Aircraft aerodynamic layout multi-objective optimization system supported by adaptive genetic algorithm

Abstract

The invention relates to the field of aircraft design, in particular to an aircraft aerodynamic layout multi-objective optimization system supported by an adaptive genetic algorithm, and the system converts aircraft parameters into binary chromosomes through a data input module, including wing leading edge sweepback, wing root and tip wingspan length positioning parameters and skin thickness parameters; the optimization region determination module and the adaptive division module determine boundary points of an optimization region according to the parameters, and perform adaptive division according to spatial distribution to obtain population initialization parameters, and the fitness calculation module obtains fitness values according to the parameters, calculates crowding distances of individuals, and calculates the crowding distances of the individuals. The chromosome selection module executes a selection operation based on the fitness value, executes an improved global elite individual selection operation in a child population in combination with a crowding distance, and performs crossover variation on a parent population of a non-maximum algebra to generate a child population meeting population quantity maintenance, so that the coding precision and the calculation efficiency are improved; and the multi-objective optimization of the aerodynamic layout of the aircraft is realized.

Description

Technical Field

[0001] The present invention relates to the field of aircraft design, and in particular to a multi-objective optimization system for aircraft aerodynamic layout supported by an adaptive genetic algorithm, which is suitable for the optimal design of aircraft aerodynamic layout in aerospace engineering. Background Art

[0002] Aircraft aerodynamic design is a key technology in the aerospace field. Traditional aerodynamic design relies primarily on the designer's experience and repeated wind tunnel testing, which is time-consuming, costly, and inefficient. With the advancement of computer technology, numerical simulation and optimization algorithms have gradually been applied to aerodynamic design, but existing technologies still have many shortcomings.

[0003] Currently commonly used genetic algorithms often suffer from slow convergence and a tendency to get stuck in local optimal solutions when solving complex, multi-objective optimization problems for aircraft aerodynamic layouts. Standard genetic algorithms use fixed parameter settings, making them difficult to adapt to dynamic changes during the optimization process. Furthermore, their single population structure makes it difficult to strike a balance between global search and local optimization. Furthermore, traditional methods oversimplify the objective function in multi-objective optimization problems, making it difficult to accurately evaluate the overall performance of an aircraft.

[0004] Therefore, there is an urgent need for an optimization system that can adaptively adjust parameters, balance global search and local optimization, and accurately evaluate multi-objective functions to improve the efficiency and quality of aircraft aerodynamic layout design. Summary of the Invention

[0005] The purpose of the present invention is to provide an aircraft aerodynamic layout multi-objective optimization system supported by an adaptive genetic algorithm, which improves the efficiency and quality of aircraft aerodynamic layout design through the adaptive performance and multi-objective evaluation capabilities of the optimization algorithm.

[0006] The present invention proposes a multi-objective optimization system for aircraft aerodynamic layout supported by an adaptive genetic algorithm, comprising:

[0007] a data input module for converting aircraft parameters into binary chromosomes, wherein the aircraft parameters include a wing leading edge sweep angle, a wing root span length positioning parameter, a wing tip span length positioning parameter, and a skin thickness parameter;

[0008] an optimization region determination module, connected to the data input module, and configured to determine the boundary points of the optimization region according to the output of the data input module;

[0009] An optimization region adaptive division module, connected to the optimization region determination module, is used to adaptively divide the boundary points of the optimization region according to spatial distribution and obtain population initialization parameters, including the number of subpopulation individuals and the number of parent population individuals;

[0010] A fitness calculation module, connected to the optimization region adaptive division module, is used to obtain a fitness value according to the population initialization parameters and calculate the crowding distance of the individual;

[0011] A chromosome selection module is connected to the fitness calculation module, and is used to perform a selection operation based on the fitness value, and perform an improved global elite individual selection operation in the offspring population in combination with the crowding distance, and perform crossover mutation on the parent population of non-maximum generation to generate a sub-population that meets the population size maintenance requirement.

[0012] Preferably, the chromosome selection module includes:

[0013] a data reading unit, configured to read the calculation results of the fitness calculation module, sort the fitness values ​​in descending order, select the fitness values ​​corresponding to the first M fitness values ​​as individuals of the subpopulation, where M is the number of individuals in the subpopulation, and then remove the first M fitness values ​​from all fitness values ​​and select fitness values ​​equal to the number of individuals in the parent population as individuals of the parent population;

[0014] a subpopulation evolution unit connected to the data reading unit, and configured to perform an improved global elite individual selection operation on the output of the data reading unit, wherein the improved global elite individual selection operation comprises: randomly generating a random number Random(0,1); if Random is greater than a preset probability threshold, the subpopulation performs a normal selection operator operation; if Random is less than the preset probability threshold, the subpopulation replaces the child individual with a higher fitness with the parent individual;

[0015] The parent population evolution unit is connected to the child population evolution unit and is used to execute the evolution operation of the parent population to generate a child population that meets the population quantity maintenance requirement.

[0016] Preferably, the system further comprises:

[0017] a first fuzzy neural network module, connected to the fitness calculation module, for obtaining optimized aircraft parameters using flight efficiency, flight stability, maneuverability, and stealth capability as input data and aircraft parameters as output data;

[0018] Among them, the flight efficiency is the sum of the aircraft's flight resistance, the aircraft's lift during cruising, and the aircraft's total lift-to-drag ratio; the flight stability is the stability of the aircraft's flight altitude; the maneuverability is the stability of the aircraft's maximum level flight speed and the launch platform's level flight speed; the stealth capability is the sum of the aircraft's fuselage area and the maximum length from the nose to the tail of the aircraft during cruising.

[0019] Preferably, the system further comprises:

[0020] The second fuzzy neural network module is connected to the first fuzzy neural network module and is used to obtain optimized flight resistance using the aircraft parameters as input data and the flight resistance as output data.

[0021] Preferably, the first fuzzy neural network module comprises:

[0022] An input layer, configured to take the input variables consisting of the flight efficiency, flight stability, maneuverability, and stealth capability as input data;

[0023] A fuzzification layer connected to the input layer, configured to fuzzify the input variables according to the fuzzy sets of the input variables and corresponding membership functions to obtain fuzzified results;

[0024] a fuzzy reasoning layer, connected to the fuzzification layer, for performing fuzzy reasoning on the fuzzification result to obtain a reasoning result;

[0025] The defuzzification layer is connected to the fuzzy reasoning layer and is used to perform defuzzification processing on the reasoning result to obtain the optimized aircraft parameters.

[0026] Preferably, the second fuzzy neural network module comprises:

[0027] An input layer, configured to take the aircraft parameters as input variables;

[0028] A fuzzification layer connected to the input layer, configured to fuzzify the input variables according to the fuzzy sets of the input variables and corresponding membership functions to obtain fuzzified results;

[0029] a fuzzy reasoning layer, connected to the fuzzification layer, for performing fuzzy reasoning on the fuzzification result to obtain a reasoning result;

[0030] The defuzzification layer is connected to the fuzzy reasoning layer and is used to perform defuzzification processing on the reasoning result to obtain the optimized flight resistance.

[0031] Preferably, when the Random is greater than 0.6, the sub-population performs a common selection operator operation, and when the Random is less than 0.6, the sub-population replaces the offspring individuals with low fitness with the parent individuals with high fitness.

[0032] Preferably, the optimization area determination module is further configured to:

[0033] Determining the maximum and minimum values ​​of the optimization region according to the maximum and minimum values ​​of the aircraft parameters, wherein the optimization region is a feasible domain of the aircraft parameters, wherein the maximum and minimum values ​​of the optimization region are in a preset multiple relationship with the maximum and minimum values ​​of the aircraft parameters;

[0034] Among them, when the number of the aircraft parameters is 7, the preset multiple is 15; when the number of the aircraft parameters is 8, the preset multiple is 20.

[0035] Preferably, the system further comprises:

[0036] The aerodynamic analysis module is used to calculate the aerodynamic performance of the current aircraft aerodynamic shape and obtain the data information required for aerodynamic design and aerodynamic shape optimization, including angle of attack, sideslip angle, Mach number, Reynolds number, angle of attack, aircraft flight attitude and aircraft shape coordinates;

[0037] The aircraft layout optimization module is connected to the aerodynamic analysis module and the data input module, and is used to provide the data input module with an initial population of genetic algorithms, including the generation of the initial aircraft shape, the description of the aircraft parameters, the determination of the aircraft shape design variables, the determination of the aircraft optimization target, the construction of the initial optimization variable group, and the determination of the aerodynamic layout optimization constraint conditions.

[0038] Preferably, the system further comprises:

[0039] The post-optimization simulation calculation module is connected to the data input module and the aircraft layout optimization module, and is used to simulate and calculate the aerodynamic performance of the aircraft according to the optimized aircraft configuration, including:

[0040] Configuration data management unit, used for parameter adjustment based on existing grids;

[0041] a geometry description file generating unit, connected to the configuration data management unit, for generating a grid description file according to the aircraft configuration data;

[0042] A computational grid generation unit, connected to the geometry description file generation unit, for inputting the generated grid description file into a CFD calculation tool to generate a computational grid;

[0043] The error analysis unit is used to compare the performance of key components in the aerodynamic performance results of the optimized aircraft configuration with expected values ​​to obtain an error value, and output the optimized aircraft configuration if the error value meets the requirements.

[0044] The present invention has the following beneficial effects:

[0045] 1. Through the adaptive chromosome encoding and decoding mechanism, the aircraft parameters are accurately mapped to the genetic algorithm chromosomes, improving the encoding accuracy and computing efficiency;

[0046] 2. A dynamically adjusted secondary population structure is used to balance global search and local optimization, preventing the algorithm from falling into a local optimal solution;

[0047] 3. Design an elite retention strategy based on probability thresholds to ensure that high-quality genes are not lost during the evolution process and improve the convergence performance of the algorithm;

[0048] 4. Introducing a fuzzy neural network evaluation engine to accurately evaluate the multi-dimensional performance indicators of the aircraft and achieve multi-objective collaborative optimization;

[0049] 5. Implement adaptive control of crossover and mutation operations, intelligently select the most suitable operator according to the optimization stage, and improve algorithm efficiency;

[0050] 6. The overall system achieves efficient multi-objective optimization of the aircraft's aerodynamic layout. Compared with traditional methods, the convergence speed is increased by 30% to 50%, and the distribution of Pareto solutions is more uniform. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is the overall architecture diagram of the aircraft aerodynamic layout multi-objective optimization system supported by the adaptive genetic algorithm of the present invention;

[0052] Figure 2 Schematic diagram of the structure of the chromosome selection module of the present invention;

[0053] Figure 3 Schematic diagram of the structure of the fuzzy neural network module of the present invention;

[0054] Figure 4 Schematic diagram of the secondary population structure in the present invention;

[0055] Figure 5 This is a flowchart of the improved global elite individual selection operation in the present invention;

[0056] Figure 6 This is a flow chart for optimizing the aerodynamic layout in the present invention. DETAILED DESCRIPTION

[0057] Please refer to the attached Figure 1-6 , the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0058] Reference Figure 1The multi-objective optimization system for aircraft aerodynamic layout supported by the adaptive genetic algorithm provided by the present invention includes a data input module 1, an optimization region determination module 2, an optimization region adaptive division module 3, a fitness calculation module 4, a chromosome selection module 5, a first fuzzy neural network module 6, a second fuzzy neural network module 7, an aerodynamic analysis module 8, an aircraft layout optimization module 9 and a post-optimization simulation calculation module 10.

[0059] In one embodiment of the present invention, the data input module 1 is used to convert aircraft parameters into binary chromosomes. Specifically, the aircraft parameters include the wing leading edge sweep angle, the wing root span length positioning parameter, the wing tip span length positioning parameter, and the skin thickness parameter. The data input module 1 first normalizes these parameters and maps the parameters of different dimensions to the [0,1] interval. Then, according to the importance and accuracy requirements of the parameters, an appropriate number of coding bits is assigned to each parameter. For example, for the key parameter of the wing leading edge sweep angle, an 8-bit binary code can be assigned, providing 256 discrete values ​​to meet the accuracy requirement of 0.1 degrees; and for the secondary skin thickness parameter, a 6-bit binary code can be assigned, providing 64 discrete values.

[0060] Optimization Region Determination Module 2 is connected to Data Input Module 1 and is used to determine the boundary points of the optimization region based on the output of Data Input Module 1. In actual applications, Optimization Region Determination Module 2 determines the maximum and minimum values ​​of the optimization region based on the maximum and minimum values ​​of the aircraft parameters. The optimization region represents the feasible domain of the aircraft parameters, and its boundaries are a multiple of the actual boundaries of the aircraft parameters to ensure that the search space is sufficiently large but not so large that it wastes computing resources.

[0061] Optimization Region Adaptive Partitioning Module 3 is connected to Optimization Region Determination Module 2 and is used to adaptively divide the optimization region's boundary points according to their spatial distribution and obtain population initialization parameters, including the number of individuals in the subpopulation and parent population. This module adaptively determines the population structure by analyzing the spatial distribution characteristics of the optimization region, providing a foundation for subsequent optimization processes.

[0062] The fitness calculation module 4 is connected to the optimization region adaptive partitioning module 3 and is used to obtain the fitness value based on the population initialization parameters and calculate the crowding distance of the individual. This module is the core of the individual evaluation and provides a basis for selection by calculating the fitness value and crowding distance.

[0063] The chromosome selection module 5 is connected to the fitness calculation module 4 and is used to perform selection operations based on fitness values. It also uses crowding distance to perform improved global elite individual selection operations in the offspring population. This module performs crossover and mutation on the parent population of non-maximum generations to produce subpopulations that meet population size maintenance requirements. This module is key to achieving population evolution, continuously improving population quality through operations such as selection, crossover, and mutation.

[0064] Reference Figure 2 The chromosome selection module 5 includes a data reading unit 51 , a child population evolution unit 52 and a parent population evolution unit 53 .

[0065] The data reading unit 51 is used to read the calculation results of the fitness calculation module 4, sort the fitness values ​​in descending order, select the fitness values ​​corresponding to the first M fitness values ​​as the individuals of the sub-population, where M is the number of individuals in the sub-population, and then remove the first M fitness values ​​from all fitness values. After that, the fitness values ​​equal to the number of individuals in the parent population are selected as the individuals of the parent population. This two-level population structure design effectively balances the relationship between global search and local optimization, improving the search efficiency of the algorithm.

[0066] Preferably, the number of individuals in the subpopulation, M, can be determined based on the problem dimension, typically set to 1.5 to 2 times the problem dimension. For example, for an optimization problem involving seven aircraft parameters, the number of individuals in the subpopulation can be set to 10-14. The number of individuals in the parent population is typically set to 2 to 3 times the number of individuals in the subpopulation to provide sufficient diversity support.

[0067] The subpopulation evolution unit 52 is connected to the data reading unit 51 and is used to perform an improved global elite individual selection operation on the output of the data reading unit 51. Specifically, the operation includes: randomly generating a random number Random(0,1); if Random is greater than a preset probability threshold, the subpopulation performs a normal selection operator operation; if Random is less than the preset probability threshold, the subpopulation replaces the low-fitness offspring individuals with the high-fitness parent individuals.

[0068] In one embodiment of the present invention, the preset probability threshold is set at 0.6. When Random is greater than 0.6, the subpopulation performs a standard selection operator, such as roulette wheel selection or tournament selection. When Random is less than 0.6, the subpopulation replaces low-fitness offspring individuals with high-fitness parent individuals. This threshold setting balances randomness and determinism, maintaining population diversity while ensuring the retention of high-quality genes.

[0069] The parent population evolution unit 53 is connected to the child population evolution unit 52 and is used to perform the evolution operation of the parent population to generate child populations that meet the population size maintenance requirements. The parent population mainly maintains the diversity of the population and generates new individuals through conventional crossover and mutation operations, providing a wider search space for the child population.

[0070] Reference Figure 3 The system of the present invention further includes a first fuzzy neural network module 6 and a second fuzzy neural network module 7 for accurately evaluating the multi-dimensional performance indicators of the aircraft.

[0071] The first fuzzy neural network module 6 is connected to the fitness calculation module 4 and is used to obtain optimized aircraft parameters using flight efficiency, flight stability, maneuverability, and stealth capability as input data and aircraft parameters as output data. The specific definitions of these four performance indicators are as follows:

[0072] Flight efficiency is the sum of the aircraft's flight drag, the aircraft's lift during cruising, and the aircraft's total lift-to-drag ratio, which can be expressed as:

[0073] ,

[0074] in: It is a flight efficiency evaluation index, a dimensionless value; is the lift-to-drag ratio, which represents the lift generated per unit drag, a dimensionless value. The lift-to-weight ratio is the ratio of the lift generated by the aircraft to its own weight, a dimensionless value; 、 is the weight coefficient, satisfying , usually desirable , to highlight the leading role of lift-to-drag ratio in flight efficiency evaluation. The drag is in the denominator, which correctly reflects the relationship that the lower the drag, the higher the efficiency. Two key dimensionless ratios (lift-to-drag ratio and lift-to-weight ratio) are used as evaluation indicators, with two weight coefficients. In actual application, both ratio indicators will be normalized so that their value range is unified to the [0,1] interval to ensure the rationality of weight distribution. Flight efficiency evaluation index The value range is [0,1], and the larger the value, the higher the flight efficiency.

[0075] Flight stability is the stability of the aircraft's flight altitude and can be expressed as:

[0076] ,

[0077] in It is a flight stability evaluation index with a dimensionless value and a range of [0,1]. The larger the value, the better the flight stability. is the normalized standard deviation of flight altitude, dimensionless, with a range of is the normalized pitch angle standard deviation, dimensionless, with a range of is the normalized roll angle standard deviation, dimensionless, with a range of are weight coefficients, all dimensionless values, satisfying , usually takes the value By using The formula correctly expresses the negative correlation between standard deviation and stability - the smaller the standard deviation, the higher the stability index. For example, when the standard deviation of the flight altitude is small ( ), the corresponding item , indicating good high stability.

[0078] Maneuverability is the maximum level flight speed of the aircraft and the stability of the launch platform's level flight speed, which can be expressed as:

[0079] ,

[0080] in: is the mobility evaluation index, and its unit is dimensionless value; is the maximum level flight speed in meters per second (m / s), after normalization; is the standard deviation of the launch platform's level flight velocity, in meters per second (m / s), after normalization; is the weight coefficient, which is a dimensionless value and is usually set to 0.7 or 0.3 to balance the influence of maximum speed and speed stability. In actual optimization, if the normalized value of the maximum level flight speed of a high-speed UAV is 0.85 and the normalized value of the standard deviation of the level flight speed of the launch platform is 0.2, then the maneuverability index , indicating that the aircraft has high maneuverability.

[0081] The stealth capability is the sum of the aircraft fuselage area and the maximum length from the nose to the tail of the aircraft during cruising, which can be expressed as:

[0082] ,

[0083] in: It is an evaluation index of stealth capability, and its unit is dimensionless value; is the fuselage area in square meters ( , after normalization; The maximum length from nose to tail during cruise, in meters (m), after normalization; 、 are weight coefficients, which are dimensionless values ​​and are usually set to 0.6 and 0.4 to balance the effects of area and length on stealth performance. In the design of stealth aircraft, smaller fuselage area and length are beneficial to reducing radar cross section. For example, the normalized value of the fuselage area of ​​a stealth drone is 0.3 (the smaller the value, the better the stealth performance), and the normalized value of the maximum length is 0.35. Then the stealth capability index is , indicating that the aircraft has good stealth performance.

[0084] The first fuzzy neural network module 6 includes an input layer 61, a fuzzification layer 62, a fuzzy inference layer 63, and a defuzzification layer 64. The input layer 61 is used to receive input variables consisting of flight efficiency, flight stability, maneuverability, and stealth capability as input data. The fuzzification layer 62 is connected to the input layer 61 and is used to fuzzify the input variables based on their fuzzy sets and corresponding membership functions to obtain fuzzified results. The fuzzy inference layer 63 is connected to the fuzzification layer 62 and is used to perform fuzzy inference on the fuzzified results to obtain inference results. The defuzzification layer 64 is connected to the fuzzy inference layer 63 and is used to defuzzify the inference results to obtain optimized aircraft parameters.

[0085] In the fuzzification process, the input variables are mapped to fuzzy sets through membership functions. For each input variable, five fuzzy control subsets are set, namely positive large, positive medium, medium, negative medium and positive small. For example, its membership function can be expressed as:

[0086] ,

[0087] ,

[0088] ,

[0089] ,

[0090] ,

[0091] in: Represents the membership function of the positive fuzzy set, with a value range of [0,1], indicating The degree of belonging to a positive fuzzy set; It represents the membership function of the positive fuzzy set, and its value range is [0,1]; It represents the membership function of the medium fuzzy set, and its value range is [0,1]; It represents the membership function of the negative middle fuzzy set, and its value range is [0,1]; represents the membership function of positive small fuzzy sets, with a value range of [0,1]; to is the central value of each fuzzy set, which can be 0.9, 0.7, 0.5, 0.3, 0.1 respectively, and the unit is Same, dimensionless value; to The width parameter of each fuzzy set controls the steepness of the membership function and is usually between 0.1 and 0.2. exp represents the natural exponential function.

[0092] For example, for flight efficiency , center value , width parameter , then its membership degree to the central fuzzy set is:

[0093] ,

[0094] This means that the flight efficiency of 0.75 has a high membership of 0.946 to the central fuzzy set.

[0095] Fuzzy reasoning uses the Mamdani inference method, implemented through a series of if-then rules, such as: if flight efficiency is positive and flight stability is medium then the wing leading edge sweep angle is neutral. These rules are determined by expert knowledge or training data.

[0096] Defuzzification uses the center of gravity method to convert the fuzzy reasoning results into clear aircraft parameter values:

[0097] ,

[0098] in: is the output value of the kth aircraft parameter, such as the sweep angle of the wing leading edge, and the unit depends on the specific parameter (such as the angle in degrees (°)); is the activation degree of the i-th rule, that is, the membership combination value of the antecedent part (IF part) of the rule, with a value range of [0,1] and dimensionless; is the output value of the kth parameter corresponding to the i-th rule, and its unit is Same; n is the total number of rules, which is a positive integer. Represents a sum operation, which accumulates all rules from 1 to n.

[0099] This formula is essentially a weighted average calculation, using the activation of each rule as a weight to calculate the final parameter value. For example, when the three rules recommend wing leading edge sweep angles of 25°, 30°, and 35°, respectively, and the corresponding activations are 0.2, 0.6, and 0.3, the final output wing leading edge sweep angle is:

[0100] ,

[0101] The second fuzzy neural network module 7 is connected to the first fuzzy neural network module 6 and is used to obtain optimized flight resistance using aircraft parameters as input data and flight resistance as output data. Its structure is similar to that of the first fuzzy neural network module 6, including an input layer 71, a fuzzification layer 72, a fuzzy inference layer 73, and a defuzzification layer 74.

[0102] Input layer 71 of second fuzzy neural network module 7 is used to take aircraft parameters as input variables. Fuzzification layer 72 is connected to input layer 71 and is used to fuzzify the input variables based on their fuzzy sets and corresponding membership functions, thereby obtaining fuzzified results. Fuzzy inference layer 73 is connected to fuzzification layer 72 and is used to perform fuzzy inference on the fuzzified results to obtain inference results. Defuzzification layer 74 is connected to fuzzy inference layer 73 and is used to defuzzify the inference results to obtain optimized flight resistance.

[0103] These two fuzzy neural network modules work together. The first module converts performance indicators into aircraft parameters, and the second module converts aircraft parameters into flight resistance, forming a complete optimization chain to achieve multi-objective optimization of the aircraft's aerodynamic layout.

[0104] In one embodiment of the present invention, the optimization region determination module 2 is further configured to determine the maximum and minimum values ​​of the optimization region based on the maximum and minimum values ​​of the aircraft parameters. The optimization region is a feasible region of the aircraft parameters, wherein the maximum and minimum values ​​of the optimization region are multiples of the maximum and minimum values ​​of the aircraft parameters.

[0105] Optimization region determination module 2 determines the maximum and minimum values ​​of the optimization region based on the maximum and minimum values ​​of the aircraft parameters. The optimization region represents the feasible domain of the aircraft parameters, and its boundaries are linked to the actual boundaries of the aircraft parameters through an expansion coefficient. Specifically, when the number of aircraft parameters is 7, the expansion coefficient is 15%; when the number of aircraft parameters is 8, the expansion coefficient is 20%. This expansion coefficient setting is based on experience and a large amount of experimental data, and is intended to provide the algorithm with an appropriately sized search space, avoiding too small a value that causes the optimal solution to fall outside the boundaries, and too large a value that causes low computational efficiency.

[0106] The calculation formula for the expanded parameter range is:

[0107] ,

[0108] in: is the expanded parameter range; is the original parameter range; is the expansion coefficient. When the number of parameters is 7, the value is 0.15. When the number of parameters is 8, the value is 0.20.

[0109] For example, for the wing leading edge sweep angle parameter, if its actual range is 15°-45°, when there are 7 parameters, the expanded calculation is:

[0110]

[0111] The final expanded range is 10.5°-49.5°. In engineering practice, this range is typically rounded to an appropriate integer, such as 10°-50°. This moderate expansion ensures that the algorithm can explore potential optimal solutions near the boundaries of the parameter's practical range without excessively expanding the search space and affecting computational efficiency.

[0112] In a specific missile aerodynamic layout optimization case, the original design's wing root span length positioning parameter range was 0.15-0.25 (indicating the relative position of the wing root on the fuselage cross-section). After the optimization region was expanded, the search range was reduced to 0.05-0.35. The final optimization results showed that the optimal value was 0.18, which was within the original design range but close to the lower limit, demonstrating the necessity of pre-setting an appropriate search space expansion.

[0113] In another embodiment of the present invention, the system further includes an aerodynamic analysis module 8 and an aircraft layout optimization module 9 .

[0114] The aerodynamic analysis module 8 calculates the aerodynamic performance of the current aircraft's aerodynamic shape and obtains the data and information required for aerodynamic design and optimization, including angle of attack, sideslip angle, Mach number, Reynolds number, angle of attack, aircraft flight attitude, and aircraft shape coordinates. This data forms the basis for aircraft aerodynamic performance analysis and provides a foundation for subsequent optimization.

[0115] For example, in analyzing the aerodynamic performance of a supersonic aircraft, Aerodynamic Analysis Module 8 first set the cruise angle of attack to a range of -2° to 8°, the sideslip angle to 0°, the Mach number to 1.5, and the Reynolds number to 2×10^7. Then, using computational fluid dynamics (CFD), it calculated the lift coefficient, drag coefficient, and pitching moment coefficient at different angles of attack. The results showed that at an angle of attack of 5°, the aircraft achieved an optimal lift-to-drag ratio of 13.5, providing an important reference for subsequent layout optimization.

[0116] The aircraft layout optimization module 9 is connected to the aerodynamic analysis module 8 and the data input module 1, and is used to provide the genetic algorithm initial population to the data input module 1. This module includes functions such as generating the initial aircraft shape, describing the aircraft parameters, determining the aircraft shape design variables, determining the aircraft optimization target, constructing the initial optimization variable group, and determining the aerodynamic layout optimization constraint conditions.

[0117] During the initial aircraft shape generation process, an initial shape that meets basic aerodynamic requirements is generated based on design requirements and parameter ranges. For a subsonic UAV, for example, the initial shape includes a conventional layout with a fuselage length of 4 meters, a wingspan of 6 meters, a wing leading edge sweep angle of 25 degrees, a wing root chord length of 1.2 meters, and a wingtip chord length of 0.6 meters. The parametric description of the aircraft converts these parameters into optimizable variables, such as representing the wing leading edge sweep angle as a parameter that varies within a range of 15°-35°. The determination of the aircraft shape design variables involves selecting the variables that have the most significant impact on aerodynamic performance, such as the wing leading edge sweep angle, wing root position, and wingtip position in this example.

[0118] Determining the aircraft's optimization objectives involves setting the optimization objective function based on mission requirements. For example, for a long-endurance UAV, the primary optimization objectives might be maximizing lift-to-drag ratio and minimizing fuel consumption; for a high-speed interceptor, the primary optimization objectives might be maximizing maneuverability and minimizing drag coefficient. Constructing the initial optimization variable population generates an initial population that satisfies the constraints, providing the foundation for initiating the genetic algorithm. Determining the constraints for aerodynamic layout optimization involves setting physical and engineering constraints, such as structural strength, center of gravity location, and stability requirements, to ensure the practicality of the optimization results.

[0119] In another embodiment of the present invention, the system further includes an optimized simulation calculation module 10, which is connected to the data input module 1 and the aircraft layout optimization module 9 and is used to simulate and calculate the aircraft aerodynamic performance according to the optimized aircraft configuration.

[0120] The post-optimization simulation calculation module 10 includes a configuration data management unit 101 , a geometry description file generation unit 102 , a calculation grid generation unit 103 and an error analysis unit 104 .

[0121] The configuration data management unit 101 is used to perform parameterized adjustments based on the existing mesh to adapt the mesh to the optimized aircraft configuration. For example, when the wing leading edge sweep angle is optimized from 25° to 30°, the configuration data management unit 101 will adjust the coordinates of the mesh nodes accordingly to maintain the quality and topology of the mesh. The geometry description file generation unit 102 is connected to the configuration data management unit 101 and is used to generate a mesh description file based on the aircraft configuration data in preparation for CFD calculations. In actual applications, a geometry file in a format such as STL or IGES may be generated to describe the surface shape of the aircraft.

[0122] The computational grid generation unit 103 is connected to the geometry description file generation unit 102 and is used to input the generated grid description file into the CFD calculation tool to generate a computational grid. For example, for a complex aircraft shape, it may be necessary to generate an unstructured grid containing millions of grid points to accurately capture the flow field characteristics. This is the basis for high-precision aerodynamic performance calculations. In the optimization case of a certain hypersonic aircraft, the computational grid contains approximately 5 million tetrahedral elements. The grid is encrypted in the boundary layer area, and the first layer of grid height is guaranteed to be y+<1 to accurately simulate the boundary layer flow.

[0123] The error analysis unit 104 compares the performance of key components in the optimized aircraft configuration's aerodynamic performance results with expected values, generating an error value. If the error value meets the requirements, the optimized aircraft configuration is output; if the error value exceeds the allowable range, further optimization is required. This closed-loop verification mechanism ensures the reliability and validity of the optimization results.

[0124] Typically, the tolerance is set to no more than 5% of the desired value. For example, if the desired lift-to-drag ratio is 15, the actual calculated result should be between 14.25 and 15.75 to meet the requirement. This range is based on engineering practice and the accuracy requirements of aircraft design, accounting for calculation errors while meeting the needs of engineering applications.

[0125] In an actual cruise missile aerodynamic layout optimization case, the optimization objective was to maximize the lift-to-drag ratio at a cruise Mach number of 1.5. The initial design had a lift-to-drag ratio of 11.2, and the goal was to achieve a lift-to-drag ratio of over 13.5. After optimization using an adaptive genetic algorithm, the optimal configuration achieved a calculated lift-to-drag ratio of 13.8. Error analysis showed that this result exceeded the expected value by 2.2%, meeting the error requirement of less than 5%, and was therefore adopted as the final design.

[0126] In a specific embodiment of the present invention, the workflow of the system is as follows:

[0127] 1. Data Input Module 1 receives aircraft parameters, including a 30° wing leading edge sweep angle, a 0.2 wing root span length location parameter (indicating the wing root is located at 20% of the fuselage cross-section), a 0.8 wing tip span length location parameter (indicating the wing tip is located at 80% of the span), and a skin thickness parameter (0.01m). It converts these parameters into binary chromosomes. For example, a 30° wing leading edge sweep angle can be encoded as 00011110 (assuming an encoding range of 0°-63.75°, represented using 8 bits of binary, and a resolution of 0.25°).

[0128] 2. Optimization region determination module 2 determines the boundary points of the optimization region based on these parameters. Taking the wing leading edge sweep angle as an example, its range may be set to 0°-60°, which is an expanded version of the original range of 15°-45° (taking into account the preset multiple of 15).

[0129] 3. Optimization Region Adaptive Partitioning Module 3 adaptively divides the boundary points according to their spatial distribution, determining the number of individuals in the subpopulation to be 12 and the number of individuals in the parent population to be 30. In high-dimensional optimization problems, setting a smaller number of individuals in the subpopulation can speed up optimization, while setting a larger number of individuals in the parent population can maintain population diversity.

[0130] 4. The fitness calculation module 4 calculates the fitness value of each individual based on these parameters, and also calculates the crowding distance of the individual to provide a basis for the selection operation. For example, the flight efficiency of an individual is calculated. , flight stability ,flexibility , stealth capability , the comprehensive fitness value is 0.73, and the crowding distance is 0.15.

[0131] 5. Chromosome selection module 5 performs selection based on fitness and crowding distance. In one generation, the randomly generated Random value is 0.4, which is less than the preset threshold of 0.6. Therefore, the subpopulation replaces the low-fitness individuals with the high-fitness parent individuals. For example, the parent individual with a fitness value of 0.85 (wing sweep angle of 32°) replaces the offspring individual with a fitness value of 0.68 (wing sweep angle of 28°).

[0132] 6. The first fuzzy neural network module 6 receives the evaluation results of flight efficiency, flight stability, maneuverability and stealth capability, and outputs the optimized aircraft parameters. For example, flight efficiency , flight stability ,flexibility , stealth capability After fuzzy neural network processing, the output optimization parameters include wing leading edge sweep angle 35°, wing root span length positioning parameter 0.18, wing tip span length positioning parameter 0.85, etc.

[0133] 7. The second fuzzy neural network module 7 receives the optimized aircraft parameters and outputs flight resistance data, such as the drag coefficient under the conditions of Mach number 1.5 and angle of attack 2°. , compared with the initial design A decrease of 21.9%.

[0134] 8. The aerodynamic analysis module 8 calculates the performance of the current aircraft aerodynamic shape and obtains the lift coefficient, drag coefficient and pitch moment coefficient under the conditions of attack angle 5°, sideslip angle 0° and Mach number 0.8, which are respectively and .

[0135] 9. The aircraft layout optimization module 9 performs layout optimization based on these data, generates a new aircraft configuration, and adjusts key parameters such as wing area, airfoil distribution, and sweep angle.

[0136] 10. The post-optimization simulation module 10 simulates and verifies the optimized configuration, analyzing its performance under various flight conditions through CFD calculations. The module calculates the error value, for example, the lift-to-drag ratio error under cruise conditions is 3.2%, which is less than the 5% tolerance. The module then outputs the final optimization result.

[0137] Through the above process, the system of the present invention can efficiently achieve multi-objective optimization of an aircraft's aerodynamic layout, providing strong support for aircraft design. In a design example of a high-speed cruising UAV, the optimization of this system improved flight efficiency by 15%, flight stability by 8%, and maneuverability by 12%, while maintaining the original stealth capability and significantly enhancing the aircraft's overall performance.

[0138] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An adaptive genetic algorithm-based multi-objective optimization system for aircraft aerodynamic layout, characterized by: include: a data input module for converting aircraft parameters into binary chromosomes, wherein the aircraft parameters include a wing leading edge sweep angle, a wing root span length positioning parameter, a wing tip span length positioning parameter, and a skin thickness parameter; an optimization region determination module, connected to the data input module, and configured to determine the boundary points of the optimization region according to the output of the data input module; An optimization region adaptive division module, connected to the optimization region determination module, is used to adaptively divide the boundary points of the optimization region according to spatial distribution and obtain population initialization parameters, including the number of subpopulation individuals and the number of parent population individuals; A fitness calculation module, connected to the optimization region adaptive division module, is used to obtain a fitness value according to the population initialization parameters and calculate the crowding distance of the individual; A chromosome selection module is connected to the fitness calculation module, and is used to perform a selection operation based on the fitness value, and perform an improved global elite individual selection operation in the offspring population in combination with the crowding distance, and perform crossover mutation on the parent population of non-maximum generation to generate a sub-population that meets the population size maintenance requirement.

2. The system according to claim 1, wherein: The chromosome selection module includes: a data reading unit, configured to read the calculation results of the fitness calculation module, sort the fitness values ​​in descending order, select the fitness values ​​corresponding to the first M fitness values ​​as individuals of the subpopulation, where M is the number of individuals in the subpopulation, and then remove the first M fitness values ​​from all fitness values ​​and select fitness values ​​equal to the number of individuals in the parent population as individuals of the parent population; a subpopulation evolution unit connected to the data reading unit, and configured to perform an improved global elite individual selection operation on the output of the data reading unit, wherein the improved global elite individual selection operation includes: randomly generating a random number Random(0,1); if Random is greater than a preset probability threshold, the subpopulation performs a normal selection operator operation; if Random is less than the preset probability threshold, the subpopulation replaces the child individual with a higher fitness with the parent individual; The parent population evolution unit is connected to the child population evolution unit and is used to execute the evolution operation of the parent population to generate a child population that meets the population quantity maintenance requirement.

3. The system according to claim 1, wherein: The system further comprises: a first fuzzy neural network module, connected to the fitness calculation module, for obtaining optimized aircraft parameters using flight efficiency, flight stability, maneuverability, and stealth capability as input data and aircraft parameters as output data; Among them, the flight efficiency is the sum of the aircraft's flight resistance, the aircraft's lift during cruising, and the aircraft's total lift-to-drag ratio; the flight stability is the stability of the aircraft's flight altitude; the maneuverability is the stability of the aircraft's maximum level flight speed and the launch platform's level flight speed; the stealth capability is the sum of the aircraft's fuselage area and the maximum length from the nose to the tail of the aircraft during cruising.

4. The system according to claim 3, characterized in that The system further comprises: The second fuzzy neural network module is connected to the first fuzzy neural network module and is used to obtain optimized flight resistance using the aircraft parameters as input data and the flight resistance as output data.

5. The system according to claim 3, wherein: The first fuzzy neural network module includes: An input layer, configured to take the input variables consisting of the flight efficiency, flight stability, maneuverability, and stealth capability as input data; A fuzzification layer connected to the input layer, configured to fuzzify the input variables according to the fuzzy sets of the input variables and the corresponding membership functions to obtain fuzzified results; A fuzzy reasoning layer, connected to the fuzzification layer, for performing fuzzy reasoning on the fuzzification result to obtain a reasoning result; The defuzzification layer is connected to the fuzzy reasoning layer and is used to perform defuzzification processing on the reasoning result to obtain the optimized aircraft parameters.

6. The system according to claim 4, characterized in that The second fuzzy neural network module includes: An input layer, configured to take the aircraft parameters as input variables; A fuzzification layer connected to the input layer, configured to fuzzify the input variables according to the fuzzy sets of the input variables and the corresponding membership functions to obtain fuzzified results; A fuzzy reasoning layer, connected to the fuzzification layer, for performing fuzzy reasoning on the fuzzification result to obtain a reasoning result; The defuzzification layer is connected to the fuzzy reasoning layer and is used to perform defuzzification processing on the reasoning result to obtain the optimized flight resistance.

7. The system according to claim 2, wherein: When the Random is greater than 0.6, the sub-population performs a common selection operator operation. When the Random is less than 0.6, the sub-population replaces the offspring individuals with low fitness with the parent individuals with high fitness.

8. The system according to claim 1, wherein: The optimization area determination module is further configured to: Determining the maximum and minimum values ​​of the optimization region according to the maximum and minimum values ​​of the aircraft parameters, wherein the optimization region is a feasible domain of the aircraft parameters, wherein the maximum and minimum values ​​of the optimization region are in a preset multiple relationship with the maximum and minimum values ​​of the aircraft parameters; Among them, when the number of the aircraft parameters is 7, the preset multiple is 15; when the number of the aircraft parameters is 8, the preset multiple is 20.

9. The system according to claim 1, wherein: The system further comprises: The aerodynamic analysis module is used to calculate the aerodynamic performance of the current aircraft aerodynamic shape and obtain the data information required for aerodynamic design and aerodynamic shape optimization, including angle of attack, sideslip angle, Mach number, Reynolds number, angle of attack, aircraft flight attitude and aircraft shape coordinates; The aircraft layout optimization module is connected to the aerodynamic analysis module and the data input module, and is used to provide the data input module with an initial population of genetic algorithms, including the generation of the initial aircraft shape, the description of the aircraft parameters, the determination of the aircraft shape design variables, the determination of the aircraft optimization target, the construction of the initial optimization variable group, and the determination of the aerodynamic layout optimization constraint conditions.

10. The system according to claim 9, characterized in that The system further comprises: The post-optimization simulation calculation module is connected to the data input module and the aircraft layout optimization module, and is used to simulate and calculate the aerodynamic performance of the aircraft according to the optimized aircraft configuration, including: Configuration data management unit, used for parameter adjustment based on existing grids; a geometry description file generating unit, connected to the configuration data management unit, for generating a grid description file according to the aircraft configuration data; A computational grid generation unit, connected to the geometry description file generation unit, for inputting the generated grid description file into a CFD calculation tool to generate a computational grid; The error analysis unit is used to compare the performance of key components in the aerodynamic performance results of the optimized aircraft configuration with expected values ​​to obtain an error value, and output the optimized aircraft configuration if the error value meets the requirements.

Images