Colloquia

Summary

The breakup of liquid jets is a complex fluid dynamics phenomenon with significant implications for applications such as inkjet printing, fuel injection, and atomisation processes. A key mechanism governing this behaviour is the Rayleigh-Plateau instability, where surface tension forces cause a liquid column to fragment into droplets. This thesis investigates jet breakup dynamics using computational fluid dynamics (CFD) simulations alongside machine learning (ML) models to accelerate predictions. Traditional CFD simulations, based on the Navier-Stokes equations for a thin axisymmetric liquid jet, are computationally expensive, posing challenges for real-time applications. To address this, surrogate ML models are explored as a faster alternative while maintaining predictive accuracy. Training data is generated through CFD simulations, with a particular focus on the Ohnesorge number, which influences droplet formation dynamics. The results are validated through comparisons with amplitude analysis based on the Rayleigh–Plateau instability and other numerical methods. Fully connected neural networks (FNNs) and long short-term memory (LSTM) networks are employed to predict droplet formation and jet breakup. The results in accuracy and computational efficiency demonstrate the potential of ML-based surrogate models for real-time monitoring of jet breakup dynamics.