Assessing and Predicting the Soil Layers Thickness and Type Using Artificial Neural Networks - Case Study in Sari City of Iran
Rahman Mohseni-Astani, Poorya Haghparast, Sahab Bidgoli-Kashani · Middle-East Journal of Scientific Research · 2010
In order to identify the soil layers structure in a project, depending on the extent of the study, it is needed to dig different holes and perform several tests which demand a lot of spending. Increasing confidence in the interpolation structure and properties of the soil layers between holes leads to improved assessment of Geology and therefore the costs will be reduced and the possibility to proper planning of construction operations will be provided. Neural networks as intelligent systems use specific features of information processing in the brain such as learning and generalizing the samples, ignoring the error in the data and parallel processing that are not accessible for conventional programming methods. The present paper focuses on the information gathered from the boreholes dug in the range of 22 square kilometers of Sari city in the north of Iran. The data were collected and classified in order to determine the soil layers characteristics. Then for classifying the different layers in different depths and for determining the thickness of each layer at a specified depth, multi-layer neural networks have been trained separately. To ensure network performance in estimating the changes in soil layers, a number of data from the examining boreholes were presented to the network for the first time and then the results of neural networks have been compared with examining data boreholes which were to a large extent consistent with each other.