JEC COMPOSITES MAGAZINE - Issue #141 - July/August 2021 - 51

Sensor
Analysing the output data of Eddy current sensors using
Artificial Intelligence for quality assurance of carbon fibres
SMC/BMC
Comparative study of competitive
magnesium oxide samples for SMC/BMC composites
Curing
Rapid-cure epoxy formulations for
next-generation composite manufacturing
51
54
56
sensor TECHNOLOGY
Analysing the output data of Eddy current sensors using
Artificial Intelligence for quality assurance of carbon fibres
Lukas LechthaLer1
JuLien PiLon1
richard kuPke2
PaweL Latasiewicz4
1
, Researcher
thomas Gries1
, Research Assistant
, Director
, Director Product management
dirk FeLtin3
, CEO
christine stamm5
3
, System development NCF Inspection
, Manager Technical Service
Institut für Textiltechnik of RWTH Aachen University
2
SURAGUS GmbH
Hightex Verstärkungsstrukturen GmbH
4
SAERTEX GmbH & Co. KG
5
Teijin Europe GmbH
The rise of carbon fibre reinforced plastics throughout its fields of
application has given place to numerous innovations in its production
process. Eddy current sensors for quality assurance of carbon
fibre rovings are one such innovation. These sensors could allow
the non-destructive analysis of carbon fibres without disrupting
the production flow. This project aims to analyse the output data
from eddy current sensors using machine learning algorithms to
identify possible defects in carbon fibres.
T
he scope of the corresponding research project as well as results
of preliminary testing was published in JEC Composites
Magazine. Vol. 57 (2020) 136, p. 59-61. Data was first collected
at the Institute für Textiltechnik (ITA) of RWTH Aachen University
in Germany using the test bench specially outfitted with the proper
measurement tools. During these tests, five different severities of
damages were created on the fibres via a mounted damaging tool.
The collected data was subsequently interpreted with a suitable machine
learning algorithm to prove its ability to find the five generated
damage grades. Initial results demonstrated that five distinct clusters
of data corresponding to the five damage grades on the carbon fibres
were identified.
Introduction
The purpose of this research project is to lay the groundwork for the
interpreting measurement signals from eddy current sensors used to
analyse carbon fibre rovings by means of machine learning algorithms.
The general aim is to develop an eddy current sensor that can be integrated
into the production lines of carbon fibre manufacturers as well
as processors. This would allow a 100% non-contact and non-destructive
inspection of the fibres, thus improving the overall quality assurance
process. The sensor should be able to detect, classify and quantify
defects present in the carbon fibres like gaps, fuzz fibre breaks, etc.
As mentioned, this part of the research project focuses on using machine
learning to evaluate the data outputted by the eddy current sensor.
Because of the relative complexity of the data, machine learning
algorithms could simplify analysis in comparison with traditional
methods and prove to be a great tool for detecting finer defects that
may have occurred on the fibre. Furthermore, using such algorithms
allows evaluation of the complex data output of eddy current sensors
to be done autonomously. Without automation, data analysis can be
rather difficult and time-consuming, since this data is generally prone
to high standard deviation.
Fig. 1: Example deviation of the complex impedance when a defect is present on the
carbon fibre roving
N°141 July - August 2021 / jec composites magazine
51
TECHNOLOGY
JEC COMPOSITES MAGAZINE - Issue #141 - July/August 2021

Table of Contents for the Digital Edition of JEC COMPOSITES MAGAZINE - Issue #141 - July/August 2021
