Identification, Diagnosis, and Control of a Flexible Robot Arm

Sammanfattning: The most important factors in manufacturing are quality, cost, and productivity. The trend is towards lighter robots with increased mechanical flexibilities, and therefore there is a need to include the flexibilities in the robot models to obtain good performance of the robot. The core theme in this thesis is modeling and identification of the physical parameters of an ABB IRB 1400 industrial robot. The approximation made is that the robot arm can be described using a finite number of masses connected by springs and dampers. It has been found that a three-mass model gives a reasonably good description of the robot when moving around axis one. The physical parameters of this model are identified using off-line and on-line algorithms. The algorithms are based on prediction error methods. For the on-line identication the Matlab System Identifiation Toolbox is used. For the on-line identication the algorithm used is a modified version of a recursive prediction error method to cope with continuous time models. The models are then used in diagnosis and control. Two ways of doing diagnosis using on-line identification are investigated. Estimating some of the physical parameters of the robot arm recursively makes it possible to monitor important aspects of the system such as friction and load. LQG control of the flexible robot arm is also studied with the aim of good disturbance rejection. Aspects that have been studied are unstable regulators and the use of accelerometers.