Peak Pressure Position Control of Four Cylinders through the Ion Current Method

The ion current sensing technique is used in this paper with available engine signals (engine speed, N, and manifold absolute pressure, MAP) to train neural networks (NN) to estimate the peak pressure position (PPP) across four cylinders of a spark ignition internal combustion engine. The stochastic nature across these four cylinders is evident; the variability in the PPP is a highly useful measure of cycle to cycle variation (CCV) of combustion since it may be determined directly and so can be used in feedback control. After experimental implementation on the engine, it is seen that the technique gives reliable PPP estimation for control feedback. In addition the PPP is known to correlate well with spark advance (SA) for maximum best torque (MBT) [1].

A constrained variance (CV) control technique is a solution to reducing the variability in the PPP and so feedback is implemented to control the SA. The PPP variation across the four cylinders is reduced and stabilized at desired settings experimentally, resulting in a decrease in the differences between indicated mean effective pressure (IMEP) values across all four cylinders and a smoother engine operation.