Optimal Design Method for Control Panel Input Values Using Deep Learning and CFD

An ideal velocity curve for the operation of a casting machine is derived based on physical laws or numerical analysis when we control pouring velocity in gravity casting, plunger injection velocity in high-pressure die casting, etc. However, even if an ideal velocity curve is used directly as the control panel input, the molten metal flow would differ from the pre-calculated result because of delays and errors by the machine. In recent years, digital transformation (DX) approaches have become more active, and research has been conducted to construct input-output models for industry machines based on deep learning. In this research, we develop an output prediction model based on deep learning and propose a decision method for control panel input values to achieve an operation with an ideal velocity curve derived through the computational fluid dynamics (CFD) optimization method on actual casting machines. Finally, we verify the effectiveness of this method through pressure control experiments using a water experimental unit for low-pressure casting.