Amin shahin - Academia.edu (original) (raw)

Papers by Amin shahin

The research presented in this thesis focuses on the settlement prediction of shallow foundations... more The research presented in this thesis focuses on the settlement prediction of shallow foundations on cohesionless soils using artificial neural techniques. The problem of estimating the settlement of shallow foundations on cohesionless soils is very complex and not yet entirely understood. Over the years, many methods have been developed to predict the settlement of shallow foundations on cohesionless soils. However, methods for such predictions that have the required degree ofaccuracy and consistency have not yet been developed. Accurate prediction of settlement is essential since settlement, rather than bearing capacity, generally controls the design process of shallow foundations. In this research, artihcial neural networks (ANNs) are used in an attempt to obtain more accurate settlement prediction. ANNs are numerical modelling techniques that are inspired by the functioning of the human brain and nerve system. ANNs use the data alone to determine the structure of the model as well as the unknown model parameters' ANNs have been applied successfully to many problems in the field of geotechnical engineering and some of their applications are demonstrated in this thesis. A large database comprising a total of 189 case records is used to develop and verift the ANN models. Five parameters are considered to have the most significant impact on the settlement of shallow foundations on cohesionless soils and are thus used as the ANN model inputs. These include the footing width, footing net applied pressure, average SPT blow count over the depth of influence of the foundation, footing geometry and footing embedment ratio. The model output is the average measured sefflement of the foundation, considered in its final state. Two types of ANNs are used for the development of ANN models. The first type is multi-layer perceptrons (MLps) that are trained using the back-propagation algorithm, whereas the second type are B-spline neurofuzzy networks that are trained with the adaptive spline modelling of observation data (ASMOD) algorithm. In relation to the multi-layer perceptrons, the feasibility of ANNs for predicting the settlement of shallow foundations on cohesionless soils is investigated. A number of issues in relation to ANN construction, optimisation and validation are also investigated and guidelines for improving ANN performance are lv Abstract v developed. The issue of data division and its impact on ANN model performance is investigated in some detail by examining four different data division methods, namely, random data division; data division to ensure statistical consistency of the subsets needed for ANN model development; data division using self-organising maps (SOMs) and a new data division method using fuzzy clustering. The success or otherwise of ANNs for settlement prediction of shallow foundations on cohesionless soils is illustrated and compared with three of the most commonly used settlement prediction methods. A hand-calculation design formula for settlement prediction of shallow foundations on cohesionless soils that is based on a more accurate settlement prediction from ANN model is presented. It was found that ANNs have the ability to predict the settlement of shallow foundations on cohesionless soils with a high degree of accuracy and ouþerform traditional methods. It was also found that the new data division method that is based on fuz,zy clustering is suitable approach for data division. In relation to the newofuzzy models, the ability of ANNs to provide a better understanding of the relationship between settlement and the factors affecting settlement is investigated. It was found that neurofuzzy networks have the abilþ to provide a transparent understanding of the relationship between settlement and the factors affecting it. Settlement analysis is often affected by considerable levels of uncertainty that are usually ignored by traditional methods. In this research, ANNs are linked with Monte Carlo simulation to provide a stochastic solution for settlement prediction that takes into account the uncertainties associated with settlement analysis. A set of stochastic design charts that provide the designer with the level of risk associated with predicted settlements are developed and provided.