Journal of the Korean Society of Mineral and Energy Resources Engineers ISSN:2288-0291(Print) 2288-2790(Online) 한국자원공학회지

Rock mass rating (RMR) is a relatively simple method for classifying rock mass with the naked eye; however, it becomes inconvenient when the number of survey sections is large. In this study, we developed a learning model and a prediction model using an artificial neural network (ANN) to classify RMR mechanically. Using robust design, 3125 big data were optimized into 25 learning data. The test results after learning through the two methods were exactly the same. Through robust design, the learning data obtained by reducing the total number of cases to less than 1% had the same learning effect as the whole data, which means that the effort and cost of acquiring the learning data can be greatly reduced. For the perfect prediction of the RMR classification system, we tuned the primary predictions within a given range of rating levels. As a result, we implemented an RMR classification ANN system that perfectly predicts the RMR of 3125 big data using 25 learning data through robust design.