A novel comparative approach on inverse heat transfer analysis of an experimental setup of an extended surface

Abstract

An extended surface was used for the thermal analysis, where different optimization algorithms were compared. The thermal conductivity, heat transfer coefficient and surface emissivity were considered non-linear, temperature-dependent. The direct analysis involved pdepe solver, which was validated with Differential Transform Method (DTM) and finite difference method (FDM). An experiment was also conducted to obtain the temperature profile. Furthermore, a sensitivity analysis was done and critical parameters Φ, β, Nc, Nr, Bi, Nbr were reported, which were estimated using the combination of regularization techniques and optimization algorithms by the inverse approach. A comparison of linear least squares, Tikhonov regularization, Lasso estimator and elasticnet regularization have been done. It has been reported that elasticnet (90% lasso and 10% Ridge) objective function was best, with α = 10−4. A case study based on computational and the experimental temperature profile, to compare optimization algorithms, namely, Differential evolution (DE), Particle swarm optimization (PSO), Whale optimization algorithm (WOA), Water cycle algorithm (WCA), hybrid of the Grey wolf optimization-Cuckoo search (GWOCS), Butterfly optimization algorithm (BOA) and Atom search algorithm (ASO) was done. The top three algorithms reported in case I, were BOA, WOA and WCA with performance parameter 0.77, 0.75, 0.74, whereas in case II, were WOA, BOA and WCA with performance parameter 0.78, 0.60 and 0.50 respectively.