Neural network control for reducing engine speed fluctuation at idle

Long term average idle speed control has been studied in most engine idle control systems in automotive engines, but they allow undesirable short-term engine speed fluctuation even under steady idle conditions. The difference in the torque production among cylinders influences the idle stability by making the fluctuation of engine speed ripples. We suggest that the control of the spark ignition timing for each cylinder based on a neural network reduces the unbalanced combustion among cylinders, and maintains uniform and stable engine speed. We apply genetic algorithms to the neural network structure with oscillatory neurons in a sensor array in order to decrease the engine speed fluctuation efficiently.