6. Modeling the properties of self-compacting concrete: an M-5 Model tree based approach

Title: 6. Modeling the properties of self-compacting concrete: an M-5 Model tree based approach
Author(s): P. Aggarwal, R. Siddique, Y. Aggarwal, S.M. Gupta
Paper category : conference
Book title: 5th International RILEM Symposium on Self-Compacting Concrete
Editor(s): G. De Schutter and V. Boel
Print-ISBN: 978-2-35158-047-9
e-ISBN: 978-2-35158-088-2
Publisher: RILEM Publications SARL
Publication year: 2007
Pages: 49 - 54
Total Pages: 6
Nb references: 23
Language: English

Abstract: This paper explores the potential of M5 model tree based approach in predicting 28-days compressive strength and slump flow of self-compacting concrete. A total of 80 data collected from the exiting literature are used in present study. To compare the performance of the technique, prediction was also done using a back propagation neural network model. A correlation coefficient of 0.908 with a root mean square error of 5.92 for strength prediction was achieved by M5 model tree approach. In comparison, a value of 0.906 as correlation coefficient and 6.01 as root mean square error was obtained by neural network approach. For slump flow prediction, correlation coefficient values of 0.901 and 0.914 (root mean square error of 9.174 and 8.778) were achieved by M5 model tree and neural network modelling approaches respectively. Results from this study suggests a comparable performance by M5 model tree based approach to neural network approach for both strength and slump prediction.
It was observed that in comparison to neural network, M5 model tree requires no user-defined parameters to be set and also involves using a small computational cost , as choice of suitable architecture has always been a problem with neural network approach and requires lot of efforts and computational cost.

Online publication: 2009-06-16
Publication type : full_text
Public price (Euros): 0.00