Communication Material
Scientific Publications
Passenger flow analysis of a future public transport hub
Public transport (PT) is important for addressing future mobility needs of a country with a growing population and limited space. To be competitive with private cars, travel with PT must be fast and comfortable. For direct connections, PT is mostly competitive with private cars regarding travel time. However, when transfers between different lines and different modes of PT are involved, the picture changes. Making these transfers as short, smooth, safe and easy as possible is one option to improve PT. Public transport hubs, connecting different PT modes, can help to achieve these smooth transfers. Thus, redesigning train stations into PT hubs is one possibility to make PT more attractive. It is important to verify beforehand, that transfers in the future hub work smoothly, which can be done by passenger flow analysis, identifying possible limitations or obstacles. This paper presents a methodology for passenger flow analysis of future PT hubs. This methodology consists of forecasting travel demand, estimating transfer counts, building a microscopic model to simulate passenger flows, assessing the simulation results and calculating transfer time matrices. The suitability of the methodology is demonstrated by applying it to the ongoing process of converting a train station in Winterthur, Switzerland into a PT hub.
AutoML-Optimized Vision Transformers for Multimodal Structural Health Monitoring from Multi-Channel Feature Tensors
Reliable structural health monitoring (SHM) of wind turbines requires methods capable of handling high-dimensional, heterogeneous multi-sensor data under both labeled and label-scarce conditions, while remaining computationally efficient for real-world deployment. This paper presents a unified SHM framework that integrates feature-level data fusion, Vision Transformer (ViT) classification, multi-objective AutoML optimization, and unsupervised anomaly detection within a single pipeline. Multi-channel signals from 26 sensors are preprocessed and encoded into a compact three-channel feature tensor combining statistical descriptors, spectral and wavelet features, and a PCA-denoised inter-sensor correlation matrix. This representation enables efficient storage of time-window information and captures both local signal characteristics and global cross-sensor dependencies. A lightweight ViT is trained for supervised fault classification, while its architecture is optimized using NSGA-II to jointly maximize predictive performance and minimize computational cost, enabling deployment on resource-constrained edge devices. To address label scarcity, an autoencoder trained solely on normal-condition data is used for anomaly detection. Comparative evaluation of reconstruction- and latent- space-based metrics shows that Mahalanobis distance in the latent space provides superior sensitivity to subtle faults. Validation on the ETH Aventa AV-7 dataset demonstrates up to 98.4% macro-F1 in classification and robust anomaly detection performance. The results confirm that the proposed multisensor fusion strategy provides a reliable and scalable pipeline for SHM of real structures, applicable to both supervised and unsupervised scenarios under practical computational constraints.
Direction-Aware Physics-Informed Autoencoder for Damage-Insensitive Synchronization of Bridge Acceleration Signals
Accurate time synchronization is essential for vibration-based structural health monitoring (SHM), particularly when acceleration responses are collected using distributed wireless sensors. Residual timing errors may remain after clock-level synchronization due to clock drift, triggering jitter, communication latency, and hardware-dependent acquisition delays, introducing phase distortion and reducing the reliability of modal identification and damage-sensitive analysis. This paper proposes a post-processing synchronization framework for multi-sensor triaxial bridge acceleration signals under ambient vibration. The method combines learned waveform representations, graph- based sensor interaction, classical delay-estimation evidence, and physics-informed consistency constraints. A shared convolutional encoder extracts sensor-level latent features, while graph message passing models inter-sensor dependencies over a complete sensor graph. In parallel, correlation- and phase-based edge features provide physically meaningful delay cues for each sensor pair. Raw pairwise delay estimates are projected into cycle-consistent sensor offsets using a graph-based least-squares formulation, and the predicted offsets are used to compensate the measured signals. A two-stage inference strategy is further introduced, where the model first estimates delays from the original asynchronous signals, synchronizes the signals using the extracted offsets, and then performs a second inference step to refine the final estimates. The model is trained using healthy and mass-varied finite element simulations with artificial timing offsets, while damaged scenarios are reserved for testing to evaluate generalization to unseen structural states. The results show that the proposed two- stage framework provides more accurate and robust synchronization than direct inference and the classical GCC-PHAT baseline, supporting its use as a damage-insensitive signal-level synchronization approach for wireless bridge SHM.
CEMBOX: Gemelos Digitales aplicados a la salud estructural y el mantenimiento de infraestructuras
El creciente deterioro de la red europea de infraestructuras requiere de sistemas innovadores para mejorar la eficiencia, calidad y precisión en procesos de inspección, mantenimiento y gestión. CEMBOX, plataforma gemelo digital de infraestructuras, orquesta entornos BIM-GIS para visualizar activos, predecir estados de condición y simular escenarios de cara a la toma de decisiones. La plataforma incluye diversos verticales especializados en activos críticos, y centrados en una arquitectura ICT distribuida basada en microservicios que emplea información histórica y de sensores IoT (p. ej. acelerómetros, clinómetros), técnicas de monitorización de la salud estructural, modelos de Machine Learning, y sistemas de reporte y alarma. CEMBOX presenta aplicaciones en activos reales como el viaducto Huertas de Mateo (LAV Madrid-Valencia), el túnel 40 en la línea ferroviaria Palencia-Coruña, un talud en Briones (línea Castejón-Bilbao) o la esclusa de Champagne en el río Sena, evidenciando su capacidad para priorizar actuaciones, optimizar OPEX y reducir riesgos
PrONe-RS: A Graph-Isomorphic Pruning with Pareto-Driven Optimization Framework for Network Sizing in Remote Sensing LULC Applications
Deep Learning (DL) models have demonstrated remarkable performance in Remote Sensing (RS) land use land cover (LULC) classification. Yet, their high computational complexity demands can limit their deployment in resource-constrained edge computing environments. Existing structured pruning approaches are often weight-dependent and lack a principled mechanism to jointly coordinate the pruning process and select the optimal network sparsity ratio. To address these limitations, this work introduces PrONe-RS, a graph-isomorphic pruning coupled with Pareto-driven optimization framework for efficient neural network sizing in RS LULC applications. Through the incorporation of Isomorphic Structured Pruning (ISP), the network's structure is modeled as a graph to identify and create isomorphic groups of structurally similar computational blocks, which are consequently pruned according to a user-defined ratio. With this principally coordinated pruning process across layers, ISP preserves the network's structural coherence making it attractive for DL models utilizing diverse computing blocks. PrONe-RS is enhanced with a Pareto-driven multi-objective optimization scheme that first identifies the optimal pruning ratios achieving the best classification performance and complexity trade-off, and then automatically selects the most efficient one through a dedicated metric, thereby guiding the pruning ratio selection process. Extensive ablation study on the effective pruning values and the selection of the optimal one balancing the performance-complexity trade-off using five widely-used DL models including EfficientNet, ResNet, ResNeXt, VGG and Vision Transformer variations across two datasets, EuroSAT-Multispectral and AID, demonstrate significant reduction in number of trainable parameters, corresponding to 64−98% for the EuroSAT-Multispectral case and 44−70% for the AID case, all while preserving >91% classification performance across all metrics. Comparisons with existing pruning approaches on AID dataset and EuroSAT-RGB, demonstrate improved classification accuracy along with a substantial reduction in parameters, achieving a 91% reduction for EuroSAT-RGB, 36% for ResNet, and 75% for VGG on AID, respectively.
Public Deliverables
Coming soon
You didn‘t find what you are looking for or have any questions?
Please, contact our project or communication managers
External Advisory Board Documents
Restricted access
At the bottom of this page you can find the documents of the MOVEO External Advisory Board.
To access please use the password provided in the email by Jesús Rodríguez.
This content is password protected, enter yours below.