Prediction of Experimental Surface Heat Flux of Thin Film Gauges using ANFIS

Precise quantification of surface heat fluxes in highly transient environment is of paramount importance from the design point of view of several engineering equipment like thermal protection or cooling systems. Such environments are simulated in experimental facilities by exposing the surface with transient heat loads typically step/impulsive in nature. The surface heating rates are then determined from highly transient temperature history captured by efficient surface temperature sensors. The classical approach is to use thin film gauges (TFGs) in which temperature variations are acquired within milliseconds, thereby allowing calculation of surface heat flux, based on the theory of one-dimensional heat conduction on a semi-infinite body. With recent developments in the soft computing methods, the present study is an attempt for the application of intelligent system technique, called adaptive neuro fuzzy inference system (ANFIS) to recover surface heat fluxes from a given temperature history recorded by TFGs without having the need to solve lengthy analytical equations. Experiments have been carried out by applying known quantity of ‘impulse heat load’ through laser beam on TFGs. The corresponding voltage signals have been acquired and surface heat fluxes are estimated through classical analytical approach. These signals are then used to ‘train’ the ANFIS model, which later predicts output for ‘test’ values. Results from both methods have been compared and these surface heat fluxes are used to predict the non-linear relationship between thermal and electrical properties of the gauges that are exceedingly pertinent to the design of efficient TFGs. Further, surface plots have been created to give an insight about dimensionality effect of the non-linear dependence of thermal/electrical parameters on each other. Later, it is observed that a properly optimized ANFIS model can predict the impulsive heat profiles with significant accuracy. This paper thus shows the appropriateness of soft computing technique as a practically constructive replacement for tedious analytical formulation and henceforth, effectively quantifies the modeling of TFGs.