BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.4.12//EN
TZID:Asia/Kolkata
X-WR-TIMEZONE:Asia/Kolkata
BEGIN:VEVENT
UID:231@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20261105T110000
DTEND;TZID=Asia/Kolkata:20261105T120000
DTSTAMP:20261029T142048Z
URL:https://cds.iisc.ac.in/events/ph-d-thesis-defense-toward-machine-learn
 ing-accelerated-simulations-of-cavity-stabilized-flames/
SUMMARY:Change in Date: Ph.D Thesis Defense: Toward machine learning-accele
 rated simulations of cavity-stabilized flames
DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nPh.D. Thesis Def
 ense\n\n\n\nSpeaker: Mr. Priyabrat Dash\nS.R. Number: 06-18-01-10-12-20-2-
 19095\nTitle: Toward machine learning-accelerated simulations of cavity-st
 abilized flames\nResearch Supervisor: Prof. Konduri Aditya\nThesis examine
 r : Dr. Shubhangi S. Bansude\, Indian Institute of Technology\, Gandhinaga
 r\, Gujarat\nDate &amp\; Time : November 05\, 2026 (Thursday)\, 11:00 AM\n
 Venue : #102\, CDS Seminar Hall\n\n\n\nABSTRACT\nDirect numerical simulati
 ons (DNS) provide a rigorous framework for resolving the multiscale intera
 ctions governing turbulent reacting flows in practical combustors. By capt
 uring all dynamically relevant spatio-temporal scales\, DNS complements ex
 perimental approaches by providing access to flow and thermochemical quant
 ities that are difficult to measure at the required spatio-temporal resolu
 tion. Cavity flameholders\, essential to gas turbines and ram/scramjet eng
 ines\, represent one such configuration where this capability is especiall
 y valuable. These devices generate low-speed recirculation zones that cont
 inuously entrain heat and radicals\, facilitating flame stabilization. Des
 pite its unparalleled detail\, the prohibitive computational cost of DNS l
 imits its adoption for complex configurations. This thesis leverages the d
 ata-rich nature of DNS to develop machine learning frameworks that can acc
 elerate simulation workflows\, augment experimental diagnostics\, and yiel
 d deeper physical insight into the dynamics of cavity-stabilized combustio
 n.\n\nTo address the computational cost of DNS\, this work first develops 
 a physics-guided self-supervised super-resolution (SR) methodology to reco
 nstruct fine-scale structures in isotropic turbulence from coarse-grained 
 flow fields. By treating SR as a state estimation problem\, this approach 
 does not require paired high-resolution training data\, making it well-sui
 ted for flow configurations where fully resolved simulations are prohibiti
 ve. Separately\, a graph neural network (GNN) based SR framework is develo
 ped for turbulent reacting flows on complex meshes\, including structured 
 non-uniform and unstructured configurations. Leveraging message passing la
 yers\, this approach enables accurate reconstruction of unresolved flow fe
 atures across geometrically complex domains\, extending the applicability 
 of data-driven SR to practical combustion configurations.\n\nThe mesh-nati
 ve nature of GNNs established in the SR framework naturally extends to oth
 er modeling challenges in reacting flow simulations. A GNN-based closure h
 as been developed to predict filtered species production rates from filter
 ed thermochemical scalars on non-uniform meshes\, with a priori evaluation
 s demonstrating strong generalization across varying hydrogen blend compos
 itions and filter widths. To further reduce/alleviate the complexity of de
 tailed chemical kinetics\, two GNN-based mechanism reduction formulations 
 are introduced\, GNN-SM and GNN-AE\, leveraging surrogate-assisted and aut
 oencoder-based strategies respectively to identify kinetically dominant pa
 thways\, achieving up to 95% reduction in species and reactions while main
 taining predictive accuracy.\n\nThe mesh-agnostic nature of GNNs establish
 ed in the SR framework naturally extends to other modeling challenges in r
 eacting flow simulations. A GNN-based closure has been developed to predic
 t filtered species production rates from filtered thermochemical scalars o
 n non-uniform meshes. A priori evaluation has demonstrated strong generali
 zation across varying fuel blend compositions and filter widths. To reduce
  the complexity of detailed chemical kinetics\, two GNN-based mechanism re
 duction formulations are introduced\, GNN-SM and GNN-AE\, leveraging surro
 gate-assisted and autoencoder-based strategies respectively to identify ki
 netically dominant pathways. Up to 95% reduction in species and reactions 
 has been observed\, while maintaining predictive accuracy.\n\nDNS data is 
 further used to augment experimental diagnostics in trapped vortex combust
 ors. Vision transformer (ViT) models are trained on simulation data with s
 ynthetic noise to infer velocity fields from planar laser-induced fluoresc
 ence (PLIF) measurements\, avoiding the intrusiveness of particle image ve
 locimetry. These models demonstrate accurate reconstruction of large-scale
  vortical structures\, enabling data fusion and digital twin development w
 here intrusive diagnostics are impractical.\n\nFinally\, DNS is employed t
 o advance our physical understanding of cavity-stabilized combustion. Mass
 ively parallel compressible flow simulations of lean premixed ethylene-air
  flames in a backward-facing step reveal how the primary recirculation zon
 es transport radicals and alter dominant chemical pathways to support flam
 e stabilization. These investigations are extended to the HIFiRE 2 superso
 nic combustor\, where dual-mode and scram-mode operations are analyzed to 
 characterize flame stabilization regimes and the evolution of chemical pat
 hways in non-premixed supersonic flames. Collectively\, this work integrat
 es high-fidelity DNS with geometric-aware machine learning frameworks\, es
 tablishing a foundation for accelerated simulations\, improved diagnostics
 \, and reduced-order modeling in cavity-stabilized reacting flows.\n\n\n\n
 ALL ARE WELCOME
CATEGORIES:Events,Thesis Defense
END:VEVENT
BEGIN:VTIMEZONE
TZID:Asia/Kolkata
X-LIC-LOCATION:Asia/Kolkata
BEGIN:STANDARD
DTSTART:20251105T110000
TZOFFSETFROM:+0530
TZOFFSETTO:+0530
TZNAME:IST
END:STANDARD
END:VTIMEZONE
END:VCALENDAR