Analysis of misfires in a direct injection engine using proper orthogonal decomposition

Abstract

Previous work demonstrated that the occasional misfired and partially burned cycles (MF) in a stratified-charge, spark-ignited direct injection engine always achieved an early flame kernel, but failed to reach and inflame the fuel in the bottom of the piston bowl. This conclusion was derived from intra-cycle crank angle resolved velocity and fuel concentration images that were recorded simultaneously using high-speed particle image velocimetry and planar laser-induced fluorescence. In this study, both ensemble average analysis, conditionally sampled on either MF or Well Burned (WB) cycles and proper orthogonal decomposition (POD) are applied separately to the velocity and fuel distributions. POD of the velocity and fuel distributions near the spark plug were performed, and the mode energy and structure of the modes are compared. This analysis is used to assess the similarity and differences between the MF and the WB cycles and to identify physical insight gained by POD. The POD modes were determined from the combined set of 200 WB and 37 MF cycles to create two sets of 237 orthogonal modes, one set for the velocity, V, and one for the equivalence ratio, ε. Then, conditionally sampled averages of the POD coefficients could be used to quantify the extent to which each mode contributed to the MFs. Also, the probability density functions of the coefficients quantified the cyclic variability of each mode’s contribution. The application of proper orthogonal decomposition to velocity and equivalence ratio images was useful in identifying and analyzing the differences in flow and mixture conditions at the time of spark between well-burning and misfiring cycles. However, POD results alone were not sufficient to identify which of the cycles were misfiring cycles, and additional information was required for conditional sampling.