Bayesian Inference for Predicting Rock Mass Rating Ahead of Tunnel-Face Excavation Integrating Multi-Source Information

Abstract

The rock mass rating (RMR14) system can effectively evaluate the quality of the rock mass surrounding a tunnel, which combines the characteristics of the rock mass strength and the discontinuity. However, it is difficult to infer the RMR14 value accurately due to the data often being limited, as well as there being extensive data sources of geological exploration. The rock mass classification systems, which are known as empirical methods, are still used in current engineering design and construction. In this paper, a quantitative prediction method of the RMR14 value on tunnel excavation is proposed that is based on the Bayesian network theory. A variety of empirical formulas between the RMR14 with the geological strength index and the basic quality method are fused in the Bayesian network fused framework; MATLAB is used to generate 500 random samples, and the conditional probability table of the Bayesian network is generated based on the expected maximum algorithm. The uncertainty of the RMR value prediction is quantified by the RMR14 Progressive Prediction Probability Model. The proposed method is applied to a rock tunnel project, i.e., the Laoying tunnel in Yunnan province, China. The results show that this method is able to provide a reasonable probability inference of the RMR14.