Numerical Strategies to Study Compression Failure in Brittle Foams with 3D Realistic Microstructures

The work describes a series of numerical steps performed to study compression failure in brittle foams. The material in the present study is manufactured from mechanical stirring of ceramic slurry. The material microstructure looks like polydisperse pores interpenetrated into each other at varying extent. The work is divided into three steps. The first step is creation of a microstructure reconstruction algorithm that generates artificial microstructures statistically equivalent to the real one obtained from micro-tomography scanning. It enables creation of microstructures with multiple realizations and of different sample sizes to perform statistical studies. The next step is to perform finite element simulations to study uniaxial compression failure of the material. The compression behavior of reconstructed microstructures is compared with that of the real one along with some experimental observations. The scatter in the results obtained by simulating microstructures of different sizes is also studied. In the last step, biaxial compression failure of the material is studied with an objective to prove that a response of a smaller microstructure volume element can be utilized to predict the response of a larger (more representative) volume element. This is accomplished through use of neural networks and transfer learning.