Internal combustion engines are being required to comply with increasingly
stringent government exhaust emissions regulations. Compression ignition (C
I) piston engines will continue to be used in cost-sensitive fuel applicati
ons such as in heavy-duty buses and trucks, power generation, locomotives a
nd off-highway applications, and will find application in hybrid electric v
ehicles. Close control of combustion in these engines will be essential to
achieve ever-increasing efficiency improvements while meeting increasingly
stringent emissions standards. The engines of the future will require signi
ficantly more complex control than existing map-based control strategies, h
aving many more degrees of freedom than those of today.
Neural network (NN)-based engine modelling offers the,potential for a multi
dimensional, adaptive, learning control system that does not require knowle
dge of the governing equations for engine performance or the combustion kin
etics of emissions formation that a conventional map-based engine model req
uires. The application of a neural network to model the output torque and e
xhaust emissions from a modern heavy-duty diesel engine (Navistar T444E) is
shown to be able to predict the continuous torque and exhaust emissions fr
om a heavy;duty diesel engine for the Federal heavy-duty engine transient t
est procedure (FTP) cycle and two random cycles to within 5 per cent of the
ir measured values after only 100 min of transient dynamometer training. Ap
plications of such a neural net model include emissions virtual sensing, on
-board diagnostics (OBD) and engine control strategy optimization.