CFC2023

Ensemble Kalman Filter Experiments with Continuous Glass Melting Process Simulation

  • Seki, Taiga (AGC Inc.)
  • Miyoshi, Takemasa (RIKEN Center for Computational Science)

Please login to view abstract download link

Controlling flow of liquidus glass melt inside the furnace is one of the key factors to achieve mass production of high quality glass products. To optimize the design and operations of glass manufacturing process, CFD has been widely used. Recent concept of Digital Twin is a promising technology also for glass industry. Further optimization of the process operation and predictive maintenance can be expected using detailed real-time insights obtained by DT. To implement DT, one should prepare the virtual model and data pipeline to reflect actual situation to the model. CFD is one of the reliable model which can work as the core engine of DT. However, most of the times it is difficult to collect all the information required to launch CFD simulation in real problem. In such situation, one should estimate these model parameters to obtain accurate simulation result. Data assimilation is known as a sequential method to estimate both state of the system and the model parameters simultaneously, which has been widely used in weather forecasting. In the present study, we applied data assimilation method to the CFD model for continuous glass melting process to implement DT. From preliminary observing system simulation experiment (OSSE) studies, several model parameters, temperature distribution and flow pattern of glass melt were successfully estimated by ensemble kalman filter (EnKF) using temperature data which are measurable in reality.