Strength characterisation of self cured concrete using AI tools

Banu, T. Udaya and Rajamane, N.P. and Awoyera, P. O. and Gobinath, R (2020) Strength characterisation of self cured concrete using AI tools. Materials Today: Proceedings.

PDF Download (2MB)

Abstract

Civil engineering experimentation process is termed to be a costly process when it involves destructive testing of materials to obtain their strength and durability. Testing of materials through destructive process is century old procedure, but recent decade science involves the prediction of strength and durability using alternative methods. One such method to predict the strength in nondestructive method is employment of Soft computing technologies, this process is gaining impetus in the recent decade due to its accuracy, reliability, and versatility. In this research, we had employed artificial intelligence tool to predict the compressive strength of concrete with available real time laboratory-based data. AI tools require a greater number of data to predict the results but in this work and attempt is made to predict using a smaller number of data with more accuracy. Compressive, flexure and tensile strength of concrete is predicted using ANN techniques (Levenberg-Marquardt (L-M) process and Bayesian regularization (B-R)). Two input parameters were only employed to check the real time accuracy with a model that has 12 input layers and 18 hidden layers incorporated. Model output shows regression values of 0.97428, 0.92865 and 0.96772, concerned with L-M algorithmic model and 0.96573, 0.95625 and 0.91787 for BR based model. Also, its observed that while using L-M algorithm the best performance was obtained at 1.3287 at epoch 2 for compressive strength and 0.12417 is achieved at epoch 1 for tensile strength and 0.021578 at epoch 3 concerned with flexural strength. Also with B-R algorithm provided best performance of 2.1488 at epoch 4 for compressive strength, a value of 0.43468 at epoch 3 for flexural strength and 0.015279 for tensile strength reached at epoch 30. Thus we propose the usage of ANN even with less number of data using this method for predicting the values of compressive strength of concrete. � 2020 The Authors. Published by Elsevier Ltd.

Item Type:	Article
Uncontrolled Keywords:	Artificial neural network Compressive strength Tensile strength Flexural strength
Subjects:	T Technology > TA Engineering (General). Civil engineering (General)
Divisions:	Faculty of Engineering, Science and Mathematics > School of Civil Engineering and the Environment
Depositing User:	Mrs Patricia Nwokealisi
Date Deposited:	22 Jan 2021 11:59
Last Modified:	22 Jan 2021 11:59
URI:	http://eprints.covenantuniversity.edu.ng/id/eprint/13801

Actions (login required)

View Item