Increasing global energy demand and depleting non-renewable resources drive the need for sustainable energy storage. Cement-based energy storage materials that combine structural load bearing and energy storage are well-received due to their low cost and scalability. Cement-based supercapacitors utilize the cement matrix to form electrical double layers as well as enable underpotential deposition (UPD) and highly reversible sorption and redox processes. Though, hardened cement paste has insulating properties, while the continual change in its pore structure is at different concentrations of ions, significantly limiting ions' transport and storage capability. Adding conductive materials to create coupled charge carriers electron and ions can be a solution. The multi-scale complexity of these composites inhibits the ability of classical empirical methods to disentangle the interaction effects among competing effects. In this study, we developed a machine learning framework that could identify hierarchical relationships among structural descriptors, polarization losses and rate conditions, to enable the multi-objective optimization and inverse design of cement-based supercapacitors.

Sixty groups of cement-based positive electrode samples with single and composite conductive phase combinations were fabricated. Raw materials included ordinary Portland cement, three conductive carbons materials (i.e., carbon black (CB), carbon fiber (CF) and carbon nanotube (CNT)), electrolytic α-MnO₂ as an active material, and polycarboxylate superplasticizer for fluidity enhancement. Electrochemical tests employed a two-electrode system in a 1 mol/L ammonium acetate electrolyte at two charge-discharge rates (C-rate), which were 1.0 C and 1.5 C. The internal resistance drop (V_IR). quantifying polarization loss, was defined as a difference between the 1.8 V cut-off and the discharge starting voltage. Electrochemical impedance spectroscopy (EIS) was carried out (i.e., 0.1-10000.0 Hz, 10 mV amplitude) to extract the ohmic resistance (R_s) and charge-transfer resistance (R_ct) of representative samples. A database of 280 valid dates was constructed to predict energy density. In terms of feature engineering, the raw composition descriptors were supplemented by the total carbon mass fraction [w(C_total)], the relative ratios of w(CB), w(CNT) and w(CF), the active-to-conductive ratio w(MnO₂)/w(C_total), and conductive network index (i.e., N_et = a·CNT_ratio + b·CF_ratio + c·CB_ratio). Interaction terms V_IR × N_et, and V_IR × C-rate (i.e., V_IR × 1.0 C and V_IR × 1.5 C) were introduced, and the C-rate was treated as a one-hot categorical variable to represent rate-polarization coupling. Three tree-based regressors, i.e., Random Forest (RF), eXtreme Gradient Boosting (XGBoost) and Extremely Randomized Trees (ExtraTrees), were trained under identical preprocessing using a 5-fold GroupKFold cross-validation scheme based on formulation and voltage descriptors to avoid information leakage. Model accuracy was assessed by the coefficient of determination (R²), the root-mean-square error (RMSE) and the mean absolute error (MAE). The best-performing model was retained for an Optuna-based outer search of the network-index weights under the ordered constraint a≥b≥c≥0, followed by hyper-parameter tuning. Permutation importance and SHapley Additive exPlanations (SHAP) were then applied to interpret the optimized model.

Among the three baseline models, ExtraTrees show the optimum generalization, with an out-of-fold (OOF) R² of 0.7724 and an OOF RMSE of 0.0939. Replacing the empirically assigned weights of the network index with data-driven values increases the OOF R² to 0.8240, and subsequent hyper-parameter tuning gives a final OOF R² of 0.8354 with an RMSE of 0.0793 mW·h·cm^-2. The optimized weights follow an order of CNT (2.084) > CF (1.335) > CB (0.006), indicating that high-aspect-ratio fillers dominate the construction of continuous electron-transport pathways, whereas a particulate CB mainly improves local interfacial contact and contributes little to overall network continuity. This ranking is supported by EIS results on samples with a fixed total carbon content of 10%, in which substituting CNT for CB reduces R_s from 47.90 Ω to 32.68 Ω and R_ct from 45.35 Ω to 28.76 Ω. The SHAP analysis identifies the V_IR as the most influential predictor of energy density and shows that its negative contribution becomes dominant at > 0.5 V. The network index acts as the main positive regulator, and shifts from a restrictive to a reinforcing role once it exceeds 0.5, corresponding to the formation of an effective percolation pathway. The rate-polarization interaction terms give stronger negative SHAP contributions at 1.5 C rather than at 1.0 C, indicating that higher C-rates amplify ohmic losses and narrow the range over that the structural advantage can be exploited. These results indicate a coupled structure-polarization-rate regulation mechanism, in which the energy density is not determined by any single descriptor, but by the matching among conductive network construction, polarization control and rate condition.

A machine learning framework, incorporating an optimizable conductive network index and rate-polarization interaction features, was developed for cement-based supercapacitors. ExtraTrees outperformed RF and XGBoost, and after weight optimization and hyper-parameter tuning, gave a cross-validated R² of 0.8354 for energy density. Data-driven weight learning quantified the relative contributions of the three conductive fillers as CNT > CF > CB, which was consistent with the EIS results. The SHAP interpretation showed that the IR drop could be the primary controlling variable, the network index could be the main positive regulator, and the rate-polarization interaction could be a rate-amplified negative modulator. The energy density of cement-based supercapacitors thus depended on the matching among conductive network construction, polarization control and operating C-rate. The proposed framework could provide a data-driven and mechanism-informed basis for the prediction and optimal design of cement-based energy storage materials.

Alkali-free aluminosilicate glasses, characterized by high hardness, high strain point, high electrical insulation, excellent chemical durability, and tunable thermal expansion coefficients, are preferred substrate materials for high-end OLED displays. In this study, glass samples with a fixed total alkaline-earth oxide content of 26.17% (molar fraction) and varying Mg/Ca/Sr ratios were prepared by the melt-quenching method. The coordination environments of ²⁹Si and oxygen atoms were characterized by magic-angle spinning nuclear magnetic resonance (MAS-NMR) and X-ray photoelectron spectroscopy (XPS), and first-principles calculations were performed to resolve atomic-scale network-structure evolution and elucidate the structural origin of the ternary MgO-CaO-SrO mixed alkaline-earth effect (MAEE). The results show that the short-range structures of network formers (Si-O, Al-O, B-O) exhibit high stability during alkaline-earth substitution. Electron localization function (ELF) analysis quantitatively shows that high-field-strength Mg²⁺ induces stronger localization of oxygen electron clouds than Sr²⁺. Partial density of states (PDOS) analysis reveals that Mg 3p does not hybridize with O 2p, whereas Sr 4p exhibits p-p hybridization with O 2p, providing an electronic-level explanation for the contrasting network constraints exerted by the two cations. Coordination number analysis further distinguishes their structural roles: Mg²⁺ acts as a network intermediate (CN = 4-5), while Ca²⁺ and Sr²⁺ behave as typical network modifiers. Under equimolar Mg/Ca mixing conditions, the nonlinear effect is most pronounced, with the NBO fraction reaching 28.6%, the free volume fraction expanding from ~1.04% to ~2.75%, and the number of low-polymerization Qⁿ units increasing. These findings suggest that the competition among alkaline-earth cations in local coordination, orbital hybridization, and charge compensation is the fundamental origin of the nonlinear structural response, providing a theoretical basis for the compositional design of high-performance electronic substrate glasses.

A series of glass samples with a fixed total alkaline-earth oxide content (26.17%) and varying Mg/Ca/Sr ratios were prepared by conventional melt-quenching at 1650 ℃ and subsequently annealed at 600 ℃. The local coordination environments of ²⁹Si and oxygen atoms were characterized using magic-angle spinning nuclear magnetic resonance (MAS-NMR) and X-ray photoelectron spectroscopy (XPS). To investigate the atomic-scale evolution of the network, ab initio molecular dynamics (AIMD) simulations were conducted using the CP2K package. The 321-atom models used the PBE exchange-correlation functional and Goedecker-Teter-Hutter (GTH) pseudopotentials. Structural descriptors-including radial distribution functions (RDFs), bond-angle distributions, coordination-number curves n(r) obtained from RDF integration, Qⁿ distributions, oxygen coordination ratios, and free volume fractions-were extracted from the optimized configurations to quantify short- and medium-range network structures. Electronic-structure analyses, including electron localization function (ELF) and partial density of states (PDOS), were performed to characterize bond ionicity and orbital hybridization.

The results indicated that while the short-range structures of Si-O, Al-O, and B-O units remained highly stable across compositions, medium-range topological connectivity was strongly influenced by the MAEE. ELF analysis revealed that in the Mg-rich N3 composition, the bond-region ELF is 0.503, lower than 0.524 in the Sr-rich N5 composition (ΔELF = -0.021). This directly confirmed that high-field-strength Mg²⁺ exerted stronger polarization and localization on oxygen electron clouds, resulting in more ionic Mg-O bonds. PDOS analysis further showed that Mg 3p was buried at -37.6 eV with an O 2p energy separation exceeding 33 eV (no hybridization), whereas Sr 4p lay at -14.3 eV with an O 2p separation of only ~11 eV (p-p hybridization). The weaker electrostatic constraint from Sr²⁺ allowed O 2p orbitals to shift toward the Fermi level, consistent with the ELF findings. In the CaO-SrO substitution sequence, larger Sr²⁺ ions acted primarily as long-range charge compensators for [AlO₄]^- units, increasing the BO fraction to 74.8% in sample N2. In Mg-rich systems, high-field-strength Mg²⁺ induced more NBOs; however, the short and strong Mg-O ionic bonds locally constrained the network. Coordination number analysis from RDF integration clearly distinguished the structural roles: Mg²⁺ exhibited a CN distribution of 4-5 (network intermediate behavior), while Ca²⁺ and Sr²⁺displayed higher and broader CN distributions (typical network modifier behavior), resolving the ambiguity in bond-angle distribution differentiation. Notably, when MgO and CaO coexist in equimolar proportions within a high-SrO matrix (sample N6), pronounced nonlinear variations emerged: the NBO content reached 28.6%, and the free volume fraction, quantitatively evaluated by probe-sphere scanning (probe radius 1.2 Å, grid spacing 0.1 Å), increased from ~1.04% (N3, Mg-rich) to ~2.75% (N5, Sr-rich). This expansion was primarily attributed to the dilation of modifier coordination environments rather than to rupture of the network-former framework.

This study systematically elucidates the regulating role of alkaline-earth cation ratios on the network structure of alkali-free aluminosilicate glasses in the MgO-CaO-SrO ternary system by combining NMR and XPS characterization with AIMD simulations. The short-range structural units (Si-O, Al-O, B-O) exhibit high stability against compositional variation (Si-O CN = 4.00 ± 0.01; B-O predominantly [BO₃] with average CN = 3.08 ± 0.12), whereas the medium-range topological structure exhibits a pronounced nonlinear MAEE. High-field-strength Mg²⁺ promotes NBO formation with strong electron localization (ELF = 0.503) and no Mg 3p/O 2p hybridization, resulting in highly ionic Mg-O bonds that locally constrain the network. In contrast, larger-radius Sr²⁺ preferentially compensates [AlO₄]^- charges with weaker electron localization (ELF = 0.524) and exhibits Sr 4p/O 2p p-p hybridization (energy separation ~11 eV), facilitating network condensation. Coordination number analysis further differentiates the structural roles: Mg²⁺ functions as a network intermediate (CN = 4-5) and Ca²⁺/Sr²⁺ function as network modifiers (higher and broader CN distributions). Under equimolar Mg/Ca mixing (sample N6), the nonlinear response is most pronounced, with the NBO fraction reaching 28.6% and the free volume fraction expanding from ~1.04% (N3) to ~2.75% (N5), accompanied by an increase in low-polymerization Qⁿ units. These findings demonstrate that competitive coordination, orbital hybridization, and charge-compensation behaviors among alkaline-earth cations are the fundamental origin of the nonlinear structural response. The close agreement between experimental observations and simulation trends confirms the reliability of AIMD for resolving the structural evolution of complex glass networks and provides a theoretical basis for the compositional design of high-performance electronic substrate glasses.

AAC combined with FRP bars has broad application potential in marine engineering. Future research should focus on the following aspects: developing advanced microscale testing techniques to analyze the ITZ microstructure; investigating the bond durability of AAC prepared with seawater and sea sand under long-term field conditions; exploring the bond performance of ultra-high-performance AAC (UHPAAC) with FRP bars; and establishing more accurate prediction models by integrating multifactor test data with machine-learning methods.

Concrete pumping technology is widely applied in major construction projects. However, under long-distance or complex conditions, concrete transport often exhibits strongly non-steady characteristics due to the coupled factors like material rheology and pipeline topologies. This instability easily induces abnormal resistance surges, evolving into severe blockages. Conventional monitoring relies on post-event management or static pressure thresholds, triggering alarms only when pressure breaches a safety limit. Essentially, concrete blockage is a progressive dynamic accumulation of rheological resistance, leaving a critical "pressure accumulation" intermediate state. Static thresholds ignore this, causing severe pre-warning lag. AI methods show a potential, while intelligent perception faces three challenges, i.e., the binary classification defect overlooking the transitional state, the "majority class trap" from extreme class imbalance, and the contradiction between prediction accuracy and physical interpretability. To address these bottlenecks, this study was to propose a novel intelligent evaluation framework for non-steady concrete transport, deeply integrating rheological mechanisms with advanced machine learning architectures.

The research relied on the Zhenglong Hongshui River Bridge project, China. An embedded monitoring system with 600 Hz dynamic pressure sensors was seamlessly welded to the straight pipeline's inner wall near the pump outlet to acquire high-quality rheological resistance signals. During feature engineering, mechanical noises from valve reversing were eliminated, retaining only effective pressure-holding signals during main cylinder propulsion. Since deriving the exact friction coefficient via true flow velocity is a challenge in field engineering, macroscopic proxy variables were utilized. Assuming constant propulsion speed, the slope of the pressure curve was extracted to represent the transient growth rate of rheological friction. Standard deviation and range were extracted to quantify slip-layer stability and flow turbulence. To address extreme class imbalance in the 691-sample dataset (78.3% normal), the Synthetic Minority Over-sampling Technique (SMOTE) was applied to reconstruct the feature space, eliminating training bias. Subsequently, four ensemble algorithms (i.e., RF, XGBoost, LightGBM, GBDT) optimized by the Tree-structured Parzen Estimator (TPE) were used as base learners. Finally, a dynamic adaptive ensemble fusion architecture calculated sample-specific dynamic weights based on prediction confidence, replacing conventional static voting mechanisms.

The extracted time-domain features effectively map the dynamic evolution of in-tube rheological resistance. After SMOTE reconstruction and TPE optimization, the base models' potential to capture minority class abnormal conditions is maximized. During the comparative evaluation of the ensemble models, the study addresses the "accuracy paradox" inherent in highly imbalanced datasets. Conventional soft voting and hard voting models achieve an overall accuracy of approximately 0.90 simply because of the overwhelming majority class, dangerously masking high misdiagnosis rates for critical anomalies. The overall accuracy improves to 0.92 via implementing the dynamic adaptive fusion strategy. More importantly, the model substantially suppresses the omission rates for high-risk conditions. The recall rate for the "pressure accumulation" state reaches 91.43%, representing a 10.3% relative improvement over the soft voting method (i.e., 82.86%). For the most hazardous "blockage" state, the recall reaches 88.00%, outperforming the soft voting method (i.e., 76.00%) with a remarkable relative improvement of 15.8%. Furthermore, a SHapley Additive exPlanations (SHAP) visualization framework is introduced to verify the model's decision logic globally and locally. The global feature importance ranking confirms that the pressure curve slope is an absolute dominant feature triggering blockage predictions, followed by range and standard deviation. The SHAP beeswarm summary plots reveal that a sharp increase in these three features heavily biases the model toward a "blockage" decision. Conversely, smaller feature values drive the model toward a "normal" prediction. Local waterfall plots for individual samples further detail the positive and negative contributions of each feature to a specific prediction. This data-driven logic perfectly coincides with the physical reality that pipeline blockages instantly cause a sharp surge in rheological friction and severe dispersion in pressure values due to slip layer disruption. SHAP effectively explains the statistical feature attribution, and its strong alignment with engineering experience proves the model's physical rationality. Verifying strict causality requires a further extensive field validation across multiple complex scenarios.

Time-domain physical features validated the ability to characterize in-tube non-steady transport. The slope, standard deviation, and range extracted from high-frequency pressure sequences could serve as effective macroscopic proxies for concrete fluid resistance and slip-layer stability. The combination of SMOTE feature space reconstruction and TPE Bayesian hyperparameter tuning effectively overcome the small-sample fault warning dilemma under extreme class imbalance, eliminating training bias toward majority classes. The dynamic adaptive ensemble architecture demonstrated a core advantage in suppressing high-risk omission rates. It leveraged a minor overall precision increment to drastically reduce crucial fault omission risks via tolerating a minimal sacrifice in majority class accuracy, delivering an ample forward-intervention window. The SHAP attribution analysis achieved a white-box deconstruction of the ensemble decision logic. The alignment between the algorithm's reliance on slope surged and the true physical laws of rheological friction endowed the model with clear physical rationality, laying a solid theoretical foundation for the trustworthy application of AI technologies in intelligent civil engineering construction.

The cement industry contributes approximately 8% of global CO₂ emissions, with each ton of clinker production for releasing nearly one ton of CO₂. Mineral admixtures such as slag, fly ash, and silica fume become essential for developing sustainable concrete, enabling a partial replacement of cement (i.e., 15%-45%, mass fraction), while enhancing durability through pozzolanic reactions. However, high-volume substitutions of exceeding 30% critically depress a pore solution alkalinity from pH value 13.3 to 12.5-12.8, destabilizing passive films on conventional carbon steel (LC) and elevating corrosion risks by 3-5 times. This alkalinity reduction has severe threats to structural integrity in marine environments and deicing salt exposure, driving global infrastructure repair costs beyond $2.5 trillion annually. A Cr10Mo1 alloy corrosion-resistant steel (LA) was developed, having a chloride threshold 10 times greater than that of LC in simulated pore solutions. However, its passivation behavior in realistic cement-mineral admixture systems with synergistic interactions of Ca²⁺, SO₄^2-, and Al³⁺ ions remains unexplored. In this study, the passivation mechanisms of the steel in complex paste extracts were investigated to provide foundational data for designing durable and sustainable infrastructure.

The passivation kinetics were analyzed by electrochemical techniques like linear polarization resistance (LPR), electrochemical impedance spectroscopy (EIS), and Mott-Schottky analysis. The LA and LC electrodes (φ10 mm×10 mm) were polished to a 2000-grit finish and immersed for 10 d in extracts derived from cement pastes containing different mineral admixtures of 0%, 15%, 30%, and 45% (slag/fly ash = 1∶1), designated as 0SF, 15SF, 30SF, and 45SF. The paste extracts were prepared with P·O 42.5 cement at a water-to-binder ratio of 1∶1, cured at 25 ℃ for 14 d, filtered through 0.45 μm membranes, and stored in nitrogen to prevent carbonation. The measured pH values reduced progressively from 13.4 (0SF) to 12.5 (45SF). Electrochemical testing utilized a three-electrode cell with a platinum counter electrode and a saturated calomel reference electrode (SCE). LPR scans (±20 mV vs. open-circuit potential at 0.166 7 mV/s) and EIS measurements (10 mV amplitude, 0.01-1.00×10⁴ Hz frequency range) were conducted after passivation. The Mott-Schottky analysis was performed at 1000 Hz with potential sweeps from -1.5 V to +0.25 V vs. SCE (50 mV/s). Equivalent circuit modeling (R_s-[CPE_dl-R_ct]-[CPE_f-R_f]) was used to fit the EIS data to resolve solution resistance (R_s), passive film properties (CPE₁, R₁), and charge transfer resistance (CPE₂, R₂). X-ray photoelectron spectroscopy (XPS) was used to characterize compositional gradients by an Axis Ultra DLD spectrometer with Al K_α radiation (1486.6 eV) and Ar⁺ sputtering (0.02 nm/s). The spectra of Fe 2p₃/₂ and Cr 2p₃/₂ were deconvoluted via the Gaussian-Lorentzian peak fitting.

The steels of LA and LC both can be effectively passivated in all extracts, while demonstrating fundamentally divergent responses to increasing mineral admixture content. The LA displays significantly an enhanced passivation, evidenced by an 81% increase in polarization resistance (i.e., R_p: 348 kΩ·cm² in 0SF to 630 kΩ·cm² in 45SF) and a positive shift in corrosion potential (i.e., E_corr: -157 mV to -128 mV). Conversely, the LC suffers a degraded passivity with a 54% reduction in R_p (i.e., 348 kΩ·cm² to 160 kΩ·cm²) and a negative E_corr shift to -208 mV. The Nyquist plots corroborate these results, revealing expanded capacitive arcs for the LA and contracted arcs for the LC in low-pH value extracts. The charge transfer resistance (R_ct) for the LA increases from 5.4×10⁵ Ω·cm² (0SF) to 6.4×10⁵ Ω·cm² (45SF), which is decreased by 44% (i.e., from 5.9×10⁵ Ω·cm² in 0SF to 3.3×10⁵ Ω·cm² in 45SF) for the LC. The Mott-Schottky analysis further distinguishes their electronic properties. The LA exhibits a n-p heterojunction behavior characterized by positive slopes in the -1.00 V to +0.25 V region and negative slopes between -1.5 V and -1.0 V, with decreasing total carrier density (N_total) from 4.96×10²⁰ cm^-3 (0SF) to 1.93×10²⁰ cm^-3 (45SF). In contrast, the LC maintains a n-type semi-conductivity with increasing N_total from 4.93×10²⁰ cm^-3 to 7.09×10²⁰ cm^-3.

XPS depth profile shows critical differences in passive film architecture. The LA forms a bilayer structure comprising an outer layer (0-5 nm) of FeOOH/Fe(OH)₃ (712.2 eV) and Fe₂O₃ (710.7 eV), and an inner layer (5-15 nm) enriched with Cr₂O₃ (576.3 eV), Cr(OH)₃ (577.1 eV), and FeO (709.5 eV). Reduced alkalinity triggers an increase of 32% in Cr-oxide content (i.e., Cr_ox+hy/Cr_metal=2.8 in 45SF vs. 2.1 in 0SF), forming a dense chromium-rich barrier. The LC develops a monolayer Fe-oxide film (0-10 nm) containing Fe₂O₃, FeOOH, and FeO, with decreasing FeO content by 67% (i.e., Fe²⁺/Fe³⁺=0.18 in 45SF vs. 0.55 in 0SF). Mechanistically, a low alkalinity elevates the oxidation potential of chromium in the LA, promoting Cr³⁺ incorporation into the inner film. The resultant Cr₂O₃ (p-type semiconductor) establishes a high-energy barrier (i.e., >1.2 eV) against anion penetration, while coexisting FeO (n-type) stabilized cation transport. For the LC, FeO decomposition (Fe²⁺→Fe³⁺) at a low pH value increases oxygen vacancy density, accelerating ionic transport through the film and diminishing chloride adsorption capacity resulted from loss of [Fe²⁺-Cl^-] coordination sites.

Mineral admixtures content variation at 15%-45% could affect the passivation behavior of Cr10Mo1 alloy corrosion-resistant steel (LA) and ordinary carbon steel (LC). The two steels exhibited a good passivation in the paste extract solutions, but had different performances. As the amount of mineral admixture increased, the passivation of the LA enhanced, while the passivation effect of the LC gradually weakened. The passivation film of the LA was a bilayer structure, with the outer layer primarily composed of FeOOH/Fe(OH)₃ and Fe₂O₃, and the inner layer mainly consisting of Cr₂O₃, Cr(OH)₃, and FeO. The passivation film of LC was a monolayer structure, mainly containing Fe₂O₃, FeOOH/Fe(OH)₃, and FeO. The alkalinity of the paste extract solution decreased, and more Cr elements involved in the formation of the passivation film of the LA with the increase in mineral admixture content, enhancing its passivation. Meanwhile, for the LC, Fe oxides in the passivation film gradually decomposed, and the content of FeO that provided a decreased protection, leading to a weakened passivation. This study indicated that the introduction of mineral admixture could not hinder the passivation effect of the LA, but weaken the passivation strength of the LC.

Stray current leakage poses a significant threat to the service life of underground electrified engineering structures. Although ultra-high-performance concrete (UHPC) exhibits superior durability, compared to ordinary concrete, the corrosion mechanism of internal steel fibers induced by stray current remains unclear, thus restricting its broader application in underground environments. Under the influence of stray current, chloride ion transport in UHPC is governed by the combined effects of free diffusion, electromigration, and changes in pore structure caused by steel fiber corrosion. The mechanisms of free diffusion and electromigration are extensively studied, but the influence of steel fiber corrosion on chloride ion transport is not thoroughly investigated. In particular, the impact of corrosion-induced microstructural changes on chloride migration is not quantitatively assessed, and the micro-morphological characteristics of steel fiber corrosion under stray current are not fully revealed.

The materials used included Portland cement (P·II 52.5), silica fume (particle size: 0.1-0.2 μm), fly ash (Type F), quartz powder (particle size: 20-100 μm), quartz sand (particle size: 450-900 μm), a superplasticizer, and steel fibers. The steel fibers were added at a volume fraction of 2%, with a length of 8 mm and a diameter of 0.12 mm. The microstructural characteristics of steel fiber corrosion and the chloride ion transport mechanism in UHPC under the influence of stray current were determined by rapid chloride migration test (RCM), water absorption test, atomic force microscopy (AFM), and X-ray computed tomography (X-CT), respectively. The electrochemical corrosion of UHPC was simulated via the RCM test, and the changes in effective porosity after corrosion were determined via the water absorption test. The micro-morphological features of corroded steel fibers were characterized by AFM, and the corrosion-induced microcrack propagation and changes in porosity within the UHPC matrix were analyzed by X-CT.

The corrosion pits on the surface of the steel fibers are primarily distributed in the range of -250 to 200 nm at an external voltage of 30 V. As the voltage increases to 50 V, the pit depth expands to -430 to 700 nm, and it further extends to-580 to 770 nm at 70 V. Correspondingly, the effective porosity of UHPC increases from 0.50% to 0.58% at 30 V, from 0.60% to 1.31% at 50 V, and from 0.58% to 1.88% at 70 V.

The expansion of corroded steel fibers induces the formation and propagation of microcracks within the UHPC matrix, leading to an increase in apparent porosity. At different applied voltages of 30 V, 50 V, and 70 V, the apparent porosity increases to 1.49-1.95%, 1.50-2.43%, and 1.47-2.96%, respectively. The effective chloride diffusion coefficient measured is 3.68×10^-12 m²/s at 30 V, increases to 9.11×10^-12 m²/s at 50 V, and further increases to 28.3×10^-12 m²/s at 70 V. These results demonstrate that stray current significantly accelerates the chloride ion transport in UHPC. Neglecting the effect of porosity variation can result in a significant overestimation of the service performance of UHPC structures within their intended design life.

Stray current induced changes in the micro-morphology of steel fibers, with maximum pit depths concentrated in the range of -580 nm to 770 nm. The corrosion became more severe closer to the exposed surface, potentially causing a fracture of the steel fibers and a complete loss of their load-bearing capacity. The results of water absorption tests indicated that stray current increased the effective porosity of UHPC. This increase could be primarily attributed to the expansion of corroded steel fibers, leading to microcrack propagation within the UHPC matrix. The apparent porosity caused by steel fiber corrosion was quantified. In specimens unaffected by stray current, the apparent porosity ranged from 1.47% to 1.56%. A higher applied voltage led to more significant changes in apparent porosity.

The western region of China is characterized by high altitude, significant diurnal temperature variations, and extensive saline soil distribution. Under the coupled effect of freeze-thaw cycles and salt attack, the service life of local concrete structures decreases significantly, resulting in substantial economic losses. Although steel fiber-reinforced ultra-high performance concrete (UHPC) exhibits an enhanced freeze-thaw resistance, it faces a challenge of underutilization of fiber strength due to the limitations imposed by fiber-matrix interface degradation and inherent matrix risks. In this study, ethylenediaminetetraacetic acid (EDTA) treated steel fibers replaced conventional steel fibers to enhance the mechanical properties and freeze-thaw resistance of UHPC via improving the fiber-concrete matrix interface performance.

In this study, steel fibers were modified using the EDTA method. First, 0.05 mol/L CaCl₂ and 0.05 mol/L Na₂CO₃ solutions were prepared separately, and 1% EDTA was added to Na₂CO₃ solution. Subsequently, the steel fibers were immersed in a mixed solution of CaCl₂, Na₂CO₃, and EDTA for 6 h. Finally, the fibers were washed with distilled water. Based on the measured concentrations of SO₄^2- and Cl- in saline soil from Golmud, this study employed a 133.6 g/L NaCl and 19.5 g/L MgSO₄ solutions to simulate the corrosive environment for freeze-thaw tests. This study investigated the effect of combined salt attack-freeze-thaw cycles on the fiber-matrix interfacial bond strength, pull-out energy, relative dynamic elastic modulus, mass loss ratio, compressive strength, and flexural strength of UHPC. The phase composition and microstructure were characterized by X-ray diffraction (XRD), mercury intrusion porosimetry (MIP) and scanning electron microscopy (SEM), revealing the microscopic mechanisms of performance degradation. In addition, a freeze-thaw damage model for UHPC was also established based on the two-parameter Weibull distribution function.

The EDTA modification induces a longitudinal exfoliation on steel fiber surfaces, resulting in a significantly increased roughness. The interfacial bond strength and pull-out energy are enhanced by 22.2% and 66.7%, respectively. Correspondingly, the compressive strength and flexural strength of UHPC are improved by 2.5% and 12%. The enhanced fiber-matrix interfacial bond strength and pull-out energy increase the ultimate tensile stress required for unstable propagation of microcracks in UHPC, effectively inhibiting the initiation and development of microcracks. This mechanism thereby enhances its freeze-thaw resistance. After 500 cycles of water-freezing and salt solution freezing tests, the EDTA-treated steel fiber-reinforced UHPC demonstrates significantly superior performance, compared to conventional steel fiber UHPC in all critical metrics i.e., interfacial bond strength, pull-out energy, compressive strength, flexural strength, relative dynamic elastic modulus, mass loss ratio, and porosity. After 500 freeze-thaw cycles in water, the bond strength of plain steel fiber UHPC and modified steel fiber UHPC are decreased by 40.4% and 27.8%, respectively. Their pullout energy are declined by 29.7% and 50.6%, respectively. The relative dynamic elastic modulus are decreased by 5.26% and 4.57%, respectively. The masses are reduced by 0.46% and 0.45%, respectively. The compressive strengths are decreased by 16.6% and 16%, respectively, and the flexural strengths are decreased by 26.1% and 20.6%, respectively. After 500 salt solution freeze-thaw cycles, the bond strength of plain steel fiber UHPC and modified steel fiber UHPC are decreased by 41.3% and 39.3%, respectively. The pullout energies are declined by 40.5% and 54.2%, respectively. The relative dynamic elastic modulus are dropped by 5.44% and 5.25%. The masses are reduced by 0.5% and 0.48%. The compressive strengths are decreased by 21.4% and 20.3%, respectively, and the flexural strengths are decreased by 29.3% and 25.2%, respectively. The XRD patterns reveal that after 500 salt solution freeze-thaw cycles, some expansive products such as ettringite and gypsum are formed within the UHPC. The expansive stresses generated by these products induce crack initiation and propagation, thereby degrading the performance of the UHPC. MIP indicates that freeze-thaw cycles result in an increased porosity in UHPC. In both water-freezing and salt solution freeze-thaw environments, the porosity of modified steel fiber UHPC is lower than that of plain steel fiber UHPC. SEM images reveal that during debonding of modified steel fibers from the matrix, a greater amount of weak matrix material adhering to the fiber surfaces is carried away. This phenomenon demonstrates that modified steel fibers establish a stronger interfacial bonding with the cementitious matrix. Based on the predictions by the Weibull model, the critical freeze-thaw cycles for EDTA-treated and conventional steel fiber-reinforced UHPC are predicted to be 1610 and 1270 cycles under water-freezing exposure, and 1220 and 1150 cycles under salt solution-freezing exposure, respectively. The results indicate that in the both environments, fiber modification treatment can effectively enhance the freeze-thaw resistance of UHPC and significantly improve its durability.

The results demonstrated that EDTA modification treatment enhanced the strength and fiber-matrix interfacial bonding performance of UHPC, inhibited micro-crack propagation, thereby slowing the deterioration rate of properties such as relative dynamic elastic modulus and strength during freeze-thaw cycles, reduced mass loss, and improved the freeze-thaw resistance of UHPC. The freeze-thaw damage model predicted that EDTA-modified steel fiber UHPC could have a 26.8% increase in freeze-thaw cycle resistance under water-freezing conditions and a 6.1% improvement under salt solution-freezing environments, compared to conventional steel fiber UHPC. This study could offer a reference for the durability de-sign of UHPC structures in cold saline soil regions of western China.

3D printing technology offers a significant potential in the construction sector via enhancing build efficiency and enabling greater design freedom. However, its widespread adoption faces challenges related to the multi-objective optimization of concrete performance and the lack of intelligent decision-making for printing parameters. To address these issues, this paper was to propose a hybrid modeling framework that could integrate physics-informed and data-driven approaches for the intelligent design of 3D printing parameters tailored to target material properties. The framework first developed a rheological performance prediction model based on the physical information equation (PIE). This model, combined with the PIE, was then used to construct an extrusion screw speed (ES) database. The framework could enable accurate determination of the printing speed (PS) via incorporating key parameters such as the target concrete flow rate (Q), printing nozzle diameter (PN), and printing layer height (PH). The proposed intelligent parameter design method could provide a quantitative tool for the collaborative control of multiple concrete properties, significantly enhancing the efficiency and reliability of process parameter optimization. This study could offer a theoretical foundation and methodological support for improving the build quality and process stability of 3D printing concrete, thus facilitating the technology's application and adoption in construction engineering.

A hybrid physics-data driven framework was developed for printing parameter optimization. This involved two distinct fusion strategies:

1. Physics-Informed Rheological Prediction ("PIE embedded in ML"): A Physics-Informed Convolutional Neural Network (PICNN) was constructed to predict rheological parameters (i.e., Yield Stress-YS, and Plastic Viscosity-PV). Physical equations describing the relationship between mix proportions and rheology were embedded as soft constraints into the CNN's loss function. The Dung Beetle Optimizer (DBO) was used for hyperparameter tuning, including the weights balancing data-driven loss and physics-informed loss.

2. Physics-Serialized Extrusion Speed Prediction ("PIE serialized with ML"): An ES prior database was built via coupling multiple physical equations (i.e., density, Bernoulli, torque) with the predicted rheological parameters from the PICNN model. A Random Forest (RF) model, optimized using DBO, was then trained on this database to predict ES intelligently based on material properties and equipment conditions.

Based on the principle of mass conservation, an explicit calculation equation for PS was derived as a function of Q, PN, and PH. This could complete the parameter optimization chain from material properties to process parameters.

The PICNN model for predicting YS and PV demonstrates a stable convergence during training. Compared to the pure data-driven CNN model, the PICNN achieves significantly lower losses (e.g., testing set average loss for YS: PICNN 198.92 Pa vs. CNN 307.80 Pa; for PV: PICNN 0.37 Pa·s vs. CNN 0.77 Pa·s), confirming that physical constraints enhance predictive accuracy and robustness. Furthermore, the predicted values from PICNN show a high consistency with the calculations from the pure PIE (e.g., low MAPE of 6.02% for YS and 2.60% for PV on testing sets), thus proving a great adherence to physical laws.

The RF model for ES prediction, trained on the physics-generated database, achieves a high accuracy with R² values of 0.99 (training set) and 0.97 (testing set), and RMSE values of 1.86 r/s (training) and 2.39 r/s (testing). The scatter plots show points closely distributed around the fit line, and error distributions are concentrated at low levels, indicating the model's effectiveness in learning the complex relationships for ES prediction.

The derived equation for PS effectively establishes the quantitative relationship with Q, PN, and PH. A case study presenting calculated PS values for various common combinations of these target parameters provides a practical reference for configuring printing processes, demonstrating the applicability of the complete optimization framework.

This study established a comprehensive physics-data hybrid modeling framework for optimizing 3D printing concrete parameters, achieving a precise matching between parameters and material rheological properties.

The physics-informed rheological prediction model (PICNN) significantly improved prediction accuracy and physical consistency for YS and PV compared to purely data-driven or purely physics-based approaches.

The physics-serialized extrusion speed prediction method, utilizing a physically-consistent database and an optimized RF model, enabled high-precision, robust intelligent recommendation of ES based on material attributes.

The complete parameter optimization solution, integrating rheological prediction, ES prediction, and PS calculation, provided a systematic and intelligent paradigm for 3D printing parameter design, reducing reliance on conventional trial-and-error methods.

This research could contribute to enhancing the forming quality and process stability of 3D printing concrete, facilitating the intelligent development of construction 3D printing technology. Future work could focus on more complex multi-objective optimization scenarios and real-time dynamic parameter control strategies.

The cement industry is one of the most CO₂-intensive industrial sectors, and reducing clinker consumption through mineral substitution is an effective strategy for developing low-carbon cementitious materials. Calcium carbonate (CaCO₃) is widely used as a clinker substitute because of its abundant resources, low cost, and compatibility with cement production. In addition to the thermodynamically stable calcite, CaCO₃ also exists in metastable polymorphs such as vaterite and aragonite, which differ in crystal structure, particle morphology, specific surface area, and dissolution behavior. These differences may produce distinct effects on the rheology, hydration, pore structure, and mechanical properties of cement pastes. This study aims to clarify how different CaCO₃ polymorphs regulate the coupled rheology-hydration-structure-strength relationship of cement paste.

Ordinary Portland cement pastes containing 15% (mass fraction) calcite, vaterite, or aragonite were prepared at a constant water-to-binder ratio of 0.42, with pure cement paste as the reference. Commercial limestone powder was used as the calcite source, while vaterite and aragonite were synthesized using a double-decomposition method. Rheological properties were characterized using a rotational rheometer, and the dynamic yield stress was determined by fitting the descending branch of the flow curve with the Herschel-Bulkley model. Early hydration behavior was evaluated by isothermal calorimetry. Flexural and compressive strengths were measured at 1, 3, 7 d, and 28 d. Hydration products and phase evolution were examined by X-ray diffraction and thermogravimetry, and the pore structure was characterized by mercury intrusion porosimetry.

The three CaCO₃ polymorphs exhibited distinct physicochemical characteristics and consequently exerted different effects on cement paste performance. Calcite mainly consisted of irregular blocky particles and mainly acted through dilution and filling, leading to reductions in both static and dynamic yield stress and only limited promotion of hydration and strength development. Vaterite, composed of loosely aggregated fine particles with relatively high surface activity, increased the dynamic yield stress, accelerated early hydration most effectively, promoted the formation of portlandite, chemically bound water, and carboaluminate phases, and refined the pore structure, and was therefore more beneficial for compressive strength development. Aragonite, with a needle-like morphology, reduced the dynamic yield stress but markedly increased the apparent viscosity and thixotropic index, which indicated a stronger effect on structural build-up and rebuilding. It also improved pore structure and showed the highest compressive strength at 1-7 d as well as the highest flexural strength and flexural-to-compressive strength ratio at 28 d. These results indicated that the effects of CaCO₃ polymorphs could not be attributed to a single parameter such as particle size, but were jointly governed by morphology, specific surface area, and reactivity.

Different CaCO₃ polymorphs significantly affect the rheology, hydration, and strength development of cement pastes. Calcite mainly shows weak dilution, filling, and nucleation effects, resulting in reduced yield stress and limited strength compensation. Vaterite markedly promotes early hydration and sustained product accumulation, leading to higher dynamic yield stress, increased portlandite and chemically bound water contents, greater formation of carboaluminate phases, refined pore structure, and improved early compressive strength. Aragonite, owing to its needle-like morphology, strongly enhances static yield stress, apparent viscosity, and thixotropic recovery, while also refining pore structure and achieving the highest flexural strength and flexural-to-compressive strength ratio. Compared with calcite, metastable CaCO₃ polymorphs are more effective in regulating the coupled “rheology-hydration-strength” behavior of cement pastes. Vaterite is more suitable for enhancing hydration and promoting compressive strength development, whereas aragonite is more advantageous for rheological control, flexural performance, and overall mechanical properties.

Although the aforementioned technologies continue to be optimized along their respective development paths, there remain insurmountable bottlenecks. There is an inherent trade-off between consistency, control precision, and production efficiency. When pursuing high precision and high consistency, production efficiency is often compromised (e.g., by using slower scanning speeds or finer layer-by-layer stacking), whereas pursuing high-speed production tends to result in reduced structural controllability (e.g., excessively fast laser scanning leading to uneven pore morphology, or excessively fast template-based molding causing disordered pore orientation). This 'engineering trilemma' represents the core contradiction that thick electrode manufacturing technologies must address as they transition from the laboratory to large-scale production. Overall, there is currently no single technology capable of simultaneously achieving high precision, high consistency, and high production efficiency. Future breakthroughs may lie in: 1) the integration of multiple technologies, such as combining rapid laser processing with the template method—first using a template to construct a macro-porous framework, then employing lasers for localized fine-tuning. 2) Process innovation, including the development of continuous ice-templating or roll-to-roll compatible guided forming processes. 3) Equipment upgrades, such as the development of new equipment for multi-beam parallel laser processing and high-speed multi-nozzle 3D printing. Therefore, identifying a scalable manufacturing technology that integrates all three elements is key to transitioning thick electrodes from laboratory research to large-scale application.

Research on thick lithium-ion battery electrodes is currently at a critical stage of development. As a bridge linking intrinsic properties at the atomic scale to macroscopic performance at the electrode scale, the rational assembly of unit cell structures will become a core technological pathway for next-generation high-energy/high-power batteries. Achieving this goal requires addressing numerous challenges, including gaining a deep understanding of transport mechanisms, enabling the controlled synthesis of electrode structures, and ensuring compatibility with existing battery manufacturing systems. The engineering of unit cell structures offers us a new research direction and a broader perspective.