Data-driven mechanics
Data-driven mechanics
Embedded Files
Unlike traditional constitutive modelling approaches, which rely on predefined mathematical formulations and phenomenological assumptions, data-driven methods aim to learn material behaviour directly from data, offering greater flexibility in capturing complex constitutive relationships. This project aims to explore machine learning-based approaches for constitutive modelling in solid mechanics, with the focus on developing and neural network architectures to learn the macro-scale material reponse from fine-scale simulations data. The developed architectures will initially be tested against benchmark problems in elasticity and plasticity, where the underlying constitutive laws are relatively straightforward and well-established. These benchmark cases will provide a foundation for assessing the accuracy and generalizability of the learned models for complex material behaviour.
Selected publication [Link]
Non local continuum model for dense granular flows
Non local continuum model for dense granular flows
Solid Mechanicians have modelled dense granular flows by invoking various yield criteria and plastic flow rules in the last few decades. The most widely accepted and tested model for dense granular flows is the nonlocal granular fluidity(NGF) model. This model has shown success in predicting the rheology in a variety of flow configurations, especially it is the only model that has been predictive in a complex flow in a split-bottom cell. Although this model has shown success in capturing the flow behavior, but it fails in capturing other observations like normal stress differences and secondary vortex-type flows. The ongoing research direction is to develop a generalized rheological model, guided by vast Discrete Element Method (DEM) simulations, that can accurately predict the stress state in addition to the rheology.
Selected publication [Link]
Coupling size-segregation and flow in bidisperse granular mixtures
Coupling size-segregation and flow in bidisperse granular mixtures
Size-based segregation in dense granular mixtures has been a longstanding problem for academicians and industry. Size-segregation and flow of granular media go hand in hand. Guided by the discrete element method(DEM) simulations, we propose a phenomenological model for segregation accounting for all the driving mechanisms of segregation. Upon integrating the segregation model with the nonlocal granular fluidity (NGF)rheological model, the coupled model is capable of predicting the segregation dynamics and flow behavior simultaneouly. The ongoing work is to develop a robust numerical tool for simulating complex flows and to determine the efficacy of the model by drawing comparisons against DEM simulations and experimental observations.
Selected Publication [Link]
Mechanics of musical instruments
Mechanics of musical instruments
String-bridge interactions in the stringed musical instruments play a key role in determining the quality of sound produced. It is a peculiar feature of every instrument. Inspired from the Indian classical instruments, for instance Sitar, this project aims to model the vibrations of a string in the presence of a finite-sized curved obstacle(bridge). The finite size of the bridge introduces an additional wrapping nonlinearity apart from the nonlinearity caused by large deformations in the string. We plan to study free as well as forced vibrations in the presence of finite-sized curved constraint and investigate its role in determining the stability of the different types of motions. Furthermore, we would like to study the effect of the bridge on the coupling between the vibrational modes.
Selected Publication [Link]
Page updated
Google Sites
Report abuse