Using Artificial Intelligence Methods to Predict the Compressive Strength of Concrete Containing Sugarcane Bagasse Ash

作     者：Pazouki, Gholamreza Kang, Won-Hee Tao, Zhong Saeed, Nariman 

作者机构：Centre for Infrastructure Engineering School of Engineering Design & Built Environment Western Sydney University PenrithNSW Australia School of Computing Mathematics and Engineering Faculty of Business Justice and Behavioural Sciences Charles Sturt University Bathurst Australia 

出 版 物：《SSRN》 

年 卷 期：2023年

核心收录：

主　　题：Compressive strength 

摘      要：Sugarcane bagasse ash is an agricultural and industrial waste material produced in millions of tonnes annually. While traditionally used as a fertilizer or buried underground, incorporating this material into concrete production not only reduces cement usage and addresses environmental concerns but also enhances the mechanical properties of concrete. This research aims to investigate the predictive capabilities of various Artificial Intelligence (AI) models, including radial basis function neural network (RBFNN), artificial neural network (ANN), and support vector regression (SVR), for estimating the compressive strength of concrete containing sugarcane bagasse ash (SCBA). A dataset comprising 1819 data points from previous studies was collected, consisting of three main groups: 340 data of concrete with SCBA, 459 data of concrete without SCMs, and 1020 concrete data incorporating other supplementary cementitious materials (SCMs). The dataset was utilised in three different ways to evaluate the influence of data on model performance. Firstly, models were trained solely on SCBA data (Group A);subsequently, SCBA data was combined with concrete data without any SCMs (Group B);finally, the entire dataset was utilised (Group C). Statistical analysis revealed that the SVR model trained on SCBA data and the RBFNN model trained on group B data demonstrated the best results and accuracy. However, further refinement of the concrete without SCMs data in group B led to improved model performance, surpassing even that of the SVR model trained on group A data. © 2023, The Authors. All rights reserved.