Optimizing pervious concrete with machine learning: Predicting permeability and compressive strength using artificial neural networks

Construction and Building Materials, Год журнала: 2024, Номер 443, С. 137619 - 137619

Опубликована: Авг. 7, 2024

This study makes a significant contribution to the field of pervious concrete by using machine learning innovatively predict both mechanical and hydraulic performance. Unlike existing methods that rely on labor-intensive trial-and-error experiments, our proposed approach leverages multilayer perceptron network. To develop this approach, we compiled comprehensive dataset comprising 271 sets 3,252 experimental data points. Our methodology involved evaluating 22,246 network configurations, employing Monte Carlo cross-validation over 20 iterations, 4 training algorithms, resulting in total 1,779,680 iterations. results an optimized model integrates diverse mix design parameters, enabling accurate predictions permeability compressive strength even absence data, achieving R² values 0.97 0.98, respectively. Sensitivity analyses validate model's alignment with established principles behavior. By demonstrating efficacy as complementary tool for optimizing designs, research not only addresses current methodological limitations but also lays groundwork more efficient effective approaches field.

Язык: Английский

---

The ultimate capacity of geopolymer recycled aggregate concrete filled steel tubular columns: Numerical and theoretical study

Journal of Building Engineering, Год журнала: 2024, Номер 96, С. 110365 - 110365

Опубликована: Авг. 10, 2024

Язык: Английский

---

Utilizing Construction and Demolition Waste in Concrete as a Sustainable Cement Substitute: A Comprehensive Study on Behavior Under Short-term Dynamic and Static Loads via Laboratory and Numerical Analysis

Journal of Building Engineering, Год журнала: 2024, Номер unknown, С. 110778 - 110778

Опубликована: Сен. 1, 2024

Язык: Английский

---

Predicting High-Strength Concrete’s Compressive Strength: A Comparative Study of Artificial Neural Networks, Adaptive Neuro-Fuzzy Inference System, and Response Surface Methodology

Materials, Год журнала: 2024, Номер 17(18), С. 4533 - 4533

Опубликована: Сен. 15, 2024

Machine learning and response surface methods for predicting the compressive strength of high-strength concrete have not been adequately compared. Therefore, this research aimed to predict (HSC) using different methods. To achieve purpose, neuro-fuzzy inference systems (ANFISs), artificial neural networks (ANNs), methodology (RSM) were used as ensemble Using an ANN ANFIS, output was modeled optimized a function five independent variables. The RSM designed with three input variables: cement, fine coarse aggregate. facilitate data entry into Design Expert, model divided six groups, p-values responses 1 6 0.027, 0.010, 0.003, 0.023, 0.002, 0.026. following metrics evaluate projection: R, R2, MSE ANFIS modeling; Adj. Pred. R2 modeling. Based on data, it can be concluded that (R = 0.999, 0.998, 0.417), 0.981 0.963), 0.962, 0.926, 0.655) good chance accurately (HSC). Furthermore, there is strong correlation between ANN, RSM, models experimental data. Nevertheless, network demonstrates exceptional accuracy. sensitivity analysis shows cement aggregate most significant effect (45.29% 35.87%, respectively), while superplasticizer has least (0.227%). RSME values in 0.313 0.453 during test process 0.733 0.563 training process. Thus, found both presented better results higher accuracy construction materials.

Язык: Английский

---

Artificial neural network, machine learning modelling of compressive strength of recycled coarse aggregate based self-compacting concrete

PLoS ONE, Год журнала: 2024, Номер 19(5), С. e0303101 - e0303101

Опубликована: Май 13, 2024

This research study aims to understand the application of Artificial Neural Networks (ANNs) forecast Self-Compacting Recycled Coarse Aggregate Concrete (SCRCAC) compressive strength. From different literature, 602 available data sets from SCRCAC mix designs are collected, and rearranged, reconstructed, trained tested for ANN model development. The models were established using seven input variables: mass cementitious content, water, natural coarse aggregate fine recycled chemical admixture mineral used in designs. Two normalization techniques visualize distribution. For each technique, three transfer functions modelling. In total, six types run MATLAB estimate 28 th day Normalization technique 2 performs better than 1 TANSING is best function. k-fold cross-validation fold k = 7. coefficient determination predicted actual strength 0.78 training 0.86 testing. impact number neurons layers on was performed. Inputs standards Apart ANN, Machine Learning (ML) like random forest, extra trees, extreme boosting light gradient adopted predict SCRCAC. Compared ML, prediction shows results terms sensitive analysis. also extended determine experimental work compared it with model. Standard have similar fresh hardened properties. average 39.067 38.36 MPa, respectively correlation 1. It appears that can validly concrete.

Язык: Английский

---

Analyzing the efficacy of waste marble and glass powder for the compressive strength of self-compacting concrete using machine learning strategies

REVIEWS ON ADVANCED MATERIALS SCIENCE, Год журнала: 2024, Номер 63(1)

Опубликована: Янв. 1, 2024

Abstract Self-compacting concrete (SCC) is well-known for its capacity to flow under own weight, which eliminates the need mechanical vibration and provides benefits such as less labor faster construction time. Nevertheless, increased cement content of SCC results in an increase both costs carbon emissions. These challenges are resolved this research by utilizing waste marble glass powder substitutes. The main objective study create machine learning models that can predict compressive strength (CS) using gene expression programming (GEP) multi-expression (MEP) produce mathematical equations capture correlations between variables. models’ performance assessed statistical metrics, hyperparameter optimization conducted on experimental dataset consisting eight independent indicate MEP model outperforms GEP model, with R 2 value 0.94 compared 0.90. Moreover, sensitivity SHapley Additive exPlanations analysis revealed most significant factor influencing CS curing time, followed slump quantity. A sustainable approach design presented study, improves efficacy minimizes testing.

Язык: Английский

---

Prediction of fresh and hardened properties of self-compacting concrete using ensemble soft learning techniques

Multiscale and Multidisciplinary Modeling Experiments and Design, Год журнала: 2024, Номер 7(5), С. 4923 - 4945

Опубликована: Апрель 9, 2024

Язык: Английский

---

Predictive Modeling for Mechanical Characteristics of Ultra High-Performance Concrete Blended with eggshell powder and nano silica Utilizing Traditional Technique and Machine Learning Algorithm

Case Studies in Construction Materials, Год журнала: 2024, Номер unknown, С. e04025 - e04025

Опубликована: Ноя. 1, 2024

Язык: Английский

---

The Prediction of Pervious Concrete Compressive Strength Based on a Convolutional Neural Network

Buildings, Год журнала: 2024, Номер 14(4), С. 907 - 907

Опубликована: Март 27, 2024

To overcome limitations inherent in existing mechanical performance prediction models for pervious concrete, including material constraints, limited applicability, and inadequate accuracy, this study employs a deep learning approach to construct Convolutional Neural Network (CNN) model with three convolutional modules. The primary objective of the is precisely predict 28-day compressive strength concrete. Eight input variables, encompassing coarse fine aggregate content, water admixture cement fly ash silica fume were selected model. dataset utilized both training testing consists 111 sample sets. ensure model’s coverage within practical range concrete enhance its robustness real-world applications, an additional 12 sets experimental data incorporated testing. research findings indicate that, comparison conventional machine method Backpropagation (BP) neural networks, developed CNN paper demonstrates higher coefficient determination, reaching 0.938, on test dataset. mean absolute percentage error 9.13%, signifying that proposed exhibits notable accuracy universality predicting regardless materials used preparation.

Язык: Английский

---

Prediction of flexural strength of concrete with eggshell and glass powders: Advanced cutting-edge approach for sustainable materials

REVIEWS ON ADVANCED MATERIALS SCIENCE, Год журнала: 2024, Номер 63(1)

Опубликована: Янв. 1, 2024

Abstract Currently, there is a lack of research comparing the efficacy machine learning and response surface methods in predicting flexural strength Concrete with Eggshell Glass Powders. This aims to predict simulate strengths concrete that replaces cement fine aggregate waste materials such as eggshell powder (ESP) glass (WGP). The methodology (RSM) artificial neural network (ANN) techniques are used. A dataset comprising previously published was used assess predictive generalization abilities ANN RSM. total 225 article samples were collected split into three subsets for model development: 70% training (157 samples), 15% validation (34 testing samples). seven independent variables improve model, whereas RSM (cement, WGP, ESP) model. k -fold cross-validation validated generalizability statistical metrics demonstrated favorable outcomes. Both effective instruments strength, according results, which include mean squared error, determination coefficient ( R 2 ), adjusted adj). able achieve an 0.7532 accuracy results 0.956 strength. Moreover, correlation between models experimental data high. However, exhibited superior accuracy.

Язык: Английский

---