Fuzzy Neural Network Models for Geotechnical Problems

[mahyarov]

Fuzzy Neural Network Models for Geotechnical Problems

Author: Jeon, JongKoo | Size: 4.03 MB | Format: PDF | Quality: Original preprint | Publisher: North Carolina State University, Raleigh | Year: 2008 | pages: 411

Uncertainty, imprecision, complexity, and non-linearity are inherently associated with many problems in geotechnical engineering. The conventional modeling of the underlying systems, tend to become quite intractable and predictions from them are very difficult and unreliable. The general nature of geotechnical problems makes them ideally amenable to modeling through emerging methods of fuzzy and neural network modeling. Piles have been used as a foundation for both inland and offshore structures. The evaluation of the load carrying capacity of a pile, setup, and its drivability are important problems of pile design. In this study, Back Propagation Neural Network (BPNN) models and Adaptive Neuro Fuzzy Inference System (ANFIS) models are developed for: i) Ultimate pile capacity, ii) Pile setup, and iii) Pile drivability. A database for ultimate pile capacity and pile setup has been developed from a comprehensive literature review. Predictions for the above are made using BPNNs as well as commonly used empirical methods, and they are also compared with actual measurements. For the pile drivability analysis, a database of a number (3,283) of HP piles is developed from the data on HP piles from 57 projects in North Carolina (with both GRLWEAP data and soil profile information and without PDA and CAPWAP analyses). All of the programs are developed within MATLAB (and its toolboxes) with its Graphical User Interface (GUI). It is found that ANFIS and BPNN models for the analyses of pile response characteristics provide similar predictions, and that both are better than those from empirical methods, and can serve as a reliable and simple tool for the prediction of ultimate pile capacity and pile setup. Also, the BPNN model developed for pile drivability analysis provides good predictions. BPNN may be considered to be more efficient than ANFIS, as the BPNN model trains much faster, while both provide equally good predictions. However ANFIS models with some additional work will be more desirable for those cases in which one or more input variables may be available only in ‘fuzzy’ terms, and when the model is developed with a limited data range, because in ANFIS extrapolation beyond the data range is made through the membership functions.

Code:

***************************************
Content of this section is hidden, You must be registered and activate your account to see this content. See this link to read how you can remove this limitation:

http://forum.civilea.com/thread-27464.html
***************************************

This post has been made by CivilEA Post-Generator ver 2.1.6