An intelligent prediction method and system for wall function used in radiator simulation

Abstract

The present invention provides an intelligent prediction method and system for wall functions for radiator simulation. The detailed grid of the near-wall area of ​​the radiator is solved to the viscous bottom layer, and the solution results are analyzed to obtain data samples. The first defense strategy of the crown porcupine optimization algorithm is used to improve the particle swarm algorithm. The fourth defense strategy of the crown porcupine optimization algorithm is used to improve the foraging behavior and random behavior of the artificial fish algorithm. The two improved methods are integrated to obtain a fused intelligent algorithm. The fused intelligent algorithm is used to optimize the improved SVM, and then the improved SVM is used to predict the data samples obtained by the near-wall area grid refinement simulation to obtain a prediction model for the radiator wall function, and the grid size is further refined. The speed of the corresponding grid position is obtained through the prediction model, and multiple regressions are used for processing to obtain the wall function for radiator simulation. The present invention can perform precise simulation analysis of the radiator and accurately evaluate the heat exchange and flow performance of the radiator.

Description

Technical Field

[0001] The present invention belongs to the technical field of engineering radiators, and in particular relates to an intelligent prediction method and system for wall functions used in radiator simulation. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The radiator is one of the most important components in the cooling system of engineering machinery. Its main function is to quickly dissipate the heat generated by the engine or other heat sources through a suitable heat dissipation area to ensure that the equipment operates within an appropriate temperature range. At present, mainstream radiators often use complex enhanced heat dissipation methods to improve heat exchange efficiency. If the heat exchange capacity of radiators with different structures is analyzed by experimental measurement methods, the radiator needs to be made, and the mold for making the radiator is time-consuming and labor-intensive, and the price is very expensive, resulting in high cost of the experimental measurement method. Compared with the experimental measurement method, the numerical simulation method can avoid the problem of long cycle and high cost, but because the radiator has a large number of fins and complex geometric structures, and the fins themselves are small in spacing, the flow state near the wall is crucial to the calculation of the heat exchange and flow performance of the entire radiator. At present, there is no wall function suitable for the simulation of the structure of engineering machinery radiators, which makes the simulation results differ greatly from the experimental values. Therefore, it is crucial to use a reliable near-wall processing method to accurately evaluate the heat exchange and flow performance of the radiator. Summary of the invention

[0004] In order to solve the above problems, the present invention proposes an intelligent prediction method and system for wall functions used in radiator simulation. The present invention can perform accurate simulation analysis of the radiator and accurately evaluate the heat transfer and flow performance of the radiator.

[0005] According to some embodiments, the present invention adopts the following technical solutions:

[0006] An intelligent prediction method for a wall function for radiator simulation comprises the following steps:

[0007] Modeling the radiator to obtain a simulation model of the radiator, dividing a near-wall area of ​​the simulation model into multiple layers of grids, solving to a viscous bottom layer, and normalizing data samples obtained from the solution results;

[0008] The first defense strategy of the crown porcupine optimization algorithm is used to improve the particle swarm algorithm, and an improved particle swarm algorithm is obtained; the fourth defense strategy of the crown porcupine optimization algorithm is used to improve the foraging behavior and random behavior of the artificial fish school algorithm, and an improved artificial fish school algorithm is obtained; the improved particle swarm algorithm is first used for optimization, and when the particle position change during the optimization process is less than the set value, the improved artificial fish school algorithm is used for optimization to jump out of the local optimum and perform global optimization, and the optimal solutions of the two optimizations are compared, and the optimal one is selected until the iteration requirements are met, and a fused intelligent algorithm is obtained;

[0009] The penalty parameters and kernel functions in the support vector machine are optimized by using the fused intelligent algorithm to obtain an improved support vector machine.

[0010] The normalized data samples are trained based on the improved support vector machine to obtain the prediction model of the radiator wall function.

[0011] The number of grid layers is further subdivided, and the prediction model of the radiator wall function is used to predict the velocity of each corresponding grid position after further subdivision. The prediction results are processed by multiple regression to obtain a new wall function.

[0012] As an optional implementation, when modeling the radiator, a low Reynolds number turbulence model is used to describe the viscous damping effect in the near-wall region, the viscosity-affected region is directly solved, and a high Reynolds number turbulence model is used to solve the turbulent core region and the outer region.

[0013] As an optional implementation method, the standard k-ε turbulence model is often used to solve the fully turbulent area in the model, and the one-equation model established by Wolfshtein is used to calculate the turbulent dissipation rate and turbulent viscosity in the viscosity-affected area, and the turbulent Reynolds number is introduced The complete turbulence area and the viscosity-affected area are divided, where v is the kinematic viscosity, k is the turbulent kinetic energy, and y is the distance between the center point of the volume grid and the nearest wall. * When Re <200, a low Reynolds number turbulence model is used and the standard k-ε turbulence model is used for solution; * When the Reynolds number is >200, a high Reynolds number turbulence model is used and solved using the one-equation model established by Wolfshtein.

[0014] As an optional implementation method, the first defense strategy of the crown porcupine optimization algorithm is used to improve the particle swarm algorithm. The improved particle swarm position update formula is:

[0015]

[0016] In the formula, denote the position of the i-th particle at time t and time t+1 respectively, represents the speed of the i-th particle at time t and time t+1 respectively, β represents the proportion of the two algorithms in the optimization process, r1 represents a random number based on normal distribution, r2, r3, r4 are random numbers between 0 and 1, P i t represents the local optimum of the particle after the tth update, G i t represents the global optimum of the particle after the tth update; c1 and c2 are learning factors, which are individual learning factor and social learning factor respectively. The inertia coefficient ω is used to balance local search and global search.

[0017] As a further step, the weight β increases the weight of the crown porcupine optimization algorithm in the later stage of iteration, enhancing its ability to jump out of the local optimum. Indicates, t is the number of iterations, t max is the maximum number of iterations.

[0018] As an optional implementation, the fourth defense strategy of the crown porcupine optimization algorithm is used to improve the foraging behavior and random behavior of the artificial fish school algorithm. The improved foraging behavior of the artificial fish school algorithm is:

[0019]

[0020] In the formula represents the position of the ith artificial fish at time t and time t+1, r5 is a random number between 0 and 1, step is the moving step of the artificial fish, QX j The status is QX i The artificial fish randomly selects a new state within its field of view, and the formula is: QX j =QX i +Visual×r6, where Visual is the visual range of the artificial fish, and r6 is a random number between 0 and 1. j The food concentration at the place is not higher than QX i The artificial fish will choose a new state QX j , after trying try_num times, no food with a concentration greater than QX is found i The location of food concentration executes the random behavior improved by the fourth defense strategy of the crown porcupine algorithm. The random behavior formula is:

[0021]

[0022] Where f(·) represents the objective function, ε is a small value to prevent the formula from being divided by zero, δ is a parameter used to control the search direction, and is expressed using the formula Definition: r7, r8, r9 are random numbers between 0 and 1, μ is the proportion of the two algorithms in the optimization process, γi is the defense factor, N is the population size, t is the number of iterations, t max is the maximum number of iterations.

[0023] As an optional implementation method, the process of obtaining the fused intelligent algorithm includes: firstly using the improved particle swarm algorithm to search for the best, It represents the difference between the particle's position after iteration t+1 and iteration t divided by the particle's position at iteration t;

[0024] When the improved particle swarm algorithm shows that λ(t+1)<1% for M consecutive times during the optimization process, M is an integer, indicating that the particle position changes very little, the improved artificial fish swarm algorithm is used to continue the optimization. When λ(t+1)<1% appears again for M consecutive times, the optimal solution obtained by the optimization is compared with the optimal solution obtained by the improved particle swarm algorithm. The best of the two is selected and N n Record the number of optimizations performed using the improved artificial fish swarm algorithm. n The iteration stops when ≥3 or the number of iterations reaches the maximum number of iterations, thus obtaining a fused intelligent algorithm.

[0025] As an optional implementation method, the specific process of optimizing the penalty parameter and kernel function in the support vector machine by using a fused intelligent algorithm to obtain an improved support vector machine includes:

[0026] When solving nonlinear problems, the support vector machine maps the data set to a high-dimensional feature space through a nonlinear function and performs a linear regression operation on it. The corresponding fitting relationship is:

[0027] f(x)=w×φ(x i )+b

[0028] in, is a nonlinear function, w is the normal vector, x i is the data set, b is the displacement, and to obtain the optimal values ​​of w and b, the relaxation variable ξ is introduced i , i * , according to the support vector machine principle, linear regression is transformed into a solvable optimization problem:

[0029]

[0030] The constraints are:

[0031]

[0032] Among them, y i represents a label with a value of 1 or -1, C is the penalty parameter, ξ i * represents a training error greater than ε, ξi Represents a training error less than ε. Since the optimal solution in nonlinear regression is related to the saddle point of the Lagrange function, the Lagrange operator is introduced to obtain the regression function of the support vector machine:

[0033]

[0034] Among them, K(x i ,x) is the kernel function, 1. i is the Lagrange multiplier, n is the number of data in the data set, and a single Gaussian kernel function is selected, and its formula is:

[0035]

[0036] The obtained fusion intelligent algorithm is used to optimize the penalty parameter C of the support vector machine and the width coefficient σ of the kernel function.

[0037] As an optional implementation, the specific process of further subdividing the number of grid layers includes, when the original number of grid layers is n, further subdividing the number of layers into Mm*n, where Mm is a positive integer.

[0038] An intelligent prediction system for wall functions for radiator simulation, comprising:

[0039] A modeling and simulation module is configured to model the heat sink, obtain a simulation model of the heat sink, divide a near-wall region of the simulation model into multiple layers of grids, solve to a viscous bottom layer, and normalize data samples obtained from the solution results;

[0040] The algorithm fusion module is configured to improve the particle swarm algorithm by using the first defense strategy of the crown porcupine optimization algorithm to obtain an improved particle swarm algorithm; to improve the foraging behavior and random behavior of the artificial fish swarm algorithm by using the fourth defense strategy of the crown porcupine optimization algorithm to obtain an improved artificial fish swarm algorithm; to first use the improved particle swarm algorithm to search for the best, and when the particle position change during the search is less than the set value, the improved artificial fish swarm algorithm is used to search for the best, so as to jump out of the local optimum and search for the global optimum, and compare the optimal solutions of the two searches, select the best one, until the iteration requirements are met, and obtain a fused intelligent algorithm;

[0041] A support vector machine improvement module is configured to optimize the penalty parameter and kernel function in the support vector machine by using a fused intelligent algorithm to obtain an improved support vector machine;

[0042] The prediction model building module is configured to train the normalized data samples based on the improved support vector machine to obtain a prediction model of the radiator wall function;

[0043] The wall function prediction module is configured to further subdivide the number of grid layers, use the prediction model of the radiator wall function to predict the speed of each corresponding grid position after further subdivision, and process the prediction results using multiple regressions to obtain a new wall function.

[0044] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the steps in the above method are completed.

[0045] An electronic device comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps in the above method are completed.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention utilizes the first defense strategy of the crown porcupine optimization algorithm to improve the particle swarm algorithm, and obtains the improved particle swarm algorithm; utilizes the fourth defense strategy of the crown porcupine optimization algorithm to improve the foraging behavior and random behavior of the artificial fish swarm algorithm, and obtains the improved artificial fish swarm algorithm. When the improved particle swarm algorithm has a small difference between before and after iteration for multiple consecutive times, the improved artificial fish swarm algorithm is used to find the optimal solution, so as to jump out of the local optimum, perform global optimization, and compare the optimal solutions of the two optimizations to obtain a fused intelligent algorithm, and the fused intelligent algorithm is used to improve the support vector machine, so that when the algorithm falls into the local optimum, it can jump out of the local optimum and perform global optimization to obtain a more accurate optimal value.

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0050] Figure 1 It is a schematic diagram of the flow chart of this embodiment. DETAILED DESCRIPTION

[0051] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0053] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0054] In the absence of conflict, the embodiments in this application and the features in the embodiments may be combined with each other.

[0055] Embodiment 1

[0056] An intelligent prediction method for radiator wall function, such as Figure 1 As shown, the following steps are included:

[0057] A simulation model is constructed for the heat sink, and the near-wall area is divided into very fine layers of n grids, and the solution is obtained to the viscous bottom layer. The data samples obtained from the solution are normalized and converted into the format required by the support vector machine (SVM) algorithm.

[0058] The first defense strategy of the crown porcupine optimization algorithm is used to improve the particle swarm algorithm, and the improved particle swarm algorithm is obtained. The fourth defense strategy of the crown porcupine optimization algorithm is used to improve the foraging behavior and random behavior of the artificial fish swarm algorithm, and the improved artificial fish swarm algorithm is obtained. When the improved particle swarm algorithm has a small difference before and after iteration for many consecutive times, the improved artificial fish swarm algorithm is used to find the optimal solution to jump out of the local optimum and perform global optimization, and compare the optimal solutions of the two optimizations to obtain a fused intelligent algorithm.

[0059] Through the fusion of intelligent algorithms, the optimal penalty parameter C and kernel function width coefficient σ in SVM are found. The improved SVM is used to predict the data samples obtained from the near-wall mesh refinement simulation to obtain the prediction model of the radiator wall function. The mesh size is further refined to 100*n, and the velocity of the corresponding grid position is predicted by the obtained prediction model. Multiple regressions are used to process the prediction results to obtain the final wall function for radiator simulation.

[0060] In this embodiment, the grid size is further refined to 100*n. In other embodiments, 100 can be adjusted or changed to other values, such as 50, 20, etc., according to specific circumstances and requirements, and these are not exhaustively listed here.

[0061] The above process is described in detail below.

[0062] When constructing a radiator simulation model, first construct the radiator model in the simulation software, define boundary conditions and set grid parameters, and divide the near-wall area of ​​the constructed radiator model into very fine grids with n layers.

[0063] A low Reynolds number turbulence model is used to describe the viscous damping effect in the near-wall region, and the area affected by viscosity is directly solved. A high Reynolds number turbulence model is used to solve the turbulent core area and the outer area.

[0064] In this embodiment, the standard k-ε turbulence model is often used to solve the fully turbulent region in the model, and the one-equation model established by Wolfshtein is often used to calculate the turbulent dissipation rate and turbulent viscosity in the viscosity-affected region, and the turbulent Reynolds number is introduced. The fully turbulent region and the viscosity-affected region are divided.

[0065] In the formula, v is the kinematic viscosity, k is the turbulent kinetic energy, and y is the distance between the center point of the volume grid and the nearest wall. * When Re <200, a low Reynolds number turbulence model is used and the standard k-ε turbulence model is used for solution; * When the Reynolds number is >200, a high Reynolds number turbulence model is used and solved using the one-equation model established by Wolfshtein.

[0066] In this way, the radiator simulation model is constructed, the flow velocity of each grid is calculated, and the data samples for intelligent algorithm simulation prediction are obtained. The data samples are normalized and converted into the format required by the SVM algorithm.

[0067] The first defense strategy of the crown porcupine optimization algorithm, whose race is initialized as Where N represents the number of individuals, is the i-th candidate solution in the search space, are the lower and upper boundaries of the search range, respectively. is a vector randomly initialized between 0 and 1.

[0068] The formula for the first defense strategy of the crown porcupine optimization algorithm is:

[0069]

[0070] In the formula They represent the position of the i-th individual after t iterations and t+1 iterations, τ1 is a random number based on normal distribution, τ2 is a random number between 0 and 1, represents the optimal solution in the t-iteration process, Represents the vector generated between the current crested porcupine and the crested porcupine randomly selected from the population, which is used to represent the position of the predator at the tth iteration. The specific formula is: Where r is a random number between 1 and the population size N.

[0071] The fourth defense strategy of the crested porcupine is characterized by a physical behavior. When the predator is very close, the crested porcupine hits the predator with a physical attack. During the physical attack, the two bodies are tightly pressed together, just like an inelastic collision in one-dimensional space. The formula is:

[0072]

[0073] In the formula, represents the position of the i-th individual after t+1 iterations, represents the optimal solution in the t-iteration process, represents the position of the i-th individual after t iterations. At the same time, because the crested porcupine collides with the predator, It also represents the position of the predator. τ4 and τ5 are random numbers between 0 and 1. It represents the average force exerted by the crested porcupine on the predator. The specific formula is: In the formula represents the final velocity of the ith individual at time t+1 and is assigned according to a solution randomly selected from the current population, so represents the initial velocity of the ith individual at time t, τ6 represents a vector that generates random values ​​between 0 and 1, m i It represents the mass of the i-th individual (predator) at the t-th iteration. The specific formula is: In the formula, f(·) represents the objective function, ε represents a small value to avoid division by zero, δ is a parameter used to control the search direction, and the formula is used definition, is a vector containing randomly generated values ​​between 0 and 1; i is the defense factor, using the formula Indicates that rand represents a random number between 0 and 1, t is the current number of iterations, and t max is the maximum number of iterations; α is the convergence speed factor, which determines the convergence speed of the algorithm. After testing, the best effect is achieved when α=0.1.

[0074] Particle swarm algorithm, assuming that the search space is D-dimensional and there are N particles, the particle swarm can be expressed as: FX = {fx1, fx2, ... fx N}, where FX represents the particle swarm, fx i represents the i-th particle in the population, then the position of the i-th particle can be expressed as: i =(a1,a2,…,a D ) T , the corresponding velocity of the ith particle is: fvi =(fv1,fv2,…,fv D ) T , the local optimal solution P of the i-th particle best For: P i =(P1,P2,…,P D ) T , the global optimal solution G of the entire particle swarm best For: G i =(G1,G2,…,G D ) T During the iteration process, the particle updates its position and velocity according to the current local optimal solution and the global optimal solution. The velocity update formula is:

[0075] f i (t+1)=ω·fv i (t)+c1r1(P i (t)-a i (t))+c2r2(G i (t)-a i (t))

[0076] The position update formula is:

[0077] a i (t+1)=a i (t)+fv i (t+1)

[0078] Where t represents the tth iteration of the particle, P i (t) represents the local optimum of the particle after the tth update, G i (t) represents the global optimum of the particle after the tth update, fv i (t+1) represents the velocity of particle i after t+1 iterations, a i (t+1) represents the position of particle i after t+1 iterations. c1 and c2 are learning factors, which are individual learning factor and social learning factor respectively. r1 and r2 are random numbers between 0 and 1, which are used to represent the randomness of each particle in the search process of the particle swarm algorithm and increase the possibility of reaching the optimal solution. The inertia coefficient ω is used to balance the local search and the global search.

[0079] Artificial fish swarm algorithm, there are N artificial fish in a d-dimensional search space, the state position of the artificial fish is represented by the vector QX = (qx1, qx2, ..., qx d ) means, ||QX i -QX j|| represents the spatial distance between the two artificial fish, the visual field of the artificial fish is Visual, the moving step of the artificial fish is step, and the food concentration at the location of the artificial fish is QY=f(qx).

[0080] Foraging is an innate instinct of all animals, and fish are no exception. When a school of fish finds an area with a high concentration of food, it will move in that direction. The foraging behavior of an artificial school of fish simulates the process of fish finding food and moving towards it. The current state of the artificial fish is QX i , randomly select a new state QX within its field of view j , use QX j =QX i +Visual×Rand1(), where Rand1() is a random number between 0 and 1. j The food concentration at the site is higher than QX i The food concentration at QX j Directional movement, i.e. foraging behavior, is expressed by the formula:

[0081]

[0082] In the formula, represents the position of the ith artificial fish at time t and time t+1, and Rand2() is a random number between 0 and 1. j The food concentration at the place is not higher than QX i The artificial fish will choose a new state QX j , when the food concentration is greater than QX after trying the preset try_num times, i The location of food concentration performs random behavior, and the random behavior formula is:

[0083]

[0084] The improved particle swarm algorithm uses the first defense strategy of the crown porcupine optimization algorithm to improve the particle swarm algorithm. The improved particle swarm position update formula is:

[0085]

[0086] In the formula, denote the position of the i-th particle at time t and time t+1 respectively, Respectively represent the speed of the i-th particle at time t and time t+1. β represents the proportion of the two algorithms in the optimization process. Since the particle swarm algorithm is prone to fall into the local optimum, the proportion of the crown porcupine optimization algorithm is increased in the later stage of iteration to enhance its ability to jump out of the local optimum. Indicates, t is the number of iterations, t maxis the maximum number of iterations. r1 represents a random number based on normal distribution, and r2, r3, and r4 are random numbers between 0 and 1. i t represents the local optimum of the particle after the tth update, It represents the global optimum of the particle after the tth update. c1 and c2 are learning factors, which are individual learning factor and social learning factor respectively. The inertia coefficient ω is used to balance local search and global search.

[0087] The improved artificial fish swarm algorithm uses the fourth defense strategy of the crown porcupine optimization algorithm to improve the foraging behavior and random behavior of the artificial fish swarm algorithm. The foraging behavior formula of the improved artificial fish swarm algorithm remains unchanged, which is:

[0088]

[0089] In the formula represents the position of the ith artificial fish at time t and time t+1, and r5 is a random number between 0 and 1. j The status is QX i The artificial fish randomly selects a new state within its field of view, and the formula is: QX j =QX i +Visual×r6, where Visual is the visual field of the artificial fish and r6 is a random number between 0 and 1. j The food concentration at the place is not higher than QX i The artificial fish will choose a new state QX j , after trying try_num times, no food with a concentration greater than QX is found i The location of food concentration executes the random behavior improved by the fourth defense strategy of the crown porcupine algorithm. The random behavior formula is:

[0090]

[0091] In the formula, f(·) represents the objective function, and ε is a small value to prevent the formula from being divided by zero. δ is a parameter used to control the search direction and is calculated using the formula Definition: r7, r8, r9 are random numbers between 0 and 1. μ is the proportion of the two algorithms in the optimization process.

[0092] γ i is the defense factor, N is the population size, t is the number of iterations, t max is the maximum number of iterations.

[0093] Integrate intelligent algorithms and use the improved particle swarm algorithm to find the optimal solution. It means the difference between the particle position after t+1 iteration and t iteration divided by The position of the particle at the iteration. When the improved particle swarm algorithm shows that λ(t+1)<1% for 30 consecutive times during the optimization process, it means that the particle position changes very little, and the improved artificial fish swarm algorithm is used to continue the optimization. When λ(t+1)<1% appears again for 30 consecutive times, the optimal solution obtained by the optimization is compared with the optimal solution obtained by the improved particle swarm algorithm. The best of the two is selected, and N n Record the number of optimizations performed using the improved artificial fish swarm algorithm. n The iteration stops when ≥3 or the number of iterations reaches the maximum number of iterations. Thus, a fused intelligent algorithm is obtained.

[0094] The SVM algorithm is widely used in nonlinear and complex samples. When solving nonlinear problems, SVM maps the data set to a high-dimensional feature space through a nonlinear function and performs a linear regression operation on it. The corresponding fitting relationship is:

[0095] f(x)=w×φ(x i )+b

[0096] Where: is a nonlinear function, w is the normal vector, and b is the displacement. To obtain the optimal values ​​of w and b, the relaxation variable ξ is introduced i , According to the SVM principle, linear regression can be transformed into a solvable optimization problem:

[0097]

[0098] The constraints are:

[0099]

[0100] In the formula, y i represents the number of labels of 1 and -1, C is the penalty parameter, ξ i * represents a training error greater than ε, ξ i Indicates a training error less than ε. Since the optimal solution in nonlinear regression is related to the saddle point of the Lagrange function, the Lagrange operator is introduced to obtain the regression function of the SVM:

[0101]

[0102] Among them, K(x i ,x) is the kernel function, 1. i is the Lagrange multiplier (s). There are many commonly used kernel functions. The single Gaussian kernel function is separable and can map data to infinite dimensions, so it is widely used. Its formula is:

[0103]

[0104] The obtained fusion intelligent algorithm is used to optimize the penalty parameter C of SVM and the width coefficient σ of the kernel function, and then the obtained penalty parameter and kernel parameter are brought back to SVM to optimize the SVM performance. The obtained data samples are trained using the optimized SVM algorithm to obtain the prediction model of the radiator wall function.

[0105] The number of grid layers is further refined to 100*n, and the wall function prediction model is introduced to predict the velocity corresponding to the grid position. The wall function is obtained by multiple regression, which can be used for accurate simulation analysis of the radiator.

[0106] The values ​​of the various parameters in the above embodiments may be set to other values ​​according to the circumstances in other embodiments.

[0107] Embodiment 2

[0108] An intelligent prediction system for wall functions for radiator simulation, comprising:

[0109] A modeling and simulation module is configured to model the heat sink, obtain a simulation model of the heat sink, divide a near-wall region of the simulation model into multiple layers of grids, solve to a viscous bottom layer, and normalize data samples obtained from the solution results;

[0110] The algorithm fusion module is configured to improve the particle swarm algorithm by using the first defense strategy of the crown porcupine optimization algorithm to obtain an improved particle swarm algorithm; to improve the foraging behavior and random behavior of the artificial fish swarm algorithm by using the fourth defense strategy of the crown porcupine optimization algorithm to obtain an improved artificial fish swarm algorithm; to first use the improved particle swarm algorithm to search for the best, and when the particle position change during the search is less than the set value, the improved artificial fish swarm algorithm is used to search for the best, so as to jump out of the local optimum and search for the global optimum, and compare the optimal solutions of the two searches, select the best one, until the iteration requirements are met, and obtain a fused intelligent algorithm;

[0111] A support vector machine improvement module is configured to optimize the penalty parameter and kernel function in the support vector machine by using a fused intelligent algorithm to obtain an improved support vector machine;

[0112] The prediction model building module is configured to train the normalized data samples based on the improved support vector machine to obtain a prediction model of the radiator wall function;

[0113] The wall function prediction module is configured to further subdivide the number of grid layers, use the prediction model of the radiator wall function to predict the speed of each corresponding grid position after further subdivision, and process the prediction results using multiple regressions to obtain a new wall function.

[0114] Embodiment 3

[0115] A computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the steps in the method provided in embodiment 1 are completed.

[0116] Embodiment 4

[0117] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps in the method provided in the first embodiment are completed.

[0118] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0119] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the 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.

[0120] 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.

[0121] 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 1The steps for the functions specified in one or more boxes.

[0122] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principle of the present invention without creative labor shall be included in the protection scope of the present invention.

Claims

1. An intelligent prediction method for wall function for radiator simulation, characterized in that: The following steps are involved: Modeling the radiator to obtain a simulation model of the radiator, dividing a near-wall area of ​​the simulation model into multiple layers of grids, solving to a viscous bottom layer, and normalizing data samples obtained from the solution results; The first defense strategy of the crown porcupine optimization algorithm is used to improve the particle swarm algorithm, and an improved particle swarm algorithm is obtained; the fourth defense strategy of the crown porcupine optimization algorithm is used to improve the foraging behavior and random behavior of the artificial fish school algorithm, and an improved artificial fish school algorithm is obtained; the improved particle swarm algorithm is first used for optimization, and when the particle position change during the optimization process is less than the set value, the improved artificial fish school algorithm is used for optimization to jump out of the local optimum and perform global optimization, and the optimal solutions of the two optimizations are compared, and the optimal one is selected until the iteration requirements are met, and a fused intelligent algorithm is obtained; The penalty parameters and kernel functions in the support vector machine are optimized by using the fused intelligent algorithm to obtain an improved support vector machine. The normalized data samples are trained based on the improved support vector machine to obtain the prediction model of the radiator wall function. The number of grid layers is further subdivided, and the prediction model of the radiator wall function is used to predict the velocity of each corresponding grid position after further subdivision. The prediction results are processed by multiple regression to obtain a new wall function.

2. The intelligent prediction method of wall function for radiator simulation according to claim 1, characterized in that: When modeling the radiator, a low Reynolds number turbulence model is used to describe the viscous damping effect near the wall, directly solving the area affected by viscosity, and a high Reynolds number turbulence model is used to solve the turbulent core area and the outer area.

3. The intelligent prediction method for wall function for radiator simulation according to claim 2, characterized in that The standard k-ε turbulence model is often used to solve the fully turbulent region in the model. The one-equation model established by Wolfshtein is used to calculate the turbulent dissipation rate and turbulent viscosity in the viscosity-affected region, and the turbulent Reynolds number is introduced. The complete turbulence area and the viscosity-affected area are divided, where v is the kinematic viscosity, k is the turbulent kinetic energy, and y is the distance between the center point of the volume grid and the nearest wall. * When Re <200, a low Reynolds number turbulence model is used and the standard k-ε turbulence model is used for solution; * When the Reynolds number is >200, a high Reynolds number turbulence model is used and solved using the one-equation model established by Wolfshtein.

4. The intelligent prediction method of wall function for radiator simulation according to claim 1, characterized in that: The first defense strategy of the crown porcupine optimization algorithm is used to improve the particle swarm algorithm. The improved particle swarm position update formula is: In the formula, denote the position of the i-th particle at time t and time t+1 respectively, represents the speed of the i-th particle at time t and time t+1 respectively, β represents the proportion of the two algorithms in the optimization process, r1 represents a random number based on normal distribution, r2, r3, r4 are random numbers between 0 and 1, P i t represents the local optimum of the particle after the tth update, represents the global optimum of the particle after the tth update; c1 and c2 are learning factors, which are individual learning factor and social learning factor respectively. The inertia coefficient ω is used to balance local search and global search.

5. The intelligent prediction method of the wall function for radiator simulation according to claim 4, characterized in that: The weight β increases the weight of the crown porcupine optimization algorithm in the later stage of iteration, and enhances its ability to jump out of the local optimum. Indicates, t is the number of iterations, t max is the maximum number of iterations.

6. The intelligent prediction method of wall function for heat sink simulation according to claim 1, characterized in that: The fourth defense strategy of the crowned porcupine optimization algorithm is used to improve the foraging behavior and random behavior of the artificial fish swarm algorithm. The improved foraging behavior of the artificial fish swarm algorithm is: In the formula represents the position of the ith artificial fish at time t and time t+1, r5 is a random number between 0 and 1, step is the moving step of the artificial fish, QX j The status is QX i The artificial fish randomly selects a new state within its field of view, and the formula is: QX j =QX i +Visual×r6, where Visual is the visual range of the artificial fish, and r6 is a random number between 0 and 1. j The food concentration at the place is not higher than QX i The artificial fish will choose a new state QX j , after trying try_num times, no food with a concentration greater than QX is found i The location of food concentration executes the random behavior improved by the fourth defense strategy of the crown porcupine algorithm. The random behavior formula is: Where f(·) represents the objective function, ε is a small value to prevent the formula from being divided by zero, δ is a parameter used to control the search direction, and is expressed using the formula Definition: r7, r8, r9 are random numbers between 0 and 1, μ is the proportion of the two algorithms in the optimization process, γ i is the defense factor, N is the population size, t is the number of iterations, t max is the maximum number of iterations.

7. The intelligent prediction method of wall function for radiator simulation according to claim 1, characterized in that: The process of obtaining the integrated intelligent algorithm includes: firstly using the improved particle swarm algorithm to search for the best solution, It represents the difference between the particle's position after iteration t+1 and iteration t divided by the particle's position at iteration t; When the improved particle swarm algorithm shows that λ(t+1)<1% for M consecutive times during the optimization process, M is an integer, indicating that the particle position changes very little, the improved artificial fish swarm algorithm is used to continue the optimization. When λ(t+1)<1% appears again for M consecutive times, the optimal solution obtained by the optimization is compared with the optimal solution obtained by the improved particle swarm algorithm. The best of the two is selected and N n Record the number of optimizations performed using the improved artificial fish swarm algorithm. n The iteration stops when ≥3 or the number of iterations reaches the maximum number of iterations, thus obtaining a fused intelligent algorithm.

8. The intelligent prediction method of wall function for heat sink simulation according to claim 1, characterized in that: The specific process of optimizing the penalty parameters and kernel function in the support vector machine by using the fused intelligent algorithm to obtain the improved support vector machine includes: When solving nonlinear problems, the support vector machine maps the data set to a high-dimensional feature space through a nonlinear function and performs a linear regression operation on it. The corresponding fitting relationship is: f(x)=w×φ(x i )+b in, is a nonlinear function, w is the normal vector, x i For the data set b, the displacement is the amount of displacement. To obtain the best values ​​of w and b, the slack variable ξ is introduced. i , i * , according to the support vector machine principle, linear regression is transformed into a solvable optimization problem: The constraints are: Among them, y i The number of labels C representing 1 and -1 is the penalty parameter, ξ i * represents a training error greater than ε, ξ i Represents a training error less than ε. Since the optimal solution in nonlinear regression is related to the saddle point of the Lagrange function, the Lagrange operator is introduced to obtain the regression function of the support vector machine: Among them, K(x i ,x) is the kernel function, is the Lagrange multiplier, n is the number of data in the data set, and a single Gaussian kernel function is selected, and its formula is: The obtained fusion intelligent algorithm is used to optimize the penalty parameter C of the support vector machine and the width coefficient σ of the kernel function.

9. The intelligent prediction method of wall function for heat sink simulation according to claim 1, characterized in that: The specific process of further subdividing the number of grid layers includes: the original number of grid layers is n, and the number of layers is further refined to Mm*n, where Mm is a positive integer.

10. An intelligent prediction system for wall functions for radiator simulation, characterized in that: include: A modeling and simulation module is configured to model the heat sink, obtain a simulation model of the heat sink, divide a near-wall region of the simulation model into multiple layers of grids, solve to a viscous bottom layer, and normalize data samples obtained from the solution results; The algorithm fusion module is configured to improve the particle swarm algorithm by using the first defense strategy of the crown porcupine optimization algorithm to obtain an improved particle swarm algorithm; to improve the foraging behavior and random behavior of the artificial fish swarm algorithm by using the fourth defense strategy of the crown porcupine optimization algorithm to obtain an improved artificial fish swarm algorithm; to first use the improved particle swarm algorithm to search for the best, and when the particle position change during the search is less than the set value, the improved artificial fish swarm algorithm is used to search for the best, so as to jump out of the local optimum and search for the global optimum, and compare the optimal solutions of the two searches, select the best one, until the iteration requirements are met, and obtain a fused intelligent algorithm; A support vector machine improvement module is configured to optimize the penalty parameter and kernel function in the support vector machine by using a fused intelligent algorithm to obtain an improved support vector machine; The prediction model building module is configured to train the normalized data samples based on the improved support vector machine to obtain a prediction model of the radiator wall function; The wall function prediction module is configured to further subdivide the number of grid layers, use the prediction model of the radiator wall function to predict the speed of each corresponding grid position after further subdivision, and process the prediction results using multiple regressions to obtain a new wall function.

Images