This paper presents a simulation model of a thermofluid system

which is based on combined neural network and process mechanism characteristics.It is an artificial neural network model

the ideal simulation speed thus is ensured.During design of the network model

the physical fundamental with which the system between input and output is considered.This network model is a "grey box" rather than a conventional "black box" to a certain degree.Besides the conventional learning function of numerical mapping between input and output

training of the network reflects the learning of physical mechanism between input and output

thereby associative ability

extending inference ability and recurrent ability of the network model are ensured.