Investigation of Droplet Shedding in an Aeroengine Bearing Chamber Using Convolutional Neural Networks

Farbrother, Katrina; Cageao, Paloma Paleo; Johnson, Kathy; Ambrose, Stephen

Abstract

More efficient aeroengines are needed to reduce the environmental impact of civil aviation. Design changes for increased efficiency mean that higher loads are transmitted through the bearings supporting the rotating shafts. Better bearing chamber design is needed to facilitate these improvements, meaning that higher design capability and therefore, and improved understanding of the two phase flow is required to support modelling of the bearing chamber for future design. Nottingham University’s Bearing Chamber Rig (BCR) is designed to provide quantitative and qualitative two phase flow data in a simplified, representative bearing chamber geometry. Oil is introduced to the chamber via a rotating cup and static configuration and this paper presents data characterising droplet shedding into the chamber. High-speed imaging data is obtained at the 3 o’clock and 9 o’clock positions. Automated image processing is necessary due to the large quantities of data. A Convolutional Neural Network (CNN) has been developed and is used to detect droplets in the images and convert the data to a numerical format including droplet diameter and speed. The CNN used was trained using images from a different but similar experimental test rig and this paper shows that it can also detect droplets with good accuracy in the BCR. The experimentally obtained droplet data provides better understanding of the two-phase flow behaviour in the test rig and by extension in an aeroengine bearing chamber. Droplet diameters, normalised against nominal film thickness immediately prior to shedding, are found to be around 1.0 and normalised Sauter Mean Diameters are between 1.1 and 1.6 with diameter not significantly affected by shaft speed or oil flowrate within the ranges investigated. Droplet velocities were found to be between 4% and 13% of the cup rim speed with the numerical value showing little variation over the investigated range. This diameter and velocity distribution information supports future design capability.

Citation

Farbrother, K., Cageao, P. P., Johnson, K., & Ambrose, S. (2022, June). Investigation of Droplet Shedding in an Aeroengine Bearing Chamber Using Convolutional Neural Networks. Presented at ASME Turbo Expo 2022: Turbomachinery Technical Conference and Exposition, Rotterdam, Netherlands

• Presentation Conference Type: Edited Proceedings
• Conference Name: ASME Turbo Expo 2022: Turbomachinery Technical Conference and Exposition
• Start Date: Jun 13, 2022
• End Date: Jun 17, 2022
• Acceptance Date: Mar 15, 2022
• Online Publication Date: Oct 28, 2022
• Publication Date: Jun 13, 2022
• Deposit Date: Apr 9, 2022
• Publisher: American Society of Mechanical Engineers
• Peer Reviewed: Peer Reviewed
• Volume: 8A
• Article Number: GT2022-82275, V08AT22A013
• Book Title: Proceedings of ASME Turbo Expo 2022: Turbomachinery Technical Conference and Exposition
• ISBN: 9780791886069
• DOI: https://doi.org/10.1115/GT2022-82275
• Public URL: https://nottingham-repository.worktribe.com/output/7734295
• Publisher URL: https://asmedigitalcollection.asme.org/GT/proceedings-abstract/GT2022/86069/V08AT22A013/1149092
• Related Public URLs: https://asme-turboexpo.secure-platform.com/a