Dr. Himanshu Dave (Post Doc)

“I am passionate for finding patterns in flow field”

Role in Vademecom:
I am currently working on two different problems. The first one is aimed at improving our understanding about MILD combustion systems. Using unsupervised machine learning (ML) techniques such as Vector Quantization Principal Component Analysis (VQPCA), our objective is to assess what are the patterns that such algorithms recognize and use that to best characterize the combustion system. For example, using thermo-chemical variables from direct numerical simulation (DNS) dataset as input to VQPCA, we able to define that one system is dominated more by heat release due to higher oxygen dilution levels. The other system is more dominated by mixing due to reduced levels of chemical reactions because of lower oxygen dilution level. In the second problem, our aim is to eventually improve our understanding and access to information in experiments. The objective is to use PCA-based methods to reconstruct additional state-space variables which are not directly measured currently. This may also help in benchmarking Reynolds-averaged Navier Stokes (RANS) and large-eddy simulation (LES) models against the experiments due to availability of additional information.

Email Himanshu (himanshu.dave@ulb.be) or check his profile on: