Implementation of a multibody model as digital twin for a robotic arm

Purpose: The aim of this thesis is to implement, develop and validate the digital twin of the collaborative robot Doosan H2515, used for handling tools and accessories during various industrial procedures. The objective is to enable precise execution of motion laws by the robotic arm and to manage the physical system from its digital counterpart, in order to facilitate its application in the medical-surgical domain, requiring utmost accuracy and movement exactitude. Methods: A multibody model of the cobot is constructed, comprising six distinct components. The parts are modeled as rigid bodies, and physical and mechanical properties are assigned to each. Key properties include mass, center of mass and inertia. To achieve a system of point masses, inertia is set to zero for all parts of the system. Junctions between components are modeled by creating joints, placed at the interface between adjacent parts. Specifically, the model consists of a system with six joints, all of which are single-degree-of-freedom revolute joints. This means that they allow only revolution around the axis of rotation of the corresponding part, preventing all other movements. Angular and torque measurements are then created for each joint, enabling monitoring of the relevant quantities of the model under examination. The defined digital twin is validated through parallel experimental tests and numerical simulations, employing the Adams View software environment, and comparing the results derived from both approaches. More specifically, to test the behavior of the cobot, static, kinematic, and dynamic simulations were performed, with an increasing degree of complexity. Statics and kinematics tests were initially studied by moving the joints separately; dynamics tests permitted to better investigate on the internal torques generated at each joints. In that sense, an experimental test was designed to verify the digital twin confidence in replicating a motion controlled dynamics. Results: The results from these analyses indicate that the static and kinematic behavior of the Doosan H2515 is accurately estimated by the digital twin. However, issues arise in dynamic simulations, particularly in terms of constraint reactions. A deviation is observed between Joint Torque experimental and numerical results. The analytical correspondence with the in silico model, arises the hypothesis that the deviation could be attributed to the presence of internal brakes (elements not replicated in the digital model) whose loads and control algorithm cannot be extracted and then difficult to be estimated. Observing the joints involved in the bending of the robotic arm (Joint 2, 3, 5) this deviation appears more pronounced closer to the base due to the increased weight of downstream components; thus, Joint 2 is identified as the ‘critical joint’. Finally, Joints 1 and 6 were not validated, due to the difficulty in controlling experimental loads along their axis. The primary focus was then maintained on the joints involved in the bending of the robotic arm.