Large-scale vortex motion could enhance the fuel-air mixing and the combustion stability inside a direct-injection engine. For in-cylinder vortex motion analysis, detection of the vortex features is usually a challenging task because of large cyclic variations of vortex structure, number, and locations. In previous study, K-means clustering has been successfully applied to detect the vortex features and quantify the cyclic variations. However, K-means algorithm is somewhat limited in analyzing the complex flow with multiple vortex zones and vortex merging. Therefore, we propose a hybrid clustering method which blends in two clustering algorithms, namely, Gaussian mixture model (GMM) clustering and K-means clustering, to optimize the analysis of vortex classification, vortex zone detection, and cyclic variation quantification under different flow conditions. In this study, in-cylinder flow fields with varying degrees of swirling behavior were recorded by high-speed particle image velocimetry (PIV) under high and low swirl ratio conditions. Owing to the probability density contour feature of GMM, the clustering results show that the hybrid clustering method improves the accuracy of vortex classification and vortex zone detection, especially for low swirl cases where there exist large cluster overlaps. Additionally, this hybrid clustering method is able to retain the quantification of cyclic variations by blending K-means clustering to quantify the distance. In summary, the transient features and cyclic variations of the in-cylinder vortex motion under different swirl ratio conditions can be accurately revealed by this hybrid clustering method.
Hybridizing Unsupervised Clustering Methods for In-Cylinder Vortex Motion Analysis under Different Swirl Ratio Conditions
Sae Technical Papers
SAE WCX Digital Summit ; 2021
2021-04-06
Aufsatz (Konferenz)
Englisch
Characterisation of Barrel Swirl Motion Under Steady Flow Conditions
SAE Technical Papers | 1995
|Characterisation of barrel swirl motion under steady flow conditions
Kraftfahrwesen | 1995
|Unsupervised Clustering of Highway Motion Patterns *
IEEE | 2019
|Numerical Estimation of Medium-Speed Engine Swirl Ratio by Cylinder Induction Modeling
SAE Technical Papers | 2023
|