Rock Mass Characterization By Artificial Neural Network With Three-Dimensional Representation Of Its Output Data

This study presents the development of a simplified artificial neural network (ANN) for the characterization of rock mass with a simultaneous display of its output in a three-dimensional interactive environment as well as in the form of a text document. The system was developed using JAVA object-oriented language, virtual-reality modeling language (VRML) and JavaScript, and it runs on the Internet. The user can walk through a three-dimensional virtual roadway, checking the underground rock mass behavior and, consequently, can design a support system. The system first characterizes the rock mass with the help of a neural network and recommends a suitable support system for the particular geomining condition. The system then represents its output in a three-dimensional environment. The neural network developed in this system was trained with 261 data sets, and its performance was evaluated with five case histories. The evaluation results showed that the network could identify the rock category and support system adaptively. The average error for the support load was 2.7%. Among the five case histories, two were represented in a three-dimensional environment. The output obtained from the overall system compared favorably with the existing underground conditions.