Advances in Signal Processing for Friction Stir Welding Temperature Control

Abstract

Precise temperature control of FSW benefits from control and manipulated signals that are responsive and smooth. Accurate measurement of tool temperature and spindle speed feedback are important to temperature control, but often noise in these signals prevents optimal control. Two different methods are developed in this paper to improve signal quality. A series of Bezier curves are used to compensate signals which exhibit a periodic but arbitrarily-shaped offset. Least-squares fitting is used to obtain quality derivatives from discrete or noisy signals. The Bezier method is used to decrease the inaccurate temperature fluctuation measurements reported by telemetry collar error and adds no time delay or phase shift. The least-squares approach is used to estimate spindle speed and temperature derivatives and adds only minimal time delay while substantially reducing noise.