Johns Hopkins University

UQpy – Python toolbox

(Jan 2018 – Present)

  • UQpy (Uncertainty Quantification with Python) is a general-purpose Python toolbox for modeling uncertainty in physical and mathematical systems
  • Developed modules with various methods for sampling techniques, surrogate modeling, and sensitivity analysis
  • Sequential learning algorithms based on the surrogate modeling are computationally more efficient for complex models, ran these models on Maryland Advanced Research Computing Center (MARCC) server using UQpy tools
  • UQpy provides tools such as Refined Stratified Sampling (RSS), Gradient Enhanced-RSS (GE-RSS) and Adaptive Kriging – Monte Carlo Simulation (AKMCS) for sequential learning algorithm

Large Scale Boundary Layer Wind Tunnel (BLWT) Experiments

(Jan 2020 – Present)

  • Developed a semi-automated framework to conduct high-volume through-put experiments using a wind tunnel facility at the University of Florida
  • A machine learning based framework sequentially guides the experiments by selecting the next roughness terrain
  • Framework has been explored with various active learning functions to identify different roughness terrains which generate wind profile with similar second-order properties
  • Various learning functions have been applied to large-scale experimental models to improve surrogate fit and sensitivity analysis has been done based on Sobol Indices.

Adaptive Surrogate Model Construction for Efficient Global Sensitivity Analysis

(July 2021 – Dec 2022)

  • Developing an algorithm to tackle the Sensitivity Analysis in case of computationally expensive models by utilizing the characteristics of surrogates and Sobol Indices
  • A proposed active learning function adaptively learns new samples based on complete available data and reduces the uncertainty in Sobol Indices

Study on Effects of Imperfections in Ship Grillage Structure using FEM model

(Dec 2019 – Dec 2020)

  • The main aim is to study the behavior of ship grillage in the nonlinear region due to different types of imperfections (plate, web, and flange imperfections)
  • Redundant imperfections can be identified, using adaptive Sensitivity Analysis, and removed to improve the model
  • GE-RSS method can reduce the uncertainty(variance) in the ultimate strength or other quantity of interest by adaptively learning samples using gradient estimates

Gradient Enhanced-Refined Stratified Sampling using Voronoi Stratification

(Aug 2018 – May 2019)

  • Developed an efficient learning algorithm to extend the existing GE-RSS method in the case of Voronoi stratification
  • The algorithm aims to adaptively generate new samples to reduce the (uncertainty) output variance
  • The critical aspect is to approximately compute the gradient at the centroid of each Voronoi cell using a surrogate model and use it to identify the contribution of each cell to the output variance

Indian Institute of Technology Kanpur (IITK)

Master’s Thesis: Physical Parameter Estimation of Non-Classical Damped System from Spatially Incomplete Mode Parameters

(May 2016 – Apr 2017)

  • Identified physical parameters of non-classically damped spring-mass systems using test vibration data, which would provide a better estimate of the system’s performance under dynamic loading
  • Identification of mode shapes in different cases of non-classically damped system with incomplete instrumentation, i.e. when not all DOFs of the system are excited or observed
  • Established the minimum instrumentation necessary for identifying the complete set of physical parameters
  • The proposed method identifies exact physical parameters for noise-free spatially incomplete modal parameters and has acceptable errors when data is corrupted by noise.

Finite Element Analysis of Concrete Wall

(Mar 2016 – Apr 2016)

  • Applied the basic concepts of FEMA to understand the behavior of concrete wall under in-plane loading
  • Used a FEMA software ABAQUS to model concrete wall with different shapes of opening. Further, observed the behavior of shape of openings on concrete walls under in-plane loading
  • Discussed about the failure mode and maximum stress developed in case of different openings

Vibration Controlling Dampers

(Mar 2016 – Apr 2016)

  • Studied about the different types of dampers and their need to dissipate seismic energy to reduce damages
  • Modeled different types of dampers - Friction, Viscous and Steel dampers for analysis of shear type buildings
  • Discussed about various limitations in modeling of above-mentioned types of dampers

Mortar-less Masonry and Interlocking Blocks

(Mar 2015 – Apr 2015)

  • Studied the disadvantages of conventional masonry and need for alternate construction method such as mortar less masonry
  • Investigated about the performance of constructed dry stack masonry walls under axial, shear and flexural loading
  • Modes of failure of masonry walls were also observed for different type of loadings

Control establishment and Mapping using GPS and Total Station

(Oct 2014 – Nov 2014)

  • Worked with a team of 6 member handled various instruments such as Total station, Levelling, GPS
  • Created a UDS program to store relative/local coordinates of almost 470 point, which represents land feature in academic area
  • Corrected elevation using Levelling from a Bench Mark and coordinates of control points was measured again using GPS
  • Multi-date data was aggregated into suitable files and imported into QGIS software to create a topographical map

Design Project – Industrial Building

(Oct 2014 – Nov 2014)

  • Worked in a team of 6 students, performing a structural analysis of the given building for different load cases.
  • Designed elements of industrial building like top and bottom chord, crane girder, base plate & connections.
  • Taking due notice of the economic feasibility and constraints, checked the structure’s engineering aspects like bending moment, flexural, torsional and web buckling using IS800 codes.