans by simulating the binary crossover operator;

[0023] S2-3-2: Update the offspring based on polynomial mutation and / or exchange mutation, obtain feasible solutions based on the update results of all individuals in the clan, and use the best solution among all feasible solutions as the fireworks solution;

[0024] S2-3-3: Calculate the number of spark solutions and explosion amplitude of each fireworks solution, the number of spark solutions is used to determine how many neighbor solutions each fireworks solution generates in the fireworks algorithm, the explosion amplitude is used to determine the displacement radius of the component x-axis coordinate, and generate a spark solution set according to the number of spark solutions and the explosion amplitude;

[0025] S2-3-4: Combine the descendants of the generated clan and the spark solution set, select the solution with the smallest target value as the elite solution in the current generation population X, and use the elite solution in the current generation population X as the optimal layout solution of the current generation population X on the x-axis;

[0026] S2-3-5: Execute the tournament selection operator to obtain the next generation population X.

[0027] In one embodiment of the present invention, the determination of whether the current population Y has obtained a feasible solution on the y-axis is specifically as follows:

[0028] If the layout solution of population Y on the y-axis satisfies the non-overlapping constraint, the system center of mass constraint, and the heat dissipation capacity constraint at this time, it is considered that the current population Y has obtained a feasible solution on the y-axis.

[0029] In one embodiment of the present invention, the updating of offspring based on polynomial mutation and / or exchange variation is specifically:

[0030] The hybrid mutation operator is obtained by combining polynomial mutation and exchange mutation. The parameters required to execute the hybrid mutation operator are calculated based on the descendants of all clans. According to the hybrid mutation operator, individuals perform PLM mutation operation with probability p and exchange mutation operation with probability 1-p in each generation.

[0031] In one embodiment of the present invention, the calculation method of the number of spark solutions is:

[0032]

[0033] Among them, n f is the number of spark solutions to be generated; n e is a constant used to control the number of spark solutions; C1(XF) is the target value of the fireworks solution, where XF represents the layout solution with the smallest target value in the population; n is the population size; is the target value corresponding to the layout solution with the largest target value in the population; C1(X i ) is the target value of the i-th layout solution; ε is a small constant used to avoid division by zero; in order to prevent the number of spark solutions from being too large or too small, the boundary is set to n f ∈[an e , bn e ], where a and b are constants and a<b<1;

[0034] The calculation method of the explosion amplitude is:

[0035]

[0036] Among them, r f Indicates the maximum adjustment of the x-axis, which is related to the interval between two adjacent heat pipes; r e To control the explosion range, It is the objective value of the layout solution with the smallest objective value within the population.

[0037] In one embodiment of the present invention, in order to avoid excessive spark decryption and improve exploration performance, an adjustment amount for the explosion amplitude is set, specifically:

[0038] The explosion amplitude is adjusted once:

[0039]

[0040] in, is the explosion amplitude after adjustment, d hp Represents the distance between two adjacent heat pipes, round(r f ) represents r f Round down;

[0041] Next, adjust the explosion amplitude twice:

[0042]

[0043] in, is the explosion amplitude after secondary adjustment; α cf , β cf , α f and β f is a constant for adjusting the dynamic explosion amplitude; FEs indicates the number of target value calculations used; MaxFEs is the total number of target value calculations.

[0044] In one embodiment of the present invention, generating a spark solution set according to the number of spark solutions and the explosion amplitude is specifically:

[0045] The generated n f Each spark solution in the spark solutions represents a neighbor layout solution generated by the best layout solution. k (k=1,...,n f ), initialize by copying the fireworks solution, and randomly select the component set Del that needs to be adjusted from all candidate components;

[0046] For the components belonging to the Del set, the x-axis coordinate of the component is displaced and adjusted according to the explosion amplitude:

[0047]

[0048] Among them, XS kj represents the x-axis coordinate of the j component in the k-th spark solution, is a and A random number between XF j represents the x-axis coordinate of the jth component in the fireworks solution XF, D sel is a set of randomly selected dimensions;

[0049] Adjust the component layout to a reasonable range to obtain the final spark solution set. The specific steps of adjusting the component layout to a reasonable range are as follows:

[0050] XS kj =max(min(XS kj ,u j ), l j ),

[0051] Among them, XS kj represents the x-axis coordinate of the j-th component in the k-th spark solution; u j and l j are the upper and lower limits of the x-axis value of the j-th component.

[0052] In one embodiment of the present invention, the execution of the tournament selection operator to obtain the next generation population X is specifically:

[0053] Assume that the size of the candidate solution set is N candidate , you need to select The solution is used to guide the next iteration; A solution, build Championships;

[0054] Randomly selected in each tournament candidate solutions, the remaining The candidate solutions will be evenly distributed to the In the tournaments, the population with the smallest target value in each tournament is selected and released into the next generation population CL. t+1 .

[0055] In one embodiment of the present invention, when updating the optimal layout solution of population X on the x-axis and updating the optimal layout solution of population Y on the y-axis, after each iterative update, the distance vector between the current optimal layout solution and the optimal layout solution of the previous iteration is calculated:

[0056]

[0057] Among them, Δfb t is the distance vector, represents the current best layout solution, represents the best layout solution of the previous generation, FEs t Indicates the number of target value calculations used up to now, FEs t-1 Indicates the number of target value calculations used up to the previous generation;

[0058] Apply the distance vector to the optimal layout solution in the current iteration to obtain a new layout solution X new :

[0059]

[0060] Unlock the new layout new The worst solution in the replacement population X.

[0061] The above technical solution of the present invention has the following advantages compared with the prior art:

[0062] The present invention transforms the problem of heat pipe constraint assembly layout into a dual-population solution problem, and on this basis realizes global search through genetic operators and increases the advantages of local search through fireworks operators, thereby achieving effective processing of the HCLO problem and improving the effectiveness, speed and accuracy of solving the optimal layout solution of the heat pipe constraint assembly. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:

[0064] Figure 1 is a flow chart of the method of the present invention.

[0065] Figure 2 It is a flow chart of the process steps of the method of the present invention.

[0066] Figure 3 It is a schematic diagram of the hybrid operator solution process in the present invention. DETAILED DESCRIPTION

[0067] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0068] The HCLO problem can be formalized as a single-objective optimization problem with continuous constraints, which requires minimizing the maximum heat generation power borne by the heat pipe under non-overlapping constraints, system center of mass constraints, component-heat pipe overlap constraints, and heat dissipation capacity constraints.

[0069] The non-overlapping constraint requires that there should be no overlap between different components or between components and layout domains, and the overlapping volume should be strictly equal to 0 to ensure the feasibility of the layout solution. Therefore, the non-overlapping constraint g1(X) can be expressed as:

[0070]

[0071] Where, ΔV ij (i, j>0, i!=j) represents the overlapping area between component i and component j, N c is the total number of components.

[0072] In order to ensure the static stability of the layout area, the system center of mass constraint must be met: in the O-xy system, the center of mass of the effective load of the layout area along the Y axis is required to be controlled within the allowable range, that is, the deviation of the system center of mass should be within the predefined offset range. Based on this, the system center of mass constraint g2(X) can be expressed as:

[0073] g2(X)=|y c -y e |-δy e ≤0,

[0074] in, is the current Y-axis system center of mass, y e is the expected Y-axis system center of mass, δy e is the maximum allowable center of mass deviation, the mass of the i-th component is mi , the center of mass is y i , N c is the number of components.

[0075] In order to ensure effective heat conduction, the foot area of ​​the component must be placed horizontally on top of any one or several heat pipes, that is, effective heat conduction can only be performed when the component completely spans the entire heat pipe. Based on this, the component-heat pipe overlap constraint g3(X) can be expressed as:

[0076]

[0077] Among them, d i represents the distance from the outer boundary of the heat pipe closest to the i-th component. If the i-th component effectively spans at least one heat pipe, then d i Will remain at 0.

[0078] The maximum heat dissipation power of each heat pipe is In order for the entire heat conduction system to work properly, it is necessary to ensure that the actual total heat dissipation power of each heat pipe is not greater than its maximum heat dissipation power. The actual heat dissipation power of a heat pipe depends on the heat accumulation of the components that effectively cross the heat pipe. If a component effectively crosses multiple heat pipes, its heat will be evenly accumulated on each crossed heat pipe. Based on this, the heat dissipation capacity constraint g4(X) can be expressed as:

[0079]

[0080] Among them, P j hp represents the actual heat dissipation power of the jth heat pipe, represents the set of components that effectively span the jth heat pipe, P i represents the heat generated by the i-th component that effectively crosses the current heat pipe, Indicates the total number of heat pipes that the component effectively spans, Indicates the maximum heat dissipation power of each heat pipe, N hp Indicates the total number of heat pipes in the current layout domain. When solving the DFH-4 problem, the heat pipe will run through the entire board along the x-axis, that is, two layout domains. The calculation of the actual heat pipe heat dissipation power should be performed based on the layout of the components in the two layout domains at the same time.

[0081] In actual engineering problems, there are often many components and multiple local optimal solutions are distributed in the objective function space. Therefore, the HCLO problem has difficulties such as high dimension, multi-peak, multi-constraint and sparse feasible domain. In view of these difficulties, the present invention designs an improved bi-population hybrid optimization method (Bi-population Clan-based Hybrid Algorithm, BCHA), which solves the better solution of the layout of heat pipe constraint components through the hybrid optimization operator of improved mutation and genetic operator (GA) and fireworks algorithm (Fireworks Algorithm, FWA).

[0082] Aiming at the difficulty of HCLO problem, the present invention proposed a dual population-clan framework in the previous study, and adopted genetic operators (GA) such as crossover and mutation as the backbone algorithm to realize population solution.

[0083] However, the present invention finds that the genetic algorithm (GA) is not good at local search, which reduces the possibility of finding a good solution, and it is difficult to escape when trapped in a local optimal solution. In order to further improve the search effect, the present invention combines GA with the Fireworks Algorithm (FWA) which has more advantages in local search, and proposes a hybrid optimization algorithm based on GA and FWA.

[0084] Considering the advantages of GA algorithm in global search, the present invention first performs GA search, and then performs local search based on FWA operator. In addition, unlike the traditional fireworks algorithm, if each individual is regarded as a firework, the number of sparks generated is too large and the similarity is very high, which not only wastes time but also leads to meaningless exploration. Therefore, when the population is divided into clans (sub-populations), only the optimal layout solution in each clan is regarded as a fireworks solution, and the new layout solution generated by the neighborhood search is called a spark solution.

[0085] Embodiment 1

[0086] Reference Figure 1 and Figure 2 As shown, the present invention discloses a hybrid optimization method for heat pipe constraint assembly layout, comprising the following steps:

[0087] S1: Establish an x-axis and a y-axis in two directions perpendicular and parallel to the direction of the heat pipe, respectively. In this embodiment, the x-axis is established in the direction perpendicular to the direction of the heat pipe, and the y-axis is established in the direction parallel to the direction of the heat pipe; according to the structure of the heat pipe constraint component, establish a population X on the x-axis and a population Y on the y-axis, wherein the population X represents a set of layout solutions of the heat pipe constraint component in the x-axis direction, and the population Y represents a set of layout solutions of the heat pipe constraint component in the y-axis direction, and initialize the population X and the population Y; in this embodiment, the initialization of the population X and the population Y is to randomly generate two populations and set variable parameters.

[0088] S2: Combine the genetic algorithm, the fireworks algorithm and the tournament selection operator to find the optimal layout solution of population X on the x-axis, find the optimal layout solution of population Y on the y-axis based on the optimal layout solution of population X on the x-axis, and combine the optimal layout solution of population X on the x-axis and the optimal layout solution of population Y on the y-axis to obtain the optimal feasible solution for the layout of the heat pipe constraint assembly.

[0089] S2-1: Set the maximum number of function evaluations and clan size;

[0090] S2-2: Evaluate population X. In this embodiment, the evaluation function C1 is used to evaluate the current solution of population X:

[0091]

[0092] in, represents the solution of the i-th individual in the t-th generation population X, express The C1 function evaluation value, express The heat dissipation capacity constraint, f(·) is the maximum heat generation power carried by a single heat pipe in the layout, express The maximum heat generation power carried by a single heat pipe under the layout, Indicates that all the populations in the tth generation The maximum value among the values;

[0093] If the number of evaluations does not reach the maximum number of function evaluations, execute S2-3; if the number of evaluations reaches the maximum number of function evaluations, execute S2-10;

[0094] S2-3: Update the clan size of population X as needed, and combine the genetic algorithm, fireworks algorithm and tournament selection operator to update the optimal layout solution of population X on the x-axis, that is, the process of one round of iteration in population X. Specifically,

[0095] S2-3-1: Get the size of population X (i.e., the t-th generation population X) at this time, set the parameters required by the operator; construct all clans in population X, and obtain the offspring of all clans by simulating the binary crossover operator (clan-based SBX);

[0096] S2-3-2: Update the offspring based on polynomial mutation and / or exchange mutation, map the decision variables of all individuals in the clan to the feasible domain, that is, obtain feasible solutions based on the update results of all individuals in the clan, and use the optimal solution among all feasible solutions as the fireworks solution;

[0097] Update offspring based on polynomial mutation and / or crossover mutation, specifically:

[0098] Spark solution will perform mutation operator to increase diversity. Considering that the use of Gaussian mutation may make the adjustment of some dimensions close to zero. In order to avoid such problems, the present invention uses a hybrid mutation operator that considers both polynomial mutation and exchange mutation to update the offspring, that is:

[0099] The hybrid mutation operator is obtained by combining polynomial mutation and exchange mutation. The parameters required to execute the hybrid mutation operator are calculated based on the descendants of all clans. According to the hybrid mutation operator, individuals perform PLM mutation operation with probability p and exchange mutation operation with probability 1-p in each generation.

[0100] S2-3-3: Calculate the number of spark solutions and explosion amplitude of each fireworks solution, the number of spark solutions is used to determine how many neighbor solutions each fireworks solution generates in the fireworks algorithm, the explosion amplitude is used to determine the displacement radius of the component x-axis coordinate, and generate a spark solution set according to the number of spark solutions and the explosion amplitude;

[0101] S2-3-3-1: Calculate the number of spark solutions and explosion amplitude of each firework solution:

[0102] Since the present invention hopes that a better layout solution can generate more neighbors, the calculation method of the number of spark solutions is:

[0103]

[0104] Among them, n f is the number of spark solutions to be generated, n e is a constant used to control the number of spark solutions, which is related to the size of each tribe. In this embodiment is the number of individuals in the current clan, such as is the target value of the fireworks solution, where XF represents the layout solution with the smallest target value in the population, and n is the population size; is the target value corresponding to the layout solution with the largest target value in the population; C1(X i) is the target value of the i-th layout solution, ε is a small constant used to avoid division by zero; in order to prevent the number of spark solutions from being too large or too small, the boundary is set to n f ∈[an e , bn e ], wherein a and b are constants adjusted according to actual conditions, and a<b<1. In this embodiment, a=0.1, b=0.2.

[0105] Since the present invention hopes to find a better layout solution in a smaller neighborhood, the calculation method of the explosion amplitude is:

[0106]

[0107] Among them, r f Indicates the maximum adjustment of the x-axis, which is related to the interval between two adjacent heat pipes; r e To control the explosion range, It is the objective value of the layout solution with the smallest objective value within the population.

[0108] S2-3-3-2: Considering r f As the number of iterations increases, the neighborhood will become smaller during the search process, resulting in the spark solution set being generated within a very small neighborhood range, making the present invention ineffective. Therefore, in order to avoid excessive spark decryption and improve exploration performance, the adjustment amount for the explosion amplitude is set, specifically:

[0109] Considering that the interval between adjacent heat pipes is fixed, the size of the square root is well suited to problems of different scales and complexities. Therefore, based on the characteristics of the problem, the explosion amplitude is adjusted to:

[0110]

[0111] in, is the explosion amplitude after adjustment, d hp Represents the distance between two adjacent heat pipes, round(r f ) represents r f Round down;

[0112] There may be multiple fireworks solutions in the population. In order to enable the better fireworks solution to generate neighbor solutions in a smaller area, thereby better realizing the local search in a small area, and the worse fireworks solution to find new layout solutions in a larger range, thus improving the exploration performance, the explosion amplitude is further adjusted according to the performance of the fireworks solution as follows:

[0113]

[0114] in, is the explosion amplitude after secondary adjustment; α cf, β cf , α f and β f is a constant that adjusts the dynamic explosion amplitude according to actual conditions; FEs represents the number of target value calculations used; MaxFEs is the total number of target value calculations; since there may be multiple clans in a population and each clan has a fireworks solution, there may be multiple fireworks solutions in a population, and core firework represents the fireworks with the lowest target value. In this embodiment, α cf =0.5, β cf =0.3, α f =1, β f =1.5, MaxFEs=5000D where D is the number of decision variables in the HCLO problem (i.e., 2*number of components).

[0115] S2-3-3-3: Generate spark solution set according to the number of spark solutions and explosion amplitude:

[0116] S2-3-3-3-1: The generated n f Each spark solution in the spark solutions represents a neighbor layout solution generated by the best layout solution. k (k=1,...,n f ), first initialize by copying the fireworks solution, and then randomly select the component set Del that needs to be adjusted from all candidate components:

[0117] S2-3-3-3-2: For the components belonging to the Del set, adjust the x-axis coordinate of the component according to the explosion amplitude:

[0118]

[0119] Among them, XS kj represents the x-axis coordinate of the j-th component in the k-th spark solution, is a and A random number between XF j represents the x-axis coordinate of the jth component in the fireworks solution XF, D sel is a set of randomly selected dimensions;

[0120] S2-3-3-3-3: Finally, since the component layout exceeds the boundary, it will lead to an infeasible solution. Therefore, it is necessary to adjust the component layout to a reasonable range to obtain the final spark solution set. The component layout is adjusted to a reasonable range to maintain feasibility. Specifically:

[0121] XS kj =max(min(XS kj ,u j ), lj ),

[0122] Among them, XS kj represents the x-axis coordinate of the j-th component in the k-th spark solution; u j and l j is the upper and lower limits of the x-axis value of the jth component. In this embodiment, u j , l j Depending on the problem, it is the layout boundary. For example, for test problem 1, u j =-40, l j =40.

[0123] S2-3-4: Combine the descendants of the clan generated by the genetic algorithm (GA) and the spark solution set, select the solution with the smallest target value as the elite solution in the current generation, i.e., the t-th generation population X, and use the elite solution in the current generation population X as the optimal layout solution of the current generation population X on the x-axis;

[0124] S2-3-5: Execute the tournament selection operator to obtain the next generation, i.e., the t+1 generation population X. Specifically:

[0125] After the spark solution set is generated, the original population and the newly generated spark solution set are used as candidates to select the next generation population. Unlike other algorithms, due to the uncertainty of the size of the spark solution set, the number of candidates in this invention changes dynamically. In view of this feature, a selection operator based on the improved tournament is designed.

[0126] Assume that the size of the candidate solution set is N candidate , you need to select The solution is used to guide the next iteration; A solution, build Championships;

[0127] First, each tournament is randomly selected. candidate solutions; secondly, the remaining The candidate solutions will be evenly distributed to the tournaments; finally, the population with the smallest target value in each tournament is selected and released into the next generation population CL t+1 .

[0128] The above selection method has two advantages. First, this strategy can ensure that the current optimal solution can be retained in the next iteration because all candidate solutions have participated in the tournament. Second, since the candidate population is distributed as evenly as possible, it helps to maintain population diversity and avoid premature convergence.

[0129] S2-3-6: When updating the optimal layout solution of population X on the x-axis, a feedback mechanism is designed to further improve the possibility of finding a better layout solution and jump out of the local optimal layout solution. After each iterative update, the distance vector between the current optimal layout solution and the optimal layout solution of the previous iteration is calculated:

[0130]

[0131] Among them, Δfb t is a distance vector, which indicates the offset between the current best layout solution and the previous best layout solution. represents the current best layout solution, that is, the solution with the smallest objective value in the tth generation, represents the best layout solution of the previous generation, that is, the best layout solution of the t-1 generation, FEs t Indicates the number of target value calculations used up to the current time, that is, up to the tth generation, FEs t-1 Indicates the number of target value calculations used up to the previous generation, that is, up to the t-1 generation;

[0132] Apply the distance vector to the optimal layout solution in the current iteration to obtain a new layout solution X new :

[0133]

[0134] Unlock the new layout new The worst solution in the replacement population X.

[0135] S2-4: If the layout solution of population X on the x-axis (i.e. the best x-axis distribution of the tth generation) ) satisfies the component-heat pipe overlap constraint, activate population Y and execute S2-5; otherwise, execute S2-2;

[0136] S2-5: Determine whether it is the first time to activate population X or whether a better layout solution of population X on the x-axis is obtained. If so, execute S2-6; otherwise, execute S2-7;

[0137] S2-6: Use the layout solution of population X on the x-axis to evaluate population Y. In this embodiment, the evaluation function C2 is used according to Evaluate the current solution of population Y:

[0138]

[0139] Among them, C2(Y i t ) indicates Y i t The C2 function evaluation value, represents the best X-axis distribution of population X in the tth generation, Y is the solution with the minimum objective value selected from the offspring of the clan generated by the t-th generation genetic algorithm (GA) and the spark solution set; i t represents the solution of the i-th individual in the t-th generation population Y (i.e., the y-axis layout solution), Represents a complete layout solution, consisting of and jointly constitute; Indicates the number of constraints violated by the current layout scheme. Indicates the amount by which the current layout scheme violates constraint j, represents the maximum violation of constraint j by all layout schemes of the tth generation, α is the penalty coefficient, and the value may be α=2. Execute S2-7;

[0140] S2-7: Determine whether the optimal feasible solution for the complete heat pipe constraint assembly layout has been obtained based on the layout solution of the current population X on the x-axis, if not, execute S2-8, if yes, execute S2-2;

[0141] S2-8: Combine the genetic algorithm and the tournament selection operator to update the optimal layout solution of population Y on the y-axis, and execute S2-9;

[0142] S2-3-1 to S2-3-5 output the best solution in the current generation population and the next generation population, which is applicable to population X. For the update of population Y, the FWA operator is not executed, so the process of combining the genetic algorithm and the tournament selection operator to update the optimal layout solution of population Y on the y axis, that is, a round of iteration in population Y, is specifically as follows:

[0143] S2-8-1: Get the size of population Y (i.e., the t-th generation population Y) at this time, set the parameters required by the operator; construct all clans in population Y, and obtain the offspring of all clans by simulating the binary crossover operator (clan-based SBX) for each clan;

[0144] S2-8-2: Update the offspring based on polynomial mutation and / or exchange mutation, map the decision variables of all individuals in the clan to the feasible domain, that is, obtain a feasible solution based on the update results of all individuals in the clan;

[0145] S2-8-3: Select the solution with the smallest C2 target value from the offspring solution set of the clan generated by the genetic algorithm (GA) as the elite solution in the current generation, i.e., the t-th generation population Y, and use the elite solution in the current generation population Y as the optimal layout solution of the current generation population Y on the y-axis;

[0146] S2-8-4: Execute the tournament selection operator to obtain the next generation, i.e., the t+1 generation population Y.

[0147] S2-8-5: When updating the optimal layout solution of population Y on the y-axis, a feedback mechanism is designed to further improve the possibility of finding a better layout solution and jump out of the local optimal layout solution. After each iterative update, the distance vector between the current optimal layout solution and the optimal layout solution of the previous iteration is calculated:

[0148]

[0149] Among them, Δfb t is a distance vector, which indicates the offset between the current best layout solution and the previous best layout solution. represents the current best layout solution, that is, the y-axis layout solution with the smallest C2 target value in the t-th generation population Y. represents the best layout solution of the previous generation, that is, the best layout solution of the t-1 generation, FEs t Indicates the number of target value calculations used up to the current time, that is, up to the tth generation, FEs t-1 Indicates the number of target value calculations used up to the previous generation, that is, up to the t-1 generation;

[0150] Apply the distance vector to the optimal layout solution in the current iteration to obtain a new layout solution Y new :

[0151]

[0152] Unlock the new layout new The worst solution in the replacement population Y.

[0153] S2-9: Determine whether the current population Y has obtained a feasible solution on the y-axis. If so, store the feasible solution, and the population Y goes into hibernation and executes S2-2; if not, directly execute S2-2;

[0154] Determine whether the current population Y has obtained a feasible solution on the y-axis, specifically:

[0155] If at this time the layout solution of population Y on the y-axis (that is, the best y-axis distribution of the tth generation ) satisfies the non-overlapping constraints, system center of mass constraints, and heat dissipation capacity constraints, then the current population Y is considered to have obtained a feasible solution on the y-axis.

[0156] S2-10: Output the solution with the smallest target value from the stored feasible solution set, thereby obtaining the optimal feasible solution for the layout of the heat pipe constraint component.

[0157] Considering that it is a waste to optimize population Y for the poor x-axis component layout, the algorithm only optimizes population X in the early stage. Only when population X reaches certain conditions of convergence (such as the algorithm reaches the function evaluation threshold, the best x-axis distribution of the tth generation If the component-heat pipe overlap constraint is satisfied, population Y can be activated and receive the heat provided by population X. After that, the two populations evolve independently in each generation, and population Y will be based on Optimize the Y-axis coordinates. Once If the non-overlapping constraint, system center of mass constraint and heat dissipation capacity constraint are satisfied, then As a feasible layout solution, population Y will be saved and will go dormant until population X finds an x-axis layout with a better target value and meets the conditions, and population Y will be activated again.

[0158] The present invention decomposes the layout of the heat pipe constraint assembly into the layout of the heat pipe constraint assembly in two directions, the x-axis and the y-axis, converts the layout of the heat pipe constraint assembly in the x-axis into the layout solution of the population X on the x-axis, and converts the layout of the heat pipe constraint assembly in the y-axis into the layout solution of the population Y on the y-axis, thereby converting the optimal layout of the heat pipe constraint assembly in the x-axis and the y-axis into two problems of finding the optimal layout solution of the population X on the x-axis and the optimal layout solution of the population Y on the y-axis.

[0159] The two problems cooperate with each other and a hybrid operator is used to search for the optimal layout solution. Figure 3 The figure shows an example of the search process based on the hybrid operator, which includes four steps: (a) determining the fireworks solution, (b) generating the spark solution set, (c) updating the population, and (d) the feedback mechanism. Taking the process of generating two sparks as an example, the information represented by the shapes and colors is shown in the legend. a) First, select the individual with the lowest fitness, that is, the lowest target value, as the fireworks in each clan; b) determine the number of generated spark solutions, the explosion amplitude, and generate the spark solution set; c) perform improved tournament selection on the population to determine the next generation of population. d) Execute the feedback mechanism to jump out of the local optimum.

[0160] Embodiment 2

[0161] The present invention also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the hybrid optimization method for the layout of the heat pipe constraint assembly in the first embodiment is implemented.

[0162] Embodiment 3

[0163] The invention also discloses a hybrid optimization system for the layout of heat pipe constraint components, which comprises a population establishment module and a heat pipe constraint component layout module.

[0164] The population establishment module is used to establish the x-axis and y-axis respectively in two directions perpendicular and parallel to the direction of the heat pipe, and establish a population X on the x-axis and a population Y on the y-axis according to the structure of the heat pipe constraint assembly. The population X represents a set of layout solutions of the heat pipe constraint assembly in the x-axis direction, and the population Y represents a set of layout solutions of the heat pipe constraint assembly in the y-axis direction.

[0165] The heat pipe constraint component layout module is used to combine the genetic algorithm, the fireworks algorithm and the tournament selection operator to find the optimal layout solution of population X on the x-axis, find the optimal layout solution of population Y on the y-axis based on the optimal layout solution of population X on the x-axis, and combine the optimal layout solution of population X on the x-axis and the optimal layout solution of population Y on the y-axis to obtain the optimal feasible solution of the heat pipe constraint component layout.

[0166] Embodiment 4

[0167] The present invention also discloses a device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the hybrid optimization method for the layout of the heat pipe constraint assembly in the first embodiment is implemented.

[0168] The present invention converts the problem of heat pipe constraint assembly layout into a dual population solution problem, and on this basis, realizes global search through genetic operators such as crossover and mutation, and increases the advantages of local search through fireworks operators, thereby realizing effective processing of the HCLO problem, and having strong global search and local search capabilities, thereby improving the effectiveness, speed and accuracy of solving the optimal layout solution of the heat pipe constraint assembly. The dual population solution problem can locate multiple local optima at the same time, and each solution only needs to consider half of the design variables and some constraints, which reduces the search space dimension and improves the effectiveness of the search on the one hand, and reduces the computational complexity on the other hand, thereby improving the solution efficiency.

[0169] In order to further demonstrate the beneficial effects of the present invention, four sets of HCLO problem benchmark examples provided by IEEE CEC are used in this embodiment for simulation experiments. The difficulty of the problems is progressive, where problem 1 contains 6 components in 1 layout area, problem 2 contains 15 components in 1 layout area, problem 3 contains 40 components in 2 layout areas, and problem 4 contains 90 components in 4 layout areas. On this basis, in order to further verify the effectiveness of the method, all components in the original 4 problems are rotated 90 degrees clockwise to generate problems 5 to 8 respectively. Considering that the layout domains of the original 4 benchmark problems are not squares, the generated new problems are not symmetrical with the original problems and can be used to test the method.

[0170] To prove the effectiveness of the present invention, the present invention (hereinafter referred to as BCHA) was compared and evaluated with the existing HCLO solving methods GAHCLO and BCGA methods. All methods were implemented using Python 3.8.2, and all calculations were performed using a 2.3GHz Intel Core i7 computer with 16GB of memory under Windows environment.

[0171] Considering the randomness of the method, the present invention conducted 30 experiments, and the stopping rule of each experiment was 5000D (D is the number of decision variables) with a maximum evaluation number, and the results are shown in Table 1. Each problem includes the following five statistical indicators: the mean, standard deviation, minimum and maximum values ​​of the objective function value, and the success rate. The best indicator value in each item is bolded.

[0172] Table 1 Simulation experiment results

[0173]

[0174] It can be seen from Table 1 that the effect of the present invention is far better than GAHCLO, and the statistical results between the two methods are significant. Among the five evaluation indicators, the values ​​of BCHA are equal to or lower than GAHCLO, indicating that BCHA performs better than GAHCLO.

[0175] BCHA and BCGA can find feasible solutions with a 100% success rate in 120 experiments on 8 problems, while GAHCLO cannot guarantee a 100% success rate. In each problem, BCHA can obtain the same or smaller target value as BCGA, indicating that BCHA has better search ability and is more likely to jump out of the local optimum, which is more prominent in the most complex problems 4 and 8. In addition, BCHA performs better than the comparison methods in both the mean and maximum values ​​of all problems, which means that BCHA has higher stability.

[0176] In summary, it is proved that the present invention improves the solution accuracy of the HCLO problem.

[0177] 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 adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0178] 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, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0179] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0180] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing 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.

[0181] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.

Claims

1. A hybrid optimization method for heat pipe constraint component layout, It is characterized in that include: S1: Establish an x-axis and a y-axis in two directions perpendicular and parallel to the direction of the heat pipe, respectively, and establish a population X on the x-axis and a population Y on the y-axis according to the structure of the heat pipe constraint assembly, wherein the population X represents a set of layout solutions of the heat pipe constraint assembly in the x-axis direction, and the population Y represents a set of layout solutions of the heat pipe constraint assembly in the y-axis direction; S2: Combine the genetic algorithm, the fireworks algorithm and the tournament selection operator to find the optimal layout solution of population X on the x-axis, find the optimal layout solution of population Y on the y-axis based on the optimal layout solution of population X on the x-axis, and combine the optimal layout solution of population X on the x-axis and the optimal layout solution of population Y on the y-axis to obtain the optimal feasible solution for the layout of the heat pipe constraint assembly; S2-1: Set the maximum number of function evaluations and clan size; S2-2: Evaluate population X using evaluation function C 1 Evaluate the current solution of population X: in, represents the solution of the i-th individual in the t-th generation population X, express C 1 Function evaluation value, express The heat dissipation capacity constraint, f(·) is the maximum heat generation power carried by a single heat pipe in the layout, express The maximum heat generation power carried by a single heat pipe under the layout, Indicates that all the populations in the tth generation The maximum value among the values; If the number of evaluations does not reach the maximum number of function evaluations, execute S2-3; if the number of evaluations reaches the maximum number of function evaluations, execute S2-10; S2-3: Combine the genetic algorithm, the fireworks algorithm and the tournament selection operator to update the optimal layout solution of population X on the x-axis; S2-3-1: Get the size of population X at this time, construct all clans in population X, and obtain the offspring of all clans by simulating the binary crossover operator; S2-3-2: Update the offspring based on polynomial mutation and / or exchange mutation, obtain feasible solutions based on the update results of all individuals in the clan, and use the best solution among all feasible solutions as the fireworks solution; Updating offspring based on polynomial mutation and / or exchange mutation, specifically: combining polynomial mutation and exchange mutation to obtain a hybrid mutation operator, calculating parameters required to execute the hybrid mutation operator based on the offspring of all clans, and individuals executing the PLM mutation operation with probability p and the exchange mutation operation with probability 1-p in each generation according to the hybrid mutation operator; S2-3-3: Calculate the number of spark solutions and explosion amplitude of each fireworks solution, the number of spark solutions is used to determine how many neighbor solutions each fireworks solution generates in the fireworks algorithm, the explosion amplitude is used to determine the displacement radius of the component x-axis coordinate, and generate a spark solution set according to the number of spark solutions and the explosion amplitude; S2-3-3-1: Calculate the number of spark solutions and explosion amplitude of each firework solution: The calculation method of the spark solution quantity is: Among them, n f is the number of spark solutions to be generated; n e is a constant used to control the number of spark solutions; C 1 (XF) is the target value of the fireworks solution, where XF represents the layout solution with the smallest target value in the population; n is the population size; is the target value corresponding to the layout solution with the largest target value in the population; C 1 (X i ) is the target value of the i-th layout solution; ε is a small constant used to avoid division by zero; in order to prevent the number of spark solutions from being too large or too small, the boundary is set to n f ∈[an e ,bn e ], where a and b are constants, and a <b<1; The calculation method of the explosion amplitude is: Among them, r f Indicates the maximum adjustment of the x-axis, which is related to the interval between two adjacent heat pipes; r e To control the explosion range, is the target value of the layout solution with the smallest target value within the population; S2-3-3-2: In order to avoid excessive spark decryption and improve exploration performance, set the adjustment amount for the explosion amplitude, specifically: The explosion amplitude is adjusted once: in, is the explosion amplitude after adjustment, d hp Represents the distance between two adjacent heat pipes, round(r f ) represents r f Round down; Next, adjust the explosion amplitude twice: in, is the explosion amplitude after secondary adjustment; α cf , β Cf , α f and β f is a constant for adjusting the dynamic explosion amplitude; FEs indicates the number of target value calculations used; MaxFEs is the total number of target value calculations; S2-3-3-3: Generate spark solution set according to the number of spark solutions and explosion amplitude: S2-3-3-3-1: The generated n f Each spark solution in the spark solutions represents a neighbor layout solution generated by the best layout solution. k For example, k = 1, ..., n f , initialized by copying the fireworks solution, and randomly selecting the set of components Del that need to be adjusted from all candidate components; S2-3-3-3-2: For the components belonging to the Del set, adjust the x-axis coordinate of the component according to the explosion amplitude: Among them, XS kj represents the x-axis coordinate of the j component in the k-th spark solution, is a and A random number between XF j represents the x-axis coordinate of the jth component in the fireworks solution XF, D sel is a set of randomly selected dimensions; S2-3-3-3-3: Adjust the component layout to a reasonable range to obtain the final spark solution set. The specific steps of adjusting the component layout to a reasonable range are as follows: Among them, XS kj represents the x-axis coordinate of the j-th component in the k-th spark solution; u j and l j are the upper and lower limits of the x-axis value of the jth component; S2-3-4: Combine the descendants of the generated clan and the spark solution set, select the solution with the smallest target value as the elite solution in the current generation population X, and use the elite solution in the current generation population X as the optimal layout solution of the current generation population X on the x-axis; S2-3-5: Execute the tournament selection operator to obtain the next generation population X, specifically: Assume that the size of the candidate solution set is N candidate , you need to select The solution is used to guide the next iteration; A solution, build Championships; Randomly selected in each tournament candidate solutions, the remaining The candidate solutions will be evenly distributed to the In the tournaments, the population with the smallest target value in each tournament is selected and released into the next generation population CL. t+1 ; S2-3-6: When updating the optimal layout solution of population X on the x-axis and the optimal layout solution of population Y on the y-axis, after each iterative update, calculate the distance vector between the current optimal layout solution and the optimal layout solution of the previous iteration: Among them, Δfb t is the distance vector, represents the current best layout solution, represents the best layout solution of the previous generation, FEs t Indicates the number of target value calculations used up to now, FEs t-1 Indicates the number of target value calculations used up to the previous generation; Apply the distance vector to the optimal layout solution in the current iteration to obtain a new layout solution X new : Unlock the new layout new Replace the worst solution in population X; S2-4: If the layout solution of population X on the x-axis satisfies the component-heat pipe overlap constraint at this time, activate population Y and execute S2-5; otherwise, execute S2-2; S2-5: Determine whether it is the first time to activate population X or whether a better layout solution of population X on the x-axis is obtained. If so, execute S2-6; otherwise, execute S2-7; S2-6: Use the layout solution of population X on the x-axis at this time to evaluate population Y and execute S2-7; Use the evaluation function C 2 according to Evaluate the current solution of population Y: in, express C 2 Function evaluation value, represents the best X-axis distribution of population X in the tth generation, is the solution with the minimum objective value selected from the offspring and spark solution set of the clan generated by the t-th generation genetic algorithm; represents the solution of the i-th individual in the t-th generation population Y, Represents a complete layout solution; Indicates the number of constraints violated by the current layout scheme. Indicates the amount by which the current layout scheme violates constraint j, represents the maximum violation of constraint j by all layout schemes of the tth generation, and α is the penalty coefficient; S2-7: Determine whether the optimal feasible solution for the complete heat pipe constraint assembly layout has been obtained based on the layout solution of the current population X on the x-axis, if not, execute S2-8, if yes, execute S2-2; S2-8: Combine the genetic algorithm and the tournament selection operator to update the optimal layout solution of population Y on the y-axis, and execute S2-9; S2-8-1: Get the size of population Y at this time and set the parameters required by the operator; construct all clans in population Y, and obtain the offspring of all clans by simulating the binary crossover operator for each clan; S2-8-2: Update the offspring based on polynomial mutation and / or exchange mutation, map the decision variables of all individuals in the clan to the feasible domain, that is, obtain a feasible solution based on the update results of all individuals in the clan; S2-8-3: Select C from the offspring solution set of the clan generated by the genetic algorithm 2 The solution with the smallest objective value is taken as the elite solution of the current generation, and the elite solution in the current generation population Y is taken as the optimal layout solution of the current generation population Y on the y-axis; S2-8-4: Execute the tournament selection operator to obtain the next generation, i.e., the t+1 generation population Y; S2-8-5: When updating the optimal layout solution of population Y on the y-axis, in order to further improve the possibility of finding a better layout solution and jump out of the local optimal layout solution, after each iterative update, calculate the distance vector between the current optimal layout solution and the optimal layout solution of the previous iteration: Among them, Δfb t is a distance vector, which indicates the offset between the current best layout solution and the previous best layout solution. represents the C in the tth generation population Y 2 The y-axis layout solution with the minimum target value, represents the best layout solution of the t-1th generation, FEs t Indicates the number of target value calculations used up to the tth generation, FEs t-1 Indicates the number of target value calculations used up to the t-1th generation; Apply the distance vector to the optimal layout solution in the current iteration to obtain a new layout solution Y new : Unlock the new layout new The worst solution in the alternative population Y; S2-9: Determine whether the current population Y has obtained a feasible solution on the y-axis. Specifically, if the layout solution of the population Y on the y-axis satisfies the non-overlapping constraint, the system center of mass constraint, and the heat dissipation capacity constraint, then the current population Y is deemed to have obtained a feasible solution on the y-axis. If yes, the feasible solution is stored, population Y goes into hibernation, and S2-2 is executed; if no, S2-2 is executed directly; S2-10: Output the solution with the smallest target value from the stored feasible solution set, thereby obtaining the optimal feasible solution for the layout of the heat pipe constraint component.