Non-circular cutting with a lathe using a three-stage intelligent controller

In this paper, a three-stage intelligent controller (TSIC) for non-circular cutting with a lathe (NCL) is developed. In the first stage, an effective neural network is used to learn the dynamics of the voice coil servomotor system. Then a modified zero-phase tracking control (MZPTC) based on the learned model is designed to obtain an acceptable tracking result. Because MZPTC is a type of feedforward control, its performance cannot be ensured as the system is subjected to external load or the aging of system components. In the second stage, an adaptation in control input generated by a fuzzy sliding-mode control is then synthesized with the previous MZPTC to ameliorate control performance. The third stage is to use a forecast compensate control to compensate the cutting error caused by the variant cutting depth (or cutting force). The experimental results of NCL using TSIC, including tracking accuracy (i.e. profile error of finished workpiece) and smoothness of control input (i.e. surface roughness of finished workpiece), are much improved.