MechE-CSE PhD Thesis Defense | Xinlin Zhong

Xinlin Zhong, MechE-CSE PhD Thesis Defense Announcement

Thesis Title: Physics-based and Machine Learning Hybrid Modeling of Oil Transport and Oil Consumption in Internal Combustion Engines – a Digital Twin

Date: Tuesday, July 29, 2025
Time: 3 PM ET
Location: 3-270 / Zoom

Thesis Committee:

• Dr. Tian Tian, Department of Mechanical Engineering, MIT (advisor and chair)
• Professor Pierre Lermusiaux, Department of Mechanical Engineering, MIT
• Professor Wai K. Cheng, Department of Mechanical Engineering, MIT

Abstract:

A considerable amount of resources is invested by the automotive industry in combating internal combustion (IC) engine emissions resulting from lubrication oil consumption (LOC). Reducing LOC requires a thorough understanding of the oil transport in the piston ring pack, which is a multi-physics process with different length and time scales. Particularly, the multi-time scale nature has never been addressed in the field.

This thesis develops a comprehensive digital twin framework to model oil transport in the piston ring pack across engine cycles and to predict LOC. A modular model architecture is employed, and a hybrid approach combining machine learning and physics-based models accurately and efficiently resolves each mechanism. The model incorporates essential physics responsible for the oil flow from the bottom of the ring pack to the top, including piston ring dynamics, gas flow, oil transfer between ring and liner, oil redistribution on piston land and grooves, oil vaporization, and ring rotation. The effects of expansive design and operating parameters are also considered. Three LOC sources, namely, oil vaporization on the liner, reverse gas flow, and the top ring up-scraping, are identified, and they are all found to be highly dependent on the ever-changing relative ring gap locations. As such, without knowing the ring gap locations, only a range of LOC can be predicted. 

To shed light on the enduring myth of ring rotation, a model is developed for the first time. An investigation into the driving mechanisms reveals piston secondary motion and bore expansion as the dominant factors in determining the rotation behavior. It was found that under a constant friction coefficient between the ring and the groove, stationary points inevitably arise for the top two rings, and a non-uniform distribution is needed to render any ring rotation patterns. By applying a gradient-based calibration procedure to the ring groove friction coefficient distribution, this work identifies the necessary ring groove lubrication condition to match experimentally observed ring rotation patterns.

The Digital Twin framework is a valuable tool for studying how the lubricating oil is supplied, distributed, released, and consumed in the piston system, and the effects of design and operating parameters. The framework integrates processes at different length scales and traces the evolution of the oil distribution over engine cycles. For the first time, the unsteadiness of the oil transport in the ring pack under a steady engine operation is modeled with a reasonable computation time. The model is a foundation for future ring pack design optimization tasks where a balance between oil consumption and proper lubrication must be reached. Based on the current framework, future improvements on ring/groove lubrication conditions and the liner finish effects can expand the predictability of the model in both time and length scales.