Expert annotation is costly for agricultural text-pair matching, whereas unlabeled pairs are more readily available. We investigate a two-stage semi-supervised model, Semi-TM, that combines a bidirectional long short-term memory (BiLSTM) encoder based on the enhanced sequential inference model (ESIM) with supervised contrastive learning and a masked-view objective. We evaluate a Chinese sentence-pair dataset containing agricultural and general-domain questions and three public benchmarks. On the experimental dataset, at a labeled-data proportion of 30%, Semi-TM achieves a macro-F1 of 0.8943±0.0034 across five runs, compared with 0.8625±0.0025 for the cross-entropy baseline. In the public-benchmark evaluations, its macro-F1 is 0.8731±0.0030 on AFQMC, 0.8991±0.0045 on LCQMC, and 0.8447±0.0034 on PAWS-X under the dataset splits and labeled subset sizes used in this study. Ablations examine supervised contrastive learning, unlabeled examples, masking, and training order. A per-pair analysis further characterizes performance across sentence lengths and identifies representative agricultural matching errors. The evaluated model has fewer parameters and a smaller storage footprint than ESIM but has higher single-pair CPU latency.
Bridge-type compliant mechanisms (BCMs) are widely utilized in micro- and nano-positioning systems for motion amplification. Balancing performance across amplification capability, stiffness, stress, and structural compactness for a BCM is challenging because of the coupled and competing effects among these performance metrics. This paper presents an integrated framework that combines surrogate modeling, multi-objective optimization, and multi-criteria decision-making for the application-specific optimization of BCMs. Geometric parameterization of the BCM is established to enable fast, automated finite element analysis. Optimal Latin hypercube sampling (OLHS) is used to generate the BCM sample set, whose pairwise correlations are assessed through correlation analysis. Based on the sample responses, the surrogate model with the highest predictive accuracy is selected from RSM, RBF, SVR, and Kriging. Four multi-objective optimization algorithms, MOPSO, SPEA2, NSGA-II, and NSGA-III, are compared in terms of convergence, solution distribution, and coverage to select the algorithm for subsequent optimization. The resulting Pareto solutions are ranked using AHP–EWM combined with TOPSIS. The framework is applied to the series design of three BCMs intended to provide graded output stiffnesses. The optimized designs are verified through finite element analysis and experimental study. The results demonstrate that the framework balances competing performance requirements and facilitates application-oriented series design of BCMs.
In this article, we develop a tractable mixed-integer linear programming framework for the annual scheduling of a hybrid offshore wind and pumped hydroelectric energy storage system. The proposed formulation uses McCormick envelopes to relax the non-convex hydraulic power relations while embedding generation dispatch within a direct current optimal power flow model. In particular, we incorporate a cumulative storage incentive into the objective function to capture the trade-off between net electricity-market revenue and intra-horizon water availability, examining its weighting multiplier through a systematic sensitivity analysis. We illustrate the performance of the proposed formulation applied to (i) a modified PJM five-bus benchmark and (ii) an IEEE 30-bus network hosting two hybrid installations, driven by historical wind profiles from Tamaulipas and locational marginal prices from the Mexican wholesale electricity market. Moreover, we assess the selected binary operating-mode sequences ex post by re-optimizing the continuous dispatch under the exact bilinear hydraulic equations. The numerical results demonstrate that coordinated dispatch enhances energy arbitrage and dynamic grid interaction. In the five-bus case, the reference cumulative storage incentive raises the average upper-reservoir elevation relative to scheduling without the incentive, but both schedules end at the minimum operating level. Thus, the incentive improves water availability within the annual horizon without preventing terminal depletion or guaranteeing long-term water reserves.
Supplier selection is a critical strategic decision in supply chain management, with a direct impact on operational performance, competitiveness, and long-term sustainability. The contemporary business environment, characterised by globalisation, climate change, and increasingly stringent regulatory requirements, necessitates an expanded supplier evaluation framework that encompasses the dimensions of resilience and sustainability. Industry 5.0 and digital twins (DT) are advanced concepts that enable improvements in the supplier selection process through the integration of advanced technologies and human knowledge. Using a five-member expert panel from different departments, both criterion importance and the qualitative supplier ratings were assessed individually by each expert; quantitative criteria were derived from a preliminary DT prototype running documented scenario simulations, in the case of an agricultural company operating in south-eastern Bosnia and Herzegovina. Interval type-2 fuzzy sets (IT2FS) were applied, with asymmetric fuzzy numbers used in the membership function. The results obtained using IT2FS Simple Weight Calculation (SiWeC) showed that total costs, recovery capability following disruptive events, and product and service quality were the most important criteria. Qualitative supplier ratings were aggregated by unweighted arithmetic mean, while criterion importance was determined through IT2FS SiWeC. The six selected suppliers were ranked using IT2FS Multi-Attributive Border Approximation Area Comparison (MABAC), which demonstrated that supplier SUPP1 achieved the best results according to the expert assessments and the DT-based scenario simulations. These findings were additionally confirmed through comparative and sensitivity analyses. This study contributes practical guidance regarding the criteria and suppliers that agricultural enterprises should select in order to improve the procurement process and reduce supply chain risks. This research presents a framework for supplier selection based on aggregated DT output data. The DT prototype used in this study is a preliminary, uncalibrated simulation system whose outputs are treated as simulation-based estimates rather than validated real-time measurements. This limitation is stated explicitly, and the calibration of the DT against actual supplier performance is identified as a direction for future research. This framework can be extended to real-time dynamic updating, which is a direction for future research.
Fixed-grid topology optimization is widely used for auxetic metamaterial design, but its discretization-dependent boundaries may require substantial refinement or post-processing, potentially altering slender ligaments and the resulting deformation mechanisms. This study develops an adaptive body-fitted bi-directional evolutionary structural optimization (BESO) framework for two-dimensional periodic auxetic metamaterials. The topology is persistently represented by an implicit field on a fixed background grid, while a geometry-conforming T6 finite-element mesh is regenerated for periodic homogenization at each iteration. Mutual-energy sensitivities are normalized by element area and transferred from the nonuniform analysis mesh back to the background design field, enabling stable evolutionary updates under repeated remeshing. The adaptive workflow reconstructs the boundary within the optimization loop, avoiding a separate post-optimization B-spline smoothing step; in the fixed-grid Q4 comparison, this smoothing reduces the magnitude of the Poisson’s ratio. The mesh-parameter study shows how the adaptive mesh-size ratio affects computational time and the optimized auxetic response. In the tested comparison, the adaptive workflow shows closer agreement between the optimized and independently verified Poisson’s ratios.
