Student Posters Repository

The solution of inverse problems is fundamental to a wide variety of applications. The data for inverse problems often contains noise which is unknown. Quantifying uncertainty in the solution of inverse problems is important for further optimization and decision making. The focus of the work is on PDE-constrained Bayesian inverse problems in multiphysics setting. The forward model is defined as a coupled set of conservation laws (PDEs) describing the physics of some manufacturing process. Solving complex PDEs in large scale means solving large scale linear systems many times which in turn requires the use of HPC resources. PETSc library is used for solving linear systems using Krylov iterative method. PETSc utilizes MPI for scaling to cores. Moreover once the solution of the forward problem is established, it has to be solved multiple times for Monte Carlo Markov Chain integration in order to explore high-dimensional probability space of inverse problem solution.