基于ELM的飞机数字化装配定位运动模型

针对飞机装配中开敞性较差环境下的串联装配机构半闭环定位运动控制问题进行研究,提出了基于极限学习机(EML)算法的飞机数字化装配定位运动模型。通过分析飞机数字化装配串联定位机构的运动学模型特点及性能要求,提出了飞机数字化装配定位运动的单隐含层前馈神经网络模型,并基于极限学习机提出了装配定位运动的数据辨识模型,且最后给出了基于极限学习机算法的定位运动离线辨识方法。通过将某大型飞机机身壁板柔性预定位工装作为试验平台进行验证,结果表明,获得的定位运动模型使直接装配定位精度达到±0.25 mm,满足某大型飞机机身壁板长桁的装配定位精度要求±0.50 mm。试验系统涉及的若干关键技术已应用于某大型飞机的壁板组件装配预定位柔性工装系统。

Kinematic model of digital assembly location for airplane based on ELM

The research of semi-closed loop positioning for the tandem assembly mechanism in open poor aircraft assembly environment is conducted and the kinematic model of aircraft digital assembly location is studied based on extreme learning machine (ELM) for the positioning movement in assembly process. By analyzing the kinematic characteristics and performance requirements of the aircraft digital assembly location, the single-hidden layer feedforward neural-network model of assembly positioning movement is proposed, the data identification model of positioning movement is presented based on ELM, and finally the offline positioning movement identification method based on ELM is proposed. Achieving testing a certain type of aircraft fuselage panels' flexible pre-positioning tooling, the results show that the obtained positioning motion model meets the directly assembly positioning accuracy by ±0.25 mm and reaches the requirements ±0.50 mm about aircraft stringer assembly location accuracy. Several key technologies involved in the test system have been successfully applied to a large aircraft assembly system.