Computational Technology Resources - CCP

Keywords: neural networks, mass concrete, predict, properties, strength, slump, data transforms.

Summary

In determining properties of mass concrete, the uncertainties of materials, temperature, site environmental situations, personal skillfulness, and errors in calculations and testing process should be considered. This kind of concrete mix proportioning was somewhat complicated, time-consuming, and uncertain. In this investigation, as a tool to minimize the uncertainties and errors, back propagation neural networks were used to predict the strength and slump of mass concrete. The Effects of major parameters such as: cement, water, fine aggregate, four ranges of coarse aggregate, concrete temperature, maximum size of coarse aggregate, air content, unit weight, additives, water / cement ratio, etc were included in model.

Various data transforms were tried in order to increase the method efficiency and it was found that models based on raw data never gave the best results.

The experimental records from 73 different mass concrete mixes of karoon-III dam were used in the training and testing process. The networks were trained based on 58 sets and then tested using the remaining 15 sets.

Neural network training can be made more efficient if certain preprocessing steps are performed on the network inputs and targets. These transformations are described as follow as: 1 - Linear transformation 2 - Normal distribution [1,2] transformation 3 - Principal component analysis. In some situations the dimension of the input vector is large, but the components of the vectors are highly correlated. It is useful in this situation to reduce the dimension of the input vectors. An effective procedure for performing this operation is principal component analysis. This technique has three effects: it orthogonalizes the components of the input vectors; it orders the resulting orthogonal components so that these with the largest variation come first; and it eliminates those components which contribute the least to the variation in the data set [3].

The training of the networks was carried out with various numbers of nodes in the single middle layer and various target accuracies. Trial and error procedures gave us the optimum number of middle layers. We defined 8 different models based on previous studies.

Apparently MSE was major factor in selecting optimum models, but network performance at end points was the other factor, which had to be considered.

The results indicated that the model could predict the strength and slump of mass concrete with adequate accuracy required for practical design purpose. Structural engineers to predict the strength and slump of mass concrete without conducting costly experimental tests can use these models by using a try & error procedure.

The following conclusions can be drawn from this work [4]:

Neural networks trained with principal component method make better predictions of strength and slump.
Without pre processing routines it cannot possible to achieve appropriate results.
Appropriate performance of networks trained with raw data instead of non-dimensional ratios.
MSE efficiency for performance evaluation of neural networks.

References

1: Ramezanianpour, A.A., Davarpanah, A., "Concrete Properties Estimation and Mix Design Optimization Based on Neural Networks" World Conference on Concrete Materials and Structures (WCCMS), Kuala Lumpur, Malaysia, May 2002.
2: Ramezanianpour, A.A., Davarpanah, A., "Concrete Properties Modeling Using Neural Networks" 6th International Conference on Concrete Technology for Developing Countries, Amman, Jordan, October 2002.
3: Demuth, H., Beale, M, Neural Network Toolbox for Use with MATLAB-User Guide, The Math Works Inc, 2001.
4: Yeh, I-Cheng; "Design of High-Performance Concrete Mixture Using Neural Networks and Nonlinear Programming", Journal of Computing in Civil Engineering, Vol. 13,No. 1, pp.36-42, 1999. doi:10.1061/(ASCE)0887-3801(1999)13:1(36)

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £95 +P&P)

	Computational & Technology Resources an online resource for computational, engineering & technology publications
	not logged in - login
Front Page Browse CCP CSETS CTR IJRT Other Authors Search Purchase Guide FAQ Contact us	Civil-Comp Proceedings ISSN 1759-3433 CCP: 80 PROCEEDINGS OF THE FOURTH INTERNATIONAL CONFERENCE ON ENGINEERING COMPUTATIONAL TECHNOLOGY Edited by: B.H.V. Topping and C.A. Mota Soares Paper 140 Modeling Mass Concrete Properties with Neural Networks A.A. Ramezanianpour+, M. Haghani+ and A.A. Mortezazadeh* +Department of Civil Engineering, Amirkabir University of Technology, Iran Faculty of Engineering, University of Tehran, Iran doi:10.4203/ccp.80.140 purchase the full-text of this paper Full Bibliographic Reference for this paper A.A. Ramezanianpour, M. Haghani, A.A. Mortezazadeh, "Modeling Mass Concrete Properties with Neural Networks", in B.H.V. Topping, C.A. Mota Soares, (Editors), "Proceedings of the Fourth International Conference on Engineering Computational Technology", Civil-Comp Press, Stirlingshire, UK, Paper 140, 2004. doi:10.4203/ccp.80.140 Keywords:* neural networks, mass concrete, predict, properties, strength, slump, data transforms. Summary In determining properties of mass concrete, the uncertainties of materials, temperature, site environmental situations, personal skillfulness, and errors in calculations and testing process should be considered. This kind of concrete mix proportioning was somewhat complicated, time-consuming, and uncertain. In this investigation, as a tool to minimize the uncertainties and errors, back propagation neural networks were used to predict the strength and slump of mass concrete. The Effects of major parameters such as: cement, water, fine aggregate, four ranges of coarse aggregate, concrete temperature, maximum size of coarse aggregate, air content, unit weight, additives, water / cement ratio, etc were included in model. Various data transforms were tried in order to increase the method efficiency and it was found that models based on raw data never gave the best results. The experimental records from 73 different mass concrete mixes of karoon-III dam were used in the training and testing process. The networks were trained based on 58 sets and then tested using the remaining 15 sets. Neural network training can be made more efficient if certain preprocessing steps are performed on the network inputs and targets. These transformations are described as follow as: 1 - Linear transformation 2 - Normal distribution [1,2] transformation 3 - Principal component analysis. In some situations the dimension of the input vector is large, but the components of the vectors are highly correlated. It is useful in this situation to reduce the dimension of the input vectors. An effective procedure for performing this operation is principal component analysis. This technique has three effects: it orthogonalizes the components of the input vectors; it orders the resulting orthogonal components so that these with the largest variation come first; and it eliminates those components which contribute the least to the variation in the data set [3]. The training of the networks was carried out with various numbers of nodes in the single middle layer and various target accuracies. Trial and error procedures gave us the optimum number of middle layers. We defined 8 different models based on previous studies. Apparently MSE was major factor in selecting optimum models, but network performance at end points was the other factor, which had to be considered. The results indicated that the model could predict the strength and slump of mass concrete with adequate accuracy required for practical design purpose. Structural engineers to predict the strength and slump of mass concrete without conducting costly experimental tests can use these models by using a try & error procedure. The following conclusions can be drawn from this work [4]: Neural networks trained with principal component method make better predictions of strength and slump. Without pre processing routines it cannot possible to achieve appropriate results. Appropriate performance of networks trained with raw data instead of non-dimensional ratios. MSE efficiency for performance evaluation of neural networks. References 1 Ramezanianpour, A.A., Davarpanah, A., "Concrete Properties Estimation and Mix Design Optimization Based on Neural Networks" World Conference on Concrete Materials and Structures (WCCMS), Kuala Lumpur, Malaysia, May 2002. 2 Ramezanianpour, A.A., Davarpanah, A., "Concrete Properties Modeling Using Neural Networks" 6th International Conference on Concrete Technology for Developing Countries, Amman, Jordan, October 2002. 3 Demuth, H., Beale, M, Neural Network Toolbox for Use with MATLAB-User Guide, The Math Works Inc, 2001. 4 Yeh, I-Cheng; "Design of High-Performance Concrete Mixture Using Neural Networks and Nonlinear Programming", Journal of Computing in Civil Engineering, Vol. 13,No. 1, pp.36-42, 1999. doi:10.1061/(ASCE)0887-3801(1999)13:1(36) purchase the full-text of this paper (price £20) go to the previous paper go to the next paper return to the table of contents return to the book description purchase this book (price £95 +P&P)
Back to top	©Civil-Comp Limited 2023 - terms & conditions