A fast method for the determination of the heat transfer coefficient of external wall

In this study, the main objective is to provide a fast method for the determination of the heat transfer coefficient of external wall based on artifical neural network and the inspection method of using cold box and hot box. Three backpropagation neural networks were modeled and the inputs of the networks for training and testing were considered as the first 3, 4 and 5 sets of data measured in field inspection. The simulation results of ANNs show that the data of the first 4 or 5 hours in measurement are enough for the prediction of the heat transfer coefficients of external wall, and this method reduces inspection duration from 48 to 4 hours with mean absolute error below 5% and can be used in the inspection to replace conventional method.