Title: Constrained variance control of peak-pressure position by spark-ionisation feedback for multi-cylinder control
Authors: N. Rivara, P.B Dickinson, A.T. Shenton
Addresses: Department of Engineering, The University of Liverpool, Brownlow Hill, Liverpool, L69 3GH, UK. ' Department of Engineering, The University of Liverpool, Brownlow Hill, Liverpool, L69 3GH, UK. ' Department of Engineering, The University of Liverpool, Brownlow Hill, Liverpool, L69 3GH, UK
Abstract: A neural network (NN) is identified in conjunction with the ionisation current from the spark plug of a spark ignition (SI) gasoline internal combustion (IC) engine as a peak-pressure position (PPP) virtual sensor. The NN is trained offline using reduced datasets from principal component analysis (PCA) and available engine signals to predict the PPP under transient load, speed and spark advance (SA) settings. An ARMAX model is identified around the NN that allows linear feedback control techniques to be applied. Constrained variance (CV) control is used to ensure the PPP tracks to a set point known to give minimum advance for best torque (MBT) by actuation of the SA. The technique is then replicated on all the cylinders of a four-cylinder 1.6l gasoline IC engine which experimentally demonstrates a reduction in cylinder-to-cylinder differences of indicated mean effective pressure (IMEP) after applying the control.
Keywords: constrained variance control; minimum variance; ionisation current; neural networks; ANNs; principal component analysis; PCA; peak pressure position; spark ionisation feedback; multi-cylinder control; spark plugs; spark ignition; internal combustion engines; gasoline IC engines; virtual sensors.
International Journal of Advanced Mechatronic Systems, 2009 Vol.1 No.4, pp.242 - 250
Available online: 06 Jun 2009 *Full-text access for editors Access for subscribers Purchase this article Comment on this article