Fraunhofer-Institut für Produktionsanlagen und Konstruktionstechnik IPK

Filters
Reset filters

Settings
Now showing 1 - 10 of 1940
No Thumbnail Available
Publication

AI-based welding process monitoring for quality control in large-diameter pipe manufacturing

2024-04-25 , Gook, Sergej , El-Sari, Bassel , Biegler, Max , Rethmeier, Michael

The paper presents the experimental results into the development of a multi-channel system for monitoring and quality assurance of the multi-wire submerged arc welding (SAW) process for the manufacture of large diameter pipes. Process signals such as welding current, arc voltage and the acoustic signal emitted from the weld zone are recorded and processed to provide information on the stability of the welding process. It was shown by the experiments that the acoustic pattern of the SAW process in a frequency range between 30 Hz and 2.5 kHz contains the most diagnostic information. In the spectrogram of the acoustic signal, which represents the time course of the frequency spectrum of the welding process, the formation of weld irregularities such as undercuts could be reliably identified. The on-line quality assessment of the weld seam produced is carried out in combination with methods of artificial intelligence (AI). From the results obtained, it can be concluded that the use of the latest concepts in welding and automation technology, combined with the high potential of AI, can achieve a new level of quality assurance in pipe manufacturing.

View
No Thumbnail Available
Publication

Volumetric Videos in Extended Reality for enhanced Trainings in Industry

2024 , Freydank, Erik , Kießling, Nora , Konkol, Kathrin

This work presents the concept and prototype of a software for a training system that utilizes volumetric video technology in extended reality environments. Taking assembly processes as an example, a comprehensive immersive video training framework is introduced that enables interaction with volumetric videos.

View
No Thumbnail Available
Publication

With synthetic data towards part recognition generalized beyond the training instances

2024 , Koch, Paul , Schlüter, Marian , Krüger, Jörg

In this work we investigate the effect of using synthetic data, generated in a simulation, in order to pre-train an AI-based image classification for industrial components. After pre-training we use real camera-captured training images to fine-tune the AI with the aim to close the Sim2Real domain gap. We compare our approach to purely using real training images of a single candidate object instance. In an exemplary case study for screw recognition, we found that a given AI classification algorithm dropped its recognition rate from 99.8% to 88.5% when testing the algorithm with known and unknown screw instances of the learned object classes, respectively. Employing our pre-training method on the basis of synthetic data, the drop in recognition rate is decreased from 99% to 96.95%. Thus, our proposed method has only a relative drop of 2.05% when shifting towards a generalized domain (including unknown part instances), while a compared approach on the basis of real camera-captured data showed a drop of 11.3%.

View
No Thumbnail Available
Publication

Characterisation and Investigation of the Mechanical Material Properties of Recycled Glass-Fibre Reinforced Polymers for Sustainable Reusability

2024 , Ludwig, Jonas , Hocke, Toni , Polte, Mitchel , Uhlmann, Eckart

Waste products from glass-fibre reinforced polymers (GFRP), which are used for wind turbine rotor blades, represent a substantial environmental and disposal challenge. According to the state of the art, thermal or chemical recycling methods require high energies and resource consumptions. Based on a new recycling approach, this study analyses the potential of mechanical recycling as a sustainable alternative for the management of GFRP waste. This paper presents a novel solution for mechanical recycling of GFRP waste and analyses the mechanical properties of recycled GFRP materials manufactured with mechanical recycled and processed GFRP waste. The study investigates the impact of particle size classifications on the mechanical properties of recycled GFRP materials, providing insights for optimising the recycling process. Tensile strength tests were carried out to measure the Young’s modulus E and ultimate tensile strength Rm to compare the properties of recycled GFRP specimens with newly manufactured GFRP materials and thus identify the suitability of these materials for specific applications. Furthermore, the GFRP material samples were analysed using computer tomography (CT) to determine the fibre lengths lf in order to analyse their influence on the mechanical material properties. First results showed a correlation between the particle length lp and the ultimate tensile strength Rm of the investigated GFRP material samples. The mechanical recycling approach enables more sustainable GFRP materials compared to the current state of the art, which significantly reduces the environmental impact of GFRP waste disposal.

View
No Thumbnail Available
Publication

Investigation on stereo-ToF data fusion for the inspection of used industrial parts

2024 , Krumpek, Oliver , Schlüter, Marian , Hügle, Johannes

Optical systems for automated or partially automated inspection have been making important contributions to ensure the quality and functionality of technical products for many years. Often used to monitor the quality of newly produced goods, vision systems also aim to play an important role in identification and condition description of used industrial parts such as aged vehicle components. In this work, a passive stereo system and a Time-of-Flight (ToF) sensor of the latest generation were used to create the desired sensor system. In the first step, the pixel-based information of both sensors was exploited to spatially calibrate the transformation between the left stereo camera and the ToF sensor by forming a 2D-3D correspondence set of detected feature points. To compensate for the resolution difference of the sensors, numerous interpolation points were randomly sampled on the reconstructed sparse surface mesh of the ToF sensor to create the missing sub-pixel information. It could be shown that the fused sensor information led to an increase in incompleteness by 7.81% on average for all components examined. The higher noise in the ToF measurement data in the fill-ins could be mitigated by using an adapted median kernel filtering. The average deviation of the measurement from a reference dataset was 1.30mm for the stereo system, 2.51mm for the ToF system, and 1.42mm for the fused result. The result of this work is promising as the quality of the surface mesh could be raised especially for critical surface areas and the underlying RGB data itself can be used for pixel-wise classification and segmentation.

View
No Thumbnail Available
Publication

Kombination des Laserhybridschweißens und UP-Engspaltschweißens für dickwandige Bauteile zur Erhöhung der Wirtschaftlichkeit

2024 , Gook, Sergej , El-Sari, Bassel , Üstündag, Ömer , Gumenyuk, Andrey , Biegler, Max , Rethmeier, Michael

Dieser Beitrag befasst sich mit der Entwicklung eines wirtschaftlichen und robusten Verfahrens zum Schweißen dicker Stahlbleche. Es werden Ergebnisse für das Fügen von Blechen im Dickenbereich von 25 mm bis 80 mm vorgestellt. Die 30 mm dicken Schweißnähte konnten im Stumpfstoß in zwei Durchgängen unter Verwendung des Laserhybridschweißens und des Unterpulverschweißens (UP) geschweißt werden. Das Laserhybridschweißen wird als Einschweißung mit einer Tiefe von ca. 25 mm ausgeführt. Die UP-Lagen werden auf der gegenüberliegenden Seite des Blechs aufgebracht. Bei richtiger Wahl der Schweißparameter wird der Nahtquerschnitt durch eine sichere Überlappung beider Lagen geschlossen. Bei Blechdicken über 30 mm ist eine Kantenvorbereitung erforderlich, die eine Ausführung der UP-Fülllagen in Engspalttechnik ermöglicht. Die Vorteile der vorgeschlagenen Prozesskombination liegen darin, dass die Nahtwurzel der Laserhybridnaht nicht geformt werden muss. Defekte in der Wurzel der Laserhybrideinschweißung können durch die UP-Lagen effektiv beseitigt werden. Das Verfahren bietet eine hohe Stabilität des Schweißprozes-ses in Bezug auf die Qualität der Blechkanten. Plasmageschnittene Kanten konnten ohne Bindefehler geschweißt werden. Die erreichte Nahtqualität wurde durch zerstörende und zerstörungsfreie Prüfungen bestätigt.

View
No Thumbnail Available
Publication

Towards anomaly detection: A feasibilty study for quality control in productions of nonwovens

2024 , Lehr, Jan , Pape, Martin , Nguyen Hoang, Viet , Philipps, Jan , Krüger, Jörg

In the face of the upcoming Corona pandemic in 2020, which continues to this day, adequate protection for all citizens has been very important and continues to be essential. An easy way to minimize the risk of infection in public, but poorly ventilated places, are FFP masks. The basic material from which FFP masks are made is nonwoven. The quality of the produced nonwoven is the basis for the correct function and thus the protection against infection. The present research work investigates methods of computer vision to detect contaminations and damages on the nonwoven. Both supervised and semi supervised methods are evaluated. For this work, an inspection system consisting of two separate acquisition systems was developed for image data acquisition. The first system is suitable for the visible wavelength and has a theoretical resolution of 2.4 μm per pixel. The second camera system is designed for the near infrared range and has a theoretical resolution of 5.5 μm per pixel. The acquisition system collects an image data set comprising 1,760 images with 920 images of defect-free nonwoven samples and 840 images of defective nonwoven samples. First, the wavelength range suitable for optical inspection of uncoated nonwoven samples is investigated. A further investigation dealt with the question of whether the reflected light or transmitted light method is more suitable for optical inspection. Finally, coated nonwoven is also inspected using the reflected light method. An investigation using the transmitted light method is not possible, as the material is almost non-transparent. Despite the small amount of data, very good results were achieved. Machine learning methods from the field of image processing are usually classified as deep learning. This means that the large network architectures require very large amounts of data in order to learn complex patterns. Publicly available datasets for method evaluation typically consist of over 1,000 images per class or defect. The nonwoven samples provided in this work and the resulting image database is about a factor of six smaller than is actually intended for the methods used. In this work it has been possible to achieve inspection accuracies of 97.5 %.

View
No Thumbnail Available
Publication

An easy-to-use and cost-efficient fingertip exchange system for robot assembly tasks

2024 , Kröger, Ole , Goppold, Jean-Paul , Mohr, Sebastian

For handling in automated (robot-based) assembly processes, different gripper solutions are often required for different parts or process steps. In addition, introducing a new grasping solution can also be a time-consuming job. Previous solutions use e.g. specialized tool changing systems where the entire gripper system is exchanged to handle the variations of different parts, shapes and geometries. However, these systems are expensive and require an entire magazine of complete gripper systems. Therefore, in this paper we present a flexible, easy and cost-efficient fingertip exchange system. In this way, only one gripper is needed, as it can be used for many different fingertips. The system consists of a mechanical fingertip changing system, which can be installed on different (user-specific) gripper systems. In addition, we introduce a Parametrized Finger Method that automates the design for custom fingertips. Important process parameters such as the type of gripper kinematics, grasping force and max. length of the gripper fingers can be taken into account. The fingertips can then be manufactured cost-effectively with a 3D printer and fit perfectly to the individual objects. The system was benchmarked by the task-board task from the World Robot Challenge 2018 and with a UR5 robot. The average success rate of the automatically generated tips in this real-world scenario was 90 %.

View
No Thumbnail Available
Publication

Insights of Anomaly Detection: How does Polluted Training Data Influence Performance?

2024 , Lehr, Jan , Pape, Martin , Günther, Samuel , Krüger, Jörg

Anomaly detection is one of the most popular fields for computer vision in industrial applications. The idea of training machine learning only on defect-free objects saves enormous amounts of integration effort. The state of the art shows that current methods on public data sets (e.g. MVTec AD data set [1]) have already solved the problem with AUROC segmentation scores of more than 99 %. In real-world applications training data is not as "clean" as in public data sets. This work investigates the changes in detection performance when outliers end up in the training data. For this purpose, the training data is enriched step by step with images of defective objects. The AUROC score and the anomaly score is used as a quality criterion for performance measurement. We show that state of the art methods can be very robust, but that in some scenarios a draw down of 15 percentage points is possible.

View
No Thumbnail Available
Publication

Energy efficiency measures for metal forming using hydraulic and mechanical presses

2024 , Preuß, Mandy , Thiele, Gregor , Krüger, Jörg

In conventional hydraulic presses only around 13 percent of the energy supplied is converted into the plastic forming of the metal. Mechanical presses achieve an overall efficiency of 13 to 30 percent. A variety of approaches aspire efficiency improvements by tuning process parameters. For the operation engineers it is a crucial task to find measures which are both technical suitable and economical promising. This paper summarizes popular tuning parameters and presents the assessment by seven experts who participated in semi-structured interviews. For hydraulic presses, the most recommend tuning parameter to be adjusted is the pressing force. For mechanical presses the continuous stroke operation is the most popular measure.

View