Linear stochastic estimation of the coherent structures in internal combustion engine flow

Abstract

A methodology for estimating the in-cylinder flow of an internal combustion engine from a number of point velocity measurements (sensors) is presented. Particle image velocimetry is used to provide reference velocity fields for the linear stochastic estimation technique to investigate the number of point measurements required to provide a representative estimation of the flow field. A systematic iterative approach is taken, with sensor locations randomly generated in each iteration to negate sensor location effects. It was found that an overall velocity distribution accuracy of at least 75% may be achieved with 7 sensors and 95% with 35 sensors, with the potential for fewer if sensor locations are optimised. The accuracy of vortex centre location predictions is typically within 2–3 mm, suggesting that the presented technique could characterise individual cycle flow fields by indicating vortex locations, swirl magnitude or tumble, for example. With this information on the current cycle, a control system may be enabled to activate in-cycle adjustment of injection and/or ignition timing, for example, to minimise emissions.