<div class="csl-bib-body">
<div class="csl-entry">Müller-Gritschneder, D., & Lieber, P. (2025, August 16). <i>Combining TinyML & Model-based Architecture Methods for Predictive Maintenance in Highly Regulated Environments</i> [Presentation]. AI Solution Days 2025, Nürnberg, Germany. http://hdl.handle.net/20.500.12708/219807</div>
</div>
-
dc.identifier.uri
http://hdl.handle.net/20.500.12708/219807
-
dc.description.abstract
This talk aims to show the full TinyML flow, from AutoML tools to automate the design of TinyML applications to Model-driven architecture (MDA) tools to integrate the tinyML components into a full system.
At first, the tinyML flow is illustrated and automation methods for model development and compression are outlined with special focus on tailoring the AI models for low-power embedded target platforms such as Micro-Controller Units. For full system development, the Hard- and Software-Architecture must span the entire cycle from “as-is” to “will-be”. Data acquisition, the integration with external system units plus the collected data for training ML-models is a typical IoT complexity. Model-driven architecture allows to gain visibility over these dependencies of data-streams, physical systems and intersections.
To illustrate the approach, we demonstrate a TinyML System for Platform Screen Door Systems (PSD) use case. PSDs are used in public transportation and separate the waiting area from the rail line, preventing passenger contact with moving vehicles. These systems require high availability (99.4%+), leading to costly preventive maintenance. PdM (Predictive Maintenance) is not yet applied to PSD. This use case discusses the innovation development of Albayrak (Aldoor/Turkiye), a designer and manufacturer of PSD. Albayrak was looking for an Al-enhanced data-driven predictive maintenance approach for condition monitoring. The goal is early online detection of mechanical faults using a TinyML edge device for PdM, aiming for 99.9% availability at minimal cost. This requires intensive data collection and processing for AI-based condition monitoring, which was realized by combining MDA and TinyML.
Albayrak and Sparx Systems CE are consortium members of the Penta project ECOMAI. The project, founded by the EU under the European Chips Act, addresses critical challenges in the semiconductor market.
en
dc.language.iso
en
-
dc.subject
tinyML
en
dc.subject
Motor Control
en
dc.subject
Embedded Machine Learning
en
dc.title
Combining TinyML & Model-based Architecture Methods for Predictive Maintenance in Highly Regulated Environments
en
dc.type
Presentation
en
dc.type
Vortrag
de
dc.type.category
Presentation
-
tuw.researchTopic.id
I2
-
tuw.researchTopic.name
Computer Engineering and Software-Intensive Systems
-
tuw.researchTopic.value
100
-
tuw.publication.orgunit
E191-02 - Forschungsbereich Embedded Computing Systems
-
tuw.author.orcid
0000-0003-0903-631X
-
tuw.event.name
AI Solution Days 2025
en
tuw.event.startdate
16-09-2025
-
tuw.event.enddate
16-09-2025
-
tuw.event.online
On Site
-
tuw.event.type
Event for scientific audience
-
tuw.event.place
Nürnberg
-
tuw.event.country
DE
-
tuw.event.presenter
Müller-Gritschneder, Daniel
-
tuw.event.track
Single Track
-
wb.sciencebranch
Informatik
-
wb.sciencebranch
Elektrotechnik, Elektronik, Informationstechnik
-
wb.sciencebranch
Mathematik
-
wb.sciencebranch.oefos
1020
-
wb.sciencebranch.oefos
2020
-
wb.sciencebranch.oefos
1010
-
wb.sciencebranch.value
50
-
wb.sciencebranch.value
40
-
wb.sciencebranch.value
10
-
item.languageiso639-1
en
-
item.grantfulltext
none
-
item.openairetype
conference presentation
-
item.openairecristype
http://purl.org/coar/resource_type/R60J-J5BD
-
item.cerifentitytype
Publications
-
item.fulltext
no Fulltext
-
crisitem.author.dept
E191-02 - Forschungsbereich Embedded Computing Systems
