BMVg, DTEC: LaiLa

LaiLa – laboratory for intelligent lightweight construction

The production environment in fiber composite lightweight construction is characterized by a high proportion of manual activities and, at the same time, a high number of variants with the highest quality standards. In aircraft construction, for example, a 100% test of all components is carried out, whereby this activity is mainly carried out manually and the documentation is stored in paper form. As a result, this environment demands a high level of expert knowledge and is heavily dependent on the individual availability of staff. On the other hand, quality-relevant processes are not available digitally, which in later life cycle phases, e.g. in the context of production or in operation, leads to considerable additional expenditure. At the same time, recurring efforts in the development and integration of solutions inhibit the comprehensive implementation of information and automation technology. As a result, modern data analysis and process optimization cannot be carried out. A major challenge for the introduction of digital technologies is the transfer and further development of research results for use in production.

Various applications of sub-symbolic artificial intelligence (AI), including for the analysis of product quality, system diagnosis and predictive maintenance. Machine learning (ML) models, for example neural networks, cannot be used directly in production, but require a lot of effort, for example for setting parameters. In the field of symbolic AI, there are approaches to formal information modeling to describe machine functions and to ensure semantic interoperability between cyber-physical systems. However, the two areas of AI are usually considered separately in automation technology. The combination of formal knowledge models with approaches from ML has great potential, because machine-readable information can reduce the training effort in the field of ML. In addition, the content of knowledge models can be optimized with “learned” content. This project therefore aims to link symbolic and subsymbolic methods of AI so that expert knowledge and machine functions are described in a machine-readable manner and methods of subsymbolic AI are made accessible.

The LaiLa project deals with the following three core challenges:
● How can interfaces between physical production and quality assurance, digital models, cloud architectures, simulations and machine learning look like?
● Can data from the areas of production be recorded uniformly and used for automated production and product monitoring, e.g. use the evaluation of the hole quality based on sensor data from processing machines? Which data, system architectures and which machine learning processes are suitable for industry?

Duration: 09/01/2020 to 08/31/2024

The LaiLa project is funded by the Federal Ministry of Defense (BMVg) in the LaiLa funding program.

HSU

Letzte Änderung: 9. October 2020