ESTIMATION OF ROCK SHEAR STRENGTH BASED ON POROSITY AND UCS PARAMETERS USING FEED-FORWARD NEURAL NETWORK

• Main Author: Fansesa Adiguna, Briam
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77163
• Institution: Institut Teknologi Bandung

Rock shear strength parameters are commonly used in the structural design process in Rock Mechanics. The strength is generally expressed in terms of two properties: cohesion (c) and friction angle (?). The rock shear strength value is generally obtained through laboratory testing, namely direct shear tests. However, there are several things that are quite hampering the process of determining the value of rock shear strength in the laboratory. The availability of good-quality rock samples is difficult to obtain, especially when facing jointed, highly fractured, and weathered rock mass. On the other hand, the development of research using artificial neural networks or ANN began to be in demand by various researchers because it can solve complex problems. Some researchers have used the ANN method and its derivatives to estimate various mechanical parameters of rocks (cohesion and friction angle) in various types of rocks and provide a fairly good level of accuracy. Therefore, estimating the value of rock shear strength using artificial neural networks is an interesting research topic because it is economical and easy to do. The dataset used in this study was divided into a training dataset and a testing dataset. Training dataset using rock samples from a coal mine in Central Borneo (Tanjung 1). Testing dataset using rock samples from coal mines in East Borneo (Tanjung 3). The Tanjung 1 sample (training dataset) consists of claystone and siltstone lithology. While the Tanjung 3 sample (testing dataset) consists of mudstone lithology. Geologically, both datasets belong to the Tanjung Formation. The accuracy of the ANN-Claystone model is +1.68 kPa for cohesion prediction and +5.63 degrees for friction angle prediction. While the accuracy of the ANNSiltstone model is +1.87 kPa for cohesion prediction and +1.39 degrees for friction angle prediction. Both ANN models are able to provide a fairly good accuracy for prediction of rock shear strength with notes, prediction data is in the distribution range of the training dataset.