Talk on "Towards CAD-based robust aerodynamic shape design for practical industrial workflows"
Title: Towards CAD-based robust aerodynamic shape design for practical industrial workflows
Practical industrial design workflows are highly complex and bound to a base master CAD model. Traditionally evolutionary algorithms are used to optimise the CAD parameters when only black-box evaluation of candidate design is possible. Evolutionary algorithms become expensive for large number of design parameters and their convergence is quite slow compared to gradient-based methods. Algorithmic differentiation (AD) tools can reliably parse sources of large and complex computer codes to produce design sensitivities for gradient-based optimisation with minimum user intervention. With the adjoint mode AD, design sensitivities can be obtained at a computational complexity that scales with the number of output and are independent of the number of design degrees of freedom. Recently at our research group we successfully
differentiated a complex open-source CAD kernel OpenCASCADE and an in-house parallel CFD code to obtain directly the design sensitivities on the CAD parameter space. This allows gradient-based optimisation directly on the CAD parameters. Application top the differentiated CAD-CFD sensitivities to the problem of minimising losses in a Turbomachinery stator vane and drag on an aircraft wing is shown. A novel uncertainty quantification method using adjoint sensitivity called FastUQ is presented with motivation towards application to robust aerodynamic shape design.
Dr. Pavanakumar Mohanamuraly is currently working as a postdoctoral researcher at CERFACS, Toulouse, France. He has earned his Ph.D. from Queen Mary University of London, United Kingdom, for the dissertation on Fast Adjoint-assisted Multilevel Multifidelity Method for Uncertainty Quantification of the Aleatoric Kind. He has a MS degree from Pennsylvania State University, USA in Aerospace Engineering with a minor in Computational Science. His BTech in Aeronautical Engineering degree is from Madras Institute of Technology (MIT), Anna University.
His area of research is on multidisciplinary design optimisation and analysis (MDO) in engineering design workflows which spans CAD-based design and optimisation, multidisciplinary coupling, high-performance computing, and uncertainty quantification (UQ) in aerospace and mechanical engineering.
Event Date: 11th March, 2020(Wednesday)
Event Time: 02:30 PM
Venue: Room No:115