Proposal of ball end-milling condition decision methodology using data-mining from tool catalog data

Machining is often performed by a machining center using various cutting tools and end-milling conditions for different shapes and materials. Recent improvements in CAM system make it easier for even unskilled engineers to generate NC programs. In the NC program, the end-milling conditions are decided by engineers. However, engineers need to decide the order of the process, cutting tool selection, and the end-milling conditions on the basis of their expertise and background knowledge because the CAM system cannot automatically decide them. Data-mining methods were attracted attention to support decisions about end-milling conditions. Our aim was to extract new knowledge by applying data-mining techniques to a tool catalog. We used both hierarchical and non-hierarchical clustering methods and also principal component regression. We focused on the shape element of catalog data and we visually clustered ball end-mills from the viewpoint of tool shape, which here meant the ratio of dimensions, by using the k - means method. Expressions for calculating end-milling conditions were derived from response surface method. We conducted end-milling experiments to validate the availability of calculated values.