Gain Tuning for High-Speed Vibration Control of a Multilink Flexible Manipulator Using Artificial Neural Network

Abstract

Flexible manipulators are associated with merits such as low power consumption, use of small actuators, high-speed, and their low cost due to fewer materials’ requirements than their rigid counterparts. However, they suffer from link vibration which hinder the aforementioned merits from being realized. The limitations of link vibrations are time wastage, poor precision, and the possibility of failure due to vibration fatigue. This paper extends the vibration control mathematical foundation from a single link manipulator to a three-dimensional, two links flexible manipulator. The vibration control theory developed earlier feeds back a fraction of the link root strain to increase the system damping, thereby reducing the strain. This extension is supported by experimental results. Further improvements are proposed by tuning the right proportion of root strain to feed back, and the timing using artificial neural networks. The algorithm was implemented online in matlab interfaced with dSPACE for practical experiments. From the practical experiment done in consideration of a variable load, neural network tuned gains exhibited a better performance over those obtained using fixed feedback gains in terms of damping of both torsional and bending vibrations and tracking of joint angles.