Prediction of coating thickness through thermal analysis

Abstract

An attempt has been made to develop a methodology to predict the uniform or non-uniform thickness of a coating material on a homogeneous solid material. At first the effects of variation of (uniform and non-uniform) coating thickness, on the heat transfer (by conduction) through the coated body have been rigorously investigated. A solid square steel block coated with copper (of various coating thicknesses) has been considered as the sample problem. A constant heat flux has been maintained at the bottom side where, the other three sides kept at constant temperature. The temperature profile built at the bottom side of the square block is different for different (uniform or non-uniform) coating thickness. Depending on the statistical parameters (Mean, Standard Deviation, Skew-ness and Kurtosis) extracted from the temperature profile (which has been found through CFD analysis), it may possible to predict the thickness of the coating. Two cases namely: uniform varying thickness and non-uniform varying thickness have been considered in the present problem. The training and test data have been collected using 2D conduction analysis with Finite Volume Method (FVM) in a fully automated way. Feed-Forward Neural Network (NN) with Back-Propagation Algorithm is used to predict the coating thickness. The NN has been modelled using ‘C’ programming Language in LINUX operating system. A Genetic Algorithm (GA) has also been coded (using ‘C’ programming language in Linux operating system) to optimize the performance of the NN. The above described methodology based on thermal analysis can also be used in thermography or in other prediction problem.