On the numerical assessment of engine exhaust plumes with a scaled cowling

The ability to predict the characteristics of engine exhaust is important to determining heat signatures on various components of the aircraft body. This is often obtained through numerical means such as RANS, which however relies closely on the choice of the turbulence model for accurate predictions. In this study, predictions of exhaust plumes from four turbulence models are compared against results from particle image velocimetry of a scaled helicopter engine exhaust. These models include the standard k - epsilon, realizable k - epsilon, shear-stress transport (SST) k - omega, and Durbin's turbulence models. All four turbulence models managed to capture the general shape of the exhaust when analyzed through the velocity contours at two measurement windows. However, the comparisons of velocity contours fail to describe the shift of the predicted plume from the experiments, which is important for fuselage/tail impingement. To obtain further insights to the shifts, a visual correlation in the form of a confidence ellipse through principal component analysis (PCA) is introduced and plotted for the predicted plumes. All the models' plume predictions showed quantifiable shifts in their mean-centers when compared to the measurements. In terms of matching the measurements' statistical maximum variance of the plume distribution at the furthermost plane, it was found that the realizable k - epsilon performed the best among the models. On the other hand, the SST k - omega and Durbin's model performed the best in predicting bivariate (x and y coordinates) distribution of the plume at the furthermost plane.

Automated summary

The ability to predict engine exhaust characteristics is important for determining the heat signatures on aircraft components. In this study, the predictions of four turbulence models were compared to particle image velocimetry results for a scaled helicopter engine exhaust. All four models were able to capture the general shape of the exhaust, but failed to accurately describe the shift of the predicted plume from the experiments. A visual correlation using principal component analysis showed quantifiable shifts in the predicted plume mean-centers compared to the measurements. The realizable k-ε model performed the best in matching the statistical variance of the plume distribution, while the SST k-ω and Durbin models performed the best in predicting the bivariate distribution of the plume.