Durable aggregates are essential for the stability and longevity of construction projects,and the Los Angeles Abrasion(LAA)value is a widely used indicator of aggregate durability.However,direct LAA testing is time-co...Durable aggregates are essential for the stability and longevity of construction projects,and the Los Angeles Abrasion(LAA)value is a widely used indicator of aggregate durability.However,direct LAA testing is time-consuming,costly,and requires specialized facilities.This study introduces a novel and efficient alternative by developing two hybrid machine learning models Artificial Neural Network integrated with Particle Swarm Optimization(ANN-PSO)and Artificial Neural Network combined with Teaching-Learning-Based Optimization(ANN-TLBO)for the first time to predict LAA values from petrographic characteristics of carbonate rocks.A total of 160 rock samples from 10 geological formations in Pakistan’s Salt Range were analyzed through petrographic examination and LAA testing.The predictive performance of the proposed models was compared with established techniques,including Multiple Linear Regression,Random Forest,Adaptive Boosting,Gradient Boosting,K-Nearest Neighbors,and Multilayer Perceptron.Results demonstrate that ANN-PSO significantly outperformed all other approaches,achieving an R^(2)of 0.9982,RMSE of 0.209,MAE of 0.159,and MAPE of 0.009.Model robustness was validated using Taylor plots,REC curves,and an external dataset.Sensitivity analysis identified quartz,calcite,and feldspar as the most influential factors affecting LAA prediction.The findings confirm that the ANN-PSO and ANN-TLBO models provide highly accurate,cost-effective,and practical alternatives to traditional LAA testing.This innovative approach advances rock mechanics and material assessment,offering engineers and geologists enhanced tools for aggregate selection and durability evaluation in infrastructure projects.展开更多
文摘Durable aggregates are essential for the stability and longevity of construction projects,and the Los Angeles Abrasion(LAA)value is a widely used indicator of aggregate durability.However,direct LAA testing is time-consuming,costly,and requires specialized facilities.This study introduces a novel and efficient alternative by developing two hybrid machine learning models Artificial Neural Network integrated with Particle Swarm Optimization(ANN-PSO)and Artificial Neural Network combined with Teaching-Learning-Based Optimization(ANN-TLBO)for the first time to predict LAA values from petrographic characteristics of carbonate rocks.A total of 160 rock samples from 10 geological formations in Pakistan’s Salt Range were analyzed through petrographic examination and LAA testing.The predictive performance of the proposed models was compared with established techniques,including Multiple Linear Regression,Random Forest,Adaptive Boosting,Gradient Boosting,K-Nearest Neighbors,and Multilayer Perceptron.Results demonstrate that ANN-PSO significantly outperformed all other approaches,achieving an R^(2)of 0.9982,RMSE of 0.209,MAE of 0.159,and MAPE of 0.009.Model robustness was validated using Taylor plots,REC curves,and an external dataset.Sensitivity analysis identified quartz,calcite,and feldspar as the most influential factors affecting LAA prediction.The findings confirm that the ANN-PSO and ANN-TLBO models provide highly accurate,cost-effective,and practical alternatives to traditional LAA testing.This innovative approach advances rock mechanics and material assessment,offering engineers and geologists enhanced tools for aggregate selection and durability evaluation in infrastructure projects.
