AN OPEN ACCESS JOURNAL
 JJCE Submission Home


Prediction Of Mechanical Properties Of Reactive Powder Concrete By Using Artificial Neural Network And Regression Technique After The Exposure To Fire Flame

Submitted2015-06-30
Last Update2015-06-30
TitlePrediction Of Mechanical Properties Of Reactive Powder Concrete By Using Artificial Neural Network And Regression Technique After The Exposure To Fire Flame
Author(s)Author #1
Author title:
Name: Mohammed Kadhum
Org: College of Engineering, Babylon University, Iraq
Country:
Email: moh_alkafaji@yahoo.com

Other Author(s)
Contact AuthorAuthor #1
Alt Email: moh_alkafaji@yahoo.com
Telephone:
KeywordsReactive powder concrete, Fire flame, Artificial neural network, Mechanical properties, Statistical analysis
AbstractAn experimental work was carried out to investigate some mechanical properties of Reactive Powder Concrete (RPC) which are particularly required as input data for structural design. These properties include compressive strength, flexural strength, tensile strength and static modulus of elasticity. A combined laboratory and modeling study was undertaken to develop a database of the estimation ability of the effects of exposure to real fire flame on the mechanical properties of reactive powder concrete using 2 different models: artificial neural network (ANN) and regression techniques. Experimental results were used in the estimation models. After being subjected to high temperatures from 200 to 500�C, the residual mechanical properties were determined, and RPC was considerably spalled under high temperature. Exposing to high temperatures from 200 to 400�C, mechanical properties were enhanced more or less, which can be attributed to further hydration of cementitious materials activated by elevated temperature. It was found that RPC can be used at elevated temperatures up to 300�C for heating times up to 1 hour, taking into consideration the loss of strength. Finally, prediction performances of reactive powder concrete single and multiple variable regression equations were developed, and ANN was compared. According to this comparison, best prediction performance which belongs to ANN was improved.
Paperview paper 3079.pdf (994KB)

http://www.just.edu.jo