Modelling energy consumption in test cells

In situ measurements of energy consumption in test cells are often carried out to predict the thermal performance of the insulating product. Design engineers need a tool that can use available in situ measurements to model the thermal performance of the test cell at any location world wide and also permit comparison of different insulation products. Towards this goal, we present GAP, Global Assimilation Process which is a neural network based meta modeling technique. The key feature of this method is the zero memory minimization routine and a regularization technique that avoids over training of the network. We present the theory and applications of GAP software to predict thermal performance of mineral wool and multi-foil insulation products. GAP based meta models are used to predict thermal performance of these insulation products at different test sites. It is shown that the properly trained neural network model of GAP can accurately predict the energy consumption in test cells of any location.