Traffic flow prediction is central to intelligent transportation systems, yet existing spatial-temporal graph neural networks often assume static topology and stable traffic distributions, limiting their ability to handle evolving road networks. To this end, this study analyzes the spatial-temporal evolution characteristics of continuously evolving road networks using two California highway sub-networks, PeMS-D3 and PeMS-D4, over multi-year and multi-month periods. Through this analysis, we identify two patterns that motivate efficient continual learning within these datasets: spatial evolution is highly concentrated in a small fraction of nodes, and temporal series exhibit strong periodicity together with hierarchical information redundancy, with the first principal component contributing 77.3% of total variance on PeMS-D3 and 92.5% on PeMS-D4. Based on these insights, we propose the EVOLVE framework, which localizes evolving regions and constrains update scope through adaptive subgraph learning, reducing computational complexity; captures periodic patterns and compresses temporal features through dual-path temporal encoding, improving processing efficiency; and builds a parameter-pattern-data knowledge preservation system to mitigate catastrophic forgetting. Across the two evaluated benchmarks, EVOLVE achieves a 6.75% MAE improvement over the strongest baseline TEAM on PeMS-D3 and an 8.07% improvement on PeMS-D4, together with a 19 & times; training speedup and a 25 & times; GPU memory reduction relative to TEAM. These results support EVOLVE within the evaluated California highway networks, while generalization to non-highway and non-California networks remains an open direction for future work. The code of EVOLVE is available at github.com/OvOYu/EVOLVE.
With the transformation of power systems and the advancement of carbon neutrality goals, the capacity and access density of renewable energy-based distributed generation in active distribution networks are increasing. Although this improves integration and grid reliability, the stochastic nature of distributed generation increases uncertainties in network topology and operating states, leading to bidirectional power flows and load fluctuations. These uncertainties pose significant challenges to traditional diagnostic methods, which depend on deterministic parameters and fixed network topologies. Moreover, most data-driven models treat fault detection, classification, and localization as isolated tasks, overlooking their inherent physical correlations. In addition, they often fail to capture spatial topology of distribution networks. As a result, current methods struggle to handle highly variable grids with strong topological dependencies. To address these challenges, this paper proposes a multi-task fault diagnosis framework based on graph attention networks. First, the model extracts time-frequency features of fault signals using wavelet transform and Hilbert transform. Second, it employs graph attention networks to capture spatial propagation patterns of fault features along the network topology. Finally, multi-task learning is applied for joint optimization to simultaneously accomplish fault detection, classification, and localization. The proposed model reduces reliance on precise network topology and line parameters, while integrating graph neural networks with multi-task learning to capture spatial topological features and inherent physical correlations among diagnostic tasks. The effectiveness of the proposed model is validated through simulations under diverse operating conditions, showing that the model maintains high diagnostic accuracy across diverse operating conditions.
In this paper, we give a new proof of the Lemmens-Seidel conjecture on the maximum number of equiangular lines with a common angle $\arccos(1/5)$. This conjecture was previously resolved by Cao, Koolen, Lin, and Yu in 2022 through an analysis involving forbidden subgraphs for the smallest Seidel eigenvalue $-5$. Our new proof is based on bounds on eigenvalue multiplicities of graphs with degree no larger than $14$. To control the maximum degree of the graph associated with equiangular lines, we employ a recent inequality of Balla derived by matrix projection techniques. Our strategy also leads to a new proof for the classical result obtained by Lemmens and Seidel in 1973 for the case where the common angle is $\arccos(1/3)$.
Transcendental Liouvillian extensions are differential fields in which one can model polylogarithmic, hyperexponential, and trigonometric functions, logarithmic integrals, and their (nested) rational expressions. For such an extension, we construct a complement, over the subfield of constants, to the subspace of derivatives and develop an algorithm that decomposes any element of the field into the sum of a derivative and a component in this complement. Consequently, an element is a derivative if and only if its complementary component vanishes. Moreover, the algorithm enables us to determine elementary integrability over the extension by computing parametric logarithmic parts. It also yields a reduction-based approach to constructing telescopers for elements of the extension, provided that an a priori order bound is available. (c) 2026 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/).
Abstract Cement asphalt (CA) mortar, widely used as the filling layer in railway ballastless track systems, is frequently subjected to repeated temperature cycling during service. To explore the property evolution patterns of the dry and water-saturated CA mortar under temperature cycling (TC), the degradation behaviors of dynamic mechanical properties were investigated by temperature cycling tests within the range of − 40 ° C to 60°C, combined with dynamic temperature scanning, mercury intrusion porosimetry (MIP) for pore structure characterization, and viscoelastic constitutive modeling. Results showed that the moisture condition critically affects the effect of TC on CA mortar. The dry CA mortar following TC exhibited an increase in storage modulus accompanied by a decrease in loss factor, whereas the water-saturated mortar showed the opposite tendency. The asphalt-to-cement (A/C) ratio has a substantial impact on the dynamic mechanical properties. Higher A/C ratios accelerate the growth rate of storage modulus in dry specimens, whereas they slow the degradation of storage modulus in water-saturated specimens until the ratio exceeds 0.7, beyond which the decay accelerates. This contrasting behavior arises from different mechanisms. The increase in dried specimens is mainly attributed to intensified asphalt aging at higher A/C ratios. Specimens in a water-saturated state benefit from the toughening effect of asphalt and the cement’s inhibition of asphalt aging at lower A/C ratios, whereas excessive asphalt content ( A / C > 0.7 ) promotes the degree of aging, increases brittleness, aggravates freeze–thaw damage, and ultimately intensifies modulus deterioration. A viscoelastic constitutive model based on the fractional differential three-parameter solid model provides an accurate fit to the dynamic test curves under different TC conditions, with parameters E 0 and E 1 effectively characterizing the elastic behavior. This study provides valuable theoretical insight into the mechanical behavior evolution of CA mortar under extreme temperature and moisture conditions, and offers practical guidance for enhancing the long-term safety and durability of railway infrastructure in harsh climates, while also supporting the prediction of their long-term service life.
