Link deformation prediction and compensation of heavy-duty manipulator for fusion reactor

The extreme operating environment of the fusion reactor (such as radiation, and magnetic fields) makes manual direct maintenance unfeasible. Remote handling has become the core means of in-vessel components maintenance. For the in-vessel components with heavy payload, the heavy-duty manipulator is a key maintenance equipment. When carrying heavy load in-vessel components, the links of the long cantilever manipulator will suffer deformation problem and therefore, makes the end position and posture accuracy low. To improve the end position and posture accuracy, a new method combines the Newton-Euler method and the neural network method was proposed. The Newton-Euler method is used for link force calculation and the neural network is adopted to learning the link deformation. The deformation model of the heavy-duty manipulator is established and the end error is compensated by the pseudo-inverse Jacobian method. The proposed methods are applied on the heavy-duty manipulator, and the link deformation prediction and compensation model are verified. The results show that the link deformation under different payloads can be predicted and compensated. After compensation, the maximum position and posture error can be reduced to 3.18 mm and 1.67 $\times {10}^{-3}$rad, respectively. The end position and posture errors can be reduced by 63.4 % and 78.1 %, respectively. The proposed model can be used for the link deformation prediction and compensation of the real heavy-duty manipulator assembly in the future to further validate the accuracy improvement.