Self-optimization in Large Scale Assembly

Abstract

Efficiency in aircraft production can be increased by using flexible robotic assembly systems instead of fixed jigs, but the flexibility can only be used in combination with efficient control algorithms. For large components which have an individual deformation, e.g. due to gravity, not only automated but self-optimizing control algorithms are required, which allow an autonomous product-specific adaptation of the systems behavior during assembly. Therefore, models are required in order to describe the products behavior to external forces and to adapt robot motions. Multiple linear regression is presented as an approach to generate a product model based on experimental data. The product model is used to generate robot motion for an automated untwisting process of large components. The depicted approach is being validated at a demonstrator consisting of two industrial robots and a CFRP panel.