Audio Signal Processing for Quantitative Moulding Material Regeneration
Philine Kerst, M.Sc.
As a natural product with limited resources, sand is of existential importance for various industries. Foundry technology in particular requires considerable quantities worldwide for sand moulds and cores. Depending on the process, the sand is recycled and thus reused. However, especially in the case of inorganically bound moulding sand, it is still frequently not recycled. Existing regeneration methods include mechanical, pneumatic, or combined processes. These have been developed, but have reached their analytical optimisation limits, as the processes are not transparent and make in situ analyses of moulding materials impossible. Within the scope of this research work, a methodology for the computer-aided processing of sound and image data with the help of convolutional neural networks (CNNs) was developed, which is to evaluate the non-measurable changes of the moulding material in the running process in real time via process acoustics. The aim is to optimise process control in terms of time, cost, and energy efficiency.
Keywords: Regeneration, Audio Signal Processing, Machine Learning