Currently, there are no commercial software systems to support the achievement of predefined surface roughness for newly machined parts. Computer Aided Manufacturing (CAM) programmers and machine operators rely on their experience to find the best trade-off between time and quality optimization.This paper presents a novel concept for predicting surface roughness based on real-time machining data, planning data, and quality measurements. By linking the data points of all the different sources to the respective manufacturing feature, like drilling or face milling, the data cleaning process is sped up significantly facilitating queries of machine learning applications.On the one hand, the relations derived from that machine learning model, e.g. a random forest, are integrated into a CAM planning software for parameter evaluation and automatic tool path optimization. On the other hand, the proposed surface roughness prediction algorithm is used in a live simulation to give machine operators immediate visual feedback on the produced finish to enhance precise feed rate adaption. Hence, a data-driven decision support system is to be developed which supports both inexperienced and experienced personnel at the planning and execution phase within the product life cycle.
Development of Self-Learning NC-Programming System for Data-Driven Surface Prediction and Adaptive Cutting Parameter Optimization: FO999891268 (FFG - Österr. Forschungsförderungs- gesellschaft mbH)