An Inverse Method to Estimate the Principal Thermal Conductivities of Composite Material

This paper reports methodology for simultaneous estimation of principal thermal conductivity of a composite made of aluminium core with aluminium face sheet bonded by using an adhesive layer, using inverse method. The direct problem consists of a three dimensional heat conduction equation in an orthotropic composite medium. The direct problem is solved by using fluent to estimate the temperature distribution on the test sample by varying the thermal conductivity at steady state condition. An artificial neural network is used as an inverse tool to retrieve the thermal conductivity. Sixty sample data temperature values are used to train a neural network by randomly varying the thermal conductivity within a range of 0.5 x, ky z independence study has been carried out to find the number of neurons required to train the network based on the mean square error and regression values. The methodology is demonstrated by retrieving thermal conductivity from the trained network by giving a temperature data that were not used for the training. The retrieved values of thermal conductivities from the trained network were found to deviate to a maximum of ±15% from the actual values. Keywords— Composite material, Inverse heat transfer, Artificial neural network.