Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Uncertainty-Aware Data-Based Method for Fast and Reliable Shape Optimization

Yunjia Yang, Runze Li, Yufei Zhang, Haixin Chen

TL;DR

This paper addresses the reliability gap in data-based aerodynamic shape optimization (DBO) by introducing uncertainty-aware DBO (UA-DBO). It replaces a purely deterministic surrogate with a Gaussian stochastic encoder–decoder (GS-ED) that quantifies predictive uncertainty, and integrates this into a model-confidence-aware objective to penalize high-uncertainty predictions. The approach combines a prior-based surrogate framework with robust optimization principles, enabling near-CFD accuracy with substantially reduced computational cost, and demonstrates improved stability and performance across drag-divergence and buffet-onset airfoil problems. The results show that UA-DBO reduces model-prediction errors, mitigates overconfident mispredictions, and achieves performance close to CFD-based optimization at a fraction of the computational burden, offering practical benefits for rapid, reliable aerodynamic design.

Abstract

Data-based optimization (DBO) offers a promising approach for efficiently optimizing shape for better aerodynamic performance by leveraging a pretrained surrogate model for offline evaluations during iterations. However, DBO heavily relies on the quality of the training database. Samples outside the training distribution encountered during optimization can lead to significant prediction errors, potentially misleading the optimization process. Therefore, incorporating uncertainty quantification into optimization is critical for detecting outliers and enhancing robustness. This study proposes an uncertainty-aware data-based optimization (UA-DBO) framework to monitor and minimize surrogate model uncertainty during DBO. A probabilistic encoder-decoder surrogate model is developed to predict uncertainties associated with its outputs, and these uncertainties are integrated into a model-confidence-aware objective function to penalize samples with large prediction errors during data-based optimization process. The UA-DBO framework is evaluated on two multipoint optimization problems aimed at improving airfoil drag divergence and buffet performance. Results demonstrate that UA-DBO consistently reduces prediction errors in optimized samples and achieves superior performance gains compared to original DBO. Moreover, compared to multipoint optimization based on full computational simulations, UA-DBO offers comparable optimization effectiveness while significantly accelerating optimization speed.

Uncertainty-Aware Data-Based Method for Fast and Reliable Shape Optimization

TL;DR

This paper addresses the reliability gap in data-based aerodynamic shape optimization (DBO) by introducing uncertainty-aware DBO (UA-DBO). It replaces a purely deterministic surrogate with a Gaussian stochastic encoder–decoder (GS-ED) that quantifies predictive uncertainty, and integrates this into a model-confidence-aware objective to penalize high-uncertainty predictions. The approach combines a prior-based surrogate framework with robust optimization principles, enabling near-CFD accuracy with substantially reduced computational cost, and demonstrates improved stability and performance across drag-divergence and buffet-onset airfoil problems. The results show that UA-DBO reduces model-prediction errors, mitigates overconfident mispredictions, and achieves performance close to CFD-based optimization at a fraction of the computational burden, offering practical benefits for rapid, reliable aerodynamic design.

Abstract

Data-based optimization (DBO) offers a promising approach for efficiently optimizing shape for better aerodynamic performance by leveraging a pretrained surrogate model for offline evaluations during iterations. However, DBO heavily relies on the quality of the training database. Samples outside the training distribution encountered during optimization can lead to significant prediction errors, potentially misleading the optimization process. Therefore, incorporating uncertainty quantification into optimization is critical for detecting outliers and enhancing robustness. This study proposes an uncertainty-aware data-based optimization (UA-DBO) framework to monitor and minimize surrogate model uncertainty during DBO. A probabilistic encoder-decoder surrogate model is developed to predict uncertainties associated with its outputs, and these uncertainties are integrated into a model-confidence-aware objective function to penalize samples with large prediction errors during data-based optimization process. The UA-DBO framework is evaluated on two multipoint optimization problems aimed at improving airfoil drag divergence and buffet performance. Results demonstrate that UA-DBO consistently reduces prediction errors in optimized samples and achieves superior performance gains compared to original DBO. Moreover, compared to multipoint optimization based on full computational simulations, UA-DBO offers comparable optimization effectiveness while significantly accelerating optimization speed.
Paper Structure (52 sections, 19 equations, 18 figures, 6 tables)

This paper contains 52 sections, 19 equations, 18 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Uncertainty-aware data-based optimization framework
  3. Overall framework
  4. Model-confidence-aware objective function
  5. Uncertainty quantification for the machine learning model
  6. Training of GS-ED model
  7. Inference with GS-ED model
  8. Uncertainty calibration
  9. Experiment setup
  10. Shape optimization problem
  11. Model setup for the optimization problem
  12. Prior-based strategy for multipoint aerodynamic performance prediction
  13. Model architecture
  14. Pretraining settings of surrogate models
  15. Airfoil database
  16. ...and 37 more sections

Figures (18)

  • Figure 1: Comparison of the DBO and UA-DBO frameworks
  • Figure 1: Architectures of VAE, PVAE, and GS-ED models
  • Figure 1: Model performance with different hyperparameters
  • Figure 1: Architecture of the model for buffet onset optimization (The EB# and DB# stand for the encoder and decoder blocks, respectively. The FC and conv stand for the fully connected layer and the 1D Convolutional layer, respectively. The values in brackets indicate the channel number and feature map sizes of the tensors.)
  • Figure 2: Lift-curve-break buffet onset criterion for buffet onset
  • ...and 13 more figures