Fraunhofer-Institut für Produktionsanlagen und Konstruktionstechnik IPK

Filters
Reset filters

Settings
Now showing 1 - 10 of 19
No Thumbnail Available
Publication

Inertial Measurement Unit based Human Action Recognition for Soft-Robotic Exoskeleton

2021 , Kuschan, Jan , Burgdorff, Moritz , Filaretov, Hristo , Krüger, Jörg

Absence from work caused by overloading the musculoskeletal system lowers the life quality of the worker and gains unnecessary costs for both the employer and the health system. Exoskeletons can present a solution. Typically, such systems struggle with stiffness and discomfort and primarily a lack of battery lifetime. Soft-robotic exoskeletons offer a possibility to overcome these problems by increasing the system flexibility, not limiting the supported DoF and being actuator and joint together. Since soft-robotic exoskeletons can be designed only using power when supporting the wearer, it is possible to increase the battery lifetime by only acting on those actions for which the wearer needs support. Dealing with controls for soft-robotic exoskeleton one major difficulty is to find a compromise between saving energy and supporting the wearer. Having an action-depending control can reduce the supported actions to cover only relevant ones and increase the lifetime of the battery. The system conditions are to detect the user actions in real-time and distinguish between actions which require support and those which do not. We contribute an analysis and modification of human action recognition(HAR) benchmark algorithms from activities of the daily living, transferred them onto industrial use cases containing short and mid-term action and reduce the models to be compatible using embedded computers for real-time recognition on soft exoskeletons. We identified the most common challenges for inertial measurement units based HAR and compare the best-performing algorithms using a newly recorded data set overhead car assembly for industrial relevance. As a benchmark data set we focused on the "Opportunity" data set. By introducing orientation estimation, we were able to increase the F1 scores by up to 0.04. With an overall F1 score without a Null-class of up to 0.883, we were able to lay the foundation to use HAR for action dependent force support.

View
No Thumbnail Available
Publication

Model-Based Systems Engineering (MBSE) as computer-supported approach for cooperative systems development

2020 , Schmidt, Marvin M. , Stark, Rainer

With rising globalization and a trend towards Cyberphysical systems (CPSs) as well as smart products the demand for cross-company and interdisciplinary collaboration increases. To handle the complexity of these systems and products Model-Based Systems Engineering (MBSE), as an enhanced form of Systems Engineering (SE), has emerged in engineering and is adopted by many companies. While this approach tries to cope with the current complexity trends, it does address the collaborative aspect of product creation only in a small scope. This paper shall address the combination of MBSE and collaboration in engineering to form a computer-supported approach for collaborative systems development.

View
No Thumbnail Available
Publication

Hand Shape Recognition Using Very Deep Convolutional Neural Networks

2018 , Rakowski, Alexander , Wandzik, Lukasz

This work examines the application of modern deep convolutional neural network architectures for classification tasks in the sign language domain. Transfer learning is performed by pre-training the models on the ImageNet dataset. After fine-tuning on the ASL fingerspelling and the 1 Million Hands datasets the models outperform state-of-the-art approaches on both hand shape classification tasks. Introspection of the trained models using Saliency Maps is also performed to analyze how the networks make their decisions. Finally, their robustness is investigated by occluding selected image regions.

View
No Thumbnail Available
Publication

Quantification and compensation of systematic errors in pressure measurements applied to oil pipelines

2018 , Thiele, Gregor , Liu, Martin , Chemnitz, Moritz , Krüger, Jörg

The monitoring of pipeline operation is an important research topic, especially for the detection and localization of leaks as well as for an efficient control. For these purposes, physical quantities in pipelines are calculated from measurement data on the basis of a mathematical model. In contrast to static models, adaptive models vary their parameters or even their structure to reach the most probable solution. But in most cases, even the best fit will hold residuals caused by discrepancies between the real system and its model. These residuals allow an estimation of travel-time delays of pressure waves and offsets in pressure values. The basic idea of our approach is to interpret these systematic, time-invariant errors of pressure measurements in pipelines either as sensor displacements or as technical defects. The proposed procedure leads to a hypothesis for a model update, regarding the sensor positions. This displacement compensation as well as a variance analysis was successfully applied to real data from a crude oil pipeline in Europe. A cross validation proves the general capability of the developed method to reduce the uncertainties.

View
No Thumbnail Available
Publication

Analysis and recycling of bronze grinding waste to produce maritime components using directed energy deposition

2021 , Müller, Vinzenz , Marko, Angelina , Kruse, Tobias , Biegler, Max , Rethmeier, Michael

Additive manufacturing promises a high potential for the maritime sector. Directed Energy Deposition (DED) in particular offers the opportunity to produce large-volume maritime components like propeller hubs or blades without the need of a costly casting process. The post processing of such components usually generates a large amount of aluminum bronze grinding waste. The aim of the presented project is to develop a sustainable circular AM process chain for maritime components by recycling aluminum bronze grinding waste to be used as raw material to manufacture ship propellers with a laser-powder DED process. In the present paper, grinding waste is investigated using a dynamic image analysis system and compared to commercial DED powder. To be able to compare the material quality and to verify DED process parameters, semi-academic sample geometries are manufactured.

View
No Thumbnail Available
Publication

Automated tool-path generation for rapid manufacturing and numerical simulation of additive manufacturing LMD geometries

2019 , Biegler, Max , Wang, Jiahan , Graf, Benjamin , Rethmeier, Michael

In additive manufacturing (AM) Laser Metal Deposition (LMD), parts are built by welding layers of powder feedstock onto a substrate. Applications for steel powders include forging tools and structural components for various industries. For large parts, the choice of tool-paths influences the build-rate, the part performance and the distortions in a highly geometry-dependent manner. With weld-path lengths in the range of hundreds of meters, a reliable, automated tool path generation is essential for the usability of LMD processes. In this contribution, automated tool-path generation approaches are shown and their results are discussed for arbitrary geometries. The investigated path strategies are the classical approaches: ""Zig-zag-"" and ""contour-parallel-strategies"". After generation, the tool-paths are automatically formatted into g-code for experimental build-up and ASCII for a numerical simulation model. Finally, the tool paths are discussed in regards to volume-fill, microstructure and porosity for the experimental samples. This work presents a part of the IGF project 18737N ""Welding distortion simulation"" (FOSTA P1140)

View
No Thumbnail Available
Publication

Morphing Detection Using a General-Purpose Face Recognition System

2018 , Wandzik, Lukasz , Kaeding, Gerald , Vicente-Garcia, Raul

Image morphing has proven to be very successful at deceiving facial recognition systems. Such a vulnerability can be critical when exploited in an automatic border control scenario. Recent works on this topic rely on dedicated algorithms which require additional software modules deployed alongside an existing facial recognition system. In this work, we address the problem of morphing detection by using state-of-the-art facial recognition algorithms based on hand-crafted features and deep convolutional neural networks. We show that a general-purpose face recognition system combined with a simple linear classifier can be successfully used as a morphing detector. The proposed method reuses an existing feature extraction pipeline instead of introducing additional modules. It requires neither fine-tuning nor modifications to the existing recognition system and can be trained using only a small dataset. The proposed approach achieves state-of-the-art performance on our morphing datasets using a 5-fold cross-validation.

View
No Thumbnail Available
Publication

Production environment of tomorrow (ProMo)

2021 , Kuschan, Jan , Müller, Vinzenz , Monchinger, Stephan , Heimann, Oliver , Niebuhr, Carsten , Kabha, Oday

Small defects in the grain or major damage to a moulded part or tool can bring production to a standstill. SMEs in particular have neither the personnel nor the equipment to repair such damage on their own, so they send it to specialised contractors. The repair process is carried out manually, depending on the accuracy requirements, and is usually completed by a finishing process. This work requires qualified personnel and, at the same time, requires a lot of time in case of larger damages. In this paper we present a way to map the Maintenance, Repair and Operations (MRO) process chain in a partially automated manner. The symbiosis of individual technologies results in a significantly increased efficiency of the MRO process chain, which continues to focus on people and their process knowledge. While Directed Energy Deposition (DED) for the MRO of moulded parts is used widely, usually a high manual effort in measuring the component geometries and teaching of the machine tool paths is necessary. However, there are clear advantages compared to the manufacture of new parts or manual laser welding repair. At the same time, the resource and energy requirements can often be significantly reduced compared to new part production. ProMo focuses on automating the time-consuming machine programming by reducing the number of necessary work steps in CAD/CAM-based program creation. Based on a subsequent robot-guided scan, a digital actual 3D model is generated. Due to intelligent path planning algorithms, no manual programming of the robot is necessary and at the same time it is possible to detect components of different sizes, shapes and covers in this system with a minimum of effort. In addition, the operator passes on elementary information, such as the approach path of the milling head, to the subsequent processes by means of finger gestures and can thus significantly reduce tedious CAM programming steps. Now, the scanned component is transferred to a 3D-CAD model and a target/actual comparison is created for the damaged areas. Those are milled out in a defined manner and then restored using DED.

View
No Thumbnail Available
Publication

Multimodal Environment Dynamics for Interactive Robots

2018 , Haninger, Kevin , Surdilovic, Dragoljub

Interactive robots offer improved performance in tasks with environmental uncertainty, but accommodating environment input weakens predictions of contact force or position trajectories, making the identification of subtask completion or faults difficult. This paper develops a task monitoring approach for complex assembly tasks that involve transitions between discrete environment dynamic modes. In semi-structured environments, these dynamic modes and their transitions are approximately known a priori, allowing task monitoring through estimation of the current mode and fault detection as a deviation from expected, desired dynamic mode transitions. This allows a more natural description of many interactive tasks, improving robustness to variations in force or position trajectories that impedance control seeks to address. The ability of impedance and admittance controlled robots to identify their environment is investigated, making consideration of joint and end-effector physical compliance. Prior information on environment dynamics and mode transitions allow recursive estimates of dynamic mode suitable for online use, under both full state knowledge and only force/position measurements. Experiments with an admittance controlled robot in a gear assembly task validate the approach.

View
No Thumbnail Available
Publication

Identification of Human Dynamics in User-Led Physical Human Robot Environment Interaction

2018 , Haninger, Kevin , Surdilovic, Dragoljub

Human dynamic models are useful in design of physical human-robot and human-robot-environment interaction: informing choice of robot impedance, motivating relaxations to passivity-based safety constraints, and allowing online inference to user intent. Designing for performance objectives such as stable well-damped contact transitions also requires nominal models, but the use of human models in controller design is limited. Established approaches to identify human dynamics apply position or force perturbation and measure the corresponding response, mostly to validate neuromuscular hypotheses on motor control, which raises questions about their transferability to human-led collaboration. Here, human dynamics are identified in a task which closely resembles the final application, where the human leads the robot into contact with a (virtual) wall. This paper investigates the impact of human dynamics on coupled system behavior, and establishes a general framework for identification in human-led scenarios, making consideration of unmeasured human input. Experiments with different stiffness environments allow inference to human dynamics, and characterize the range of human dynamics which can be modulated by the user.

View