Workpiece dynamic analysis and prediction during chatter of turning process

In turning operations, a common problem that can drastically degrade the quality of a machined part is regenerative chatter. Proper machine design, such as increased stiffness and damping of the machine tool structure can broaden the range of stable operating conditions. However, the inherent geometry of the tool and workpiece can sometimes severely limit the range of stable cutting. In this case, active control is needed in order to allow for a sufficiently broad range of stability. This work investigates turning relatively compliant workpieces, therefore it is the body that undergoes the bulk of the motion during chatter. Since it is highly impractical to instrument the workpiece, a Neural Network trained with Particle Swarm Optimization is used to transform a radial displacement measurement made at the cutting tool to an estimation of the radial displacement of the workpiece. This could serve as an observer in a real time control system that could mitigate chatter by appropriately actuating an active toolholder such as a fast tool servo. The workpiece displacement was predicted with an average RMSE of 1.41 and 1.70 μm for the two testing datasets. This current approach differs from other chatter detection investigations because with the direct displacement measurement of the toolholder and the output from the Neural Network observer, there is information about both bodies’ motions. In this way, direct conclusions can be made about the stability of cutting in a chatter detection scheme. In addition, the nature of the transition from stable cutting to chatter is investigated by experimentally measuring the variation in uncut chip thickness over time.