2015 Volume 20 Issue 2 Pages 23-28
It is important for the geotechnical engineers to predict the amount of the settlements as exactly as possible for the safe design of earth and hydraulic structures. The simple technique called mv method is often used to calculate the settlement for the small scale structure, and the parameter mv is properly necessary for the technique. Some researchers have presented empirical formulas for the estimation of mv that use single or multiple soil parameter models, such as natural water content, unconfined compressive strength, static penetration resistance and others. However, the estimated values of mv vary widely and there is not necessarily an extremely strong correlation. Moreover, if we try to present a new formula, it is difficult to determine some input soil parameters necessary to overcome complex multicollinearity problems. This study has proposed a simple method that estimate mv from some soil parameters using the artificial neural network (ANN) with batch learning algorithm developed by Kobayashi et al instead of performing the oedometer test. The optimum combination of input soil parameters has been evaluated and the results of the proposed method have been compared with 4 empirical formulas on the viewpoint of learning efficiency.
