12-25-2013, 03:52 PM
This thesis-Project report is submitted as a part of M Tech in Civil Engineering.You can take help of this thesis to prepare your M Tech B Tech Final year project report.
Abstract:-
The evolution of computational geotechnical engineering analyses closely follows the development in computational methods. The soil is considered as a complex material produced by the weathering of solid rock. Due to its uncertain behavior, modeling the behavior of such
materials is complex by using more traditional forms of mechanistic based engineering methods like analytical and finite element methods etc. Very often it is difficult to develop theoretical/statistical models due to the complex nature of the problem and uncertainty in soil parameters. These are situations where data driven approach has been found to more appropriate than model oriented approach. To take care of such problems in artificial intelligence (AI) techniques has been developed in the computational methods. Though AI techniques has proved to have the superior predictive ability than other traditional methods for modeling complex behavior of geotechnical engineering materials, still it is facing some criticism due to the lack of transparency, knowledge extraction and model uncertainty. To overcome this problem there are developments of improvised AI techniques. Different AI techniques as ‘black box’ i.e artificial neural network (ANN), ‘grey box’ i.e Genetic programming (GP) and ‘white box’ i.e multivariate adaptive regression spline (MARS) depending upon its transparency and knowledge extraction. Here, in this study of GP and MARS ‘grey box’ and ‘white box’ AI techniques are applied to some geotechnical problems such as prediction of lateral load capacity of piles in clay, pull-out capacity of ground anchor, factor of safety of slope stability analysis and ultimate bearing capacity of shallow foundations. Different statistical criteria are used to compare the developed GP and MARS models with other AI models like ANN and support vector machine
(SVM) models. It was observed that for the problems considered in the present study, the MARS and GP model are found to be more efficient than ANN and SVM model and the model equations are also found to be more comprehensive. But as every numerical method has its own
advantages and disadvantages and are also problem specific, there is a need to apply these techniques to other Geotechnical engineering problems to draw final conclusions regarding its efficacy.
Abstract:-
The evolution of computational geotechnical engineering analyses closely follows the development in computational methods. The soil is considered as a complex material produced by the weathering of solid rock. Due to its uncertain behavior, modeling the behavior of such
materials is complex by using more traditional forms of mechanistic based engineering methods like analytical and finite element methods etc. Very often it is difficult to develop theoretical/statistical models due to the complex nature of the problem and uncertainty in soil parameters. These are situations where data driven approach has been found to more appropriate than model oriented approach. To take care of such problems in artificial intelligence (AI) techniques has been developed in the computational methods. Though AI techniques has proved to have the superior predictive ability than other traditional methods for modeling complex behavior of geotechnical engineering materials, still it is facing some criticism due to the lack of transparency, knowledge extraction and model uncertainty. To overcome this problem there are developments of improvised AI techniques. Different AI techniques as ‘black box’ i.e artificial neural network (ANN), ‘grey box’ i.e Genetic programming (GP) and ‘white box’ i.e multivariate adaptive regression spline (MARS) depending upon its transparency and knowledge extraction. Here, in this study of GP and MARS ‘grey box’ and ‘white box’ AI techniques are applied to some geotechnical problems such as prediction of lateral load capacity of piles in clay, pull-out capacity of ground anchor, factor of safety of slope stability analysis and ultimate bearing capacity of shallow foundations. Different statistical criteria are used to compare the developed GP and MARS models with other AI models like ANN and support vector machine
(SVM) models. It was observed that for the problems considered in the present study, the MARS and GP model are found to be more efficient than ANN and SVM model and the model equations are also found to be more comprehensive. But as every numerical method has its own
advantages and disadvantages and are also problem specific, there is a need to apply these techniques to other Geotechnical engineering problems to draw final conclusions regarding its efficacy.