A flexible tooling system for aero-engine pipelines with complex components based on human-robot collaboration

Abstract

The demand for the manufacturing multi-variety, small batch, and customized pipelines in the aero-engine manufacturing industry is continuously increasing. Driven by advanced manufacturing technologies, the key factor affecting product quality has shifted from various manufacturing stages to the final assembly stage. Currently, automated assembly solutions are constrained by technical bottlenecks and cost pressures, making them difficult to implement; as a result, assembly tasks still rely heavily on manual operations. To adapt to fierce market competition and production changes, manufacturers must strive to automate pipeline assembly. For this purpose, this paper proposes an automated flexible tooling system for pre-welding positioning and assembly quality evaluation. The system adopts a multiple collaborative robots’ architecture and performs reconfigurable assembly in a human-robot collaboration (HRC) mode. The specific implementation steps are as follows: first, workers install connectors on the fixtures of each robot's end effector, adjust them to an appropriate posture, and obtain the spatial posture data of the connectors through scanning; subsequently, a method for determining registration schemes is developed on the basis of the structural characteristics of the connectors. The feature elements are further extracted, and data registration is completed through human-computer interaction (HCI); then, an improved NSGA-III method that integrates the Levenshtein Distance Congestion Elimination (LCE) method is proposed. This method incorporates three types of constraints: obstacle avoidance constraints for the collaborative movement of multiple collaborative robots, feasibility constraints for robot movement, and rotational angle constraints for connectors. It solves the multi-objective optimization problem among total assembly time, uniformity of time allocation, and energy consumption, enabling rapid and efficient robot posture reconstruction; finally, simulation and experimental verification of the system are conducted. The field assembly verification results demonstrate that the assembly quality is significantly improved compared with that of traditional and recent representative algorithms. The proposed assembly method has an accuracy ranging from 0.0452 mm to 0.0807 mm, with an assembly precision of approximately 0.0607 mm, and ensures the stability and predictability of the assembly quality.