Numerical Inspection of Heterogeneity: Homogeneous Intrusion in Bulk

Abstract

Composites are often constructed by combining two or more anisotropic materials for various objectives. But thermal cracking may occurred along interfaces due to the mismatch of thermal properties between contiguous materials. Therefore, this type of thermal cracking is one of the most crucial concerns in the use of such composites. During manufacture of any product (of particular material) through casting or any other process, a homogeneous intrusion of any other material may occurred with the bulk of main material. Under thermal loadings, these manufacturing defects may even propagate and eventually cause failure of components. Therefore it is clear that the study on the thermal conduction across those defective components plays a crucial role to provide accurate assessment of possible damaging or cracking due to the defect.
Thermography is a very effective tool for early detection of breast cancer and for estimation of Tumor location and parameter. Numerical modelling of heat transfer within a woman breast is being an attractive tool that may reveal the conditions under which tumours can be detected in a thermogram. However, to make a proper diagnosis, thermograms alone will not be sufficient. To analyze the thermogram objectively, analytical tools like statistical methods and artificial neural network are recommended to be incorporated in various journals.
The present work investigates the effect on heat conduction due to the homogeneous intrusion in a bulk. The temperature-variation on the surface of the bulk has been used here to detect the position and size of the intrusion. Then a model is proposed to predict the position (within the bulk) and size of the intrusion by knowing the temperature distribution on the surface. The basic goal is to correlate the temperature distribution on the surface with the size and position of the intrusion. The principle developed here may also be used for estimation of tumour location and parameter.
Attempt has been made to develop methodologies for prediction in heat conduction problem using ‘C’ programming language and commercial software: Gambit and Fluent. The performances of the developed approaches have been tested to predict the location of a circular intrusion within a square slab from some known temperature values collected elsewhere at the bottom side (bottom wall) of the slab. The data (temperature history along the bottom wall for different positions of the intrusion) have been numerically evaluated by the forward calculation using commercial CFD software: Gambit and Fluent in a fully-automated way by an indigenously developed journal file. Two approaches have been developed to solve the prediction problem. In Approach 1, a multi-layer feed-forward NN with BP algorithm has been used, where the NN-parameters have been determined through a thorough parametric study by varying them one after another in stages. In approach 2, the NN-parameters have been optimized simultaneously by utilizing one global optimizer like GA and a local optimization tool, namely BP algorithm.