Session: 02-10-01: Variation Simulation and Tolerancing

Paper Number: 70212

Start Time: Friday, 12:05 PM

70212 - Towards a Digital Twin Setup for Individualized Production of Fabricated Components 

One of the challenges of high precision manufacturing is the variation that affects the process between each individual manufactured part. Variation comes in many shapes and forms, and can cause problems with tolerances and quality in finely tuned production lines that are assumed to run under nominal conditions.

Currently, this problem is approached using a mix of robust design methodology and manual individual adjustments based on experience and measurements. Robust design and geometry assurance are important tools for creating a manufacturing process that is as insensitive as possible to disturbances and noise, but cannot possibly predict and mitigate every type of process variation. Manual process adjustments done by skilled and experienced operators can be an effective way of compensating variation, but it is costly and time-consuming while also being vulnerable to human error.

In order to reliably and efficiently predict and compensate process variation that cannot be mitigated through robust design, a system is required which can detect variation and implement a process adjustment to counteract it. This way, a satisfactory process outcome can be achieved even during non-nominal conditions.

This type of functionality requires that the system contains a model of the real process. By feeding the model with measurements of real conditions, the outcome of the process under these exact conditions can be predicted. The prediction can then be used to calculate how the process should be adjusted in order to produce a correct outcome under current working conditions. This type of system is sometimes referred to as a digital twin.

In order to do its job, a digital twin for a manufacturing process requires three things. First, it needs measurement data that describes a process parameter which varies from its nominal value. Second, a simulation of some aspect of the manufacturing process is required to predict how this measured parameter will affect final product quality. Third, the digital twin has to have control over a process setting that can be adjusted according to the calculated prediction to achieve satisfactory quality. The effectiveness of the digital twin will depend on what is measured, how it is analyzed, and how accurately the manufacturing process can be adjusted to counteract the predicted loss of quality under current conditions.

This paper specifies and highlights a set of technologies and functionalities which could be used to implement a functional digital twin into a high precision manufacturing process involving sheet metal forming, machining, and welding. A framework is proposed, showing how these functionalities can allow a digital twin to interact with the manufacturing process to shift it from mass production to individualized production.

Presenting Author: Hugo Hultman GKN Aerospace Engine Systems

Authors:

Hugo Hultman GKN Aerospace Engine Systems
Stefan Cedergren GKN Aerospace Engine Systems
Rikard Söderberg Chalmers University of Technology
Kristina Wärmefjord Chalmers University of Technology