|About this Abstract
||Materials Science & Technology 2020
||Additive Manufacturing of Ceramic-based Materials: Process Development, Materials, Process Optimization and Applications
||Uncertainty Quantification in Additive Manufacturing of Piezocomposites through Physics-informed Data-driven Modelling
||Zhuo Wang, Li He, Chen Jiang, Zhen Hu, Xuan Song, Lei Chen
|On-Site Speaker (Planned)
A significant challenge for additively manufactured (AM-ed) piezocomposites is the presence of heterogeneous sources of uncertainty that lead to variability in the properties. We aims to develop a deep learning (DL)-enabled data-driven uncertainty quantification (UQ) method that leverages (1) extensive physics-based simulation data, and (2) a limited amount of experimental data. A two-scale computational model is developed to concurrently account for the porous microstructure in the ceramic grain scale and the complex 3D polymer-ceramic interfacial geometry to generate the data for UQ analysis. To reduce the effect of model-based uncertainty that is presented in the DL model, a Bayesian calibration framework will be utilized to calibrate and correct the DL-based model using limited expensive-but-realistic experimental data. With the reduced effect of epistemic uncertainty by using sufficient data and/or experimental calibration and validation, the data-driven GSA approach will then be used to quantify the physics-based uncertainty with a high confidence.