Analyzing In-Cylinder Flow Variations in a Motored Spark Ignition Engine using Proper Orthogonal Decomposition.

Abstract

The development of clean and efficient internal combustion engine technologies is inhibited by the current limitations in understanding cycle-to-cycle combustion variations. Cycle-to-cycle in-cylinder flow variations are thought to be one of the leading causes of cycle-to-cycle combustion variations. In this study, high-speed particle image velocimetry (PIV) data was acquired in an optical research engine with varying spatial resolution and dynamically varying time separations between PIV images for optimal velocity dynamic range throughout the engine cycle. Proper orthogonal decomposition (POD) was then used to quantitatively examine the cycle-to-cycle flow variations and intra-cycle flow evolution in these data sets. One of the causes of in-cylinder flow variations was found to be the oscillatory motion of the intake valve during its opening and closing. The scaling of in-cylinder flow with engine speed was also studied by measuring in-cylinder velocities at three different engine speeds. Further, the use of POD as a tool for differentiating between flow patterns in different data sets was demonstrated by comparing experimental data with two different large-eddy simulation data sets.

It was found that the level of cycle-to-cycle variability in intake valve oscillations influences in-cylinder flow patterns during the intake stroke. Changes in intake valve oscillations may be triggered by engine speed transients, but may also occur between different engine runs. POD was used to show that the direction and magnitude of the flow patterns during intake scale, on average, with the horizontal position of the intake valve. However, it was not possible to establish a one-on-one connection between intake valve motion and intake flow for individual cycles. Neither was a clear link found between variations in intake flow pattern and flow close to top dead center compression using two-component velocity data from the central tumble plane of the optical engine. POD was also shown to be an effective and quantitative tool for the comparison of large experimental data sets at three different engine speeds, and large combined experimental and computational data sets at 800 rpm, accurately reflecting differences in in-cylinder flow evolution and variation between the data sets.