Method for aerodynamic drag reduction and insulation performance synergistic optimization of car roof insulator under high-speed airflow

Abstract

The application discloses a kind of high-speed airflow under car roof insulator aerodynamic drag reduction and insulation performance synergic optimization method, belong to motor car roof insulator technical field, this method includes according to modeling target and initial sample set, establishes car roof insulator approximate model;According to car roof insulator approximate model, establish radial basis neural network model;According to the radial basis neural network model, the optimal parameters of concave surface are calculated using multi-island genetic algorithm;According to the optimal parameters of concave surface, construct concave non-smooth structure on the upper and lower surfaces of car roof insulator umbrella skirt, realize high-speed airflow under car roof insulator aerodynamic drag reduction and insulation performance synergic optimization.The application solves the problem that existing motor car roof insulator is difficult to further reduce resistance and has insufficient insulation performance.

Description

Technical Field

[0001] This invention belongs to the field of high-speed train roof insulator technology, and particularly relates to a method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow. Background Technology

[0002] The resistance experienced by high-speed trains during operation is mainly mechanical friction resistance and aerodynamic drag. As train speed increases, the proportion of aerodynamic drag increases, reaching approximately 80% of the total resistance when the train reaches 360 km / h. Exposed external insulation equipment on the roof, including roof insulators, surge arresters, and insulating bushings, has a similar structure, and the air resistance acting on these external insulation devices accounts for 5% to 8% of the total aerodynamic drag. Currently, traditional train drag reduction technologies are maturing, and existing optimization methods have reached a bottleneck. Constructing a non-smooth structure on the surface of the roof insulator skirts will become a new approach to drag reduction for high-speed trains.

[0003] The main function of external insulation equipment such as roof insulators is electrical insulation. During train operation, roof insulators often experience discharge phenomena, which endanger the safe operation of the train. However, due to the limitations of the roof space and the mechanical properties of the insulators, the structural height and skirt size of the roof insulators cannot be increased indefinitely. Summary of the Invention

[0004] To address the aforementioned shortcomings in the existing technology, this invention provides a method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow, which solves the problems of existing high-speed train roof insulators being unable to achieve further drag reduction and having insufficient insulation performance.

[0005] To achieve the aforementioned objectives, the technical solution adopted by this invention is as follows: a method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow, comprising the following steps:

[0006] S1. Based on the modeling objectives and the initial sample set, establish an approximate model of the roof insulator;

[0007] S2. Based on the approximate model of the roof insulator, establish a radial basis function neural network model;

[0008] S3. Based on the radial basis function neural network model, the optimal parameters of the concave surface are calculated using the multi-island genetic algorithm;

[0009] S4. Based on the optimal parameters of the concave surface, a non-smooth concave structure is constructed on the upper and lower surfaces of the roof insulator skirt to achieve coordinated optimization of aerodynamic drag reduction and insulation performance of the roof insulator under high-speed airflow.

[0010] The beneficial effects of this invention are as follows: The optimization of the roof insulator structure differs from traditional optimization methods. A concave, non-smooth structure is constructed on the insulator surface. The presence of the concave surface optimizes the flow field distribution on the insulator surface, thereby reducing the pressure drag of the insulator and the viscous drag of the shed surface, ultimately achieving aerodynamic drag reduction optimization of the roof insulator. At the same time, the concave surface of the insulator increases the creepage distance, increases the flashover voltage of the insulator, and improves the insulation performance. This invention achieves synergistic optimization of aerodynamic drag reduction and insulation performance improvement of the EMU roof insulator, providing a new approach to insulator structure optimization.

[0011] Further, step S1 specifically includes:

[0012] S101. Based on the modeling objective and the anti-interference conditions between the insulator boundary layer thickness, skirt size, airflow velocity, and concave unit cells, the range of values ​​for the pit parameters is obtained.

[0013] S102. Based on the range of values ​​for the pit parameters, an initial sample set is obtained using the Latin square sampling method.

[0014] S103. Based on the modeling objective and the initial sample set, obtain the drag coefficient of each initial sample point;

[0015] S104. Using the initial sample set and the resistance coefficient of each initial sample point, obtain the initial approximate model of the roof insulator;

[0016] S105. Based on the initial approximate model of the roof insulator, obtain the accuracy evaluation index data;

[0017] S106. Determine whether the accuracy evaluation index data meets the set threshold. If yes, obtain the approximate model of the roof insulator; otherwise, proceed to step S107. The expression of the objective function of the approximate model of the roof insulator is:

[0018] L c =Min(f(a,b,h,l1,l2))

[0019] Among them, L c is the objective function of the approximate model of the roof insulator; Min(·) is the minimum value function; f(·) is the function for calculating the drag coefficient of the roof insulator; a is the horizontal axis length of the non-smooth elliptical concave structure of the insulator skirt; b is the vertical axis length of the non-smooth elliptical concave structure of the insulator skirt; h is the depth of the non-smooth elliptical concave structure of the insulator skirt; l1 is the horizontal spacing of the non-smooth elliptical concave structure of the insulator skirt; l2 is the vertical spacing of the non-smooth elliptical concave structure of the insulator skirt.

[0020] S107. Add parameter sample points to the initial sample set based on the addition criterion, and return to step S103; the expression for the addition criterion is:

[0021]

[0022] Where E[I(x)] is the point addition criterion; I(x) is the newly added sample point; x is the pit parameter vector; G min This represents the minimum drag coefficient of the initial approximate model of the roof insulator; φ(·) represents the predicted response value of the roof insulator's drag coefficient; φ(·) represents the standard normal distribution function related to the insulator's drag coefficient. δ(x) is the normal probability density function; δ(x) is the mean square error of the model at the pit parameter vector x.

[0023] The beneficial effect of the above-mentioned further scheme is that establishing an approximate model of the roof insulator prepares for the establishment of a radial basis function neural network model.

[0024] Further, step S2 specifically includes:

[0025] S201. Based on the Gaussian radial basis function, the radial basis function is obtained:

[0026]

[0027] Where ρ(·) is the radial basis function; x is the input layer vector; e is the logarithmic constant; β i Let c be the input parameter vector of the i-th pit; i is the index of the pit parameter vector; ||·|| is the Euclidean norm; i The radial basis function center represents the Gaussian function center vector of the pit parameter;

[0028] S202. Based on the radial basis functions, obtain the variance of the radial basis functions:

[0029]

[0030] Where σ is the variance of the radial basis functions; a j This represents the expected output value of the roof insulator resistance coefficient under the pit parameter sample; y j This is the output of the j-th pit parameter vector; p is the number of input samples; n is a constant value for the sample values; j is the index of the input pit parameter vector.

[0031] S203. Combining the backpropagation algorithm and the least squares method, the weights from the hidden layer to the output layer are obtained:

[0032]

[0033] Where w is the weight from the hidden layer to the output layer; m is the number of hidden layer nodes; X P This is the P-th vector input to the roof insulator recess parameter input layer; c maxThe maximum distance between the centers of the pit parameter sample;

[0034] S204. Based on the weights from the hidden layer to the output layer, the variance of the radial basis function, and the center of the radial basis function, the radial basis neural network model is obtained:

[0035]

[0036] Where ψ(·) is the radial basis function neural network model; q is the total number of samples; w i The hidden layer to output layer weights are the parameter vectors of the i-th pit.

[0037] The beneficial effect of the above-mentioned further scheme is that the establishment of the radial basis neural network model prepares for obtaining the optimal parameters of the concave surface.

[0038] Furthermore, the expression for the loss function of the radial basis function neural network model in step S204 is as follows:

[0039]

[0040] Among them, R 2 y is the loss function for the radial basis function neural network model; i This is the calculated value of the insulator resistance coefficient; This is the predicted value of the insulator resistance coefficient; y i This is the average value of the calculated insulator resistance coefficient.

[0041] The beneficial effects of the above-mentioned further scheme are: the loss function can be used to verify the fitting accuracy of the calculated results of the roof insulator resistance coefficient, thereby improving the accuracy of the calculated results of the insulator resistance coefficient.

[0042] Further, step S3 specifically includes:

[0043] S301. Based on the radial basis function neural network model, obtain the value range of the pit parameter variable, and obtain the initial population according to the value range of the pit parameter variable;

[0044] S302. Divide the initial population into several subgroups to obtain several islands;

[0045] S303. Calculate the fitness value of the pit parameter vector in each island;

[0046] S304. Determine whether the number of iterations has reached the set value. If so, take the pit parameter vector with the largest fitness value as the optimal parameters of the concave surface. Otherwise, proceed to step S305.

[0047] S305. Based on the pit parameter vectors in each island, use a genetic algorithm to calculate a new initial population and return to step S302.

[0048] The beneficial effects of the above-mentioned further scheme are: the multi-island genetic algorithm maintains the diversity of the optimal solutions for the pit parameters, and can better find the globally optimal combination of pit parameters for the roof insulator.

[0049] Furthermore, in step S4, the concave non-smooth structure is elliptical and arranged laterally on the umbrella skirt surface along the airflow direction; the major axis of the concave surface is consistent with the airflow direction.

[0050] The beneficial effects of the above-mentioned further scheme are as follows: Constructing a concave non-smooth structure on the surface of the insulator optimizes the flow field distribution on the surface of the insulator, thereby reducing the pressure difference drag of the insulator and the viscous drag of the shed surface, and ultimately achieving optimized aerodynamic drag reduction of the roof insulator; at the same time, the concave surface of the insulator increases the creepage distance, increases the flashover voltage of the insulator, and improves the insulation performance. Attached Figure Description

[0051] Figure 1 This is a flowchart of the method of the present invention.

[0052] Figure 2 This is a full view of the roof insulator structure and a schematic diagram of the concave surface structure design of the umbrella skirt in this invention. Detailed Implementation

[0053] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0054] Example 1

[0055] like Figure 1 As shown, in one embodiment of the present invention, a method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow includes the following steps:

[0056] S1. Based on the modeling objectives and the initial sample set, establish an approximate model of the roof insulator;

[0057] S2. Based on the approximate model of the roof insulator, establish a radial basis function neural network model;

[0058] S3. Based on the radial basis function neural network model, the optimal parameters of the concave surface are calculated using the multi-island genetic algorithm;

[0059] S4. Based on the optimal parameters of the concave surface, a non-smooth concave structure is constructed on the upper and lower surfaces of the roof insulator skirt to achieve coordinated optimization of aerodynamic drag reduction and insulation performance of the roof insulator under high-speed airflow.

[0060] Step S1 specifically involves:

[0061] S101. Based on the modeling objective and the anti-interference conditions between the insulator boundary layer thickness, skirt size, airflow velocity, and concave unit cells, the range of values ​​for the pit parameters is obtained.

[0062] S102. Based on the range of values ​​for the pit parameters, an initial sample set is obtained using the Latin square sampling method.

[0063] S103. Based on the modeling objective and the initial sample set, obtain the drag coefficient of each initial sample point;

[0064] S104. Using the initial sample set and the resistance coefficient of each initial sample point, obtain the initial approximate model of the roof insulator;

[0065] S105. Based on the initial approximate model of the roof insulator, obtain the accuracy evaluation index data;

[0066] S106. Determine whether the accuracy evaluation index data meets the set threshold. If yes, obtain the approximate model of the roof insulator; otherwise, proceed to step S107. The expression for the objective function of the approximate model of the roof insulator is:

[0067] L c =Min(f(a,b,h,l1,l2))

[0068] Among them, L c is the objective function of the approximate model of the roof insulator; Min(·) is the minimum value function; f(·) is the function for calculating the drag coefficient of the roof insulator; a is the horizontal axis length of the non-smooth elliptical concave structure of the insulator skirt; b is the vertical axis length of the non-smooth elliptical concave structure of the insulator skirt; h is the depth of the non-smooth elliptical concave structure of the insulator skirt; l1 is the horizontal spacing of the non-smooth elliptical concave structure of the insulator skirt; l2 is the vertical spacing of the non-smooth elliptical concave structure of the insulator skirt.

[0069] S107. Add parameter sample points to the initial sample set based on the addition criterion, and return to step S103; the expression for the addition criterion is:

[0070]

[0071] Where E[I(x)] is the point addition criterion; I(x) is the newly added sample point; x is the pit parameter vector; G min This represents the minimum drag coefficient of the initial approximate model of the roof insulator; φ(·) represents the predicted response value of the roof insulator's drag coefficient; φ(·) represents the standard normal distribution function related to the insulator's drag coefficient. δ(x) is the normal probability density function; δ(x) is the mean square error of the model at the pit parameter vector x.

[0072] Step S2 specifically involves:

[0073] S201. Based on the Gaussian radial basis function, the radial basis function is obtained:

[0074]

[0075] Where ρ(·) is the radial basis function; x is the input layer vector; e is the logarithmic constant; β i Let c be the input parameter vector of the i-th pit; i is the index of the pit parameter vector; ||·|| is the Euclidean norm; i The radial basis function center represents the Gaussian function center vector of the pit parameter;

[0076] S202. Based on the radial basis functions, obtain the variance of the radial basis functions:

[0077]

[0078] Where σ is the variance of the radial basis functions; a j This represents the expected output value of the roof insulator resistance coefficient under the pit parameter sample; y j This is the output of the j-th pit parameter vector; p is the number of input samples; n is a constant value for the sample values; j is the index of the input pit parameter vector.

[0079] S203. Combining the backpropagation algorithm and the least squares method, the weights from the hidden layer to the output layer are obtained:

[0080]

[0081] Where w is the weight from the hidden layer to the output layer; m is the number of hidden layer nodes; X P This is the P-th vector input to the roof insulator recess parameter input layer; c max The maximum distance between the centers of the pit parameter sample;

[0082] S204. Based on the weights from the hidden layer to the output layer, the variance of the radial basis function, and the center of the radial basis function, the radial basis neural network model is obtained:

[0083]

[0084] Where ψ(·) is the radial basis function neural network model; q is the total number of samples; w i The hidden layer to output layer weights are the parameter vectors of the i-th pit.

[0085] The expression for the loss function of the radial basis function neural network model in step S204 is as follows:

[0086]

[0087] Among them, R 2 y is the loss function for the radial basis function neural network model; i This is the calculated value of the insulator resistance coefficient; This is the predicted value of the insulator resistance coefficient; This is the average value of the calculated insulator resistance coefficient.

[0088] Step S3 specifically involves:

[0089] S301. Based on the radial basis function neural network model, obtain the value range of the pit parameter variable, and obtain the initial population according to the value range of the pit parameter variable;

[0090] S302. Divide the initial population into several subgroups to obtain several islands;

[0091] S303. Calculate the fitness value of the pit parameter vector in each island;

[0092] S304. Determine whether the number of iterations has reached the set value. If so, take the pit parameter vector with the largest fitness value as the optimal parameters of the concave surface. Otherwise, proceed to step S305.

[0093] S305. Based on the pit parameter vectors in each island, use a genetic algorithm to calculate a new initial population and return to step S302.

[0094] In step S4, the concave non-smooth structure is elliptical and arranged laterally on the umbrella skirt surface along the airflow direction; the major axis of the concave surface is consistent with the airflow direction.

[0095] Example 2

[0096] A method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow involves constructing concave non-smooth structures on the upper and lower surfaces of the roof insulator skirts without affecting the creepage distance of the insulator.

[0097] like Figure 2As shown, the surface profile of the concave non-smooth structure is elliptical. The concave non-smooth structure is arranged laterally along the airflow direction on the surface of the umbrella skirt, and the major axis of the ellipse is consistent with the airflow direction. Taking the arrangement spacing (l1, l2), the horizontal and vertical axes (a, b), and the depth h of the concave non-smooth structure as variables, and the minimum drag coefficient as the objective, the response values ​​of the concave non-smooth structure and other related parameters are calculated by selecting samples through the Latin square sampling method. Then, an approximate model of the roof insulator is established. Finally, the optimal parameters of the concave surface are obtained by global optimization using a multi-island genetic algorithm.

[0098] In this embodiment, the final arrangement parameters of the elliptical concave non-smooth structure of the roof insulator skirt are l1 = 3.7 mm and l2 = 4.5 mm; the horizontal axis length a = 5.8 mm, the vertical axis length b = 3.4 mm, and the depth h = 0.8 mm of the elliptical concave non-smooth structure of the insulator skirt.

[0099] In this embodiment, the roof insulator can achieve a drag reduction effect of 19.62% under high-speed airflow, and the surface discharge flashover voltage under high-speed airflow is higher than that of the smooth surface insulator. This can achieve synergistic optimization of the aerodynamic drag reduction and insulation performance of the train roof insulator under high-speed airflow.

[0100] In this embodiment, the aerodynamic drag on the roof insulator consists of two parts: pressure difference drag caused by the pressure difference between the front and rear sides due to airflow separation on the insulator surface; and frictional drag caused by the viscosity of the gas and friction with the insulator surface. Adding a non-smooth surface structure to the insulator skirt can reduce the loss of turbulent kinetic energy and thus reduce drag by controlling the skirt boundary layer, but the drag reduction effect should be achieved by ensuring that the depth is no greater than the boundary layer thickness.

[0101] Calculation of Reynolds number and boundary layer thickness for flow field of roof insulator:

[0102] Re(l) = vl / ν

[0103] δ(l) = 0.035l / Re(l) 1 / 7

[0104] In the formula: v is the incoming flow velocity, taken as 80 m / s; l is the average diameter of the umbrella skirt, taken as 0.18 m; ν is the kinematic viscosity coefficient, taken as 0.0722 m. 2 / s. The calculated maximum depth of the concave surface should not exceed 1.24mm.

[0105] In aerodynamics, the effectiveness of drag reduction is typically evaluated using the drag reduction ratio ΔC. D / C D To evaluate:

[0106] ΔC D / CD =(C D1 -C D2 ) / C D2

[0107] In the formula: C D1 To add the aerodynamic drag coefficient before the concave face; C D2 The aerodynamic drag coefficient after adding the concave surface.

[0108] Roof insulators play a crucial role in electrical insulation. A simple streamlined design may reduce the creepage distance, thereby lowering the insulation performance.

[0109] In this embodiment, the good electrical performance of epoxy resin roof insulators under high-speed airflow conditions is predicated on maintaining stable mechanical strength, i.e., the wind pressure resistance of the skirts, preventing them from being lifted by the airflow and causing bending deformation. Therefore, when designing the concave parameters, it is necessary to consider the insulator's ability to withstand mechanical bending loads. An unreasonable design will affect its local mechanical strength and hardness. According to standard TB / T 3077-2017, the maximum depth of the skirt surface of the composite roof insulator should not exceed 1mm. Considering the insulator boundary layer thickness, skirt size, airflow velocity, and anti-interference conditions between concave unit bodies, the range of values ​​for the concave parameters l1, l2, a, b, and h with drag reduction effect is determined by preliminary calculations to be [2.0, 5.0]mm, [2.0, 5.0]mm, [3.0, 6.5]mm, [3.0, 6.5]mm, and [0.3, 1.0]mm.

[0110] Using concave parameters l1, l2, a, b, and h as experimental variables, the minimum aerodynamic drag, i.e., the minimum drag coefficient C of the insulator, is determined. D To optimize the target and obtain the best combination of parameters for drag reduction, the following optimization process is adopted: (1) Determine the range of values ​​for the optimization parameters and select multiple sample points using the Latin square sampling method; (2) Calculate the response values ​​of the sample points using numerical simulation and construct an approximate model of the roof insulator; (3) Select several new sample points to verify the accuracy of the approximate model of the roof insulation; (4) Under the premise of ensuring the reliability of the accuracy of the approximate model of the roof insulation, use the optimization algorithm to globally search for the optimal parameter design of the concave surface; (5) Substitute the design parameters back into the simulation model to verify the accuracy of the search results.

[0111] This invention employs a multiple island genetic algorithm (MIGA) to optimize the concave surface parameters. Based on the constraints described above, the minimum drag coefficient C of the roof insulator is used as the optimization criterion. DTo achieve the desired result, a multi-island genetic algorithm was used for global optimization of the approximate model. The initial population size was set to 60, the number of islands to 20, and the maximum number of iterations to 100. The optimal concave surface parameters for the insulator skirt were finally obtained as follows: l1 = 3.7 mm, l2 = 4.5 mm, a = 5.8 mm, b = 3.4 mm, and h = 0.8 mm. The optimal solution was verified using a simulation model. Adding the optimal concave surface to the roof insulator skirt achieved a maximum drag reduction rate of 15.62%. Furthermore, within the airflow velocity range of 0–90 m / s, the flashover discharge voltage of the optimized insulator was higher than that of the insulator with a smooth surface, indicating a significant improvement in insulation performance.

Claims

1. A method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow, characterized in that, Includes the following steps: S1. Based on the modeling objective and the initial sample set, establish an approximate model of the roof insulator; step S1 specifically includes: S101. Based on the modeling objective and the anti-interference conditions between the insulator boundary layer thickness, skirt size, airflow velocity, and concave unit cells, the range of values ​​for the pit parameters is obtained. S102. Based on the range of values ​​for the pit parameters, an initial sample set is obtained using the Latin square sampling method. S103. Based on the modeling objective and the initial sample set, obtain the drag coefficient of each initial sample point; S104. Using the initial sample set and the resistance coefficient of each initial sample point, obtain the initial approximate model of the roof insulator; S105. Based on the initial approximate model of the roof insulator, obtain the accuracy evaluation index data; S106. Determine whether the accuracy evaluation index data meets the set threshold. If yes, obtain the approximate model of the roof insulator; otherwise, proceed to step S107. The expression for the objective function of the approximate model of the roof insulator is: in, The objective function is the approximate model of the roof insulator. It is a minimum value function; A function for calculating the resistance coefficient of roof insulators; a The horizontal axis length of the non-smooth elliptical concave structure of the insulator skirt; b The length of the longitudinal axis of the non-smooth elliptical concave structure of the insulator skirt; h The depth of the non-smooth structure of the elliptical concave surface of the insulator skirt; l 1 represents the lateral spacing of the non-smooth, elliptical concave structure of the insulator skirt; l 2 represents the longitudinal spacing of the elliptical concave non-smooth structure of the insulator skirt; S107. Add parameter sample points to the initial sample set based on the addition criterion, and return to step S103; the expression for the addition criterion is: in, For the addition of points criteria; For newly added sample points; This is the pit parameter vector; This represents the minimum drag coefficient of the initial approximate model of the roof insulator; This is the predicted response value of the roof insulator resistance coefficient; This is the standard normal distribution function related to the insulator resistance coefficient; It is the probability density function of a normal distribution; Let x be the mean square error of the model at the pit parameter vector. S2. Based on the approximate model of the roof insulator, establish a radial basis function neural network model; S3. Based on the radial basis function neural network model, the optimal parameters of the concave surface are calculated using the multi-island genetic algorithm; S4. Based on the optimal parameters of the concave surface, a non-smooth concave structure is constructed on the upper and lower surfaces of the roof insulator skirt to achieve coordinated optimization of aerodynamic drag reduction and insulation performance of the roof insulator under high-speed airflow.

2. The method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow as described in claim 1, characterized in that, Step S2 specifically involves: S201. Based on the Gaussian radial basis function, the radial basis function is obtained: in, These are radial basis functions; The input layer vector; It is a logarithmic constant; Let i be the input parameter vector for the i-th pit; The index of the pit parameter vector; It is the Euclidean norm; The radial basis function center represents the Gaussian function center vector of the pit parameter; S202. Based on the radial basis functions, obtain the variance of the radial basis functions: in, The variance of the radial basis functions; This represents the expected output value of the roof insulator resistance coefficient under the pit parameter sample. This is the output of the parameter vector for the j-th pit; The number of input samples; The sample numerical values ​​are constants; Input the pit parameter vector index; S203. Combining the backpropagation algorithm and the least squares method, the weights from the hidden layer to the output layer are obtained: in, Weights from hidden layer to output layer; This represents the number of hidden layer nodes. The P-th vector input to the parameter input layer for the roof insulator recess; The maximum distance between the centers of the pit parameter sample; S204. Based on the weights from the hidden layer to the output layer, the variance of the radial basis function, and the center of the radial basis function, the radial basis neural network model is obtained: in, It is a radial basis function neural network model; The total number of samples; The hidden layer to output layer weights are the parameter vectors of the i-th pit.

3. The method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow as described in claim 2, is characterized in that... The expression for the loss function of the radial basis function neural network model in step S204 is as follows: in, This is the loss function for the radial basis function neural network model; This is the calculated value of the insulator resistance coefficient; This is the predicted value of the insulator resistance coefficient; This is the average value of the calculated insulator resistance coefficient.

4. The method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow as described in claim 1, characterized in that, Step S3 specifically involves: S301. Based on the radial basis function neural network model, obtain the value range of the pit parameter variable, and obtain the initial population according to the value range of the pit parameter variable; S302. Divide the initial population into several subgroups to obtain several islands; S303. Calculate the fitness value of the pit parameter vector in each island; S304. Determine whether the number of iterations has reached the set value. If so, take the pit parameter vector with the largest fitness value as the optimal parameters of the concave surface. Otherwise, proceed to step S305. S305. Based on the pit parameter vectors in each island, use a genetic algorithm to calculate a new initial population and return to step S302.

5. The method for synergistic optimization of aerodynamic drag reduction and insulation performance of roof insulators under high-speed airflow as described in claim 1, characterized in that, In step S4, the concave non-smooth structure is elliptical and arranged laterally on the umbrella skirt surface along the airflow direction; the major axis of the concave surface is consistent with the airflow direction.

Images