An innovative approach for prediction of aerodynamic coefficients in shock tunnel testing with soft computing techniques

Abstract

A blunt bi-conic aluminum model, integrated with a three component accelerometer force balance system, is tested in the IITB-Shock Tunnel at various angle of inclinations with an intention to measure aerodynamic coefficients. Initially, Genetic Algorithm (GA) is employed to deduce the orthogonal inputs and their responses from distributed point loads applied on the test model during calibration experiments. Time histories of these loads and their responses are then used to train the architecture of Adaptive Neuro Fuzzy Inference System, ANFIS. This ANFIS architecture is further employed for force prediction from the acquired acceleration responses during shock tunnel experiments. Testing of the experimental acceleration responses of 0° and 10° angle of attack experiments showed encouraging agreement for recovered aerodynamic coefficients with the accelerometer balance theory based predictions. These results clearly showed the necessity to consider multiple point loads and their acceleration responses for training the soft computing algorithm or calibration of the force balance. Further, use of only one point force and its responses, for training ANFIS, is not only seen to have discrepancy in prediction but also is noted to be non-unique due to choice of the loading point. Moreover, it is recommended to choose the loading point near the center of pressure for the tested experimental conditions to incur lower discrepancy in prediction using single point loading data for ANFIS training.