Investigation Into Key Technologies For The Development Of CNC System Of A Novel 5-DOF Hybrid Robot

This dissertation investigates some key technologies for the development of CNC system of a novel 5-DOF(degrees of freedom)hybrid robot known as Tri Mule,especially designed and developed for on-site manufacturing of large and complex structural components.The main contents are concerned with kinematic and rigid-body dynamic modeling,non-singular tool path planning,controller parameters off-line iterative tuning as well as on-line adaptive fuzzy-logic tuning with the goal to improve the actuated joint tacking accuracy of the parallel mechanism within the robot.The following contributions have been made.The canonical models suited to displacement,velocity,acceleration and rigid-body dynamic analysis of the Tri Mule robot are developed.A criterion for choosing the appropriate solution of inverse displacement analysis is presented,that enables the avoidance of mechanical interference and enlargement of the task workspace.As a counterpart of inverse displacement analysis,a semi-analytical algorithm for forward displacement analysis is developed with high computational efficiency.Meanwhile,the model developed for rigid-body dynamic analysis shows explicitly the couplings between the joint forces and joint movements of parallel mechanism within the hybrid robot.These works lay a comprehensive and solid foundation for the rest contents related to the development of the CNC system of the robot at hand.By taking the Tri Mule robot as an example,it reveals that at a singular configuration that varies with the system configurations,sudden changes occur in both rotation of the C-axis of the A/C wrist and lengths of three telescopic legs within the 3-DOF parallel mechanism.This finding leads to two effective methods for non-singular tool path planning of the hybrid robots having A/C wrist.The first method deals with the singularity problem by directly modifying the C-axis commands in the joint space using linear interpolation technique.The second method resolves the same problem using the property that the singular axis and tool axis rotate about the same axis in the opposite direction.The use of this property enables an effective algorithm for non-singular tool path generation to be developed by modifying a partial set of the control points of B-splines parameterized in the operation space.Both simulation and experimental results on a prototype machine show that in the neighborhood of singularity,continuous and smooth joint motions and acceptable tracking accuracy can be ensured after the modification.The proposed method is general such that it can be employed to generate non-singular tool path of other type hybrid robots having A/C wrist.By combining the invariant principle of compound control and rigid body dynamics,a data-driven based iterative approach is developed for feedback and feedforward controller parameter tuning of the parallel mechanism within the Tri Mule robot by considering cross-talk torque disturbances.The use of this method allows the formulation of the plant-free identification Jacobian in such a way that it is only related to the feedback/feedforward controller parameters tuned in the previous iteration cycle and the joint tracking errors associated with sequential perturbed statuses in the current iteration cycle.Both simulation and experimental results on a prototype machine show that the parameters tuned by the algorithm converge at a satisfactory rate and provide good extrapolation capability.The results also show that the root mean square of joint tracking errors can be effectively reduced compared with the case where no joint couplings are considered when the system operates at high speed.However,it is unnecessary to consider joint couplings if the system operates at relatively low speed.The proposed approach is general and it is thereby suitable for controller parameter tuning at any arbitrary configuration,and applicable to other types of parallel mechanisms whenever their joint couplings are not negligible.Considering that joint inertial effects of the 3-DOF parallel mechanism within the Tri Mule robot varies with system configurations,an adaptive fuzzy-logic control strategy is proposed for real time feedback and feedforward controller parameters tuning.This approach features the generation of a group of specific configurations representing different inertial levels via cluster analysis,and the creation of a membership function that matches the joint inertial distributions across the entire task workspace.Merging these two threads allows the controller parameters at any arbitrary configuration to be estimated by taking the parameters off-line tuned at the specific configurations as the inputs of the membership function.Both simulation and experimental results on a prototype machine show that only eight specific configurations are required,at which the controller parameters of three actuated joints of the 3-DOF parallel mechanism need to be tuned for implementing the adaptive control strategy.It also concludes that it is unnecessary to use the adaptive control strategy unless the robot moves in the region where the gradient of joint inertial effects varies sharply.The outcomes of this dissertation have been partially used for the development of CNC system of the Tri Mule robot,and laid a solid foundation for the performance enhancement of the product and its engineering applications.