Journal of building engineering最新文献

{"title":"Multi-scale Pore Structure and Mechanical Anisotropy of PVA Fiber-Reinforced 3D-Printed Seawater Sea-Sand Sulfoaluminate Cement","authors":"Jingwei Ying, Wanfu Qiao, Hao Wang, Thi Loan Pham","doi":"10.1016/j.jobe.2026.117231","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117231","url":null,"abstract":"Driven by the green construction demands of deep-sea infrastructure, the 3D-printed seawater sea-sand sulfoaluminate cement (3D-printed-SWSS-SAC) system exhibits immense application potential. However, multi-scale composite pore defects induced by the vibration-free layer-by-layer extrusion process and complex internal physicochemical reactions severely restrict the further improvement of its mechanical properties. To overcome this bottleneck, this study systematically investigated the regulatory effects of polyvinyl alcohol (PVA) fiber dosage (0-0.5 vol%) on the printability, 3D evolution of the pore network, and mechanical anisotropy of this system. The results indicated that an appropriate amount of PVA fibers significantly enhanced the shape retention capacity of the extruded filaments. Their surface-abundant hydroxyl groups effectively improved the internal hydration environment via hydrogen bond adsorption, promoting the massive early formation of amorphous AH<ce:inf loc=\"post\">3</ce:inf> gel and thereby densifying the matrix. At the optimal dosage (0.4 vol%), the microscopic pore structure achieved global optimization: the total porosity decreased significantly from 21.40% to 18.97%, and the proportion of highly harmful pores (&gt;200 nm) that are prone to inducing stress concentration dropped sharply from 53.1% to 35.2%. Additionally, at this dosage the 3 d flexural strength of the printed specimens along the transverse filament direction (Y direction, 10.22 MPa) and the elastic modulus along the printing direction (X direction, 4.85 GPa) were already essentially equivalent to those of the vibrated cast specimens under the same conditions. Digital image correlation (DIC) coupled with specific crack area analysis confirmed that micropore optimization effectively shifted the failure mode from brittle interlayer delamination to quasi-ductile multiple microcrack propagation. This study elucidates the synergistic regulatory mechanism of PVA fibers on the microscopic pores and macroscopic mechanical properties of the 3D-printed-SWSS-SAC system, providing a scientific and quantifiable engineering design basis for marine-oriented 3D-printed composites.","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"85 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884059","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Thermally activated Algerian diatomite as a supplementary cementitious material: calcination optimization, rheology and mortar strength","authors":"Fatiha Ait Ahmed, Chouaib Aribi, Noureddine Mesboua, Issam Aouari, Arbaoui Ahcene, Ramdani Soumia","doi":"10.1016/j.jobe.2026.117233","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117233","url":null,"abstract":"Natural diatomite is a promising supplementary cementitious material, but its performance depends strongly on thermal activation, which must remove organic and carbonate impurities while preserving reactive amorphous silica and porous frustule structure. This study presents the systematic investigation of Algerian diatomite from the Sig-Mascara region, evaluating three calcination temperatures (850, 900, and 950 °C) combined with four cement replacement levels (5, 10, 15, and 20 wt.%).","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"5 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884058","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Effect of cross-section and stirrup configuration on the torsional performance of geopolymer concrete columns under compression","authors":"Cheng Peng, Hyeon-Jong Hwang, Yunxing Du, Dong Ding, Xiang Hu, Shuisheng Li, Caijun Shi","doi":"10.1016/j.jobe.2026.117257","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117257","url":null,"abstract":"To clarify the compression-torsion behavior of slag-based geopolymer concrete (GPC) columns with different cross-sections and stirrup details, quasi-static tests and finite element analyses were conducted on nine solid and three hollow-section columns. The test parameters were axial compression ratio, stirrup spacing, stirrup diameter, and cross-section configuration. The results showed that the specimens mainly failed in torsion-bending, shear-torsion, or compression-torsion modes, and a higher axial compression ratio promoted rapid propagation of diagonal cracks and intensified brittle failure. Hollow-section specimens showed faster stiffness degradation after cracking, and their peak torques were only 59.4%-71.4% of those of the corresponding solid specimens. Axial compression delayed cracking and enhanced torsional capacity up to an axial compression ratio of about 0.6; beyond this value, premature compression damage weakened the beneficial effect and promoted brittle post-peak degradation. The most favorable strength-deformation balance was obtained when the longitudinal bar-to-stirrup strength ratio was approximately 1.20. The finite element model reasonably reproduced the crack distribution and stress redistribution, but certain deviations remained in the curve slope and post-peak response. Current design models for ordinary concrete generally overestimated the torsional capacity of GPC columns, especially for hollow-section specimens. Based on these findings, a modified method applicable to C30 grade solid-section GPC columns was proposed, while the hollow-section tests provided preliminary evidence of the influence of section form on post-cracking torsional resistance. The mean ratio of predicted to experimental results was 1.014, indicating reasonable accuracy with a certain safety reserve.","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"19 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884056","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AeroPDSF: An attention-enhanced surrogate framework for aerodynamic optimization of perforated double-skin facades in super-tall buildings","authors":"Kun Lin, Siyao Yang, Annan Zhou, Hongjun Liu","doi":"10.1016/j.jobe.2026.117249","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117249","url":null,"abstract":"Super-tall buildings are highly sensitive to wind-induced responses, and perforated double-skin facades (PDSFs) provide a passive control strategy for improving aerodynamic performance without substantially changing the main building form. However, the aerodynamic effectiveness of PDSFs is jointly affected by opening ratio, opening pattern, opening location, and wake interaction, making large-scale optimization computationally expensive. This study proposes AeroPDSF, an attention-enhanced surrogate framework for aerodynamic optimization of PDSFs. The framework integrates selective facade-unit representation, a computational fluid dynamics (CFD)-informed response database, an attention-enhanced convolutional surrogate model, and sensitivity-informed non-dominated sorting genetic algorithm III (NSGA-III) optimization to efficiently optimize PDSF layouts and reduce structural responses, including peak top acceleration, peak top displacement, and maximum inter-story drift ratio. The results show that the proposed model accurately captures the nonlinear relationship between facade layouts and structural responses. Sensitivity analysis indicates that effective openings are mainly concentrated in the central facade region, and vertical openings are more effective in weakening fluctuating lift and wake vortex shedding. Compared with conventional NSGA-III, the sensitivity-informed strategy achieves faster convergence and lower final objective values. Compared with the fully non-perforated baseline, the optimized layouts reduce peak top displacement, maximum inter-story drift ratio, and peak top acceleration by 54.5%, 57.4%, and 92.2%, respectively. CFD flow-field analysis further confirms that the optimized openings reduce structural responses by forming pressure-release passages and weakening coherent wake structures. The proposed framework provides an efficient and interpretable pathway for passive aerodynamic optimization and high-performance facade layout design of super-tall buildings.","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"43 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884057","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CO2-Driven Inverse Estimation for HVAC Energy Simulation and Control Screening: A Case Study of a Multi-Floor University Library in Shanghai, China","authors":"Longbo Chang, Yaolong Cheng, Kunge Hou, Guocheng Liu, Xingang Liu, Jianyuan Hou, Renxi Zhang","doi":"10.1016/j.jobe.2026.117229","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117229","url":null,"abstract":"Occupancy and ventilation-related boundary conditions are difficult to observe directly in real public buildings, which limits the reliability of heating, ventilation and air-conditioning (HVAC) energy simulation and control assessment. This study develops a low-intrusion CO<ce:inf loc=\"post\">2</ce:inf>-driven grey-box framework for reconstructing simulation-ready boundary conditions under limited-observation conditions. Multi-floor indoor CO<ce:inf loc=\"post\">2</ce:inf> measurements and turnstile-derived whole-building occupancy data are used to infer floor-resolved occupancy schedules and equivalent air-change rate boundaries, which are then imported into an OpenStudio/EnergyPlus model. A university library in Shanghai, China, was used as a real multi-floor case study over a 37-day cooling-season period. Compared with the baseline model, the proposed model reduced the hourly Normalized Mean Bias Error (NMBE) from 11.20% to 1.17%, while the Coefficient of Variation of the Root Mean Square Error [CV(RMSE)] changed from 25.41% to 24.46%. These results indicate that the reconstructed boundaries mainly help mitigate systematic model bias and improve the representativeness of simulation inputs. Based on the calibrated model, a preliminary chilled-water supply temperature reset strategy was further evaluated, yielding energy savings of 4.88% in simulation and 5.35% in field implementation. The findings suggest that CO<ce:inf loc=\"post\">2</ce:inf>-driven boundary reconstruction can provide practical support for HVAC model calibration and preliminary control screening in real multi-floor public-building operation.","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"18 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884088","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Generalizable deep learning-based hysteresis analysis of varying post-cast precast beam-column joints","authors":"Shunyao Wang, Shan He, Jianwei Li, Peiyuan Xu, Ruiyang Zhang","doi":"10.1016/j.jobe.2026.117082","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117082","url":null,"abstract":"Accurate prediction of the full hysteresis response of post-cast precast reinforced concrete beam–column joints is essential for seismic performance assessment and design-oriented applications. However, conventional numerical approaches such as finite element methods are computationally expensive. Recently, artificial intelligence (AI) has drawn tremendous attention and interest in hysteresis modeling, while most existing AI-based models are typically developed for a single joint configuration and exhibit limited generalization across varying structural parameters, which greatly limits their application in engineering practice. To tackle this generalizability issue, this study proposes a Hysteresis Generalized Network (HystGNet) for predicting the hysteresis responses of joints with diverse structural configurations using a single model. HystGNet learns the nonlinear mapping from joint structural parameters and imposed displacement histories to restoring force responses by effectively integrating loading temporal dependencies with structural features. The model is trained and tested on a numerical dataset comprising 5000 post-cast precast beam–column joints developed in OpenSees, with systematic consideration of varied structural parameters and loading protocols. Its generalization capability is further evaluated under both unseen loading amplitudes and unseen joint parameter combinations. Additionally, HystGNet is validated using independent cyclic tests on three post-cast precast joint specimens. Results show that the proposed model can accurately reproduce hysteresis curves and key engineering indices, including backbone curves, secant stiffness degradation, and cumulative energy dissipation. The results demonstrate that HystGNet can efficiently predict hysteresis behaviors of diverse joints and effectively address the generalizability challenge of existing data-driven methods.","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"58 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884090","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Toward rational selection and design of concrete set retarders: Insights from DFT-enhanced QSAR modeling","authors":"Qingqing Xu, Zhenhua Wei, Han Liu, Beatrice Pomaro","doi":"10.1016/j.jobe.2026.117262","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117262","url":null,"abstract":"Concrete admixtures are essential for controlling properties of cementitious materials, yet their choice relies mainly on trial-and-error testing that offers limited insight into the mechanisms governing efficient and rational admixture selection. In this study, three sets of molecular descriptors, including structural fingerprints and physically informed descriptors from density functional theory (DFT) calculations, are explored and compared to establish a quantitative structure-activity relationship (QSAR) framework examining a small dataset of concrete set retarders. The results reveal that carboxyl and hydroxyl groups act as electronically privileged anchoring sites, driving strong adsorption onto calcium-rich cement surfaces. The molecular surface area and adsorption energy calculated by DFT indicate that retarders with multiple polar functional groups and flexible backbones exhibit stronger interfacial interactions, which directly correlate with longer setting times. DFT analysis further reveals that electrons are depleted at the carboxyl and hydroxyl sites and accumulated at surface Ca atoms, providing atomic-level validation and a clear mechanistic link between molecular structure and retarding performance. This study highlights the importance of integrating structural fingerprints with DFT calculations to understand and screen for effective retarders under data-constrained conditions – a promising approach for rational design and targeted optimization of chemical admixtures for advanced cementitious materials.","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"40 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884054","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Mini-Mason 3D Printing Orthoconcrete: Preparation, Properties, and Microstructure","authors":"Weiguo Shen, Jinyue Wang, Gelong Xu, Jiangtao Sun, Bo Tian, Zhitang Li, Zhe He, Haftamu Brhanu Amare, Michaele Alem Gebreyohannes, Guiming Wang","doi":"10.1016/j.jobe.2026.117261","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117261","url":null,"abstract":"A novel Mini-Mason 3D printing concrete method is developed, representing an innovative approach to casting orthoconcrete. This method involves the layer-by-layer pouring of fluid mortar and coarse aggregate. The present study investigates the optimum volume fraction of coarse aggregate in fresh mortar and its influence on the mechanical and durability properties of hardened Mini-Mason concrete. The mechanism of this method is analyzed using microhardness testing, BSE imaging, and pore structure characterization. The findings indicate that the fluidity of the mortar is the primary parameter that governs the preparation of the concrete mixture. The mortar, with a fluidity of 400 mm, allows the maximum coarse aggregate volume fraction of 53.3% when using well-graded aggregate. Increasing the aggregate fraction enhances compressive strength, penetration resistance, and volumetric stability. The Mini-Mason method produces orthoconcrete, characterized by a uniform distribution of aggregate, thinner and harder interfacial transition zones (ITZs), and improved compactness due to reduced total porosity and fewer harmful pores.","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"43 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884055","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Investigation of UHPC fracture performance with different steel fiber fractions based on acoustic emission","authors":"Changdi Li, Huigang Xiao, Zhenhao Mao, Xin Wang, Min Liu, Qiang Gui, Li Chen","doi":"10.1016/j.jobe.2026.117194","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117194","url":null,"abstract":"Three-point bending tests on notched beams were performed to investigate the effect of steel fiber volume fraction &lt;ce:italic&gt;V&lt;/ce:italic&gt;&lt;ce:inf loc=\"post\"&gt;&lt;ce:italic&gt;f&lt;/ce:italic&gt;&lt;/ce:inf&gt; (1.0%, 1.5%, 2.0%, and 2.5%) on the fracture properties of ultra-high performance concrete (UHPC). The initial cracking fracture toughness (&lt;ce:math altimg=\"si1.svg\"&gt;&lt;/ce:math&gt;) and unstable fracture toughness (&lt;ce:math altimg=\"si2.svg\"&gt;&lt;/ce:math&gt;) of UHPC were calculated using double-K fracture model (DKFM). The influence of &lt;ce:italic&gt;V&lt;/ce:italic&gt;&lt;ce:inf loc=\"post\"&gt;&lt;ce:italic&gt;f&lt;/ce:italic&gt;&lt;/ce:inf&gt; on &lt;ce:math altimg=\"si1.svg\"&gt;&lt;/ce:math&gt;, &lt;ce:math altimg=\"si2.svg\"&gt;&lt;/ce:math&gt; and the fracture energy &lt;ce:italic&gt;G&lt;/ce:italic&gt;&lt;ce:inf loc=\"post\"&gt;&lt;ce:italic&gt;F&lt;/ce:italic&gt;&lt;/ce:inf&gt; were analyzed. Acoustic emission (AE) was used to characterize the microstructural evolution and damage development during bending and to capture the crack-propagation patterns of notched UHPC specimens with different &lt;ce:italic&gt;V&lt;/ce:italic&gt;&lt;ce:inf loc=\"post\"&gt;&lt;ce:italic&gt;f&lt;/ce:italic&gt;&lt;/ce:inf&gt; values. Within the investigated &lt;ce:italic&gt;V&lt;/ce:italic&gt;&lt;ce:inf loc=\"post\"&gt;&lt;ce:italic&gt;f&lt;/ce:italic&gt;&lt;/ce:inf&gt; ranging from 1.0% to 2.5%, the fracture toughness and &lt;ce:italic&gt;G&lt;/ce:italic&gt;&lt;ce:inf loc=\"post\"&gt;&lt;ce:italic&gt;F&lt;/ce:italic&gt;&lt;/ce:inf&gt; increased with increasing fiber content, because of the crack resistance and bridging effects of steel fibers. As &lt;ce:italic&gt;V&lt;/ce:italic&gt;&lt;ce:inf loc=\"post\"&gt;&lt;ce:italic&gt;f&lt;/ce:italic&gt;&lt;/ce:inf&gt; increased from 1.0% to 2.5%, &lt;ce:math altimg=\"si1.svg\"&gt;&lt;/ce:math&gt;, &lt;ce:math altimg=\"si2.svg\"&gt;&lt;/ce:math&gt;, and &lt;ce:italic&gt;G&lt;/ce:italic&gt;&lt;ce:inf loc=\"post\"&gt;&lt;ce:italic&gt;F&lt;/ce:italic&gt;&lt;/ce:inf&gt; increased by 78.81%, 122.13% and 77.52%, respectively, showing that steel fibers could improve the UHPC fracture performance. Meanwhile, the AE parameters provided useful information for interpreting microcrack initiation and propagation during UHPC fracture. The AE ringing count, cumulative energy, and b-value could be used to identify the three typical stages in UHPC fracture tests. The b-value served as an indicator for microcrack initiation and propagation, whereas the cumulative ringing count and accumulated energy reflected the toughening effect of steel fibers. A clustering algorithm was adopted to classify the peak frequency into three frequency bands. The low (0–100 kHz), medium (100-220kHz), and high frequency signals (220-400 kHz) were mainly associated with matrix cracking, fiber pullout, and fiber-matrix debonding, respectively. The Kneedle algorithm was utilized to determine the average frequency/rising angle for analyzing the proportion of tensile and shear cracks. AE source localization was further used to quantify the fracture process zone width, which increased from 22 to 32 mm as &lt;ce:italic&gt;V&lt;/ce:italic&gt;&lt;ce:inf loc=\"post\"&gt;&lt;ce:italic&gt;f&lt;/ce:italic&gt;&lt;/ce:inf&gt; increased from 1.0% to 2.5%. AE signals exhibited good consistency with the time-dependent development of mechanical pr","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"23 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884089","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Data-driven energy performance analysis and prediction of VRF systems in a block-style commercial building based on long-term operational data","authors":"Guodan Liu, Yafan Yang, Debao Li, Xiaojie Zhou, Ning Wang, Jixin Zou, Guiting Liu, Xucheng Zhuang, Songtao Hu","doi":"10.1016/j.jobe.2026.117230","DOIUrl":"https://doi.org/10.1016/j.jobe.2026.117230","url":null,"abstract":"Energy consumption forecasting plays an important role in building energy management. Although variable refrigerant flow (VRF) systems have been widely studied in office and residential buildings, research on their energy consumption characteristics and prediction methods in block-style commercial buildings remains limited because of diverse tenant functions and heterogeneous operating patterns. In this study, the energy consumption characteristics of a VRF system in a block-style commercial building located in Nanning, China, were investigated using long-term monitored operational data. The results indicate that the annual VRF energy consumption intensity was 41.57 kWh/(m<ce:sup loc=\"post\">2</ce:sup>·a), of which 83.8% was consumed during the cooling season. A segmented multiple linear regression (MLR) model was developed to predict VRF energy consumption under heating and cooling conditions. The model achieved a coefficient of determination (R<ce:sup loc=\"post\">2</ce:sup>) of 0.9464, a root mean square error (RMSE) of 0.0206, and a coefficient of variation of RMSE (CV-RMSE) of 16.55% for the test dataset. Sensitivity analysis based on standardized regression coefficients (SRC) revealed that indoor unit operation hours (T<ce:inf loc=\"post\">IU</ce:inf>) was the most influential factor affecting cooling energy consumption, followed by the indoor–outdoor temperature difference (t<ce:inf loc=\"post\">DBT</ce:inf> - t<ce:inf loc=\"post\">re</ce:inf>) and the thermostat setpoint temperature (t<ce:inf loc=\"post\">set</ce:inf>). Furthermore, the predictive performance of several data-driven models, including MLR, random forest (RF), back propagation neural network (BPNN), and long short-term memory (LSTM), was compared for different commercial store types. Among these models, MLR demonstrated reliable predictive performance. For the clothing store subsystem, the model achieved an R<ce:sup loc=\"post\">2</ce:sup> of 0.908, an RMSE of 6.225, and a CV-RMSE of 6.0%. The findings suggest that simple and interpretable data-driven models based on operational data can provide accurate predictions of VRF energy consumption and support practical energy management in block-style commercial buildings.","PeriodicalId":15064,"journal":{"name":"Journal of building engineering","volume":"57 1","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884060","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}