Predicting Geometrical Variation in Fabricated Assemblies Using a Digital Twin Approach Including a Novel Non-Nominal Welding Simulation

Artikel i vetenskaplig tidskrift, 2022

The aerospace industry faces constantly increasing demands on performance and reliability, especially within the vital area of engine development. New technologies are needed in order to push the limits of high precision manufacturing processes for the next generation of aircraft engines. An increased use of in-line data collection in manufacturing is creating an opportunity to individualize each assembly operation rather than treating them identically. Welding is common in this context, and the interaction between welding distortion and variation in part geometries is difficult to predict and manage in products with tight tolerances. This paper proposes an approach based on the Digital Twin paradigm, aiming to increase geometrical quality by combining the novel SCV (Steady-state Convex hull Volumetric shrinkage) method for non-nominal welding simulation with geometrical data collected from 3D scanning of parts. A case study is presented where two parts are scanned and then welded together into an assembly. The scan data is used as input for a non-nominal welding simulation, and the result of the simulation is compared directly to scan data from the real welded assembly. Three different welding simulation methods are used and assessed based on simulation speed and ability to predict the real welding result. The segmented SCV method for welding simulation shows promising potential for this implementation, delivering good prediction accuracy and high simulation speed.