EPP Europe P2.2023
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
» TEST & QUALITY ASSURANCE<br />
Source: Göpel electronic<br />
AI evaluation of a QFN pin solder joint in a 3D X-ray image with AI advisor<br />
of prismatic battery cells for foreign objects or the<br />
AXI inspection of sheath thermocouples directly in<br />
the production line. As a human being, you often see<br />
certain defect features quickly – but the machine<br />
needs an expert who has mastered the parameterisation<br />
of an inspection function in order to be able<br />
to evaluate a feature correctly. This is where AI can<br />
help. The AI system is trained using example images.<br />
This results in an AI model that carries out the evaluation<br />
without any setting parameters. This is exemplified<br />
in the evaluation of welded joints on sheath<br />
thermocouples.<br />
Zusammenfassung<br />
Noch sind wir nicht bei autonomen Inspektionssystemen<br />
angekommen, doch KI-basierte Technologien in der Inspektion<br />
ermöglichen bereits heute die vollautomatische<br />
Erstellung sowie Optimierung von Prüfprogrammen.<br />
Résumé<br />
Bien que nous n’ayons pas encore de système<br />
d’inspection automatique, les technologies basées sur l’IA<br />
permettent déjà aujourd’hui de créer et d’optimiser des<br />
programmes de contrôle de manière entièrement<br />
automatique.<br />
Резюме<br />
Хотя мы еще не пришли к автономным системам<br />
контроля, тем не менее технологии на основе<br />
искусственного интеллекта в области контроля уже<br />
сегодня обеспечивают полностью<br />
автоматизированное изготовление и оптимизацию<br />
программ проверки.<br />
Further AI applications are in the works. In x-ray<br />
inspection, AI solutions for void detection, BGA<br />
head-in-pillow detection and short-circuit control<br />
are already being developed. AI approaches are<br />
emerging in predictive maintenance and for the<br />
analysis of results data from inspection systems<br />
across the whole SMT line. “There’s a long way to go<br />
before we achieve fully-automated inspection but<br />
we shouldn‘t forget how far we’ve come,” says Tuerk.<br />
AI at the verification station<br />
AI is not only saving human resources in the creation<br />
of inspection programmes. There is also the potential<br />
to save time and money at verification and repair<br />
stations. Goepel electronic relies upon the AI advisor<br />
software module for this purpose. Based on pretrained<br />
models, the AI advisor makes its own decision<br />
for each anomaly found by the inspection system. The<br />
AI‘s decision is based on the verification decisions<br />
made by humans in comparable cases in the past.<br />
Now there are different options for deploying AI decision-making.<br />
Anomalies that are classified with high<br />
confidence as pseudo errors can be evaluated directly<br />
by the AI – without human intervention. This saves<br />
time. The AI advisor can also warn us about whether<br />
a human error could lead to a so-called human slip. If<br />
an operator incorrectly classifies a real error as a<br />
pseudo error, the user is asked to reconsider his decision.<br />
Especially in the case of X-ray images – where<br />
error interpretation is often more demanding – the AI<br />
advisor can be a welcome aid. Its functionality is constantly<br />
being expanded with the aim of a fully autonomous<br />
classification of all abnormalities found.<br />
Trust is good – control is better!<br />
“Often there is still a sense of unease about the<br />
use of artificial intelligence in the field of inspection<br />
systems” notes Tuerk. “Why did the AI decide one<br />
way and not another? Will the AI find previously discovered<br />
defects even after it has been trained with<br />
new image samples? These are all legitimate questions.<br />
Luckily we already have some answers. The AI<br />
must be trustworthy – decisions must be comprehensible.<br />
With this in mind, Goepel electronic is<br />
working on ‘Comprehensible AI’ to increase trust and<br />
acceptance. We shouldn’t forget AI is only as good as<br />
the human who trained it, and this depends entirely<br />
on data selection, its completeness and consistently<br />
correct labelling and categorisation,” says Tuerk.<br />
There may still be some way to go before we<br />
achieve fully-autonomous inspection but, thanks to<br />
AI, the next paradigm shift is already well underway.<br />
Productronica, Booth A2.239<br />
www.goepel.com<br />
54 <strong>EPP</strong> <strong>Europe</strong> » 11 | 2023