arXiv subjects

Bryan T. Adey

Publications and source records attributed to Bryan T. Adey.

3 recordsLinked to original sources

arXiv2026

Assessing flood-conditioned power dependency of urban rail transit and its effects on resilience under climate change

Urban rail transit (URT) faces increasing flood risk while relying on traction power systems whose failures can trigger cascading disruption. Existing studies on URT flood resilience often oversimplify power dependency by overlooking traction power redundancy and rarely consider flood-conditioned cascading impacts, providing limited insight into how flood-induced traction power failures propagate through operations and affect system resilience. This study marks the first quantitative assessment of flood-conditioned power dependency of URT and its effects on system resilience. The methodology integrates a double-layer URT network model, spatial flood exposure assessment, service disruption simulation, a flood-conditioned power dependency index, and a journey-based performance metric that captures resilience properties of robustness and redundancy. It is applied to the London rail transit network under surface water flood scenarios across current and RCP8.5 climate conditions, two traction power feeding mechanisms, and four substation flood-failure thresholds. Results show that traction power redundancy fundamentally shapes the severity of cascading impacts. Under double-end feeding, power-attributable performance loss remains minimal, whereas single-end feeding leads to substantially greater loss and is highly sensitive to substation failure thresholds. The proposed dependency index is strongly correlated with dependency-related performance loss, supporting its use as a proxy for dependency-induced additional consequences in strategic planning.

physics.soc-ph↗

arXiv2019

Implicit supervision for fault detection and segmentation of emerging fault types with Deep Variational Autoencoders

Data-driven fault diagnostics of safety-critical systems often faces the challenge of a complete lack of labeled data associated with faulty system conditions (i.e., fault types) at training time. Since an unknown number and nature of fault types can arise during deployment, data-driven fault diagnostics in this scenario is an open-set learning problem. Most of the algorithms for open-set diagnostics are one-class classification and unsupervised algorithms that do not leverage all the available labeled and unlabeled data in the learning algorithm. As a result, their fault detection and segmentation performance (i.e., identifying and separating faults of different types) are sub-optimal. With this work, we propose training a variational autoencoder (VAE) with labeled and unlabeled samples while inducing implicit supervision on the latent representation of the healthy conditions. This, together with a modified sampling process of VAE, creates a compact and informative latent representation that allows good detection and segmentation of unseen fault types using existing one-class and clustering algorithms. We refer to the proposed methodology as "knowledge induced variational autoencoder with adaptive sampling" (KIL-AdaVAE). The fault detection and segmentation capabilities of the proposed methodology are demonstrated in a new simulated case study using the Advanced Geared Turbofan 30000 (AGTF30) dynamical model under real flight conditions. In an extensive comparison, we demonstrate that the proposed method outperforms other learning strategies (supervised learning, supervised learning with embedding and semi-supervised learning) and deep learning algorithms, yielding significant performance improvements on fault detection and fault segmentation.

cs.LG↗

arXiv2016

Generation of Spatially Embedded Random Networks to Model Complex Transportation Networks

Random networks are increasingly used to analyse complex transportation networks, such as airline routes, roads and rail networks. So far, this research has been focused on describing the properties of the networks with the help of random networks, often without considering their spatial properties. In this article, a methodology is proposed to create random networks conserving their spatial properties. The produced random networks are not intended to be an accurate model of the real-world network being investigated, but are to be used to gain insight into the functioning of the network taking into consideration its spatial properties, which has potential to be useful in many types of analysis, e.g. estimating the network related risk. The proposed methodology combines a spatial non-homogeneous point process for vertex creation, which accounts for the spatial distribution of vertices, considering clustering effects of the network and a hybrid connection model for the edge creation. To illustrate the ability of the proposed methodology to be used to gain insight into a real world network, it is used to estimate standard structural statistics for part of the Swiss road network, and these are then compared with the known values.

physics.soc-ph↗