Neural network control of automotive fuel-injection systems

A neural network methodology is developed for air-to-fuel (A/F) ratio control of automotive fuel-injection systems. The dynamics of internal combustion engines and fuel-injection systems are extremely nonlinear, impeding methodical application of control theories. Thus, the design of standard production controllers relies heavily upon calibration and look-up tables. A neural network-type controller is developed in this article for its function-approximation abilities and its learning and adaptive capabilities. A cerebellar model articulation controller (CMAC) neural network is implemented in a research automobile to demonstrate the feasibility of this control architecture. Experimental results show that the CMAC fuel-injection controller is very effective in learning the engine nonlinearities and in dealing with the significant time-delays inherent in engine sensors.<>