• Physics-informed neural networks for elastoplasticity in strong and weak forms. • Strong form supports forward prediction and parameter inversion with data loss. • Data-free weak form captures plastic evolution via energy minimization. • Kolmogorov–Arnold network is compared with multilayer perceptron in both forms. Physics-informed neural networks have recently achieved remarkable success in solving elastic problems by embedding governing equations into the training of neural networks. Building upon these advances, this study extends physics-informed neural networks to material nonlinearity and develops two computational frameworks for small-strain von Mises elastoplasticity. The strong-form framework enforces governing equations through pointwise residual minimization, enabling unified forward–inverse modeling of field variables and unknown material parameters. In contrast, the weak-form framework derived from total potential energy minimization allows data-free learning of elastoplastic evolution through incremental loading, yielding stable and physically consistent predictions. The recently developed Kolmogorov–Arnold network is further incorporated and compared with the conventional multilayer perceptron. Results show that the Kolmogorov–Arnold network alleviates the gradient oscillation and convergence instability in the weak-form framework but performs less effectively in the strong form. Furthermore, validations against reference solutions from conventional numerical methods demonstrate the potential of the developed physics-informed neural network frameworks as mesh-free surrogates for elastoplastic analysis.

Speech perception relies on multiple acoustic cues whose relative weighting varies across languages. The present study examines how long-term language experience shapes cue weighting in second-language (L2) speech perception, refining an attentional-learning account of cross-linguistic transfer. Native speakers of English, Dutch, Spanish, Korean, and Mandarin completed a cue-weighting task targeting English lexical stress, in which vowel quality, pitch, and duration were orthogonally manipulated. Results revealed robust, dimension-specific differences across first-language (L1) groups that could not be explained solely by the presence or absence of lexical stress or lexical tone in the L1. Instead, cue weighting reflected how acoustic dimensions function within the L1 cue ecology, including their relative contribution to lexical distinctions and the stability and interpretability of these mappings across contexts. Cue redundancy constrained relative cue strength without eliminating attentional sensitivity to secondary dimensions. Machine-learning classification further showed that L1-linked attentional profiles were sufficiently structured to support prediction, even among L2 listeners with substantial English proficiency, demonstrating the persistence of L1-shaped attentional tuning. These findings support a view of cue weighting as reflecting durable, multidimensional attentional priors shaped by long-term experience and highlight the importance of L1 cue ecologies in understanding cross-linguistic transfer in speech perception.

Accurate and efficient rock mass quality classification is a prerequisite for assessing slope stability, designing support schemes, and ensuring mining safety in open-pit mines. However, traditional empirical classification methods rely heavily on expert judgment and often struggle to capture the complex, nonlinear relationships among factors influencing slope stability. Existing intelligent classification models also suffer from limitations, including sensitivity to incomplete data, insufficient feature interaction learning, and unstable performance on small-scale datasets. To address these issues, this study develops a deep forest (DeepForest) model optimized by three metaheuristic algorithms—brown bear optimizer (BBO), tuna swarm optimizer (TSO), and sparrow search algorithm (SSA)—to intelligently classify slope rock mass quality. A rock mass quality dataset containing 204 groups of slope and non-slope cases was established to train and evaluate the classification performance of the DeepForest models. Six influencing factors were set as input parameters: uniaxial compressive strength (UCS) of rock, rock quality designation (RQD), spacing of discontinuities (Sd), rock mass integrity coefficient (Kv), groundwater conditions (W), and site type (St). Multivariate imputation by chained equations (MICE), isolation forest (IsoForest), and synthetic minority over-sampling technique (SMOTE) were used to handle missing values, outliers, and imbalance in the dataset, respectively. The performance of the proposed models was evaluated using five metrics: accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC). The experimental results indicate that the BBO-DeepForest model performed best on the independent test set, with accuracy, precision, recall, F1-score, and average AUC values of 0.878, 0.682, 0.678, 0.678, and 0.961, respectively. A comparison with seven well-known imputation algorithms revealed the superiority of the selected imputation algorithm in recovering incomplete rock mass quality datasets. Model interpretation results showed that RQD and UCS are critical feature parameters for classifying slope rock mass quality. At last, the proposed BBO-DeepForest model was employed to verify the rock mass quality of three slopes at the Luming molybdenum mine, resulting in classifications consistent with on-site observations. It demonstrates that combining DeepForest with metaheuristic optimization algorithms is a feasible and accurate approach for intelligently classifying the rock mass quality of slopes.

Distance-decay models are fundamental to accessibility modeling; yet their alignment with actual travel behavior remains insufficiently examined in empirical terms. To help address this gap, we propose a dual-component Tanner–Gaussian decay model that seeks to mitigate two key limitations of traditional accessibility frameworks by simultaneously describing the phenomena of full decay and local peaks. The model is calibrated using survey data from an urban park in Hangzhou, China, and subsequently assessed on two additional datasets from Wuhan and Shanghai. Results indicate that: (1) traditional Gaussian functions may overestimate short-distance and underestimate long-distance accessibility, while Tanner functions tend to capture long-tail decay more effectively and appear more suitable for long-distance accessibility estimation; (2) the proposed dual-component model performs favorably in large samples, though its stability appears sensitive to sample size. By comparing the application of multiple accessibility models across three datasets, we highlight how their suitability varies depending on the specific context. This comparison may offer a reference that could assist designers in identifying underserved areas and supporting more equitable access to urban green spaces in the context of rapid urbanization.

• Fourteen studies show that fine phonetic detail is systematically regulated in shaping sound systems. • Fine phonetic detail links production, perception, and learning, sustaining contrasts and enabling sound change. • Sound systems emerge from controlled allocation of continuous phonetic parameters within prosodic structure. • Prosodic structure guides segmental and suprasegmental realization across languages and domains. • Phonetic grammar is a language-specific, system-internal control system shaped by motor, perceptual, and cognitive pressures. This special issue examines how fine phonetic detail participates in the shaping of sound systems. Across fourteen studies, the central theme is that subtle temporal, spectral, and articulatory patterns are not incidental by-products of articulation, but are systematically regulated aspects of speakers’ phonetic knowledge. They provide the means through which phonological contrasts and prosodic structure are realized, maintained, and sometimes reorganized. The contributions show how languages allocate continuous phonetic parameters—such as timing, coordination, voice quality, and nasality—within prosodic domains (e.g., phrases, words, and syllables) and under general biomechanical and communicative pressures. Studies of Irish, Hawaiian, Japanese, and Mandarin illustrate how prosodic structure guides segmental and suprasegmental realization. Work on English, German, Danish, and Cantonese demonstrates how fine phonetic detail underlies patterns of variation and creates potential pathways for change. Production connects naturally to perception and learning: findings from English accent adaptation and Samoan iterated learning reveal how listeners stabilize or reinterpret detail, linking individual processing to community-level patterning. A set of studies on Italian, Korean, English, and L2 German show how prominence reorganizes cues across articulation, interaction, and acquisition, shaping how speakers signal and listeners recover linguistic structure. These studies converge on a view in which fine phonetic detail arises from a central phonetic component (or the phonetic grammar) of linguistic structure—controlled by speakers, shaped by universal motor and perceptual constraints, and continually adjusted through perception and learning. In this perspective, sound systems emerge from the interplay of these regulated patterns, which sustain contrasts, support communication, and open principled routes for change.

This study investigates chloride ingress and the attainment of steel depassivation conditions in fully saturated concrete exposed to seawater under long-term temperature scenarios. A coupled reactive-transport model is implemented in Toughreact by integrating multi-ion diffusion, thermodynamic aqueous speciation/mineral equilibria, kinetic dissolution-precipitation of major hydrates, chloride binding, and porosity feedback on the effective diffusion coefficient D e . The model is benchmarked against published long-term submerged chloride profiles to ensure realistic coupled transport-reaction behavior. Results show that non-Fickian near-surface features (chloride “drop” and subsurface peaks) can persist even under permanently submerged conditions, driven by a thin altered surface layer with reduced transport porosity/D e and diminished binding capacity (reduced bound chloride). Parametric simulations over 5-40°C (up to 50 years) demonstrate that higher temperature accelerates mineral alteration, alkalinity loss, and the advance of depassivation indicators. A consistent comparison of three criteria (free chloride, total chloride, and [Cl - ]/[OH - ]) shows that they may diverge when alkalinity, binding, and porosity evolve; thus, [Cl - ]/[OH - ] provides a chemistry-consistent depassivation index. For a representative cover depth of d=50 mm, depassivation conditions are reached within decades under warm exposure.

As a crucial agricultural crop in China, litchi exhibits a biennial bearing pattern with alternating high-yield and low-yield cycles, known as on-year and off-year respectively. Research has identified unstable floral initiation as the primary cause of irregular fruiting in mid-to-late maturing cultivars. Rapid and accurate quantification of female to male flower ratios during the flowering phase enables targeted management strategies to optimize floral development and enhance fruit-setting rates. This study proposes Flower Quantification and Gender Recognition Network (FQGR-Net), a three-branch neural network architecture for simultaneous classification and counting of female and male flowers. Through module-level optimization, FQGR-Net improves both counting accuracy and computational efficiency, achieving average MAE of and RMSE of across categories in experiments conducted on the self-constructed dataset. Comparative experiments with other deep neural network models on public datasets show the proposed method achieves optimal performance. A regression analysis between predictions and ground truth produces values of and for female and male flower quantification respectively. A dedicated litchi flower phenotyping analyzer was developed to address the technological gap in automated floral census systems. Field trials demonstrated over accuracy in female/male flower counting.

The frequency data is generated through ANSYS modeling. The model is created through an eigenvalue analysis of a cantilever beam with different distributed damages.

This study investigates how second-language (L2) listeners from five first-language (L1) backgrounds—English, Dutch, Mandarin, Spanish, and Korean—perceive English lexical stress, focusing on their use of vowel quality, pitch, and duration cues. Participants completed a cue-weighting perception task (Tremblay et al., 2021) in which two acoustic dimensions were manipulated orthogonally while the third was neutralized. Data for Dutch listeners come from the original study. Predictions about cross-linguistic transfer were based on the functional weight of each cue in the L1. The following L1 effects were predicted: For vowel quality: English, Mandarin > Dutch > Spanish, Korean; for pitch: Mandarin > Korean > Dutch, Spanish > English; for duration: English, Mandarin > Dutch, Spanish > Korean. Bayesian mixed-effects models tested the effects of cues and L1 with L2 proficiency (Lemhöfer & Broersma, 2012) as a covariate. The results aligned broadly with our predictions: for vowel quality, English > Mandarin > Dutch > Korean > Spanish; for pitch: Mandarin > Korean, Dutch > Spanish > English; for duration: English, Mandarin, Dutch > Spanish > Korean. These findings support a cue-weighting typology shaped by L1-specific cue prominence, with implications for theories of transfer and perceptual learning in L2 acquisition.

This collection of articles aims to provide a pioneering introduction in this journal to the use of artificial intelligence in ground improvement research and applications, coupled with experimental testing and computational methods. Using optimisation algorithms, probabilistic modelling, and machine learning (ML), they advance prediction of settlement, strength, and failure behaviour across soil–cement systems, fibre-reinforced composites, jet-grouting, alkali-activated binders, and embankments.The paper by Nima et al. (2025) evaluates the performance of soil–cement (SC) columns in improving the seismic stability of soft soils through combined experimental and numerical analyses. The study demonstrates that SC columns substantially reduce settlement under both rigid and ductile overburdens and that a grey wolf optimisation framework, supported by artificial neural networks, can reliably identify optimal design parameters. While the proposed model provides a comprehensive tool for seismic soil improvement, further field validation is recommended to enhance its practical applicability.Collico et al. (2025) present a probabilistic Bayesian framework for predicting unconfined compressive strength (UCS) and diameter properties of jet-grouted columns during preliminary design. Drawing on a regional data set of soil and system parameters, the method combines local and most-similar site information to improve prediction accuracy under data-limited conditions. The approach offers a more reliable alternative to conventional empirical or theoretical correlations, supporting cost-effective and informed decision-making in early project phases, although further data enrichment is needed to enhance prediction robustness.Martins et al. (2025) apply a novel design of experiments (DOE) methodology to evaluate fibre-reinforced cement-stabilised soils, enabling prediction of both strength and failure mode while quantifying the influence of key parameters. Results show that binder content is the dominant factor governing UCS, while polypropylene fibres outperform sisal in improving ductility. The brittleness index was validated as a reliable criterion for distinguishing failure modes, and the DOE model demonstrated high accuracy with reduced testing effort, offering a robust framework for predictive analysis, though further validation across wider soil, binder, and fibre conditions is recommended.Tinoco et al. (2025) investigate the use of ML to predict the UCS of soils stabilised with one-part alkali-activated binders, offering a sustainable alternative to Portland cement. Despite a limited data set, random forest, neural networks, and support vector machines achieved high predictive accuracy, with water and soil content identified as the most influential parameters. The study demonstrates the potential of ML as a pre-design tool to optimise soil stabilisation, reduce reliance on laboratory testing, and support sustainable construction, while underscoring the need for broader data sets and further model validation.Jones et al. (2025) apply Bayesian updating to assess embankment performance on soft ground, comparing models with different soil layering and random variables. Results show that while both four- and nine-layer models can reproduce monitoring data, the nine-layer model yields more realistic posterior parameters, especially when only surface settlement data are available. Incorporating magnetic extensometer and piezometer data significantly improves prediction quality, underscoring the importance of high-quality field monitoring, while highlighting that computational efficiency depends more on hardware and software optimisation than on model simplification.These contributions reveal an emerging direction in ground improvement, where artificial intelligence takes a central role, enhanced by experimental studies and computational modelling. Together, they point to smarter, data-driven design tools that enhance sustainability and bridge laboratory insights with field performance, although broader data sets and large-scale validation are still needed.