Optimal Surface Pressure Sensor Placement for Lift Prediction of an Airfoil Subject to Gust

This work aims to study the optimal placement of surface pressure sensors and the prediction methods for the lift coefficient. We explored various sensor placements, including uniformly spaced sensors, and other configurations identified through data-driven methods such as QR pivoting of proper orthogonal decomposition (POD) modes and Bayesian inference. The interactions of the airfoil with both uniform and gusty inflow at multiple angles of attack are considered, where the gusty flow is generated by placing a cylinder upstream of the airfoil. The investigated flow fields are simulated using large-eddy simulations at Re_c of 10⁴. Deep learning models including convolutional neural networks (CNNs) and long short-term memory (LSTM) models are trained using simulation data to predict lift. The performance of the trained models and the optimality of the sensor locations are assessed by comparing the L2 errors of the predicted lift coefficient. The results indicate that the trained models can offer reasonable predictions of the lift coefficient even with sparse sensor information. They also demonstrate a certain level of robustness to the changes in training models and conditions. However, the predictions are found to be more sensitive to the quantity and placement of the sensors.