Neural network and zone modelling of a gas-fired furnace operating under non-steady state conditions

摘要

This paper initially compares predictions and measurements from a range of gas-fired furnaces to demonstrate that mathematical models,based on hte zone mehtod of radiation analysis,can simulate satisfactorily the transient behaviour of these installations.However the computational times associated with these models are generally too great for on-line ocntrol purposes and it has been roposed that neural network models are an alternative for this purpose.Consequently the second part of the paper is ocncernedwith representation of a zone model of a metal reheating furance starting up from "cold" by an appropriate artificial neural netowkr.The predictions fromthe two models are good agreement with respect to hte duration of the inital start up period,the energy consumption and,to a lesser extent,the load surface temperature at discharge.