An Artificial Neural Network Approach to Assess the Weathering Properties of Sancaktepe Granite

Weathering process of rocks leads continuous changes in their physico-chemical state which affects their use in engineering applications. Many tests are used to identify the grade of weathering for classification purposes. Numerous descriptive engineering classifications are developed on weathering of rock and rock masses. These classifications include five or six grades of weathering schemes with crisp boundaries and linguistic terms. The transitions between grades are naturally ambiguous and gradual. Further, the engineering grades of a classification scheme are not suitable for all rock types due to the difference on weathering process of each rock type. Applications of artificial intelligence methods became a practical instrument in the solution of this kind of complex problems in rock mechanics and engineering geology. This study proposes implementation of artificial neural network (ANN) to evaluate weathering properties of Sancaktepe granite crops out around Gebze District in Kocaeli Province in Turkey. For this purpose, uniaxial compressive strength, bulk density, ultrasonic velocity, quick absorption, and point load tests are conducted on drill cores, road cuts and field exposure samples of different weathering grades of Sancaktepe granite. Since the relationship between the rock properties and the weathering index value is nonlinear, an ANN classification model is developed-based on the description of weathering grades using a database consisting of 60 dataset. It is concluded that implication of ANN provides higher performance in classifying than it does in classical method.