Plant-Inspired Metaheuristic Optimization of Extreme Learning Machines for Porosity Prediction in High-Performance Concrete

Accurate estimation of the porosity of high-performance concrete is crucial for evaluating its long-term performance and mechanical strength under diverse environmental conditions. To achieve this accuracy, this study employed and analyzed five hybrid extreme learning machine (ELM) models using artificial root foraging (ARFA), flower pollination (FPA), invasive weed optimization (IWO), phototropic optimization (POA), and waterwheel plant (WWPA) algorithms. Each model was trained (TR) and tested (TS) on 80% and 20% of 240 high-performance concrete specimens. The water-to-binder ratio (WB), aggregate ratio (AR), binder content (B), and fly ash (FA) were determined as the most significant features using the cosine amplitude sensitivity method. Later, the variance inflation factor revealed that these features have weak-to-considerable multicollinearity in estimating the porosity of concrete. The root mean square error ( RMSE = 0.4422 / 0.4730 for TR/TS), mean squared logarithmic error ( MSLE = 0.0024 / 0.0028 for TR/TS), and performance ( R = 0.9913 / 0.9885 for TR/TS) metrics identified the WWPA_ELM model as the most accurate model. In addition, the prediction capabilities of models were visualized using the Taylor plot and the regression error characteristics (REC) curve. Moreover, the Anderson–Darling test analyzed the normality hypothesis for the predicted porosity using each model. Finally, the curve-fitting issue was analyzed using a feature multicollinearity–sensitivity relationship plot. The ARFA_ELM, FPA_ELM, POA_ELM, and WWPA_ELM models overfitted due to the considerable multicollinearity of the WB feature, which made a significant contribution to the accurate assessment of porosity.

Authors

Institutions

Publication Details

Journal
Journal of structural design and construction practice.
Published
2026-11-18
DOI
https://doi.org/10.1061/jsdccc.sceng-2222
Primary Topic
Machine Learning and ELM
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Plant-Inspired Metaheuristic Optimization of Extreme Learning Machines for Porosity Prediction in High-Performance Concrete

Asma Yahiaoui, Jitendra Khatti
1 links
Journal of structural design and construction practice.
Machine Learning and ELM
article

Plant-Inspired Metaheuristic Optimization of Extreme Learning Machines for Porosity Prediction in High-Performance Concrete

Asma Yahiaoui, Jitendra Khatti
article en

Abstract

Accurate estimation of the porosity of high-performance concrete is crucial for evaluating its long-term performance and mechanical strength under diverse environmental conditions. To achieve this accuracy, this study employed and analyzed five hybrid extreme learning machine (ELM) models using artificial root foraging (ARFA), flower pollination (FPA), invasive weed optimization (IWO), phototropic optimization (POA), and waterwheel plant (WWPA) algorithms. Each model was trained (TR) and tested (TS) on 80% and 20% of 240 high-performance concrete specimens. The water-to-binder ratio (WB), aggregate ratio (AR), binder content (B), and fly ash (FA) were determined as the most significant features using the cosine amplitude sensitivity method. Later, the variance inflation factor revealed that these features have weak-to-considerable multicollinearity in estimating the porosity of concrete. The root mean square error ( RMSE = 0.4422 / 0.4730 for TR/TS), mean squared logarithmic error ( MSLE = 0.0024 / 0.0028 for TR/TS), and performance ( R = 0.9913 / 0.9885 for TR/TS) metrics identified the WWPA_ELM model as the most accurate model. In addition, the prediction capabilities of models were visualized using the Taylor plot and the regression error characteristics (REC) curve. Moreover, the Anderson–Darling test analyzed the normality hypothesis for the predicted porosity using each model. Finally, the curve-fitting issue was analyzed using a feature multicollinearity–sensitivity relationship plot. The ARFA_ELM, FPA_ELM, POA_ELM, and WWPA_ELM models overfitted due to the considerable multicollinearity of the WB feature, which made a significant contribution to the accurate assessment of porosity.

Journal of structural design and construction practice.Vol. 32(2)
University of Sciences and Technology Houari Boumediene (DZ), Columbia University (US)
Machine Learning and ELM
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon