Intelligent Techniques Applied in Infrastructure, Engineering, and Construction

1. Introduction

The accelerating convergence of artificial intelligence (AI) and smart construction [1] technologies is transforming every phase of infrastructure delivery—from concept [2] and design [3] to construction [4], operation, and renewal. In recent years, breakthroughs in predictive analytics [5], computer vision [6], and real-time sensing have begun to alter long-standing engineering workflows, enabling decisions [7,8] that are simultaneously faster, more accurate, and more sustainability-oriented than traditional heuristics permit [9,10,11,12]. Motivated by this profound shift, the present Special Issue of Buildings assembles 23 original papers that collectively illustrate how data-driven intelligence can unlock unprecedented levels of efficiency, resilience, and environmental stewardship across civil engineering domains. The contributions cover a range of topics, including underground construction, bridge and tunnel engineering, railway systems, concrete technology, cost optimization, risk governance, and Fourth Industrial Revolution (4IR) adoption, reflecting the wide thematic net cast by the guest editors. Beyond merely cataloging applications, the papers reveal emerging best practices for integrating domain knowledge with advanced algorithms, establishing robust data pipelines, and validating digital predictions against field measurements. Collectively, they confirm that intelligent techniques are no longer experimental add-ons but essential, increasingly standardized tools for twenty-first-century infrastructure planning and management.

2. Underground Construction and Geo-Infrastructure Intelligence

Snap-fit optimization for folding steel arches (Contribution 1) and stability evaluation of layered multi-stage fill slopes (Contribution 2) provide complementary advances in data-calibrated numerical modeling and field verification. Contribution 3 compares pile–anchor and double-row-pile supports for foundation pits adjacent to metro tunnels, while Contribution 4 integrates vehicle–track coupling dynamics with void detection in ballastless CA-mortar, illustrating how physics-based simulations can be fused with vibration signatures for rapid condition assessment. For TBM tunneling, Contribution 19 develops a boreability-aware rock-mass classification using ensemble learning. Contribution 14 predicts deformation when new tunnels overcross existing lines, and Contribution 15 introduces a self-developed drilling-test system linked to random-forest models for real-time rock-grade identification. These studies show that intelligent analytics can reduce uncertainty in subsurface construction, supporting safer and greener underground networks.

3. Smart Structures, Materials, and Performance Prediction

High-temperature debonding of FRP-strengthened beams is modeled analytically (Contribution 5), while Contribution 6 quantifies the dual effects of traffic-induced vibration on high-strength concrete curing and old-to-new interfaces—information critical to bridge widening without traffic closure. Contribution 7 uses refined finite elements to capture nonlinear stresses in spherical hinges during swivel-bridge construction, and Contribution 8 couples multi-field simulations with on-site crack-control trials for pier concrete in high-altitude climates. Contribution 9 enhances thermal-parameter inversion for mass concrete via a mixed-strategy sparrow-search algorithm, exemplifying how hybrid optimization accelerates calibration against sensor data. Contribution 10 blends laboratory and ABAQUS analysis to clarify load transfer in stepped DX piles, whereas Contribution 18 extends the classical method to mechanically connected precast piles under horizontal load. Contributions 21 and 22 explore seismic behavior of precast beam–column joints and fatigue of ultra-short studs in UHPC, both supported by detailed numerical or S–N modeling. Together, these 10 papers affirm that ML-enhanced analytics, surrogate modeling, and digital twins are maturing as reliable instruments for performance-critical decision-making.

4. Cost, Optimization, and Lifecycle Sustainability

Contribution 13 presents an adaptive self-explanatory CNN that predicts the cost of Huizhou replica vernacular dwellings to within 0.6% MAPE and makes feature importance fully transparent via SHAP values—evidence that trustworthy AI can penetrate cost-management workflows. Contribution 23 extends optimization to steel-plate fitting, reducing fabrication errors by up to 75% through a minimum-error interpolation method and thereby lowering both cost and embodied carbon.

From a materials perspective, Contribution 20 quantifies how wet–dry cycles alter shear strength and cohesion in unsaturated clayey sands, data essential for moisture-resilient roadway subgrades. Contributions 17 and 8 address plateau concrete cracking under harsh environmental gradients, offering deformation-compensated mix designs that double long-term compressive reserves. These results emphasize that intelligent techniques are key enablers of whole-life-cycle sustainability—from material selection to asset maintenance.

5. Risk Assessment, Digitalization, and 4IR Adoption

Contribution 12 couples triangular-fuzzy theory with Bayesian networks to rank construction-stage risks in cantilever-casting arch bridges, whereas Contribution 11 fuses UAV panoramas with a Swin Transformer–enhanced YOLOv8 to achieve 98.7% accuracy in bridge defect localization. Contribution 16 provides rare survey data on 4IR technology uptake in an emerging-economy construction sector, revealing virtualization as a gateway to broader AI adoption. Contributions 9 and 19 independently confirm that hybrid or ensemble algorithms outperform single-heuristic approaches for parameter inversion and classification, reinforcing the risk-management benefits of algorithmic diversity.

6. Outlook

Taken together, the articles in this Special Issue crystallize three interrelated trajectories. First, data-informed simulation loops are becoming the normative backbone of design: sensor-calibrated finite-element models, surrogate ML surrogates, and hybrid optimization routines now shorten iteration cycles while providing confidence bounds that were previously unattainable. Second, the rise of explainable intelligence is dismantling the “black-box” barrier; methods such as SHAP analyses, Bayesian inferences, and transformer-based visualizations are empowering practitioners to interrogate algorithmic reasoning, thereby enhancing adoption in safety-critical contexts. Third, an explicit focus on lifecycle resilience is redirecting innovation from short-term performance gains toward long-term durability and carbon reduction, as illustrated by studies on plateau-concrete cracking, UHPC fatigue, and AI-guided maintenance scheduling.

Looking ahead, further progress will depend on tighter coupling between domain-specific physics and general-purpose AI architectures, the establishment of standardized and interoperable data schemas that facilitate cross-project learning, and the development of governance frameworks that embed ethical considerations such as fairness, transparency, and data sovereignty. Equally important will be interdisciplinary collaboration—bringing together civil engineers, computer scientists, materials specialists, and policymakers—to translate laboratory advances into deployable tools that can scale across diverse regulatory and climatic contexts. With these foundations in place, intelligent techniques will not simply augment conventional practice but will redefine the baseline expectations for how infrastructure is conceived, built, and sustained.