Application of Feed Forward Neural Network along with Different Input Parameters for Predicting the Mechanical Properties of SCC

Document Type : Original Article

Authors

1 M.Sc. Student of Construction Engineering and Management, Department of Civil Engineering, Tabari Institute of Higher Education, Babol, Iran

2 Assistant Professor and Head of the Tabari Institute of Higher Education, Babol, Iran

Abstract

In this article, artificial neural network, as a time – and budget – saving tool, was applied for predicting the compressive, tensile and flexural strength of Self-Compacting Concrete (SSC). Feed Forward ANN models were designed using the available test data of different concrete mix-designs of SCC were obtained from different sources. In order to show the effect of the selected input parameters on the amounts of the test error in the prediction of the desired properties, data used in ANN models were arranged in two formats of 8 and 140 input parameters. According to the results, the optimized Feed Forward networks are able to predict the properties of SCC with suitable accuracy. In addition, for all of the desired properties, the accuracy of the networks with 140 inputs is higher than networks with 8 inputs. In fact, the effective data that are selected as inputs in the ANNs can significantly increase the prediction precision by more simulation the prediction conditions to the experimental conditions.

Keywords

Asas Journal
Volume 19, Issue 49
March 2018
Pages 36-50

Files

History

  • Receive Date: 09 December 2017
  • Revise Date: 21 January 2018
  • Accept Date: 06 May 2018

Share

How to cite

Statistics