Techno Press
Tp_Editing System.E (TES.E)
Login Search
You logged in as...

cac
 
CONTENTS
Volume 35, Number 6, June 2025
 

Abstract
Strain-hardening ultra-high-performance concrete (SHUHPC) is prized for its exceptional ductility and strength, offering diverse applications. Understanding the intricate relationships between fiber constituents, geometric parameters, and the matrix has been challenging. This study employs the Connection Weight Approach (CWA) with two calibrated Artificial Neural Network (ANN) models to elucidate these complexities. This study presents a novel approach by integrating ANN and CWA techniques to explore the complex, non-linear interactions of SHUHPC that enhance the understanding of its mechanical properties and offer insights beyond previous research. The aim is to predict SHUHPC's energy absorption capacity (g) and strain at peak stress (epc) under direct tensile stress. Findings reveal that a 15% increase in the fiber reinforcement index enhances energy absorption, yet excessive levels limit epc. Determined through rigorous testing, optimal mixes include an 8% silica fume dosage, resulting in a notable 12% increase in compressive strength. Deformed steel fibers, particularly twisted variants, significantly boost energy absorption metrics by 18%. Fiber-matrix interactions play a pivotal role in achieving these results. This study clarifies ANN model predictions' ambiguity, offering actionable insights driven by data. These findings advance SHUHPC understanding and propose strategies for its optimized applications.

Key Words
ANN; connection weight approach; energy capacity absorption; strain at peak tensile stress; strain hardening UHPC; uniaxial tensile behavior

Address
Joaquin Abellan-Garcia: Department of Civil and Environmental Engineering, Universidad Del Norte, Barranquilla, Colombia
Yassir M. Abbas and M. Iqbal Khan: Department of Civil Engineering, College of Engineering, King Saud University, Riyadh 12372, Saudi Arabia
Vicente Martínez-Lirón: Universidad Católica de Murcia (UCAM), Av. de los Jerónimos, 135, 30107 Guadalupe de Maciascoque, Murcia, Spain

Abstract
Exploring new friendly filler applicable for self-compacting concrete/mortar is always encouraged to persuade a phenomenon of sustainable development. The current study aims at assessing synergistic influences of limestone powder (LP) and unground rice husk ash (URHA) additions respectively as partial replacements of slag and fine aggregate (FA) in ranges of 10 to 50 mass.% and 10 to 50 vol.% on comprehensive performances of a slag-cement self-compacting mortar (SCM). Experimental results indicated that an increment of either LP or URHA amount remarkably increased viscosity of the fresh SCMs. Sole addition of LP partially replacing slag at 10 mass.% led to the hardened SCMs with the enhanced flexural and compressive strengths, strength efficiency (SE) of cement, and durability performances except a slight increase in drying shrinkage. By using LP as partial substitution of slag at 10 mass.%, addition of URHA partially replacing FA at 10 vol.% substantially improved both the mechanical strengths and durability in terms of ultrasonic wave velocity (UPV) and sulfate resistance of the modified SCMs, which was attributed to the additional hydration products detected from X-Ray diffraction technique. Compared with the reference SCM without additive, the modified SCM containing mixture of LP and URHA respectively replacing slag and FA at 10 mass.% and 10 vol.% had compressive strength increased at 15.3%, flexural strength increased at 27.6%, UPV increased at 3.8% at 28 days. Especially, in this study, hybrid addition of LP and URHA respectively replacing slag and FA at 10 mass.% and 10 vol.% was also considered as the optimum amounts for maximizing the SE of cement at up to 0.252 MPa/(kg/m3) in SCMs.

Key Words
durability; engineering properties; limestone powder; self-compacting mortar; unground rice husk ash

Address
Faculty of Civil Engineering, College of Engineering, Can Tho University, Can Tho City 900000, Vietnam

Abstract
The steel-UHPC composite system has recently used in some bridges due to its light weight and long-term durability compared with steel-concrete composite system. However, the cold joint is required in many applications due to construction or design requirements. The cold joint at the hogging moment area is typically subjected to high tensile stress that may create cracks and therefore the cold joint shape is the key factor that influences the performance. The research numerically investigated the performance of the UHPC composite system using finite element simulation. The investigated parameters were cold joint shapes, reinforcement ratio and UHPC thickness. The FE results agreed very well with the filed results in load-deflections, load-strains, load-slips, and failure modes. It was found that the cold joint shape had negligible effect on the ultimate load with the differences within 4% among the investigated joints and the FE results confirmed that using T shaped joint had better performance. Meanwhile, the reinforcement ratio played a significant role on shifting the crack from the interface to the mid span. This would be beneficial as the fiber is typically less at the interface and as results the crack propagation will be faster at the interface. As the reinforcement ratio increases from 0.98% to 3.91%, the ultimate load increases up to 12% for the joints. The UHPC slab thickness of 55 mm is sufficient as in increase the thickness to 75 mm had negligible effects on the load capacity with a maximum increase of only 1.5%.

Key Words
cold joint; finite element; hogging-moment regions; steel-UHPC composite bridge; UHPC

Address
Yanping Zhu: Department of Civil Engineering, Montana Technological University, Butte, MT, United States
Abdul Ridah Saleh: College of Engineering, University of Babylon, Babil, Iraq
Majid M.A. Kadhim: College of Engineering, University of Babylon, Hilla, Iraq
Yang Zhang: Key Laboratory for Wind and Bridge Engineering of Hunan Province, College of Civil Engineering, Hunan University, Changsha, 410082, China
Ali A. Semendary: Department of Civil Engineering, University of Manitoba, Winnipeg, MB, R3T 5V6, Canada

Abstract
The inclusion of fibers in concrete mix significantly changes the functionality and strength properties of hardened concrete. However, it is quite challenging to determine the specific contribution of fiber in concrete particularly, because of their non-uniform dispersion, and wide variation of material properties compared to that of the actual concrete ingredients. This study investigates the complex domain of predictive modeling for compressive strength of fiberized concrete (FBC), utilizing data-driven fourteen machine learning (ML) techniques. A comprehensive dataset comprising 608 test results of fiberized concrete properties is meticulously collected, organized, and utilized for the training and evaluation of ML models. The input features of the proposed ML models are water-to-cement ratio (W/C), coarse aggregate-cement ratio (CA/C), fine aggregate-cement ratio (FA/C), admixture utilization, percentage of fiber, types of fiber (nine different types), fiber aspect ratio (l/d), and fiber tensile strength (MPa). This study employs a diverse range of fourteen different ML algorithms, including Linear Regression (LR), Ridge Regression (RR), Lasso Regression (Lasso), Decision Tree (DT), Extra Trees Regression (ET), Random Forest (RF), Support Vector Regression (SVR), Artificial Neural Network (ANN), K-Nearest Neighbors (KNN), Adaboost (AB), Catboost (CB), Gradient Boost (GB), Extreme Gradient Boosting (XGB), and Light Gradient Boosting Machine (LGBM). The findings indicate that the XGB model demonstrates the best level of accuracy in predicting the compressive strength of FBC by achieving over 70% of the test data points with minimal error. The feature importance and SHAP value revealed that apart from the W/C ratio, CA/C, and FA/C, the fiber category and its tensile strength (MPa) were identified as crucial parameters that have a substantial effect on the compressive strength of FBC. The analysis also claimed the presence of fiber with a high tensile strength (11%) is important for improving the compressive strength rather than its higher volume percentage.

Key Words
compressive strength; fiber strength; fiber; machine learning; XG

Address
Tanvir H Tusher, Khondaker S Ahmed: Department of Civil Engineering, Military Institute of Science and Technology (MIST), Dhaka, Bangladesh
Avijit Pal and Nur Yazdani: Department of Civil Engineering, The University of Texas at Arlington, Arlington, Texas, USA
Md. Shahjalal: Department of Civil Engineering, University of Calgary, Calgary, Alberta, Canada

Abstract
Analyzing the mechanical behavior of concrete structures under elevated temperatures is critical for ensuring fire safety, structural integrity, and damage detection. Geopolymer concrete (GPC), a sustainable alternative to Portland cement concrete, is known for its superior thermal resistance. However, accurately predicting its compressive strength after exposure to high temperatures, ranging from 25 oc to 1100 oc, remains a challenge due to the complex interactions of material properties under thermal stress. In this study, machine learning (ML) algorithms are employed to forecast the compressive strength of GPC using a comprehensive dataset of 332 experimental data points gathered from an extensive literature review. Six different ML models—Artificial Neural Networks (ANN), Support Vector Machines (SVR), Gradient Boosting (GBoost), Random Forest, XGBoost, and LightGBM—were trained and evaluated based on their performance. The results indicate that GBoost and LightGBM models outperformed others, delivering the most accurate predictions with the lowest errors. These findings highlight the effectiveness of ML models in predicting the residual compressive strength of GPC after high-temperature exposure, offering valuable insights for fire-resistant construction applications.

Key Words
elevated temperature; geopolymer concrete; machine learning; residual compressive strength

Address
Ahmet Emin Kurtoglu: Department of Civil Engineering, Engineering Faculty, Igdir University, Igdir, 76000, Türkiye
Muhammed Kaya: Department of Computer Engineering, Engineering Faculty, Igdir University, Igdir, 76000, Türkiye
Necip Altay Eren: Department of Construction, Technical Vocational School, Gaziantep University, Gaziantep, 27310, Türkiye

Abstract
In seismically active regions, accuracy in predicting the behavior of buildings with reinforced concrete walls is crucial. The demand for reliable simulations has driven the development of two nonlinear analysis approaches: force-controlled and displacement-controlled. Traditional force-controlled iterative algorithms, when faced with complex nonlinear load-displacement patterns, often have convergence problems, especially near critical points such as post-peak instabilities or during the softening phase of the response. In such cases, the use of the unidirectional displacement control algorithm proposed by Batoz and Dhatt (1979) has proven to be more effective. However, the need to understand the behavior of structural walls with irregular cross sections has increased experimental studies on this type of specimen subjected to multidirectional cyclic loads, especially in countries such as Spain, Switzerland, Germany, and Japan. Despite the efforts, reproducing such complex behaviors in numerical models is still a challenge with the algorithms currently available, which is why this work introduces a new formulation and convergence strategy that extends the traditional unidirectional displacement control algorithm to a multidirectional approach. The key innovation of this work lies in its ability to perform simultaneous displacement control over multiple degrees of freedom, enabling accurate nonlinear analysis of more complex demand patterns. The proposed algorithm provides a robust solution for simulating advanced nonlinear behavior, addressing a significant gap in current numerical modeling practices. This extension has been rigorously validated by analyzing reinforced concrete walls with irregular cross-sections, demonstrating its superior performance under multidirectional loading.

Key Words
displacement-control; multidirectional displacements; multidirectional loads; non-linear solvers

Address
Fabián R. Rojas and Leonardo M. Massone: Department of Civil Engineering, University of Chile, Chile
Betzabeth J. Suquillo: School of Civil Engineering, Pontificia Universidad Católica de Valparaíso, Chile

Abstract
This research investigates the possible outcomes of partially replacing cement in concrete with marble sludge powder (MSP). Both freshly mixed concrete and concrete that had been given time to cure were used in studies to look at various fiber-reinforced (FR) concrete qualities. Two water-cement ratios, 0.35 and 0.40, were used to get the desired results. Marble sludge powder and 0.5% polypropylene 3S fiber were substituted at percentages of 0%, 5%, 10%, 15%, 20%, and 25%. To assess the samples' quality, a series of mechanical tests were conducted on them after 7, 14, 28, and 56 days of a cure. A split tensile and flexural strength prediction model was developed using an artificial neural network (ANN). Plotting the experimentally recorded split tensile and flexural strengths versus the statistical analysis of regression strength was done after 56 days for the artificial neural network. The results of the studies suggest that employing powder generated from leftover marble might have a positive economic impact and reduce the environmental harm that concrete causes. By substituting up to 15% of the typical aggregate with dry marble sludge powder, the split tensile and flexural strengths increased to 6.42 MPa and 7.98 MPa respectively.

Key Words
fiber-reinforced concrete; machine learning technology (MLT); marble sludge powder; polypropylene 3s

Address
Meena Devi G: Department of Mathematics, St. Joseph's College of Engineering, Chennai, Tamilnadu 600119, India
Sumant Nivarutti Shinde: Department of Civil Engineering, MIT World Peace University, Pune, Maharashtra 411038, India
Leo Amalraj J: Department of Mathematics, R.M.K. College of Engineering and Technology, Thiruvallur, Tamilnadu 601206, India
Samson Isaac: Department of Biomedical Engineering, Karunya Institute of Technology and Sciences, Coimbatore, Tamilnadu 641114, India
Arti Chouksey: Department of Civil Engineering, Deenbandhu Chhotu Ram University of Science and Technology, Murthal, Haryana 131039, India
Zunaithur Rahman D: Department of Civil Engineering, Aalim Muhammed Salegh College of Engineering, Chennai, Tamilnadu 600055, India

Abstract
Alkali-activated materials (AAMs), featured by its cementless characteristic, received broad acceptance from numerous researchers these years. However, it still remains challenging to manufacture AAM mixtures with high mechanical performance and low drying shrinkage. A significant cause of the difficulty in controlling the mechanical properties of AAMs is the absence of an appropriate model for predicting compressive strength and drying shrinkage. In this study, a comparison study was conducted to predict the compressive strength and drying shrinkage of AAMs using an extreme gradient boosting (XGBoost) and light gradient boosting machine (LightGBM) based on the key factors of the AAM mix design. Supervised machine learning algorithms were employed for the XGBoost and LightGBM methods to process the datasets through training, validation, modeling, and testing. In addition, the K-fold cross-validation method was adopted for the training dataset, in which the K values ranged from 2 to 10. The results showed that the R2 values of the best performed XGBoost and LightGBM models for strength prediction were 0.8787 and 0.8583, while for the drying shrinkage prediction R2 values of the XGBoost and LightGBM methods were 0.9906 and 0.8989, respectively. The above four models were all obtained with 10-fold cross validation. Moreover, for proving the applicability of the proposed models in the real construction work, a validation experiment for the proposed models was carried out in laboratory, and the model was able to estimate with an error of 15%. This research largely helps the decision-makers to properly use the XGBoost and LightGBM algorithms.

Key Words
alkali-activated materials; compressive strength; drying shrinkage; LightGBM; machine learning; XGBoost

Address
Y.K. Kong: 1) Division of Sustainable Resources Engineering, Graduate School of Engineering, Hokkaido University, Japan,
2) Faculty of Science and Engineering, Waseda University, Japan
Kiyofumi Kurumisawa: Division of Sustainable Resources Engineering, Faculty of Engineering, Hokkaido University, Japan

Techno-Press: Publishers of international journals and conference proceedings.       Copyright © 2025 Techno-Press ALL RIGHTS RESERVED.
P.O. Box 33, Yuseong, Daejeon 34186 Korea, Email: admin@techno-press.com