A digital twin model of ceramic powder manufacturing process using computer simulation and machine learning

J. Park, B. Kim, G-A. Ryu, H.S. Jang, S. Hyun, S. Baik
Korea Institute of Ceramic Engineering and Technology,

Keywords: powder compaction, sintering, multiphysics simulation, AI, realtime digital twin model


As a virtual characterization method, computer simulations by theoretical approaches based on the Cam-Clay and Cap plasticity models have been applied to the manufacturing process for various forms of powder products. In particular, the modified Drucker-Prager Cap model has been successively introduced and proven highly effective in simulating nonlinear behavior and related physical phenomena of ceramic powder. Recently, the digital twin(DX) approach combining computational method(CAE) with artificial intelligence (A.I.) has been developed to analyze and control real-time manufacturing parameters such as pressure, friction, and density profile of output products in this powder compaction process. A.I. algorithms, particularly machine learning, can be trained and predict efficiently the manufacturing conditions using virtual data and experimental data from the powder compaction process. In this study, we introduce a digital and automated workflow that simulates the compaction process of products with complex geometries using Abaqus and a sintering process simulation using COMSOL Multiphysics. The digitized and automatic workflow by multi-physics simulations and A.I. algorithms can successively characterize and optimize the manufacturing profile of powder density, temperature, and deformation of the final products in the two sequential processes. We suggest that data-driven digital transformation methods can be widely utilized for various powder manufacturing industries. The ceramics manufacturing DX platform(VECTOR) is also being developed to provide optimization digital tools customized for fundamental study as well as industrial needs.