Artificial Neural Networks for Predicting the Pressure Coefficients in a Naturally Ventilated Test Room

Abstract

The objective of this work is to investigate the possibility of using artificial neural networks for the prediction of air pressure coefficients across the openings in a light weight single-sided naturally ventilated test room. Experimental values have been used for the training of the network. The outside local temperature, wind velocity and direction are monitored. The pressure coefficients at the top and bottom of the openings have been estimated from the recorded data of air pressures and velocities across the openings together with indoor air temperatures. The collected data together with the air heater load and a factor indicating whether the opening is in the windward (1) or leeward (0) direction are used as input to the neural network and the estimated pressure coefficients as the output. A general regression neural network (GRNN) was employed with one hidden slab. The training was performed with satisfactory accuracy and correlation coefficients of 0.9539 and 0.9325 have been obtained for the two coefficients respectively. Satisfactory results have been obtained when unknown data were used as input to the network with correlation coefficients of 0.9575 and 0.9320 respectively.