Fraunhofer Austria

Abstract In 2021, economic losses due to supply chain disruptions in the Eurozone amounted to €112.7 billion Euro: Natural catastrophes (e.g., hurricanes, earthquakes, floods, and wildfires) multiply in the wake of climate change. The risks of pandemics, trade restrictions, cyber-attacks, and geopolitical conflicts are omnipresent. Hence, the threat of disruptive events to the European transport network grows ever greater. Additionally, GHG emissions can drastically increase by up to 50% if disruptions are not treated efficiently. Here, the project “Resilient multimodal freight transport network” (ReMuNet) comes into play: ReMuNet will enable the multimodal freight network to react and respond 20% more quickly to disruptive events and help to reduce European inland transport emissions on the main run by over 50% by 2026. As a trailblazer for the Physical Internet, the project pursues the vision to enable and incentivize synchro-modal relay-transport on European rail, road, and inland waterways to increase holistic network resilience. ReMuNet can identify and signal disruptive events and assess their impact on multimodal transport corridors. It reacts quickly to disruptive events in real time. It supports TMS providers to improve route planning resilience by communicating alternative, pre-defined, multimodal transport routes. ReMuNet orchestrates route utilization, suggests transshipment points, and optimizes capacity allocation, minimizing damage and shortening the recovery time. This paper introduces ReMuNet and its vision, objectives, and expected results.

Ensuring the correctness of annotations in training datasets is one way to increase the trustworthiness and reliability of Machine Learning.This study aims to detect semantic shifts in datasets using Feature-Based Out-Of-Distribution and outlier detection methods, assuming Out-Of-Distribution samples are far from In-Distribution data.The experiments began with distance-based methods, such as k-Nearest Neighbours and Mahalanobis, followed by feature pyramids and dimensionality reduction techniques to address high-dimensional challenges.The results showed that the k-Nearest Neighbours detector performed robustly, achieving 100% AUROC when using ResNet50 on the Caltech-101 dataset, while the Mahalanobis detector showed unstable results with scores close to 50%.Moreover, selecting the right backbone model and feature levels, particularly low-level features from ResNet50, improved performance achieving AUROC score of 96% on the DelftBikes dataset for both k-Nearest Neighbours and Local Outlier Factor.The study highlights that k-Nearest Neighbours, Local Outlier Factor, alongside feature pyramids and dimensionality reduction constitute an effective setup for Out-of-Distribution detection, but optimal performance depends on tailored configurations across varying data conditions.

Abstract Purpose: The aim of this paper is to present a computational method for identifying bottlenecks in the rail transportation network under the assumption that a modal shift from road to train should be emphasised. Methods: From a methodological point of view, the contribution is twofold: First, an estimation of the overall freight transport in Europe is done. Basic statics, as available from countries and the European Union, are utilised to estimate the actual freight transport demand between NUTS3 areas in Europe. Second, the (theoretically) needed train transports are computed by finding the shortest route between two NUTS3 regions. If the capacity of the train network is reached on a specific strech, alternatives are searched. Results: To assess, where bottlenecks are, we used the European rail network (from OpenStreetMap) as well as intermodal terminals (provided by a project partner) for our computations. The results show that our approach is working. Conclusion: The method presented within this paper is easy to use (only a few input data are needed) and fast (with respect to computational effort) such that it can be easily used for infrastructural planning tasks and assessing whether envisaged extensions of the rail network contribute to the higly desired modal shift.

The recent advances on utilizing Generative Artificial Intelligence (GenAI) and Knowledge Graphs (KG) enforce a significant paradigm shift in data-driven maintenance management. GenAI and semantic technologies enable comprehensive analysis and exploitation of textual data sets, such as tabular data in maintenance databases, maintenance and inspection reports, and especially machine documentation. Traditional approaches to maintenance planning and execution rely primarily on static, non-adaptive simulation models. These models have inherent limitations in accounting for dynamic environmental changes and effectively responding to unanticipated, ad hoc events. This paper introduces a maintenance chatbot that enhances planning and operations, offering empathetic support to technicians and engineers, boosting efficiency, decision-making, and on-the-job satisfaction. It optimizes shift scheduling and task allocation by considering technicians' skills, physical stress, and psychological state, thus reducing cognitive stress. The approach ultimately improves human performance and reliability, embodying a human-centricity in the domain of maintenance and health management. The practical impact of the maintenance chatbot is illustrated through its application in maintenance of railway cooling systems. The presented use case demonstrates the chatbot's potential as a transformative tool in maintenance management. Finally, the paper discusses the theoretical and practical considerations, in particular in the light of regulative frameworks such as EU AI ACT, highlighting the future pathways for complying with responsible AI requirements.

In the context of twin transition, companies encounter complex challenges. In the manufacturing sector, efficient utilisation of economic resources is crucial. Climate change and environmental pressures are driving companies to adopt Circular Economy (CE) business models that emphasise remanufacturing and product-service systems (PSS).However, specific guidelines for implementing these business models are lacking. In this study, we analyse CE business models based on remanufacturing and PSS to identify key elements needed to set up these specific models. We conduct a case study analysis and draft a comprehensive list of configuration options for both remanufacturing and PSS business models. Our results can assist manufacturing companies in reshaping their value proposition, delivery, creation, and capture toward a CE based on remanufacturing and PSS. The model integrates perspectives from providers, customers, and society. Numerical evaluations demonstrate that economic and environmental value creation can be simultaneously achieved and optimised through specific service arrangements.