Implementation of Soft Computing Technique for Recovery of Impulsive Heat Loads

Abstract

The knowledge of surface heat flux over aerodynamic surfaces is highly desirable for high-speed applications. Impulse test facilities like shock tubes and shock tunnels are invariantly employed for this where the aerodynamic test models experience step/ramp heat loads. Contrary to conventional methods, the usage of an advanced soft computing technique through an adaptive neuro-fuzzy inference system (ANFIS), for recovery of such surface heat loads, is theme of this paper. A coaxial thermal sensor is fabricated in house from chromel and constantan alloy. This E-type thermal probe is subjected to known heat flux (2–3.5 W) of laser light in an exclusive experimental setup, and the temperature responses are recorded. The simulations are also performed to get the temperature history for these heat loads. The experimental and computational results, either separately or together, are used to train the ANFIS network. The time-averaged values of heat flux obtained from ANFIS-based recovery shows excellent agreement in trend and magnitude (uncertainty band of $\pm 5\%$ ) with the applied heat load. The present studies demonstrate the possible use of a soft computing technique for heat flux recovery in short-duration experiments within a desired accuracy level by using training data obtained experimentally or computationally or both.