Project description

Background:

Against the backdrop of highly volatile markets, existing highly automated series production (“lights-out manufacturing”) is to be expanded into autonomous and universal manufacturing with new, flexible, and intelligent production machines (“universal lights-out manufacturing”). Universal lights-out manufacturing makes it possible to incorporate new materials and technologies into series production without delay and to manufacture highly customized products. To this end, universal manufacturing systems will independently develop and optimize manufacturing processes and process chains.

Goal:

In this context, the research project aims to develop approaches in which the entire process chain is carried out autonomously based on product requirements formulated by humans - from the planning and parameterization of the manufacturing processes to the orchestration of possible tests to the production and evaluation of the final product. Thanks to the machines' ability to learn and the mutual exchange of knowledge among them, these systems become continuously more efficient over time. This leads to excellent, constantly improving production quality with low costs, increased flexibility, and productivity. This strengthens Baden-Württemberg as a production location, making it resistant to the shortage of skilled labor and competitive in the global environment in the long term. The solutions developed will be demonstrated using examples of applications from laser-based manufacturing.

Contribution:  

As part of the “AutoUniManu” research project, the IFL is developing explainable learning systems (xAI) for the autonomous selection of relevant process parameters in laser-based manufacturing processes. Until now, the selection of relevant parameters and parameter combinations has often been carried out manually, which is very demanding and leads to errors and expensive trials. Here, xAI methods can use data-based process models to gain insights into influencing variables and interactions of underlying physical processes, the analytical modeling of which involves considerable effort.