Forschung & Transfer

Ein Forschungsteam unter Leitung von Patrick Huber von DESY und der TU Hamburg hat ein überraschendes Phänomen in einem nanoskopischen Silikatglas mit einem "Nanoschwamm" entdeckt

The Zero C project aims to create necessary conditions in higher education to provide the shipping industry and governmental institutions in Albania and Montenegro.

Modeling Cyber-Physical Systems (CPS) requires knowledge from various domains, including computer science, electrical and mechanical engineering, and control theory. One conventional approach to model CPS is to describe the physical relationships as a classical system of formulas. This requires, a solid understanding of the application domain is required to ensure relevant and accurate models. The behavior of a CPS can be estimated by numerical simulations using e.g. Matlab, Simulink, or Modelica. However, these approaches require that the system’s internals are sufficiently and accurately known, which often is not the case CPS. Thus we employ data driven learning approaches to automatically generate CPS models. We develop Flowcean, which offers a toolbox for modeling CPS from various industrial domains [1]. The project’s consortium spans three distinct application domains, maritime systems, energy grids, and intralogistics, represented by industrial partners KALP GmbH, VIVAVIS AG, SICK AG and KION GROUP AG. The company KALP deploys self-sufficient automated twistlock handling platform based on a hydraulic pressure system. VIVAVIS provides smart IoT solutions especially for the efficient control of smart electricity grids [2]. Associated partners bring further expertise in sensor technology (SICK) and robotics (KION) for intralogistic scenarios [3]. All partners provide either software or hardware solutions to demonstrate how Flowcean can improve testing, operation, and monitoring of CPS. Flowcean applies data driven learning to understand and replicate the individual system behavior. The modeling process is structured into four steps: 1. Loading recorded data or starting a simulation 2. Transforming data to a suitable format 3. Learning models using data driven techniques 4. Evaluating the model’s performance via various metrics Flowcean encompasses both online and offline learning techniques [4],[5]. To be able to model CPS of various domains, the entire pipeline has a modular structure. Thus, each transforming or learning step is composable with others to create distinct pipelines. So far, the design of the framework’s architecture and the implementation of basic examples prove a functioning application. Our upcoming goals are • the analysis of more complex CPS from the domains of the project’s consortium, • finalizing the integration of online and offline learning strategies as well as • the development of tools to use learned models for testing [6], monitoring, or behavioral prediction. Partners: • Fraunhofer Center for Maritime Logistics and Services (CML) • Institute of Embedded Systems, Hamburg University of Technology • Institute of Technical Logistics, Hamburg University of Technology • OFFIS – Institut für Informatik • VIVAVIS AG • KALP GmbH • SICK AG • KION GROUP AG The project is funded by the Federal Ministry of Education and Research (BMBF). Contact: Hendrik Rose & Markus Knitt Institute for Technical Logistics hendrik.wilhelm.rose@tuhh.de, markus.knitt@tuhh.de Maximilian Schmidt & Swantje Plambeck & Görschwin Fey Institute of Embedded Systems maximilian.schmidt@tuhh.de, swantje.plambeck@tuhh.de, goerschwin.fey@tuhh.de Bibliography [1] M. Knitt et al., “Towards the Automatic Generation of Models for Prediction, Monitoring, and Testing of Cyber-Physical Systems”, in International Conference on Emerging Technologies and Factory Automation (ETFA), 2023. [2] L. Fischer, J.-M. Memmen, E. M. S. P. Veith, and M. Tröschel, “Adversarial Resilience Learning - Towards Systemic Vulnerability Analysis for Large and Complex Systems”, ArXiv, 2018. [3] M. Knitt, Y. Elgouhary, J. Schyga, H. Rose, P. Braun, and J. Kreutzfeldt, “Benchmarking for the Indoor Localization of Autonomous Mobile Robots in Intralogistics”, Logistics Journal : Proceedings, no. 1, 2023. [4] J. Schyga, S. Plambeck, J. Hinckeldeyn, G. Fey, and J. Kreutzfeldt, “Decision Trees for Analyzing Influences on the Accuracy of Indoor Localization Systems”, in International Conference on Indoor Positioning and Indoor Navigation (IPIN), 2022. [5] E. Veith et al., “palaestrAI: A Training Ground for Autonomous Agents”, in European Simulation and Modelling Conference (ESM), 2023. [6] S. Plambeck and G. Fey, “Data-Driven Test Generation for Black-Box Systems From Learned Decision Tree Models”, in International Symposium on Design and Diagnostics of Electronic Circuits and Systems (DDECS), 2023.

System-on-Chip (SoC) technology is a driving force behind the growth and advancement of various digital technologies in our daily lives.  Focussing on cyber-physical systems (CPS), the past decade has seen massive growth in demand for different kinds of SoCs, encompassing increased computational power, energy efficiency, and cost-effectiveness. To ensure the reliable functionality of SoC devices, post-silicon validation plays a pivotal role. This is one of the most intricate and costly stages of the SoC design cycle, primarily because the post-silicon validation process generates a large amount of data (e.g. trace files, electrical test reports, oscilloscope images, etc., see Figure 1). While the complexity of SoCs is growing, the amount of test data is growing too and there is pressure to reduce the post-silicon validation time amidst fierce market competition. To overcome this issue, we propose AI (artificial intelligence) for smart post-silicon validation. This is a  collaborative venture between the massively parallel systems group, the smart sensors group, and NXP Hamburg. Our project introduces an AI-powered method to automatically detect anomalies in the test traces and oscilloscope images, which provides several benefits including a reduction in validation time, errors, and accelerated time-to-market. One of the standout features of our models lies in their training on real SoC project data, thanks to NXP Hamburg for the collaboration!  Training our models on 8,044 labeled oscilloscope images—deemed 'good'—we further evaluated their performance using the Reconstruction Error (RCE) metric. Although RCE is a prevalent metric, we introduce the use of Kernel Density Estimate (KDE) to refine anomaly detection accuracy. The decision whether a  given oscilloscipe image is anomalous or not is made by identifying a suitable threshold for the RCE (RCETh) and KDE (KDETh) metrics. Figure 2 shows the model’s performance to detect anomalies in oscilloscope images. Our goal is to minimize false negatives (predicted label 0, actual label 1) to ensure that critical anomalies are not overlooked and reliable SoCs are delivered to users, while simultaneously aiming to maintain false positives (predicted label 1, actual label 0) within an acceptable range to reduce human effort. While the combination of metrics greatly reduces the number of false negatives (68%) compared to using only the RCE metric, our quest remains to drive the false negatives to zero, ensuring airtight SoC reliability.  Our journey unveils the potential for AI to revolutionize the complex and crucial process of post-silicon validation, thus bringing fast and more reliable SoCs to market to meet the growing demands of digitalization of our world by CPS. Contact Info:  Kowshic Ahmed Akash (kowshicahmed.akash@nxp.com)  NXP Hamburg Prof. Dr.-Ing. Sohan Lal (Tel.: +49 40 42878 2037, sohan.lal@tuhh.de)  Massively Parallel Systems Group (E-EXK5) Prof. Dr.-Ing. Ulf Kulau (Tel.: +40 42878 2601, ulf.kulau@tuhh.de) Smart Sensors (E-EXK3) Hamburg University of Technology (TUHH)  Am Schwarzenberg-Campus 3, 21073 Hamburg

Correctly placing hydropower plants in a river is one of many examples where good knowledge of the bottom topography, also called bathymetry, is needed. While direct measurement of the bathymetry is possible, for example with a side scan sonar operated by a boat or an underwater remotely operated vehicle, this is very time consuming and expensive. Therefore, methods that can infer the bathymetry from the easier to measure surface height of the water are attractive. Mathematically speaking, this is an inverse problem where unknown parameters of a system are reconstructed from typically incomplete and noisy measurements of the system state. One approach to solve such inverse problems is so-called partial differential equation constrained optimisation, where system parameters are computed that reproduce the measurements but also satisfy physical constraints like mass or momentum conservation. Researchers from TUHH’s Institute of Mathematics (E-10) and Institute of Mechanics and Ocean Engineering (M-13) as well as from the the Department of Mathematics at the University of Hamburg (UHH) have published a joint paper that provides the first demonstration that this approach can reconstruct a real-world bathymetry. In their experiment, they placed a small hill, manufactured from skate board ramps, at the bottom of a 12 m long wave flume. The water at rest had a depth of 30cm and waves were being generated by a wave flap. Four sensors were installed that measure wave heights. This measured data was used to reconstruct the manufactured bathymetry by numerically solving a minimisation problem with the shallow water equations as constraints. The mathematical algorithm was implemented in Python using the Dedalus software. It could generate a qualitative reconstruction of the hill, even though the change in wave height caused by the bathymetry was only in the range of a few millimetres. Contact: Judith Angel judith.angel(a)tuhh.de Prof. Daniel Ruprecht ruprecht(a)tuhh.de Institute of Mathematics (E-10)

Der SFB lud vom 15. bis 17. Mai 2024 zum dritten internationalen Workshop zum Thema „Tailor-made Multiscale Materials Systems“ an der TUHH ein