TA391 : Behavioral Modeling of Foundation Using a Hybrid Neural Network and Simulated Annealing Algorithm
Thesis > Central Library of Shahrood University > Civil & Architectural Engineering > MSc > 2017
Authors:
Abstarct: Recently, in order to avoid spending too much time and money for the implementation of different experiments, the tendency toward computer tools that are similar to the biologic (Quasi-biological) system has increased. In this study, a new quasi-biological technique which has been developed by the combination of the artificial neural network (ANN) and simulated refrigeration (SA), has been used to predict the final bearing capacity of gravel soil. The proposed ANN / SA model has been developed baxsed on the use of laboratory results that have been collected after extensive technical literature studies. The data set includes experiments which performed on small and large scale square, circular, and bandwidths on a sand bed. The new presented model formulates the final loading capacity of the surface peaks baxsed on parameters related to the geometry of the pitch (such as width, depth and shape of the pit) and mechanical properties of the soil (such as average specific gravity and internal friction angle of soil).
In order to obtain optimal models, extensive attempts and errors have been made for using effective parameters on bearing capacity. The role of effective parameters in predicting load bearing capacity has been discussed for using the sensitivity analysis. A comparative study carried out by the well-known equations which was provided by Trezaghi, Meirov, Hansen and Wessiak proves the possibility of using the ANN / SA model as a reliable alternative for such classical equations. The results indicate a better performance of the proposed model than other artificial intelligence models in the technical literature. It is worth noting that ANN baxsed models are often recognized as black box models due to the inability to explain the underlying principles for prediction.
In other words, although ANN are successful in predicting, they are not capable of creating predictive equations. To overcome this limitation, a new process has been defined baxsed on the optimal model from ANN / SA which is returned to the relatively simple design equation. The calculation process is baxsed on the weights and biases of the best structure and in the form of a spreadsheet program.
Keywords:
#Ultimate Bearing Capacity #Artificial Neural Network #Simulated Refrigeration #Surface Paths #Grain Soils #Prediction Link
Keeping place: Central Library of Shahrood University
Visitor:
Keeping place: Central Library of Shahrood University
Visitor: