Control and Trajectory Generation Algorithms for high dynamics CNC machines

Nowadays, modern manufacturing relies on high dynamics robots and precise CNC machines to improve production efficiency and machining quality. Unfortunately, those two requirements are frequently at odds with each other. In fact, higher dynamics stress machine components, inducing vibrations that harm machining precision. Thus, meticulous mechanical design and component selection are essential, but cost constraints pose difficulties. However, software methods can enhance machine performances without changing the mechanical design and are categorized into two groups. The first optimizes the "motion reference" to prevent flexible mode excitement and/or reduce cycle time. The second employs "feedback control actions" to actively dampen vibrations and enhance tracking/positioning performances. This Ph.D. project was developed within an apprenticeship for higher training and research promoted by Salvagnini S.p.A and investigates both software approaches within an industrial application background. In the context of planning approaches, we developed a motion planning algorithm to effectively split the workspace movements between the actuators of a redundant Macro-micro manipulator used in 2D laser cutting machines. In order to reduce the problem complexity, it was divided into two separate algorithms: path and trajectory planning. The path planner computes a minimum curvature spline Macro joint path, enhancing its smoothness and improving end-effector admissible speed, while accommodating micro actuator limited range. Two different trajectory planning strategies were formulated. The first one defines a profile composed by 2 continuous jerk limited S-profiles which satisfy up to 2nd order derivative linear and joint constraints. The second one defines a spline-based velocity profile that fulfill linear and joint constraints up to 3rd order derivative. The approach shows a cycle time reduction, less end-effector vibrations and lower energy consumptions. In the context of control approaches,we developed a control action to actively suppress load-side vibrations on a two-mass system. The approach employs a full-state feedback controller, including also estimated load-side variables obtained through a sensor fusion algorithm. In fact, since the installation of position encoders on the load side is not always feasible, due to cost or space reasons, the developed sensor fusion algorithm combines motor-side position encoder measurements with those provided by a small low-cost load sideMEMS accelerometer. As industrial background, non linear aspects, like Coulomb friction, have necessarily been taken into account to increase estimation accuracy. The designed control action collaborates with a traditional cascade structure to dampen oscillations and remains robust against model parameter mismatches. The effectiveness was verified on an industrial test-bench emulating loading/unloading devices, demonstrating improved vibration suppression and reducedsettling time, even in presence of system model parameters variations.