| In order to handle the complex maintenance tasks of CFETR’s divertor,a kind of multi-purpose over-load remote handling robot has been designed by the remote handling team of ASIPP(Institute of Plasma Physics,Chinese Academy of Sciences),The robot system,called CMOR(CFETR multipurpose overload robot),adopts the long cantilever with nine link combination,it can cover a±45°sector and manage a submodule of the divertor first wall weighing up to 2 tons.The remote handling robots with redundant degrees of freedom and heavy-load characteristics all have problems such as difficulty in calculating collision-free trajectory,reducing the influence of structural deformation on the operational safety factor,and improving the control accuracy.For the requirement of high-efficiency maintenance inside CFETR,this paper do research on intelligent algorithm of CMOR.Firstly,the motion planning algorithm is optimized.The trajectory reaching any location in the narrow D-shaped CFETR vacuum vessel can be calculated quickly.Secondly,the research of real-time structural deformation visualization has been conducted.The deformation data is updated by 2 Hz to detect the state of the operation.Thirdly,the structural deformation is predicted and compensated to improve the control accuracy,and the error is reduced within±5mm.Finally,CMOR realizes efficient,safe and accurate maintenance of CFETR core components.The research topics are based on Comprehensive Research Facilities in Support of CFETR with program number of "2017YFE0300503" and Comprehensive Research Facility for Fusion Technology Program of China under Contract No.201 8-000052-7301-001228.The main research topics are as follows:1.Set objective functions.As the assembly position in the middle maintenance port,the link length of each joint directly affect the workspace and operational flexibility of CMOR.In order to ensure that all the design parameters of CMOR are reasonable,the workspace and operational flexibility are used as objective functions to optimize the installation position and the link length.The final optimized parameters provide a reference for the development of the CMOR prototype.2.Form training datasets for NN.The CMOR has nine degrees of freedom,and its operation environment is a narrow D-shaped CFETR vacuum vessel,it requires a collision-free trajectory during the remote handling tasks.Traditional motion planning algorithms have problems such as low computational success rate and long computational time.In this paper,the shortest paths without collision are used as the training datasets,a supervised neural network is trained by the training datasets to form the motion planning model,the model can calculate the collision-free trajectory quickly with a high computational success rate.3.Vibration control and deformation estimation.After modeling the flexibility of CMOR links and joints and performing rigid-flexible coupling dynamic analysis,the analysis results have been obtained.The results show that there is residual vibration at the start and stop states,and the vibration tends to remain stable during operation with a lot of structural deformation.In this paper,the rigid dynamics model of the manipulator is established by the Newton-Euler equation,and the residual vibration of the manipulator is eliminated by nonlinear control,so that the manipulator runs smoothly during the whole maintenance process.The deformation of the manipulator in different poses is used as the training dataset,while the deformation prediction model using the MLP-Transformer neural network is trained by training datasets,and the deformation prediction model can quickly calculate the structural deformation.The OpenGL library is used to build up the structural simulator to realize the real-time visualization of CMOR structural deformation during the whole maintenance process.4.Construct the control software.The CMOR deformation prediction model is used to build the structural simulator in order to monitor the safe state.The deformation prediction model of the end points,combined with the fast motion-planning model,forms the CMOR precision control algorithm.The structure simulator and the precision control are integrated into the control software,which is connected with the controller of the manipulator to perform feedback control and realize precise motion planning before operation and real-time status monitoring during operation.5.Conduct experiments and analyze results.A flexible manipulator experimental platform is built,and the deformation prediction model and precision control algorithm of CMOR are verified through this platform.The experimental results show that the deformation prediction model can quickly calculate the deformation of the manipulator.The precision control algorithm can correct the motion error of the manipulator in real time and ensure the accurate position of the end effector. |
