Direct adaptive neural control for turning complex rotating profiles

During the machining process of a complex rotating proÞle, numerous nonlinear characteristics of the process come into effect. It is thus difficult to control such a system efficiently. In this paper we propose a method to control such systems. We develop a direct adaptive control scheme for such processes, with a dynamic neural network as a controller. According to the idea of a direct adaptive control, the neural network controller can be directly adjusted based on the system-output error. The training algorithm is based on a modiÞed generalized delta rule and the only a priori knowledge required is that of response behavior of the controlled system. The proposed scheme is used to control the machining process of turning complex rotary proÞles, and the experiments of turning a typical engine piston demonstrate good real-time control performance of this system.