Hochschule Bremen - University of Applied Sciences

Hochschule Bremen

2. Exkursion "Künstliche Intelligenz in Technik und Gesellschaft"

Wann: Mittwoch, 11.12.2019, 16.00 bis ca. 17:30 Uhr
Wo: Firma TESTIA, Cornelius-Edzard-Straße 15,
Treffpunkt vor dem EcoMaT-Gebäude
Wer: alle Angemeldeten! *(s.u.)
Die Firma TESTIA gehört zur Airbus Group und beschäftigt sich seit über 25 Jahren mit den Themen zerstörungsfreies Testen sowie Inspektion und Qualitätssicherung im Luft- und Raumfahrtbereich.

In diesem Zusammenhang gewinnt der Einsatz von Augmented Reality sowie von KI-Technologien wie z.B. Predictive Maintenance immer stärker an Bedeutung. Predictive Maintenance ist eine Technologie für die vorausschauende Wartung bzw. die Überwachung des „Gesundheitsstatus“ von Maschinen bzw. deren Produkte, die zu diesem Zweck große Mengen an kontinuierlichen Sensordaten aus dem Produktionsprozess analysiert.
TESTIA hat seit kurzem seinen Sitz im neu gegründeten EcoMaT - “Center for Eco-efficient Materials & Technologies”. Im Anschluss an den Vortrag (in Englisch) ist ein Rundgang geplant.
*ANMELDUNG: bis zum 6.12.2019
Anmeldung / Ansprechpartnerin
Name Telefon E-Mail
Bohnebeck, Uta, Prof. Dr.-Ing. +49 421 5905 5096  senden
Please find below the abstracts of the presentations:

Title: A data driven approach to the monitoring of the additive manufacturing process

Abstract: Process monitoring in additive manufacturing (AM), i.e. in laser powder bed fusion (LPBF) of metal parts, has been identified as the crucial bottleneck in accelerating the AM industrialization process. Despite the growing process understanding and the technological enhancements in the last years the process stability alone cannot guarantee the high defect avoidance criteria in aerospace. Therefore a quality control of the manufactured parts becomes necessary. As of today, this quality control is performed by computed tomography (CT) following the production process. Consequently a possible defect in a part can only be detected in a post process step. This is not only time and cost expensive but also prevents the in process reaction to possible anomalies.

To reduce the cost and time needed to produce and qualify an AM part, an online monitoring system of the manufacturing process is desirable. There exist a variety of different monitoring systems from different printer manufacturers and research facilities. While these systems capture a large amount of process data they are not able to interpret the acquired data adequately.

This work uses machine learning and a data correlation approach to better understand and analyze the monitoring data of LPBF. By correlating the online monitoring with post-process testing data, i.e. computed tomography data, the effect of monitored anomalies on the finished sample are evaluated. For this purpose the datasets are registered to a common reference frame. The features detected in CT are flagged and their spatial location transferred into the monitoring data space. This way labelled data for a subsequent machine learning pipeline is generated. On the basis of the localized features a neural network is trained to detect similar features in the monitoring data. The neural networks shows a high performance on a sample dataset with artificially produced artifacts.

The approach itself is independent of the monitoring system and can be applied to a variety of input signals. The presented results are obtained using a melt pool monitoring system.

Title: Introduction to structure health monitoring (SHM) – The importance of data and analytics

Abstract: The future of inspection, especially for aerospace structured aims to identify technological advances that can continuously and autonomously monitor the critical components leading to less scheduled and unscheduled maintenance. Structure health monitoring (SHM) systems, in the context, caters to this aspect by monitoring the physical condition (and thereby the structural integrity) of a structure by the means of embedded or attached sensors with a minimal manual intervention. The presentation would introduce the concept of SHM, its benefits and how it fits into the global need for predictive maintenance of structures. Furthermore, the importance of data in SHM would be emphasized together with its importance in the technological roadmap of SHM.

We should meet at the lobby of the EcoMaT – Cornelius-Edzard-Straße 15

veröffentlicht am 2019-11-20 13:42



Access Keys: