S. Bosse, Learning Damage Event Discriminator Functions with Distributed Multi-instance RNN/LSTM Machine Learning - Mastering the Challenge, Proc. of the 5th International Conference on System-Integrated Intelligence Conference, 11.11-13.11.2020,
Bremen, Germany, 2020 Paper PDFPresentation HTMLPresentation Video
Common Structural Health Monitoring systems are used to detect past damages occurred in structures with sensor networks and external sensor data processing. The time of the damage creation event is commonly unknown. Numerical methods and Machine Learning are used to extract relevant damage information from sensor signals that is characterised by a high data volume and dimension. In this work, distributed multi-instance learning applied to raw time-series of sensor data is deployed to predict the event of the occurrence of a hidden damage in a mechanical structure using typical vibrations of the structure. The sensor processing and learning is performed locally on sensor node level with a global fusion of prediction results to estimate the damage location and the time of the damage creation. Recurrent neural networks with a long-short-term memory architecture are considered implementing a damage discriminator function. The sensor data required for the evaluation of the proposed approach is generated by a multi-body physics simulation approximating material properties.
[c20.2]
S. Bosse, Self-organising Urban Traffic control on micro-level using Reinforcement Learning and Agent-based Modelling, Proc. of the SAI IntelliSys Conference, 3-4.9.2020,
Amsterdam, Netherlands, 2020 Paper PDF
[c20.3]
S. Bosse, Self-adaptive Traffic and Logistics Flow Control using Learning Agents and Ubiquitous Sensors, Proc. of the 5th International Conference on System-Integrated Intelligence Conference, 11.11-13.11.2020,
Bremen, Germany, 2020 Paper PDFPresentation HTMLPresentation Video
Traffic flow optimisation is a distributed complex problem. Traditional traffic and logistics flow control algorithms operate on a system level and address mostly switching cycle adaptation of traffic signals and lights. This work addresses traffic flow optimisation by self-adaptive micro-level control by combining Reinforcement Learning and rule-based agent models for action selection with a new hybrid agent architecture. I.e., long-range routing is performed by agents that adapt their decision making for re-routing on local environmental sensors. Agent-based modelling and simulation are used to study emergence effects on urban city traffic flows with learning agents. The approach and the proposed agent architecture can be generalised and applied to a broader range of application fields, e.g., logistics and general transport phenomena.
[c20.4]
S. Bosse, E. Kalwait, Fatigue and damage diagnostics with predictor functions for new advanced materials by Machine Learning, MAPEX SpaceMat 2020 Symposium 31.8 -1.9.2020,
Bremen, Germany Paper PDFPoster PDF
There is an emerging field of new materials highly related to space applications like fibre-metal laminates (DFG FOR3022). Typically, material properties are determined from tensile tests. We investigate approximating predictor functions by Machine Learning (ML) for inelastic and fatigue prediction by history data measured from simple tensile tests within the elastic range of the material. We show some preliminary results from a broad range of materials and outline the challenges to derive such predictor functions by ML.
[c20.5]
S. Bosse, E. Kalwait, Damage and Material-state Diagnostics with Predictor Functions using Data Series Prediction and Artificial Neural Networks, ECSA 2020 MDPI, 15.11 -30.11.2020,
Basel, Switzerland Paper PDFPresentation HTMLPresentation Video
There is an emerging field of new materials highly related to space applications like fibre-metal laminates (DFG FOR3022). Typically, material properties are determined from tensile tests. We investigate approximating predictor functions by Machine Learning (ML) for inelastic and fatigue prediction by history data measured from simple tensile tests within the elastic range of the material. We show some preliminary results from a broad range of materials and outline the challenges to derive such predictor functions by ML.
[p20.1]
U. Engel, S. Bosse, et al., Blick in die Zukunft: Wie künstliche Intelligenz das Leben verändern wird Ergebnisse eines Umfrageprojekts in der Wissenschaft, Politik und Bevölkerung der Freien Hansestadt Bremen, Universität Bremen, Methodenzentrum,
Februar 2020, Bericht Report PDF
Die Freie Hansestadt Bremen ist ein bedeutender Wissenschafts- und Wirtschaftsstandort in Deutschland und für ihre Bürgerinnen und Bürger eine ebenso bedeutende Stadt zum Leben. Wir möchten einen Blick in die Zukunft wagen und fragen, wie sich diese Zukunft aus Sicht der in Bremen beheimateten Wissenschaft und Politik und der Bremer Bevölkerung darstellt. Auch bitten wir die in Bremen ansässigen Medien um Mitwirkung an dieser Studie.
[p20.2]
S. Bosse, Anwendung von Maschinellen Lernalgorithmen auf SHM/NDT Daten. Eingeladener Überblicksvortrag mit praktischen Erfahrungen und Anwendung im Rahmen der DFG Forschungsgruppe 3022, Deutsche Gesellschaft für zerstörungsfreie Prüfung,
Fachausschusssitzung SHM, Dresden, 30.9.2020 Script PDFPresentation HTML
In diesem Vortrag werden die Möglichkeiten von verteilten Multiinstanzlernen für SHM Anwendungen an Beispielen der Dehnungs- und Ultraschallsensorik gezeigt und analysiert. Dabei liegt ein Fokus auf den Einsatz auf Systemen mit eingeschränkten Ressourcen und Robustheit. Dabei werden neben Neuronalen Netzen und Deep Learning auch Entscheidungsbaumlerner diskutiert.
Interdisciplinary Science & Education Laboratory
