Periodic Reporting for period 4 - MILESTONE (Multi-Scale Description of Non-Universal Behavior in Turbulent Combustion)
Période du rapport: 2020-12-01 au 2021-05-31
The main overall objectives are: 1) Establish a new multi-scale framework to analyze and model turbulent combustion phenomena based on a new way to describe turbulence using so-called dissipation elements, which are space-filling regions in a scalar field allowing to capture its small-scale morphology and non-universality. 2) Create new unprecedented datasets using direct numerical simulations (DNS) and provide new analysis methods to develop and validate combustion models; this will include automatically reducing and optimizing chemical kinetic mechanisms for use in DNS and developing an on-the-fly chemistry reduction technique. 3) Apply new modeling approaches to complex and highly non-linear modeling questions, such as pollutant formation in turbulent spray combustion. The successful outcome of the project will provide new and unprecedented datasets, a quantitative description of the impact of non-universality in small-scale turbulence on different aspects of turbulent combustion, and the basis for an entirely new multi-scale closure.
These simulations generated extremely large amount of data that should be processed in a systematic way. It would be desirable to switch from an approach which heavily relies on experience and intuition to a formal assessment based on methodical approaches. A rigorous data-driven approach allows to infer the deficiency of existing models with the final goal of quantifying and reducing model uncertainty. We combined a number of different approached such as the Dissipation Element analysis and the Optimal Estimator concept, which have been employed in different configurations.
In addition, modern methodologies developed in the context of Big Data analysis, Artificial Intelligence and Machine Learning are starting to be popular in the field of turbulence and combustion. Our team is exploiting and adapting these approaches for the analysis of complex reactive systems and for the development of models. One example is given by the use of Neural Networks, which provided a convenient framework for the statistical analysis of combustion-generated particulate in a turbulent flame.
The understanding gained applying the Dissipation Element analysis will be employed to refine and develop new strategies for modeling of turbulent combustion. The same type of analysis is being applied to pollutants and multiphase phenomena and is expected to clarify some of the mechanisms governing their behavior.