Tittu Mathew: Bayesian uncertainty quantification and model selection

Machine Learning Seminar presentation

Topic: Bayesian uncertainty quantification and model selection

Speaker: Tittu Mathew, Indian Institute of Technology Madras | IIT Madras · Department of Mechanical Engineering

Time: Wednesday 2020.11.11, 10:00 CET

How to join: Please contact Jakub Lengiewicz

Abstract:

Adaptive Importance Sampling based Neural Network framework for Reliability and Sensitivity prediction for Variable Stiffness Composite Laminates with hybrid uncertainties

In this work, we propose to leverage the advantages of both the Artificial Neural Network (ANN) based Second-Order Reliability Method (SORM) and Importance sampling to yield an Adaptive Importance Sampling based ANN, with specific application towards failure probability and sensitivity estimates of Variable Stiffness Composite Laminate (VSCL) plates, in the presence of multiple independent geometric and material uncertainties. The performance function for the case studies is defined based on the fundamental frequency of the VSCL plate. The accuracy in both the reliability estimates and sensitivity studies using the proposed method were found to be in close agreement with that obtained using the ANN-based brute-force Monte Carlo Simulations (MCS) method, with a significant computational savings of 95%.

Moreover, the importance of taking into account the randomness in ply thickness for failure probability estimates is also highlighted quantitatively under the sensitivity studies section.

Additional material:

[1] Tittu Varghese Mathew, P. Prajith, R.O. Ruiz, E. Atroshchenko, S. Natarajan, Adaptive importance sampling based neural network framework for reliability and sensitivity prediction for variable stiffness composite laminates with hybrid uncertainties, Composite Structures, 2020 https://doi.org/10.1016/j.compstruct.2020.112344