Asas Journal

Asas Journal

Evaluation of artificial intelligence methods based on metaheuristic algorithms for estimating the compressive strength of lightweight structural concrete containing leca

Document Type : Original Article

Authors
Ali Ashrafian 1
Mohammad Javad Taheri Amiri 2
Niloofar Maleki 3
Milad Jamali 4
1 M.Sc, Tabari university, Babol, Iran
2 Pardisan university, Fereidoon kenar, Iran
3 M.Sc student, Pardisan university, Freidoonkenar, iran
4 phd candidate, Babol Nooshirvani University
Abstract
Compressive strength (CS) is an important mechanical characteristic of concrete, which is highly perceptible from concrete mix design. The main objective of this study is investigate metaheuristic algorithms, combined with artificial intelligence methods (hybrid models), so as to provide relationships to estimate compressive strength of lightweight expanded clay aggregate (LECA) concrete, by comparing other Artificial inteligence models (AI models), assessing effective parameters and model accuracy and finally i validatt the result of the developed models. In this study, 405 laboratory samples in their 7, 28 and 90 days old age were used. Presented models in this study were expanded using artificial neural network (ANN), multivariate adaptive regression splines (MARS) and improved MARS with crow search algorithm (MARS-CSA). Statistical indicator named R, RMSE and MAE investigated results of the developed models. In testing performance, predicted compressive strength values of MARS-CSA (R= 0.913, RMSE= 39.002) indicated significant accuracy in comparison with MARS and ANN models.
Keywords
ARTIFICIAL INTELLIGENCE METHOD
MARS
ARTIFICIAL NEURAL NETWORK
CROW SEARCH ALGORITHM
LECA CONCRETE
Asas Journal
Volume 22, Issue 58
Spring 2020
Spring 2020
Pages 19-34

  • Receive Date 07 September 2018
  • Revise Date 05 August 2019
  • Accept Date 13 February 2021