Force-Supervised Compliant Assembly with Robot Trajectory Compensation
Ionuţ-Valentin Cristea, Theodor Borangiu, Silviu Răileanu, Andrei-Alexandru CristeaProduct-driven automation and robotics have established an essential development course for production lines in flow-shop manufacturing, particularly in the major household appliances industry characterized by high volumes and severe quality requirements. The assembly of big components with tight tolerances and deformable mechanical conduct calls for flexible robot implementation to ensure short operation cycles. This paper describes a solution for compliant, simultaneous assembly of lateral panels on a refrigerator housing by two vertical articulated robots with a multiprocessor control system and edge-computing networking framework. In the proposed solution, the trajectory of the components inserted by the robot is adapted during the contact stage in terms of path and speed-profile corrections. The assembly strategy combines the passive, natural compliance of two snap-fit insertion mating elements—elastic clips into flexible spring-type fastening claws—with the active compliance from six-axis force/torque sensors to control the clip insertion trajectory of the robots. An automated vacuum monitoring system ensures secure grasping, detects leaks and seal losses, and limits panel deformation throughout the assembly process. The evaluation of the elastic deformation map of the thin-sheet panel during insertion, performed through RobotStudio simulation combined with Matlab load model computation and von Mises stress analysis, showed a maximum structural deformation of 0.516 mm caused by the vacuum gripping force and of 0.171 mm caused by the clip insertion forces, with a very low risk of permanent plastic deformation, i.e., a safety factor of 10.41. Experiments have validated in all trials (40/40) the proposed assembly strategy in terms of robot clip insertion, for panel pick-up position in the stack and appliance housing stop position on the shop floor conveyor in the precision range (0–1) mm, relative to the robot locations learned, respectively, for panel take-over and assembly.