Computational Modeling of the Soot Size Distribution in Turbulent Reacting Flows: Leveraging Data Science Tools to Rapidly Accelerate Physics-Based Model Development and Validation

Princeton University, Princeton NJ

Investigators

Abstract

Soot is an undesirable by-product of combustion formed during incomplete fuel-rich combustion of hydrocarbon fuels and has adverse effects on human health and the environment. These adverse effects vary with particle size, so controlling the distribution of soot particle sizes emitted from practical combustion systems is of increasing interest. Unfortunately, even the basic qualitative features of the soot size distribution in turbulent combustion are unknown. As a result, computational capabilities do not currently exist for efficiently predicting the soot size distribution in turbulent combustion. This project will develop fundamental understanding of the evolution of the soot size distribution in turbulent combustion using detailed, full-fidelity computational simulations, accelerated using deep learning. This knowledge will then be leveraged to develop a new computationally efficient modeling framework for predicting the evolution of the soot size distribution in turbulent combustion applicable to practical combustion systems. The new modeling framework will include a novel statistical model for the soot size distribution, a fundamentally new approach to capture the relatively slow chemistry of gaseous soot precursors, and a new turbulent transport model. The new modeling framework will be ultimately validated against experimental measurements. The software implementation of the new modeling framework will also be made publicly available for use by other researchers and in industry. Computational approaches for predicting the soot size distribution in turbulent reacting flows are extremely limited, generally restricted to oversimplified and inaccurate soot models yet still computationally very expensive. While some limited evidence has suggested that the soot size distribution may sometimes be unimodal in turbulent flames and may preferentially generate very large soot particles, the influence of turbulence on the soot size distribution is fundamentally poorly understood. This project will develop a computationally efficient modeling framework for predicting the evolution of the soot size distribution in turbulent combustion. An Adaptive Sectional-Moment (ASM) model will be developed that combines a coarse sectional method in one internal coordinate with an advanced moment method in two internal coordinates within each section. ASM will be used to conduct Direct Numerical Simulations (DNS), accelerated using a combination of deep learning and data-derived manifolds, to provide unprecedented insights into the influence of turbulence on the soot size distribution. These databases will be further used to address two key modeling challenges in Large Eddy Simulation (LES) relevant to the soot size distribution. The slow chemistry of Polycyclic Aromatic Hydrocarbons (PAH) will be captured in an accurate yet consistent manner though a new partially non-equilibrium manifold model, and new models for the small-scale turbulent transport of soot will be developed utilizing physics-based and data-based approaches. Ultimately, the new LES modeling framework with ASM will be validated against the DNS databases and experimental measurements of the soot size distribution in turbulent flames. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.