SIMULATION OF A RANGE OF THERMAL SYSTEMS BY ARTIFICIAL NEURA1 NETWORKS

摘要

The Mechanical & Manufacturing Engineering Research Unit at the University of Glamorgan has been involved for the last 5 years in the application of neural networks to simulate the behaviour of a range of high temperature systems. Consequently this paper presents results from a number of studies concerned with boilers, furnaces and heat treatment processes. The first of these studies involves the prediction of heat transfer rates from air-assisted atomised water sprays for cooling aerospace forgings from high temperatures using the nozzle air and water pressures as inputs to the network. The second case is concerned with prediction of pollutant emissions from a coal-fired boiler using previous measured emission concentrations and an indication of the current boiler load and combustion air flow rate. In both these cases experimental results were used to train the networks. The final example deals with the prediction of load temperatures in an intermittently operated, gas-fired, metal reheating furnace. In contrast, in this example, data generated from a previously validated mathematical model of the furnace were employed for training purposes. Artificial neural networks were found to provide adequate representation of all three systems.