A fuzzy-net-based multilevel in-process surface roughness recognition system in milling operations

This paper describes a fuzzy-nets approach for a multilevel in-process surface roughness recognition (FN-M-ISRR) system, the goal of which is to predict surface roughness (R-a) under multiple cutting conditions determined by tool material, workpiece material, tool size, etc. Surface roughness was measured indirectly by extrapolation from vibration signal and cutting condition data, which were collected in real-time by an accelerometer sensor. These data were analyzed and a model was constructed using a neural fuzzy system. Experimental results showed that parameters of spindle speed, feedrate, depth of cut, and vibration variables could predict surface roughness (R-a) under eight different combinations of tool and workpiece characteristics. This neural fuzzy system is shown to predict surface roughness (R-a) with 90% prediction accuracy during a milling operation.