Dynamic Mode Decomposition for Extracting Cycle-to-Cycle Variation of SIDI Engine In-Cylinder Flow Under Motoring Condition

Abstract

Abstract Cycle-to-cycle variation (CCV) of in-cylinder flow strongly affects the performance and efficiency of spark ignition direct injection (SIDI) engines. In order to achieve a precise flow control inside the engine, the underlying dynamic features of flow field CCV must be thoroughly investigated. In this work, large-eddy simulations (LES) with 50 consecutive cycles are employed for high fidelity numerical realizations of engine flow under motoring condition. To supplement the numerical analysis, time-resolved particle image velocimetry (PIV) measurements are also conducted in several cutting planes. Although the velocity root mean square (RMS) is calculated to quantify the cyclic variation intensity of simulation and experiment results, some important dynamic characteristics cannot be observed directly from velocity data. Therefore, dynamic mode decompositions (DMD), which is a widely used modal decomposition algorithm on fluid study, is used to decompose flow fields into modes with specific frequencies and provide growth rates of corresponding flow structures. This spectral information of in-cylinder flow field is ponderable for uncovering dynamic features of engine CCV. In this study, DMD algorithm is applied on both LES and PIV datasets. The frequency and growth rate differences are employed to elucidate the CCV feature deviations captured by LES and PIV. This research provides a guideline for extracting engine flow field cyclic variability feature using DMD algorithm. Based on the discussion for spectral features and potential sources of flow field variation, the capability of LES to capture CCV features is evaluated. The DMD spectrum differences between PIV and LES can guide the boundary condition perturbations used for simulation fidelity improvements.