Physics-Guided Machine Learning for CNC Milling

Selection of CNC machining parameters for computer-aided manufacturing (CAM) software is based on tool manufacturer recommendations, experience, and/or trial and error. This leads to reduced productivity, increased lead times, scrapped parts, and high cost.

The team will develop a solution that combines physics-based models, data collection, and machine learning that will optimize CNC parameters for an internal blade rotor, or blisk design. The team will model the blade in three different materials: 6061-T6 aluminum, 316L stainless steel, and 6Al4V titanium, and demonstrate how the Physics-Guided Machine Learning approach can increase productivity.

While CNC machining has largely moved from an analog to digital approach due to advances in process planning software and machine tool controllers, the selection of machining parameters remains an experience-based, trial and error process. Without optimization, the outcome is reduced productivity, increased cost, and a significant disadvantage for U.S. manufacturers with higher labor rates than most international competitors. This project addresses this critical industry challenge through the application of artificial intelligence to precision part manufacture.