Improved automatic kernel construction for Gaussian process regression in small sample learning for predicting lift body aerodynamic performance

This paper introduces an intelligent identification method for self-excited aerodynamic equations. The method is based on advanced sparse recognition technology and equipped with a new sampling strategy designed for weak nonlinear dynamic systems with limit cycle characteristics. Considering the complexity of the experiment condition and the difficult a priori selection of hyperparameters, a method based on information criteria and ensemble learning is proposed to derive the global optimal aerodynamic self-excited model. The proposed method is first validated by simulated data obtained from some well-known equations and then applied to the identification of flutter aerodynamic equations based on wind tunnel experiments. Finally, reasons for the different sparse recognition results under different sizes of candidate function space are discussed from the perspective of matrix linear correlation and numerical calculation.

Identification of the form of self-excited aerodynamic force of bridge deck based on machine learning

Laima,

Zhang,

Jin

et al. 2024

An active learning SPH method for generalized Newtonian free surface flows

Dong,

Wang,

Ouyang

2024

This paper presents an active learning smoothed particle hydrodynamics (ALSPH) method to simulate generalized Newtonian free surface flows. First, an improved smoothed particle hydrodynamics (ISPH) method is established to obtain more reliable results for free surface flows by coupling the modified kernel gradient, the artificial viscosity, the density diffusive term, and the optimized particle shifting technique. Second, based on data and Gaussian process regression (GPR), an active learning strategy is developed to provide an effective constitutive relation. It is the first time that the ISPH method is combined with GPR to simulate generalized Newtonian free surface flows. Not only can the constitutive relation of any generalized Newtonian fluid in nature be accurately predicted, but a small amount of sampling data is also able to ensure accuracy over a wide range of the shear deformation rate. The challenging droplet impact and dam break are first modeled to validate the ISPH method. Due to the lack of an analytical constitutive relation for an arbitrary generalized Newtonian fluid in nature, the Cross model is then adopted and offers the required data to validate the ALSPH method. The results indicate that the learned constitutive relation is quite consistent with the analytical one and the simulation results match well. In addition, predictive accuracy and time consumption are proven. Furthermore, to verify the applicability of the learned constitutive relation, the jet buckling case and the jet entering the static fluid case are modeled. The good performance demonstrates the ALSPH method has a promising prospect of applications in simulating complex flows in nature.

Fast flow field prediction of three-dimensional hypersonic vehicles using an improved Gaussian process regression algorithm

Yang,

Xue,

Zhao

et al. 2024

Conducting large-scale numerical computations to obtain flow field during the hypersonic vehicle engineering design phase can be excessively costly. Although deep learning algorithms enable rapid flow field prediction with high-precision, they require a significant investment in training samples, contradicting the motivation of reducing the cost of acquiring flow field. The combination of feature extraction algorithms and regression algorithms can also achieve high-precision prediction of flow fields, which is more suitable to tackle three-dimensional flow prediction with a small dataset. In this study, we propose a reduced-order model (ROM) for the three-dimensional hypersonic vehicle flow prediction utilizing proper orthogonal decomposition to extract representative features and Gaussian process regression with improved automatic kernel construction (AKC-GPR) to perform a nonlinear mapping of physical features for prediction. The selection of variables is based on sensitivity analysis and modal assurance criterion. The underlying relationship is unveiled between flow field variables and inflow conditions. The ROM exhibits high predictive accuracy, with mean absolute percentage error (MAPE) of total field less than 3.5%, when varying altitudes and Mach numbers. During angle of attack variations, the ROM only effectively reconstructs flow distribution by interpolation with a MAPE of 7.02%. The excellent small-sample fitting capability of our improved AKC-GPR algorithm is demonstrated by comparing with original AKC-GPRs with a maximum reduction in a MAPE of 35.28%. These promising findings suggest that the proposed ROM can serve as an effective approach for rapid and accurate vehicle flow predicting, enabling its application in engineering design analysis.