Development of the EAST articulated maintenance arm and an algorithm study of deflection prediction and error compensation

Abstract

Experimental Advanced Superconducting Tokamak (EAST) is the world's first fully
\nsuperconducting tokamak with a non-circular cross-section. In recent years, with
\nincreased device performance and experimental parameters, EAST has achieved a series
\nof important research results and scientific discoveries. However, EAST inner
\ncomponents of the first wall will also be facing an increasingly tough operating
\nenvironment with higher heat loads. Although plasma facing components (PFCs) have
\nbeen updated and upgraded several times, the high heat flux during experiments will
\ncause damage or even failure to local small parts in the EAST vacuum vessel (VV). Any
\nfailure of the internal components might influence the obtaining of high-quality plasma
\nor even lead to a plasma discharge and cause safety problems of the operation device.
\nTherefore, it is essential to have timely maintenance based on the condition of damaged
\ninternal components in the experimental period. In conventional manual maintenance,
\neven if small sign of damage is seen, the device has to be shut down, which seriously
\naffects the efficiency of physical experiments, and increases the time and economic costs.
\nTherefore, to ensure adequate running time, the demand for remote handling maintenance
\nof the EAST device during physical experiments is urgent.
\nThe EAST articulated maintenance arm (EAMA) system is developed for real-time
\ndetection and rapid repair operations to damaged internal components during plasma
\ndischarges without breaking the EAST ultra-high vacuum (UHV) condition. To achieve
\nthe desired objectives, the EAMA system design should guarantee that the robot can
\nstably run in the harsh environments of high temperature (80-120 oC) and high vacuum
\n(~ 10-5Pa). Meanwhile, the errors caused by the deformation of long flexible robot arms
\nshould also be predicted and compensated in real-time to obtain high accurate
\nmaintenance operations. The main contributions of this dissertation include: design,
\ndevelopment and analysis of the whole EAMA hardware system; study of two different
\nmethods of matrix structural analysis (MSA) and a back-propagation neural network (BPNN) to predict flexible deformations of a manipulator system; the eventual
\ndeveloping of an EAMA dynamic error compensation algorithm to precisely control the
\nEAMA system operation.
\nFirstly, during the EAST experiments, environment conditions inside the vacuum vessel
\nwere very harsh with complex geometry; from the design view, the requirements for an
\nEAMA robot were more critical than those for a conventional industrial robot arm.
\nBeyond the conventional techniques in robot design, many unique techniques were
\nconsidered in the EAMA manipulator arm design: (1) Redundant articulated
\nconfiguration with modular techniques were utilized in robot arm design; (2) All joint
\nactuators and high-speed transmission components were set up in a sealed box, using a
\nplanetary roller screw to change rotation into linear motion, which was transferred outside
\nby the bellows; (3) Solid lubrication using MoS2-Ti-C composite coatings were
\ndeveloped for the low-speed transmission parts exposed to the vacuum environment; (4)
\nA parallelogram link mechanism was developed to provide more than 50,000 times of
\nreduction ratio and more than 1000Nm of drive torque. Secondly, as the total length of the EAMA system reaches more than 10 m, it will produce
\nsignificant flexible deformation under the effects of torques and gravity. The deformation
\nvalues will always be constantly changing as the arm movements and postures change. If
\nthis flexibility could not be accurately predicted and reasonably compensated, it would
\nseriously affect the position accuracy of the manipulator system. Based on the