Soft Computing Model for Inverse Prediction of Surface Heat Flux From Temperature Responses in Short-Duration Heat Transfer Experiments

Abstract

Abstract Aerodynamic experiments in the high-speed flow domain mainly rely on precise measurement of transient surface temperatures and subsequent quantification of heat flux. These experiments are primarily simulated in high-enthalpy short-duration facilities for which test flow durations are in the order of a few milliseconds, and the thermal loads resemble the nature of step/impulse. This study focuses on a specially designed fast-response coaxial surface junction thermal probe (CSTP) with the capability of capturing transient temperature signals. The CSTP, with a 3.25 mm diameter and 13 mm length, incorporates a precisely examined sensing junction (20 µm thickness) and EDX, FESEM verified surface characterization. The short-duration calibration experiments are realized to mimic the simulated flow conditions of high-enthalpy test facilities. The classical one-dimensional heat conduction modeling has been used to deduce surface heat flux from the acquired temperature responses. It demonstrates a commendable accuracy of ±2.5% when compared with known heat loads of calibration experiments. Departing from traditional heat conduction models, an advanced soft-computing technique, the Adaptive Neuro-Fuzzy Inference System (ANFIS), is introduced for short-duration heat flux predictions. This methodology successfully recovers known (step or ramp) heat loads within a specific experimental time frame (0.2 s). The results exhibit excellent agreement in the prediction of trend and magnitude, carrying uncertainties of ±3% for radiative and ±5% for convective experiments. Consequently, the CSTP appears as a rapidly responsive transient heat flux sensor for real-time short-duration experiments. The soft-computing approach (ANFIS) offers an alternative means of heat flux estimation from temperature history irrespective of the mode of heat transfer and nature of heat load, marked by its prediction accuracy, diminished mathematical intricacies, and reduced numerical requisites.