Skip to Main Content
Spark ignition (SI) engines operating at very lean conditions demonstrate significant nonlinear behavior by exhibiting cycle-to-cycle bifurcation of heat release. Past literature suggests that operating an engine under such lean conditions can significantly reduce NO emissions by as much as 30% and improve fuel efficiency by as much as 5%-10%. At lean conditions, the heat release per engine cycle is not close to constant, as it is when these engines operate under stoichiometric conditions where the equivalence ratio is 1.0. A neural network controller employing output feedback has shown ability in simulation to reduce the nonlinear cyclic dispersion observed under lean operating conditions. This neural network (NN) output controller consists of three NNs: a) an NN observer to estimate the states of the engine such as total fuel and air; b) a second NN for generating virtual input; and c) a third NN for generating actual control input. The uniform ultimate boundedness of all closed-loop signals is demonstrated by using the Lyapunov analysis without using the separation principle. Persistency of the excitation condition, the certainty equivalence principle, and the linearity in the unknown parameter assumptions are also relaxed. The controller is implemented for a research engine as a program running on an embeddable PC that communicates with the engine through a custom hardware interface, and the results are similar to those observed in simulation. Experimental results at an equivalence ratio of 0.77 show a drop in NO emissions by around 98% from stoichiometric levels with an improvement of fuel efficiency by 5%. A 30% drop in unburned hydrocarbons from uncontrolled case is observed at this equivalence ratio of 0.77. Similar performance was observed with the controller on a different engine.