I am a research scientist at Leidos Research Support Team (LRST) which is involved in executing a mulit-year and multi-million dollar research service support (RSS) contract with Department of Energy (DOE) at National Energy Technology Laboratory (NETL), pittsburgh. My work is primarily focused towards developing efficient methods for intelligent monitoring of natural gas pipelines at NETL as part of its multi-disciplinary research project, Natural Gas Infrastucture Program.
In my work as a postdoctoral research scientist at Max Planck Institute, Germany i was actively involved in developing innovative methods for discovery of parsimonious nonlinear dynamical models using the concepts from scientific machine learning. I was also part of Max Planck research network on big data-driven material science BiGmax. Prior to MPI, I worked as a research scientist at TCOMS under Prof Chan Eng Soon and developed fast and accurate methods for reconstruction and propagation of multidirectional ocean wave fields. My work at TCOMS broadly touched upon concepts ranging from compressed sensing, sparse representation, model order reduction, proper orthogonal decomposition, physics informed A.I and custom made ML networks such as Fourier neural operators. In my doctoral work under Prof A R Magee and Prof R K Jaiman, I focused on developing data-driven methods for stability analysis and prediction of fluid-structure interaction. As part of my PhD research, I pursued topics ranging from system identification, ERA, Recurrent neural networks, Convolutional neural networks, Proper orthogonal decompisition and data-driven reduced order models.
I am always open for any kind of discussion on anything related to my work or any interesting ideas or topics.
In addition to the above, i try to do… 🏃 🥋 🏎️ 🏏 🎦 in my free time.
PhD in Data-driven computational fluid mechanics, 2020
National University of Singapore
M.tech in Applied mechanics, 2015
Indian Institute of Technology, Madras
B.tech in Naval architecture and Ocean engineering, 2014
Indian Institute of Technology, Madras
Intelligent sensing and monitoring of natural gas piplelines
Physics Enhanced Machine Learning
Development of digital twin of ocean wave environment
Data-driven computing for stability analysis and prediction of fluid-structure interaction
In this paper, we present two deep learning-based hybrid data-driven reduced-order models for prediction of unsteady fluid flows. These hybrid models rely on recurrent neural networks (RNNs) to evolve low-dimensional states of unsteady fluid flow. The first model projects the high-fidelity time series data from a finite element Navier–Stokes solver to a low-dimensional subspace via proper orthogonal decomposition (POD). The time-dependent coefficients in the POD subspace are propagated by the recurrent net (closed-loop encoder–decoder updates) and mapped to a high-dimensional state via the mean flow field and the POD basis vectors. This model is referred to as POD-RNN. The second model, referred to as the convolution recurrent autoencoder network (CRAN), employs convolutional neural networks (instead of POD) as layers of linear kernels with nonlinear activations, to extract low-dimensional features from flow field snapshots. The flattened features are advanced using a recurrent (closed-loop manner) net and up-sampled (transpose convoluted) gradually to high-dimensional snapshots. Two benchmark problems of the flow past a cylinder and the flow past side-by-side cylinders are selected as the unsteady flow problems to assess the efficacy of these models. For the problem of the flow past a single cylinder, the performance of both the models is satisfactory and the CRAN model is found to be overkill. However, the CRAN model completely outperforms the POD-RNN model for a more complicated problem of the flow past side-by-side cylinders involving the complex effects of vortex-to-vortex and gap flow interactions. Owing to the scalability of the CRAN model, we introduce an observer-corrector method for calculation of integrated pressure force coefficients on the fluid–solid boundary on a reference grid. This reference grid, typically a structured and uniform grid, is used to interpolate scattered high-dimensional field data as snapshot images. These input images are convenient in training the CRAN model, which motivates us to further explore the application of the CRAN-based models for prediction of fluid flows.
In this paper we present a stability analysis of passive suppression devices for the vortex-induced vibration (VIV) in the laminar flow condition. A data-driven model reduction approach based on the eigensystem realization algorithm is used to construct a reduced-order model in a state-space format. From the stability analysis of the coupled system, two modes are found to be dominant in the phenomenon of self-sustained VIV namely, the wake mode, with frequency close to that of the wake flow behind a stationary cylinder; and the structure mode, with frequency close to the natural frequency of the elastically mounted cylinder. The present study illustrates that VIV can be suppressed by altering the structure mode via shifting of the eigenvalues from the unstable to the stable region. This finding is realized through the simulations of passive control devices, such as fairings and connected-C devices, wherein the presence of appendages breaks the self-sustenance of the wake–body interaction cycle. A detailed proper orthogonal decomposition analysis is employed to quantify the effect of a fairing on the complex interaction between the wake features. From the assessment of the stability characteristics of appendages, the behaviour of a connected-C device is found to be similar to that of a fairing, and the trajectories of the eigenspectrum are nearly identical, while the eigenspectrum of the cylinder–splitter arrangement indicates a galloping behaviour at higher reduced velocities. Finally, we introduce a stability function to characterize the influence of geometric parameters on VIV suppression.