CN-122020865-A - Method for determining morphological parameters of high lift device and computing equipment

CN122020865ACN 122020865 ACN122020865 ACN 122020865ACN-122020865-A

Abstract

A method for determining morphological parameters of a high-lift device and computing equipment acquire a first sample set and a second sample set, wherein the first sample set consists of a plurality of first training samples comprising low-precision performance labels, the second sample set consists of a plurality of second training samples comprising high-precision performance labels, the number of the second training samples is smaller than that of the first training samples, a performance prediction model is trained by the first sample set and the second sample set, the low-precision prediction branch and an output layer are used for forward propagation for the first sample set, the high-precision prediction branch and the output layer are used for forward propagation for the second sample set, the performance prediction model is used for determining recommended morphological parameters in the parameter space after training is completed, and the recommended morphological parameters can be determined in a preset parameter space with lower computation cost and time cost.

Inventors

MA SHOUDONG
Shao gan
He Ruichen
FENG YANG
ZHANG YANYAN
YAN XUFEI

Assignees

天目山实验室

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. A method of determining a high lift device morphology parameter, the method involving a performance prediction model comprising at least a feature extraction module and an output layer, the feature extraction module comprising a high precision prediction branch and a low precision prediction branch, the output of the performance prediction model being a prediction of the high lift device performance parameter, the method comprising: Acquiring a first sample set and a second sample set, wherein the first sample set comprises a plurality of first training samples, the first training samples are composed of morphological parameters corresponding to first sample points in a parameter space of morphological parameters of a lifting device and low-precision performance labels, the low-precision performance labels are determined according to the morphological parameters by using a fluid solver based on simplifying assumptions, the second sample set comprises a plurality of second training samples, the second training samples are composed of morphological parameters corresponding to second sample points and high-precision performance labels, the high-precision performance labels are determined according to the morphological parameters by using a three-dimensional simulation model, and the number of the second sample points is smaller than that of the first sample points; And training the performance prediction model by using the first sample set and the second sample set, wherein the low-precision prediction branch and the output layer are used for forward propagation for the first sample set, the high-precision prediction branch and the output layer are used for forward propagation for the second sample set, and the performance prediction model is used for determining recommended morphological parameters in the parameter space after the training is completed.
2. The method of claim 1, wherein each first sample point is obtained by sampling the parameter space using an optimal latin hypercube method.
3. The method of claim 1, wherein each second sample point is screened from each first sample point based on a low precision performance tag.
4. The method of claim 1, training the performance prediction model with the first and second sample sets, comprising: Determining hidden features corresponding to morphological parameters of a training sample by using a prediction branch corresponding to the training sample aiming at any training sample, wherein the training sample comprises a first training sample and a second training sample, and the prediction branch comprises a low-precision prediction branch and a high-precision prediction branch; determining the prediction performance corresponding to the training sample according to the hidden characteristics by utilizing the output layer; And according to the performance label corresponding to the predicted performance and the training sample, at least carrying out parameter adjustment on the predicted branch corresponding to the training sample, wherein the performance label comprises a low-precision performance label and a high-precision performance label.
5. The method of claim 1, after training the performance prediction model with the first and second sample sets, further comprising: Carrying out a plurality of rounds of search on the parameter space iteration by using a preset optimization algorithm, and determining a plurality of recommended morphological parameters according to the prediction performance of each target sample point determined in each round of search process and the morphological parameters respectively corresponding to each target sample point, wherein the search process of any round comprises: determining a plurality of target sample points of the current round in the parameter space by utilizing the optimization algorithm; for any target sample point of the current round, determining a first hidden characteristic corresponding to the target sample point by utilizing the low-precision prediction branch, and determining a second hidden characteristic corresponding to the target sample point by utilizing the high-precision prediction branch; Fusing the first hidden feature and the second hidden feature to obtain a comprehensive feature; and determining the prediction performance corresponding to the target sample point according to the comprehensive characteristics by utilizing the output layer.
6. The method of claim 5, fusing the first hidden feature and the second hidden feature to obtain a composite feature, specifically comprising: And fusing the first hidden feature and the second hidden feature according to a preset fusion weight to obtain a comprehensive feature, wherein the fusion weight corresponding to the first hidden feature is smaller than the fusion weight corresponding to the second hidden feature.
7. The method of claim 5, the performance prediction model further comprising a fusion layer; before performing a number of rounds of search for the parameter space iterations using a preset optimization algorithm, the method further comprises: For any second training sample, determining a first hidden characteristic corresponding to the second training sample by utilizing a low-precision prediction branch, and determining a second hidden characteristic corresponding to the second training sample by utilizing a high-precision prediction branch; fusing the first hidden feature and the second hidden feature corresponding to the second training sample by using the fusion layer to obtain a fusion feature corresponding to the second training sample; Determining comprehensive prediction performance corresponding to the second training sample according to the fusion characteristics corresponding to the second training sample by using an output layer; according to the performance labels of the comprehensive prediction performance and the second training sample, at least carrying out parameter adjustment on the fusion layer; fusing the first hidden feature and the second hidden feature to obtain a comprehensive feature, which specifically includes: and fusing the first hidden feature and the second hidden feature by using the fusion layer to obtain a comprehensive feature.
8. The method of claim 1, wherein the parameter space comprises, in particular, a propeller layout parameter, a flap geometry parameter, a slot design parameter, and a flap deflection angle.
9. The method of claim 1 wherein the performance parameters include in particular a takeoff configuration lift coefficient and a landing configuration lift coefficient.
10. A computing device comprising a memory having executable code stored therein and a processor, which when executing the executable code, implements the method of any of claims 1-9.

Description

Method for determining morphological parameters of high lift device and computing equipment Technical Field The embodiment of the specification belongs to the technical field of data processing, and particularly relates to a method for determining morphological parameters of a high lift device and computing equipment. Background Usually, the runway length of domestic civil airports is mostly more than 3000 meters. And limited by construction conditions, a part of remote or more complex terrain areas can only construct a sprint airport with a runway length of between 1000 meters and 2000 meters. To accommodate the shorter runway length, fixed wing aircraft operable at a sprint airport may require a corresponding ultra Short Take Off and Landing (STOL) capability. At present, technicians often adopt a technical approach of combining a distributed propulsion system and a wing lift-increasing device to promote the ultra-short-distance take-off and landing capability of an aircraft. Therefore, the lift-increasing performance of the aircraft in the take-off and landing stage depends on whether the design of the lift-increasing device can effectively regulate and control and fully utilize the slipstream generated by the front propeller so as to obtain the maximum lift-increasing benefit. At present, the design method of the high lift device mainly optimizes the two-dimensional section of the wing profile. The method simplifies the three-dimensional wing into a plurality of representative two-dimensional sections, and carries out independent pneumatic analysis and structural design on each two-dimensional section. However, for a distributed propulsion aircraft with multiple sets of propulsion systems arranged at the leading edge of the wing, the slipstream generated by its propellers (i.e. the propulsion systems) has a strong three-dimensional character and a significant spanwise velocity component. Fig. 1 shows a schematic diagram of the relative positions of a propeller, a wing and a high lift device, as shown in fig. 1, a being the propeller, B being the wing, C being the flap (one type of high lift device), there being part of the slip flow tangential or radial to the propeller blades, in addition to the main slip flow directed by the propeller towards the wing (i.e. the x-direction in the drawing) when the propeller is rotated. Such complex slipstream effects are currently difficult to fully express using empirical formulas, which also results in that the design results obtained on the basis of the simplified two-dimensional cross-section for high lift devices comprising propellers fail to achieve the desired aerodynamic performance objective in a real flight environment. In order to accurately simulate the three-dimensional effect of the slip stream, three-dimensional simulation is theoretically required by adopting computational fluid dynamics (Computational Fluid Dynamics, CFD). However, the high-fidelity simulation has high calculation cost, and the optimization design problem of the high-lift device of the aircraft also relates to a plurality of variables such as the layout parameters of the propeller, the geometry of the flap, the width of the slot, the deflection angle of the flap and the like, so that the optimization problem has the characteristics of high design dimension and high calculation cost. Therefore, the computational cost required for developing the optimal design of the high lift device of the distributed propulsion aircraft by directly using three-dimensional simulation is not acceptable in engineering. However, if part of disturbance items in the existence of a real scene are ignored, the three-dimensional effect of the slipstream is simplified, a fluid solver is constructed based on the simplifying assumption, and then the fluid solver based on the simplifying assumption is utilized to develop the optimal design of the high lift device, although the calculation cost is reduced, the prediction accuracy of the aerodynamic performance is still difficult to guarantee. Therefore, the specification provides a technical scheme for determining the morphological parameters of the high lift device so as to at least partially solve the problems. Disclosure of Invention Embodiments of the present disclosure are directed to a method and a computing device for determining morphological parameters of a high lift device, including: A first aspect of the present disclosure provides a method for determining a morphological parameter of a high lift device, the method involving a performance prediction model, the performance prediction model at least comprising a feature extraction module and an output layer, the feature extraction module comprising a high-precision prediction branch and a low-precision prediction branch, an output result of the performance prediction model being a prediction result for the performance parameter of the high lift device, the method comprising: Acquiring a first sample set a