The goal of this experiment was to reduce the scrap rate of the foundry process. A service was needed which could predict the behaviour of the current batch of metal to reduce the amount of waste. If it were possible to adjust the process in real time, the foundry could lower energy consumption and production costs. Using historical data from previous production runs was seen as the key to this problem, but this entails data analytics requiring computing resources far in excess of those available to foundries or consultants like ProService.
Noesis and ProService collaborated to develop a new adaptive process for controlling the liquid-iron stage of the foundry-casting process, based on Noesis’ Optimus software and ProService ITACA technologies. The two companies worked together to create a model of the process based on historical foundry data, and provided the required expertise to deploy the solution on a Cloud-based HPC platform. Thermal analysis software is used to generate production data, which is sent to the Fortissimo Cloud-based HPC infrastructure. The correction model is updated by the HPC resources based on the production data, and then returned to the foundry system, where the correction is calculated and applied.