Multi-parameter Joint Design Method, System and Storage Medium for Aircraft Based on Deep Learning

Abstract

The present invention provides a multi-parameter joint design method, system and storage medium for aircraft based on deep learning, which relates to the technical field of aircraft parameter design. The specific steps include: collecting multiple groups of original parameters of the fuselage, wings and engines, constructing an aircraft performance evaluation model, inputting the parameter combination matrix into the trained model to obtain the climb coefficient, performing data processing and correlation analysis, generating the aircraft cruise coefficient and landing risk coefficient, and calculating the comprehensive performance coefficient based on these coefficients. With the goal of maximizing the comprehensive performance coefficient, the initial population is optimized through a genetic algorithm to obtain the optimal design parameters. The present invention can comprehensively evaluate the performance of the aircraft at different flight stages, ensuring that the design can better understand the mutual influence between each stage, and then optimizing the overall design. By using the multi-objective optimization method, the aircraft can find the best balance among multiple performance indicators.

Description

Technical Field

[0001] The present invention relates to the technical field of aircraft parameter design, and specifically provides a multi-parameter joint design method, system and storage medium for aircraft based on deep learning. Background Art

[0002] In modern aircraft design, the performance evaluation and optimization of aircraft are complex and critical technical issues. Traditional design methods often rely on experience and experiments, resulting in long design cycles, high costs and limited optimization effects. In addition, as the number of aircraft design parameters increases, the computational workload and time required for manual optimization also increase significantly, making it difficult to effectively meet complex design requirements. Therefore, how to use advanced computing technologies and deep learning methods to improve the efficiency and accuracy of aircraft design has become a technical challenge to be solved urgently.

[0003] In the prior art, a method for predicting aircraft design parameters based on machine learning with the publication number CN115048874B includes the following steps: according to the type of aircraft for which rapid parameter prediction is to be performed, determine the dependent variable parameters to be predicted and the independent variable parameters to be used as inputs, collect relevant data to form a parameter data set, and after data preprocessing and feature engineering data processing of the obtained parameter data set, use the determined independent variable parameters as input parameters and train using a regression algorithm based on machine learning to form a parameter predictor; and according to the formed parameter predictor, change the relevant input parameters to obtain the parameter prediction results of this type of aircraft. This method solves the problems of the need for a large amount of empirical data accumulation, complex iterative process and difficulty in quickly obtaining aircraft parameter combinations in traditional aircraft parameter prediction, and can quickly realize the aircraft parameter combination space in aircraft system simulation.

[0004] However, there are still the following deficiencies. From the above statements, it can be seen that the prior art only focuses on one state design parameter and performance index of the aircraft, lacks a comprehensive evaluation of the aircraft in different flight stages (climb, cruise, landing), and the parameter prediction method mainly targets a single objective and cannot effectively handle multi-objective optimization problems. This limitation will lead to performance imbalance problems in actual operation of the aircraft.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a multi-parameter joint design method, system and storage medium for aircraft based on deep learning to solve the problems raised in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for jointly designing multiple parameters of an aircraft based on deep learning, the specific steps include:

[0009] S1. Collect multiple groups of original parameters of the fuselage, wing and engine, including the fuselage diameter, fuselage weight, wing width, wing surface area, wing thickness, engine thrust and engine speed respectively. Use the parameters of multiple groups of known fuselages, wings and engines to construct individuals in the initial population;

[0010] S2. Construct an aircraft performance evaluation model. The aircraft performance evaluation model is composed of a deep learning network based on a multi-layer perceptron. Combine the fuselage diameter, fuselage weight, wing surface area, wing thickness and engine thrust of individuals in multiple groups of known initial populations to construct a parameter combination matrix. Use the known parameter combination matrix as the input, and determine the aircraft climb coefficient through the expert group scoring method. Use the aircraft climb coefficient as the label to train the aircraft performance evaluation model;

[0011] S3. Process the data of the wing width, fuselage weight, wing surface area and engine speed of individuals in the initial population, and perform correlation analysis to generate an aircraft cruise coefficient for evaluating the cruise performance of the aircraft. Process the data of the wing thickness, fuselage weight, wing surface area and engine thrust of individuals in the initial population, and perform correlation analysis to generate a landing risk coefficient for evaluating the landing risk of the aircraft. Process the aircraft climb coefficient, aircraft cruise coefficient and landing risk coefficient to generate a comprehensive performance coefficient for evaluating the comprehensive performance of the aircraft;

[0012] S4. Take the maximization of the comprehensive performance coefficient as the objective function, and construct the constraint conditions of the fuselage, wing and engine parameters. Under the constraint conditions, iteratively optimize the individuals in the initial population through the genetic algorithm to obtain the optimal individuals. Based on the optimal individuals, extract the optimal values of the fuselage, wing and engine parameters.

[0013] Further, use the parameters of multiple groups of known fuselages, wings and engines to construct individuals in the initial population. The specific process is as follows:

[0014] Collect the original parameters of the known fuselage, wing and engine, including the fuselage diameter D, fuselage weight M, wing width W, wing surface area A, wing thickness H, engine thrust T and engine speed V, and form an initial population labeled as Q. And the initial population Q = {Q1, Q2, …, Q j , …, Q m}, Q j is the jth individual in the initial population, j is the index of the initial population, and j ∈ [1, m], m is the number of the initial population, and Q j={D j ,M j ,W j ,A j ,H j ,T j ,V j}, D j ,M j ,W j ,A j ,H j ,T j ,V j are respectively the fuselage diameter, fuselage weight, wing width, wing surface area, wing thickness, engine thrust and engine speed of the j-th individual.

[0015] Furthermore, the aircraft performance evaluation model is composed of a deep learning network based on a multi-layer perceptron. The deep neural network of the multi-layer perceptron includes an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer. The first hidden layer, the second hidden layer and the third hidden layer all have at least two neurons and all use ReLU as the activation function;

[0016] The process of training the aircraft performance evaluation model is as follows:

[0017] According to the fuselage diameter, fuselage weight, wing surface area, wing thickness and engine thrust of individuals in multiple groups of known initial populations, the corresponding aircraft climb coefficient is determined based on the expert scoring method. The larger the aircraft climb coefficient, the better the climbing performance of the aircraft;

[0018] Using the known parameter combination matrix as the input quantity and the aircraft climb coefficient as the output label for training, and using the mean square error as the loss function. When the mean square error is within the range of [0, 0.1], the training of the aircraft performance evaluation model is completed.

[0019] Furthermore, the wing width, fuselage weight, wing surface area and engine speed are processed and correlation analysis is carried out to generate an aircraft cruise coefficient for evaluating the cruise performance of the aircraft. The constructed functional form is as follows:

[0020]

[0021] Among them, XS XH is the aircraft cruise coefficient, A is the wing surface area, W is the wing width, V is the engine speed, M is the fuselage weight, α is the weight coefficient of the wing surface area and the wing width, which is used to adjust the influence degree of the wing surface area and the wing width on the cruise performance, and β is the weight coefficient of the fuselage weight, which is used to adjust the influence degree of the fuselage weight on the cruise performance.

[0022] Further, data processing is performed on the wing thickness, fuselage weight, wing surface area, and engine thrust, and correlation analysis is carried out to generate a landing risk coefficient for evaluating the landing risk of the aircraft. The functional form constructed is as follows:

[0023]

[0024] Among them, XS JF is the landing risk coefficient, T is the engine thrust, H is the wing thickness, M is the fuselage weight, ω1 is the weight coefficient of the wing surface area, ω2 is the weight coefficient of the engine thrust, ω3 is the weight coefficient of the wing thickness, ω4 is the weight coefficient of the fuselage weight, 0 < ω4 < ω3 < ω2 < ω1 < 1, and ω4 + ω3 + ω2 + ω1 = 1.

[0025] Further, data processing is performed on the aircraft climb coefficient, aircraft cruise coefficient, and landing risk coefficient to generate a comprehensive performance coefficient for evaluating the comprehensive performance of the aircraft. The formula is as follows:

[0026] ZX = ω PS ·XS PS + ω XH ·XS XH - ω JF ·XS JF

[0027] Among them, ZX is the comprehensive performance coefficient, XS PS is the aircraft climb coefficient, ω PS is the weight coefficient of the aircraft climb coefficient, ω XH is the weight coefficient of the aircraft cruise coefficient, ω JF is the weight coefficient of the landing risk coefficient, ω PS > ω XH > ω JF , ω PS + ω XH + ω JF = 1.

[0028] Further, the specific process of step S5 is as follows:

[0029] Seek a balance among the aircraft climb coefficient, the aircraft cruise coefficient, and the landing risk coefficient to maximize the comprehensive performance coefficient. Then, with the maximization of the comprehensive performance coefficient ZX as the optimization goal, perform iterative optimization on the initial population Q, that is, perform selection, crossover, and mutation operations on the individuals in the initial population Q. During the iterative optimization process, set constraint conditions, that is, set the maximum and minimum values of the fuselage diameter D, the fuselage weight M, the wing width W, the wing surface area A, the wing thickness H, the engine thrust T, and the engine speed V respectively. Within the constraint ranges of the fuselage diameter D, the fuselage weight M, the wing width W, the wing surface area A, the wing thickness H, the engine thrust T, and the engine speed V, perform iterative optimization on the initial population Q. Specifically, select the individuals with the top comprehensive performance coefficients as the parents, where the top refers to the individuals in the top 50% of the comprehensive performance coefficients. Through the crossover operation, exchange and combine the genes of the parent individuals to generate new individuals. During the iterative optimization process, keep the constant workpiece characteristic parameters unchanged, and then perform mutation operations on the genes of the fuselage diameter D, the fuselage weight M, the wing width W, the wing surface area A, the wing thickness H, the engine thrust T, and the engine speed V in the newly generated individuals, and repeat the selection, crossover, and mutation operations until the predetermined number of iterations is reached;

[0030] After performing iterative optimization on the initial population Q, label the optimal individual as Q j1 ={D j1 ,M j1 ,W j1 ,A j1 ,H j1 ,T j1 ,V j1}, and the optimal values of the fuselage, wing, and engine parameters are the fuselage diameter D j1 , the fuselage weight M j1 , the wing width W j1 , the wing surface area A j1 , the wing thickness H j1 , the engine thrust T j1 , and the engine speed V j1 .

[0031] A multi-parameter joint design system for an aircraft based on deep learning, which is used to execute any one of the above-mentioned multi-parameter joint design methods for an aircraft based on deep learning, includes:

[0032] A data acquisition module, which is used to collect multiple groups of original parameters of the fuselage, wing, and engine, including the fuselage diameter, the fuselage weight, the wing width, the wing surface area, the wing thickness, the engine thrust, and the engine speed respectively, and use the parameters of multiple groups of known fuselages, wings, and engines to construct individuals in the initial population;

[0033] A performance evaluation module, which is used to construct an aircraft performance evaluation model. The aircraft performance evaluation model is composed of a deep learning network based on a multi-layer perceptron. Multiple groups of combinations of the fuselage diameter, fuselage weight, wing surface area, wing thickness, and engine thrust of individuals in the known initial population are combined to construct a parameter combination matrix. The known parameter combination matrix is used as the input, and the aircraft climb coefficient is determined by the expert group scoring method. Using the aircraft climb coefficient as the label, the aircraft performance evaluation model is trained;

[0034] A data processing and analysis module, which is used to process the data of the wing width, fuselage weight, wing surface area, and engine speed of individuals in the initial population, and conduct a correlation analysis to generate an aircraft cruise coefficient for evaluating the cruise performance of the aircraft. The data of the wing thickness, fuselage weight, wing surface area, and engine thrust of individuals in the initial population are processed and a correlation analysis is conducted to generate a landing risk coefficient for evaluating the landing risk of the aircraft. The aircraft climb coefficient, aircraft cruise coefficient, and landing risk coefficient are processed to generate a comprehensive performance coefficient for evaluating the comprehensive performance of the aircraft;

[0035] A parameter optimization module, which takes the maximization of the comprehensive performance coefficient as the objective function, constructs the constraint conditions of the fuselage, wing, and engine parameters, and iteratively optimizes the individuals in the initial population through a genetic algorithm under the constraint conditions to obtain the optimal individuals. Based on the optimal individuals, the optimal values of the fuselage, wing, and engine parameters are extracted.

[0036] A storage medium, which is used to store a computer program. When the computer program is executed by a processor, it implements any one of the above-mentioned multi-parameter joint design methods for aircraft based on deep learning.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] The present invention constructs individuals in the initial population by collecting multiple sets of original parameters of known fuselages, wings, and engines, and using the parameters of multiple sets of known fuselages, wings, and engines. The diameters, weights, wing surface areas, wing thicknesses, and engine thrusts of multiple sets of known fuselages are combined to construct a parameter combination matrix, and the aircraft climb coefficient is obtained. The individuals in the initial population are processed to generate the aircraft cruise coefficient and landing risk coefficient. The aircraft climb coefficient, aircraft cruise coefficient, and landing risk coefficient are processed to generate a comprehensive performance coefficient. Taking the maximization of the comprehensive performance coefficient as the objective function, the individuals in the initial population are iteratively optimized through a genetic algorithm to obtain the optimal individuals. Based on the optimal individuals, the optimal values of the fuselage, wing, and engine parameters are extracted. Therefore, the present invention can comprehensively evaluate the performance of the aircraft at different flight stages. This comprehensiveness ensures that the design can better understand the mutual influence between various stages, and then optimize the overall design. The multi-objective optimization method makes the design of the aircraft more flexible, can find the best balance among multiple performance indicators, and avoids the performance imbalance problem caused by single-objective design. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic flow chart of the overall method of the present invention;

[0040] Figure 2 It is a block diagram of the module composition of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to specific embodiments.

[0042] It should be noted that unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meaning understood by those of ordinary skill in the field to which the present invention belongs. The "first", "second", and similar terms used in the present invention do not indicate any order, quantity, or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or objects appearing before this term cover the elements or objects listed after this term and their equivalents, without excluding other elements or objects. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left", "right", etc. are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0043] Embodiment 1:

[0044] Please refer to Figure 1 , the present invention provides a technical solution:

[0045] A multi-parameter joint design method for aircraft based on deep learning, the specific steps include:

[0046] S1. Collect multiple groups of original parameters of the fuselage, wing and engine, including fuselage diameter, fuselage weight, wing width, wing surface area, wing thickness, engine thrust and engine speed respectively. Use the parameters of multiple groups of known fuselages, wings and engines to construct individuals in the initial population;

[0047] The original parameters of the fuselage, wing and engine can be collected from an aviation database;

[0048] S2. Construct an aircraft performance evaluation model. The aircraft performance evaluation model is composed of a deep learning network based on a multi-layer perceptron. Combine the fuselage diameter, fuselage weight, wing surface area, wing thickness and engine thrust of individuals in multiple groups of known initial populations to construct a parameter combination matrix. Use the known parameter combination matrix as the input, and determine the aircraft climb coefficient through the expert group scoring method. Use the aircraft climb coefficient as the label to train the aircraft performance evaluation model;

[0049] S4. Process the wing width, fuselage weight, wing surface area and engine speed of individuals in the initial population, and conduct correlation analysis to generate an aircraft cruise coefficient for evaluating the cruise performance of the aircraft. Process the wing thickness, fuselage weight, wing surface area and engine thrust of individuals in the initial population, and conduct correlation analysis to generate a landing risk coefficient for evaluating the landing risk of the aircraft. Process the aircraft climb coefficient, aircraft cruise coefficient and landing risk coefficient to generate a comprehensive performance coefficient for evaluating the comprehensive performance of the aircraft;

[0050] S5. Take the maximization of the comprehensive performance coefficient as the objective function, and construct the constraint conditions of the fuselage, wing and engine parameters. Under the constraint conditions, iteratively optimize the individuals in the initial population through the genetic algorithm to obtain the optimal individuals, and based on the optimal individuals, extract the optimal values of the fuselage, wing and engine parameters.

[0051] Based on the above embodiments, use the parameters of multiple groups of known fuselages, wings and engines to construct individuals in the initial population. The specific process is as follows:

[0052] Collect the original parameters of the known fuselage, wing and engine, including fuselage diameter D, fuselage weight M, wing width W, wing surface area A, wing thickness H, engine thrust T and engine speed V, and form the initial population labeled as Q, and the initial population Q = {Q1, Q2, …, Q j , …, Q m}, Q jis the j-th individual in the initial population, where j is the index of the initial population and j ∈ [1, m], m is the number of individuals in the initial population, and Q j = {D j , M j , W j , A j , H j , T j , T j}, where D j , M j , W j , A j , H j , T j , V j are respectively the fuselage diameter, fuselage weight, wing width, wing surface area, wing thickness, engine thrust, and engine speed of the j-th individual.

[0053] Based on the above embodiments, the aircraft performance evaluation model is composed of a deep learning network based on a multi-layer perceptron. The deep neural network of the multi-layer perceptron includes an input layer, a first hidden layer, a second hidden layer, a third hidden layer, and an output layer. The first hidden layer, the second hidden layer, and the third hidden layer each have at least two neurons and all use ReLU as the activation function.

[0054] In this embodiment, the input features of the deep learning network of the multi-layer perceptron include: fuselage diameter, fuselage weight, wing surface area, wing thickness, and engine thrust, a total of 5 features.

[0055] The structure of the deep learning network of the multi-layer perceptron is as follows:

[0056] Input layer: Receives the input of 5 features;

[0057] First hidden layer: Has 64 neurons and uses ReLU as the activation function;

[0058] Second hidden layer: Has 32 neurons and also uses the ReLU activation function;

[0059] Third hidden layer: Has 16 neurons and uses the ReLU activation function;

[0060] Output layer: Has a single neuron, which is the aircraft climb coefficient.

[0061] The process of training the aircraft performance evaluation model is as follows:

[0062] Based on the fuselage diameter, fuselage weight, wing surface area, wing thickness, and engine thrust of the individuals in multiple groups of known initial populations, the corresponding aircraft climb coefficient is determined based on the expert scoring method. For example, scoring is carried out within the range of 0 - 10. The larger the aircraft climb coefficient, the better the climbing performance of the aircraft.

[0063] Using the known parameter combination matrix as the input quantity and the aircraft climb coefficient as the output label for training, and using the mean squared error as the loss function. When the mean squared error is within the range of [0, 0.01], the training of the aircraft performance evaluation model is completed.

[0064] On the basis of the above embodiments, the correlations between the wing width, fuselage weight, wing surface area, engine speed, and aircraft cruise performance are as follows:

[0065] The wing width and the aircraft cruise performance are positively correlated. Because an increase in the wing width means that the aerodynamic shape of the wing can generate more lift. During the cruise phase, sufficient lift is the key to ensuring the aircraft maintains its flight altitude and stability. When the wing width increases, the wing can generate sufficient lift at a relatively lower speed, thus improving the aircraft's cruise performance.

[0066] The fuselage weight and the aircraft cruise performance are negatively correlated. Because when the aircraft is cruising, it needs to generate sufficient lift to counteract gravity. An increase in the fuselage weight means that more lift is required to maintain the same flight altitude, which usually requires the aircraft to fly at a higher angle of attack, resulting in an increase in air resistance and a decrease in the flight cruise performance.

[0067] The wing surface area and the aircraft cruise performance are positively correlated. Because the main function of the wing is to generate lift. An increase in the wing surface area means that the ability to generate lift is enhanced at the same flight speed and air density, which enables the aircraft to maintain flight at a lower speed, thus improving the cruise efficiency.

[0068] The engine speed and the aircraft cruise performance are positively correlated. Because an increase in the engine speed means that the engine can generate greater thrust. During the cruise phase, sufficient thrust is crucial for maintaining and increasing the aircraft speed. Higher thrust can help the aircraft overcome air resistance, thus improving the cruise performance.

[0069] Therefore, according to the correlations between the wing width, fuselage weight, wing surface area, engine speed, and aircraft cruise performance, the wing width, fuselage weight, wing surface area, and engine speed are now processed for data and correlation analysis is carried out to generate an aircraft cruise coefficient for evaluating the aircraft cruise performance. The constructed functional form is as follows:

[0070]

[0071] Among them, XSXH is the aircraft cruise coefficient, A is the wing surface area, W is the wing width, V is the engine speed, M is the fuselage weight, α is the weight coefficient of the wing surface area and wing width, which is used to adjust the influence of the wing surface area and wing width on the cruise performance, and β is the weight coefficient of the fuselage weight, which is used to adjust the influence of the fuselage weight on the cruise performance;

[0072] Aircraft cruise coefficient XS XH The range of is [0.1,10];

[0073] When XS XH When it is close to 0.1, it means that the lift of the aircraft is insufficient in the cruising state, unstable flight will occur, and even effective cruising cannot be maintained;

[0074] When XS XH When it is close to 10, it means that the aircraft has achieved excellent aerodynamic efficiency during cruising, which means that the aircraft can maintain a higher flight speed and stability with low fuel consumption.

[0075] Therefore, the aircraft cruise coefficient XS XH The larger it is, the better the aircraft's cruising performance.

[0076] The reason for using the above functional form to express the relationship between wing width, fuselage weight, wing surface area, engine speed and aircraft cruise coefficient is as follows:

[0077] First, wing surface area (A) and wing width (W): These two parameters directly affect an aircraft's lift and aerodynamic performance. A larger wing surface area (A) or wing width (W) generally increases an aircraft's lift. Therefore, expressing the combined effect of these two parameters in the form of a product (A·W) better reflects the impact of wing design on cruise performance.

[0078] Second, engine speed V: Engine speed V has a significant impact on the aircraft's cruise performance, but its impact is usually not linear, and as the speed increases, the improvement rate of cruise performance will slow down. The introduction of the logarithmic function can show the diminishing marginal effect between engine speed and aircraft cruise performance. Setting (V+1) is to avoid lg(V+1) being valid within the definition domain.

[0079] Third, fuselage weight M: Fuselage weight M has a negative impact on the aircraft's cruise performance. Increasing weight usually increases lift demand, thereby affecting fuel consumption and flight stability. Therefore, expressed in the form of a fraction, the M in the denominator can effectively reflect the downward trend in flight cruise performance as the fuselage weight M increases.

[0080] Fourth, under different design or operating conditions, the influence degrees of the wing surface area A, wing width W, and fuselage weight M on the cruise performance will be different. By setting the weight coefficients α and β, the contributions of these parameters in the calculation can be flexibly adjusted. Moreover, the performance of the aircraft is jointly determined by multiple interrelated factors. By introducing the weight coefficients α and β, the influences of each parameter can be comprehensively considered, avoiding a single factor from overly affecting the final calculation result and ensuring the comprehensiveness and accuracy of the analysis.

[0081] The magnitude relationship settings of the weight coefficient α of the wing surface area and wing width and the weight coefficient β of the fuselage weight are as follows:

[0082] The wing surface area A and wing width W directly affect the lift generation ability of the aircraft. Increasing the wing surface area and width can significantly improve the lift, thereby promoting the aircraft to maintain stable flight during cruising. Therefore, these two parameters play a more crucial role in the cruise performance;

[0083] When evaluating the cruise performance of the aircraft, the role of the wing is more direct and obvious, while the fuselage weight usually indirectly affects through other means, that is, the correlation between the wing surface area, wing width, and aircraft cruise coefficient is relatively large, and the correlation between the fuselage weight and aircraft cruise coefficient is relatively small;

[0084] Therefore, a higher weight coefficient is given to the wing-related parameters, that is, on the basis of β + α = 1, set 0 < β < α < 1.

[0085] On the basis of the above embodiments, the correlations between the wing thickness, fuselage weight, wing surface area, engine thrust, and aircraft landing risk are as follows:

[0086] The wing thickness and aircraft landing risk are negatively correlated. Because increasing the wing thickness can increase the aircraft lift, help the aircraft maintain flight during low-speed landing, and reduce the risk of stall.

[0087] The fuselage weight and aircraft landing risk are positively correlated. Because increasing the fuselage weight means that greater lift and higher landing speed are required during landing, which will increase the impact force when the aircraft touches the ground, thereby increasing the landing risk.

[0088] The wing surface area and aircraft landing risk are negatively correlated. Because increasing the wing surface area can provide higher lift. During the landing process, the aircraft needs to land safely at a lower speed. The large-surface-area wing can generate sufficient lift at a lower angle of attack, reducing the risk of stall.

[0089] The engine thrust and the aircraft landing risk are negatively correlated. Because an increase in engine thrust can provide the necessary lift during landing, helping the aircraft maintain control in case of emergencies. Sufficient thrust allows the aircraft to descend smoothly at a lower speed before touchdown, thus reducing the landing risk.

[0090] Therefore, based on the correlations among wing thickness, fuselage weight, wing surface area, engine thrust, and aircraft landing risk, the data of wing thickness, fuselage weight, wing surface area, and engine thrust are now processed and correlation analysis is carried out to generate a landing risk coefficient for evaluating the aircraft landing risk. The constructed functional form is as follows:

[0091]

[0092] Among them, XS JF is the landing risk coefficient, T is the engine thrust, H is the wing thickness, M is the fuselage weight, ω1 is the weight coefficient of the wing surface area, ω2 is the weight coefficient of the engine thrust, ω3 is the weight coefficient of the wing thickness, and ω4 is the weight coefficient of the fuselage weight.

[0093] The landing risk coefficient XS JF has a value range of (0, 10];

[0094] When XS JF is close to 0, it indicates that the aircraft has a high design safety factor and a low landing risk;

[0095] When XS JF is close to 10, it indicates that the aircraft has a low design safety factor and a high landing risk;

[0096] Therefore, the larger the landing risk coefficient XS JF , the greater the aircraft landing risk.

[0097] Since the wing surface area A is directly related to the lift and power control ability of the aircraft during landing, when the wing surface area A is larger, the aircraft can generate more lift during landing, reducing the stall risk. Therefore, among the above four parameters, the wing surface area A is considered to have the greatest impact on reducing the landing risk and the highest correlation with the aircraft landing risk. Thus, the weight coefficient ω1 of the wing surface area should be the largest;

[0098] An increase in engine thrust T can provide more power, helping the aircraft better control speed and attitude during landing, thus reducing the landing risk. Although it is also very important, compared with the wing surface area A, its role is more in controlling speed and attitude based on the existing lift. Then the correlation between the engine thrust T and the aircraft landing risk is lower than that of the wing surface area A. Therefore, its weight coefficient ω2 is lower than the weight coefficient ω1 of the wing surface area;

[0099] Although the wing thickness H has a direct effect on lift, its impact on landing risk is more complex. Wings of different thicknesses may behave differently at different speeds and angles, increasing uncertainty. Compared with the engine thrust T, the influence of the wing thickness H on landing risk is not as obvious as that of the engine thrust T. Therefore, the correlation between the wing thickness H and the aircraft landing risk is lower than that of the engine thrust T. Thus, the weight coefficient ω3 of the wing thickness is lower than the weight coefficient ω2 of the engine thrust;

[0100] The main function of the wing is to generate lift. The wing thickness H directly affects its lift characteristics. A thicker airfoil can generate more lift at lower speeds. During flight and landing, changes in the wing thickness H have a more significant impact on the controllability and stability of the aircraft. Compared with the wing thickness, adjustments to the fuselage weight M usually involve very small design changes, and most often improve flight performance by optimizing parameters such as the wing thickness. Therefore, the weight coefficient ω4 of the fuselage weight is lower than the weight coefficient ω3 of the wing thickness.

[0101] In summary, on the basis of ω4 + ω3 + ω2 + ω1 = 1, the weight coefficients of the wing thickness, fuselage weight, wing surface area, and engine thrust are set such that 0 < ω4 < ω3 < ω2 < ω1 < 1.

[0102] The reasons for using the above functional form to express the relationship between the wing thickness, fuselage weight, wing surface area, engine thrust, and aircraft landing risk are as follows:

[0103] First, this functional form combines multiple key parameters (wing surface area A, engine thrust T, wing thickness H, and fuselage weight M) to form a comprehensive landing risk coefficient XS JF , and this comprehensiveness can more fully reflect the contribution of each factor to the risk during aircraft landing.

[0104] Second, based on the correlation between the wing thickness H, fuselage weight M, wing surface area A, engine thrust T, and aircraft landing risk, a fractional function is used to express the relationship between the wing surface area A, engine thrust T, wing thickness H, fuselage weight M, and the landing risk coefficient. The wing thickness H, wing surface area A, and engine thrust T are placed in the denominator, and the larger the wing thickness H, wing surface area A, and engine thrust T, the smaller the landing risk coefficient XS JF , the fuselage weight M is placed in the numerator, and the larger the fuselage weight M, the larger the landing risk coefficient XS JF becomes.

[0105] Thirdly, the weight coefficients (ω4, ω3, ω2, ω1) are used to quantify the relative importance of different parameters in the landing risk assessment. Different parameters contribute differently to the landing risk. Through the weight coefficients, this relative importance can be accurately reflected. During aircraft design and performance evaluation, some parameters are more critical than others. By setting the weight coefficients, designers can focus on those parameters that have a greater impact on safety, thereby optimizing the design scheme.

[0106] Based on the above embodiments, data processing is performed on the aircraft climb coefficient, aircraft cruise coefficient, and landing risk coefficient to generate a comprehensive performance coefficient for evaluating the comprehensive performance of the aircraft. The basis formula is as follows:

[0107] ZX = ω PS ·XS PS + ω XH ·XS XH - ω JF ·XS JF

[0108] Among them, ZX is the comprehensive performance coefficient, XS PS is the aircraft climb coefficient, ω PS is the weight coefficient of the aircraft climb coefficient, ω XH is the weight coefficient of the aircraft cruise coefficient, ω JF is the weight coefficient of the landing risk coefficient.

[0109] Since when evaluating the comprehensive performance of the aircraft, more attention is focused on the climb performance of the aircraft, followed by the cruise performance, and finally the landing safety, the weight coefficients are set according to the following principles:

[0110] Priority for climb performance: Since the climb performance is the most important, a relatively high weight coefficient ω PS ;

[0111] Secondly for cruise performance: The importance of the cruise performance is second only to the climb performance, and a medium weight coefficient ω XH ;

[0112] Lowest for landing performance: Although the landing safety is also important, a relatively low weight coefficient ω JF should be given in the comprehensive evaluation because the higher the value of the landing risk coefficient, the greater the risk.

[0113] The reason for adopting the above functional form to express the relationship between the aircraft climb coefficient, aircraft cruise coefficient, landing risk coefficient, and comprehensive performance coefficient is as follows:

[0114] In the form of a linear combination, each performance index (climb, cruise, and landing risks) can be directly added or subtracted, which is convenient for quantification and calculation, and can simply and clearly reflect the contribution of each factor to the comprehensive performance.

[0115] Landing risk coefficient XS JF It is an index representing risk. The higher its value, the greater the risk. In the comprehensive performance evaluation, the risk coefficient is subtracted (negative weight), which is in line with the actual situation because high risks will reduce the overall safety and performance. Therefore, directly deducting the landing risk coefficient can effectively reflect the negative impact of risks on the comprehensive performance.

[0116] This formula can comprehensively consider the climb performance, cruise performance, and landing safety of the aircraft, enabling the comprehensive performance coefficient to comprehensively evaluate the overall performance of the aircraft, rather than just a single aspect.

[0117] Based on the above embodiments, the specific process of step S5 is as follows:

[0118] Seek a balance among the aircraft climb coefficient, aircraft cruise coefficient, and landing risk coefficient to maximize the comprehensive performance coefficient. Then, with the maximization of the comprehensive performance coefficient ZX as the optimization goal, perform iterative optimization on the initial population Q, that is, perform selection, crossover, and mutation operations on the individuals in the initial population Q. During the iterative optimization process, set constraint conditions, that is, set the maximum and minimum values of the fuselage diameter D, fuselage weight M, wing width W, wing surface area A, wing thickness H, engine thrust T, and engine speed V respectively. Within the constraint ranges of the fuselage diameter D, fuselage weight M, wing width W, wing surface area A, wing thickness H, engine thrust T, and engine speed V, perform iterative optimization on the initial population Q. Specifically, select the individuals with the top comprehensive performance coefficients as the parents. The top refers to the individuals in the top 50% of the comprehensive performance coefficients. Through the crossover operation, exchange and combine the genes of the parent individuals to generate new individuals. During the iterative optimization process, keep the constant workpiece characteristic parameters unchanged, and then perform mutation operations on the genes of the fuselage diameter D, fuselage weight M, wing width W, wing surface area A, wing thickness H, engine thrust T, and engine speed V in the newly generated individuals, and repeat the selection, crossover, and mutation operations until the predetermined number of iterations is reached;

[0119] After performing iterative optimization on the initial population Q, label the optimal individual as Q j1 ={D j1 ,M j1 ,W j1 ,A j1 ,H j1 ,T j1 ,V j1}, and the optimal values of the fuselage, wing, and engine parameters are the fuselage diameter Dj1 、The fuselage weight M j1 、The wing width W j1 、The wing surface area A j1 、The wing thickness H j1 、The engine thrust T j1 and the engine speed V j1 。

[0120] Please refer to Figure 2 ,The present invention also provides a technical solution:

[0121] A multi-parameter joint design system for an aircraft based on deep learning, the system is used to execute any one of the above-mentioned multi-parameter joint methods for an aircraft based on deep learning, including:

[0122] A data acquisition module, which is used to collect multiple groups of original parameters of the fuselage, wings and engines that are known, including the fuselage diameter, fuselage weight, wing width, wing surface area, wing thickness, engine thrust and engine speed respectively, and use the parameters of the multiple groups of known fuselages, wings and engines to construct individuals in the initial population;

[0123] A performance evaluation module, which is used to construct an aircraft performance evaluation model. The aircraft performance evaluation model is composed of a deep learning network based on a multi-layer perceptron. Combine the fuselage diameter, fuselage weight, wing surface area, wing thickness and engine thrust of the individuals in the multiple groups of known initial populations to construct a parameter combination matrix, use the known parameter combination matrix as the input, and determine the aircraft climb coefficient through the expert group scoring method. Use the aircraft climb coefficient as a label to train the aircraft performance evaluation model;

[0124] A data processing and analysis module, which is used to process the wing width, fuselage weight, wing surface area and engine speed of the individuals in the initial population, and perform correlation analysis to generate an aircraft cruise coefficient for evaluating the cruise performance of the aircraft. Process the wing thickness, fuselage weight, wing surface area and engine thrust of the individuals in the initial population, and perform correlation analysis to generate a landing risk coefficient for evaluating the landing risk of the aircraft. Process the aircraft climb coefficient, aircraft cruise coefficient and landing risk coefficient to generate a comprehensive performance coefficient for evaluating the comprehensive performance of the aircraft;

[0125] A parameter optimization module, which is used to take the maximization of the comprehensive performance coefficient as the objective function, and construct the constraint conditions of the fuselage, wing and engine parameters. Iteratively optimize the individuals in the initial population through the genetic algorithm under the constraint conditions to obtain the optimal individuals, and extract the optimal values of the fuselage, wing and engine parameters based on the optimal individuals.

[0126] A storage medium for storing a computer program, which when executed by a processor implements any one of the above-described multi-parameter joint design methods for aircraft based on deep learning.

[0127] All of the above formulas are dimensionless and take their numerical values for calculation. The formula is obtained by collecting a large amount of data for software simulation to obtain a formula that is closest to the actual situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0128] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0129] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0130] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all of them should be covered by the protection scope of the present application.

[0131] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0132] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all of them should be covered by the protection scope of the present application.

Claims

1. A multi-parameter joint design method for aircraft based on deep learning, characterized in that, The specific steps are as follows: S1. Collect multiple groups of original parameters of the fuselage, wings, and engines, including the fuselage diameter, fuselage weight, wing width, wing surface area, wing thickness, engine thrust, and engine speed respectively. Use the parameters of multiple groups of known fuselages, wings, and engines to construct individuals in the initial population; S2. Construct an aircraft performance evaluation model, which is composed of a deep learning network based on a multi-layer perceptron. Combine the fuselage diameter, fuselage weight, wing surface area, wing thickness, and engine thrust of individuals in multiple groups of known initial populations to construct a parameter combination matrix. Use the known parameter combination matrix as the input, and determine the aircraft climb coefficient through the expert scoring method. Use the aircraft climb coefficient as the label to train the aircraft performance evaluation model; S3. Process the data of the wing width, fuselage weight, wing surface area, and engine speed of individuals in the initial population, and conduct a correlation analysis to generate an aircraft cruise coefficient for evaluating the cruise performance of the aircraft. Process the data of the wing thickness, fuselage weight, wing surface area, and engine thrust of individuals in the initial population, and conduct a correlation analysis to generate a landing risk coefficient for evaluating the landing risk of the aircraft. Process the aircraft climb coefficient, aircraft cruise coefficient, and landing risk coefficient to generate a comprehensive performance coefficient for evaluating the comprehensive performance of the aircraft; S4. Take the maximization of the comprehensive performance coefficient as the objective function, and construct the constraint conditions for the fuselage, wing, and engine parameters. Iteratively optimize the individuals in the initial population through the genetic algorithm under the constraint conditions to obtain the optimal individuals, and based on the optimal individuals, extract the optimal values of the fuselage, wing, and engine parameters.

2. The multi-parameter joint design method of an aircraft based on deep learning according to claim 1, characterized in that: Use the parameters of multiple groups of known fuselages, wings, and engines to construct individuals in the initial population. The specific process is as follows: Collect the original parameters of the known fuselage, wing and engine, including the fuselage diameter S, the fuselage weight M, the wing width W, the wing surface area A, the wing thickness H, the engine thrust T and the engine speed V, and form the initial population designated as Q, and the initial population Q = {Q1, Q2, …, Q j , …, Q m}, Q j is the j-th individual in the initial population, j is the index of the initial population, and j ∈ [1, m], m is the number of the initial population, and Q j = {D j , M j , W j , A j , H j , T j , V j}, D j , M j , W j , A j , H j , T j , V j are respectively the fuselage diameter, the fuselage weight, the wing width, the wing surface area, the wing thickness, the engine thrust and the engine speed of the j-th individual.

3. The multi-parameter joint design method of an aircraft based on deep learning according to claim 2, wherein: The aircraft performance evaluation model is composed of a deep learning network based on a multi-layer perceptron. The deep neural network of the multi-layer perceptron includes an input layer, a first hidden layer, a second hidden layer, a third hidden layer, and an output layer. The first hidden layer, the second hidden layer, and the third hidden layer all have at least two neurons, and all use ReLU as the activation function; The process of training the aircraft performance evaluation model is as follows: Based on the fuselage diameter, fuselage weight, wing surface area, wing thickness, and engine thrust of individuals in multiple groups of known initial populations, determine the corresponding aircraft climb coefficient through the expert scoring method. The larger the aircraft climb coefficient, the better the climbing performance of the aircraft; Use the known parameter combination matrix as the input quantity and the aircraft climb coefficient as the output label for training. Use the mean square error as the loss function. When the mean square error is within the range of [0, 0.1], complete the training of the aircraft performance evaluation model.

4. The method for joint design of multiple parameters of an aircraft based on deep learning according to claim 3, characterized in that: Process the data of the wing width, fuselage weight, wing surface area, and engine speed, and conduct a correlation analysis to generate an aircraft cruise coefficient for evaluating the cruise performance of the aircraft. The constructed functional form is as follows: Among them, XS XH is the aircraft cruise coefficient, A is the wing surface area, W is the wing width, V is the engine speed, M is the fuselage weight, α is the weight coefficient of the wing surface area and the wing width, used to adjust the influence degree of the wing surface area and the wing width on the cruise performance, and β is the weight coefficient of the fuselage weight, used to adjust the influence degree of the fuselage weight on the cruise performance.

5. The method for joint design of multiple parameters of an aircraft based on deep learning according to claim 4, characterized in that: Data process the wing thickness, fuselage weight, wing surface area, and engine thrust, and perform a correlation analysis to generate a landing risk coefficient for evaluating the landing risk of an aircraft. The constructed function form is as follows: Among them, XS JF is the landing risk coefficient, T is the engine thrust, H is the wing thickness, M is the fuselage weight, ω1 is the weight coefficient of the wing surface area, ω2 is the weight coefficient of the engine thrust, ω3 is the weight coefficient of the wing thickness, ω4 is the weight coefficient of the fuselage weight, 0 < ω4 < ω3 < ω2 < ω1 < 1, and ω4 + ω3 + ω2 + ω1 = 1.

6. The multi-parameter joint design method for an aircraft based on deep learning according to claim 5, characterized in that: Data process the aircraft climb coefficient, aircraft cruise coefficient, and landing risk coefficient to generate a comprehensive performance coefficient for evaluating the comprehensive performance of the aircraft. The formula is as follows: ZX = ω PS ·XS PS + ω XH ·XS XH - ω JF ·XS JF Among them, ZX is the comprehensive performance coefficient, XS PS is the aircraft climb coefficient, ω PS is the weight coefficient of the aircraft climb coefficient, ω XH is the weight coefficient of the aircraft cruise coefficient, ω JF is the weight coefficient of the landing risk coefficient, ω PS > ω XH > ω JF , ω PS + ω XH + ω JF = 1.

7. The method for joint design of multiple parameters of an aircraft based on deep learning according to claim 6, characterized in that: The specific process of step S5 is as follows: Seek a balance among the aircraft climb coefficient, aircraft cruise coefficient, and landing risk coefficient to maximize the comprehensive performance coefficient. Then, with the maximization of the comprehensive performance coefficient ZX as the optimization goal, perform iterative optimization on the initial population Q, that is, perform selection, crossover, and mutation operations on the individuals in the initial population Q. During the iterative optimization process, set constraint conditions, that is, set the maximum and minimum values of the fuselage diameter D, fuselage weight M, wing width W, wing surface area A, wing thickness H, engine thrust T, and engine speed V respectively. Iteratively optimize the initial population Q within the constraint ranges of the fuselage diameter D, fuselage weight M, wing width W, wing surface area A, wing thickness H, engine thrust T, and engine speed V. Specifically, select the individuals with the top comprehensive performance coefficients as the parents. The top refers to the individuals in the top 50% of the comprehensive performance coefficients. Through the crossover operation, exchange and combine the genes of the parent individuals to generate new individuals. During the iterative optimization process, keep the constant workpiece characteristic parameters unchanged. Then, perform mutation operations on the genes of the fuselage diameter D, fuselage weight M, wing width W, wing surface area A, wing thickness H, engine thrust T, and engine speed V in the newly generated individuals, and repeat the selection, crossover, and mutation operations until the predetermined number of iterations is reached; After iteratively optimizing the initial population Q, the optimal individual is labeled as Q j1 ={D j1 ,M j1 ,W j1 ,A j1 ,H j1 ,T j1 ,V j1}, and the optimal values of the fuselage, wing, and engine parameters are the fuselage diameter D j1 , the fuselage weight M j1 , the wing width W j1 , the wing surface area A j1 , the wing thickness H j1 , the engine thrust T j1 , and the engine speed V j1 .

8. A multi-parameter joint design system for aircraft based on deep learning, the system being used to execute a multi-parameter joint design method for aircraft based on deep learning according to any one of claims 1-7, characterized in that: Including: A data acquisition module for collecting multiple groups of original parameters of the known fuselage, wings, and engines, including the fuselage diameter, fuselage weight, wing width, wing surface area, wing thickness, engine thrust, and engine speed respectively, and constructing individuals in the initial population using the multiple groups of known fuselage, wings, and engine parameters; A performance evaluation module for constructing an aircraft performance evaluation model. The aircraft performance evaluation model is composed of a deep learning network based on a multi-layer perceptron. Combine the fuselage diameter, fuselage weight, wing surface area, wing thickness, and engine thrust of the individuals in the multiple groups of known initial populations to construct a parameter combination matrix. Use the known parameter combination matrix as the input, and determine the aircraft climb coefficient through the expert group scoring method. Use the aircraft climb coefficient as the label to train the aircraft performance evaluation model; A data processing and analysis module is used to process the wing width, fuselage weight, wing surface area, and engine speed of individuals in the initial population, conduct correlation analysis, generate an aircraft cruise coefficient for evaluating the cruise performance of the aircraft, process the wing thickness, fuselage weight, wing surface area, and engine thrust of individuals in the initial population, conduct correlation analysis, generate a landing risk coefficient for evaluating the landing risk of the aircraft, and process the aircraft climb coefficient, aircraft cruise coefficient, and landing risk coefficient to generate a comprehensive performance coefficient for evaluating the comprehensive performance of the aircraft; A parameter optimization module is used to take the maximization of the comprehensive performance coefficient as the objective function, construct the constraint conditions for the fuselage, wing, and engine parameters, iteratively optimize the individuals in the initial population through a genetic algorithm under the constraint conditions, obtain the optimal individuals, and extract the optimal values of the fuselage, wing, and engine parameters based on the optimal individuals.

9. A storage medium, characterized in that: For saving a computer program, the computer program, when executed by a processor, implements a multi-parameter joint design method for an aircraft based on deep learning according to any one of claims 1-7.

Images