An estimated method of intact rock strength using Gaussian process regression

During excavation of rock tunnels or construction of underground openings, estimating and determining the intact rock strength is of vital importance. Theoretical and practical methods for estimating intact rock strength were developed during past rock engineering practices, for example, determining s1 from s2 and s3. While, random uncertainty inherently exists in intact rock strength. The existing methods for estimating intact rock strength seldom take this kind uncertainty into consideration. In the paper an estimated method of intact rock strength using Gaussian process regression (GPR) is proposed to consider the uncertainty. GPR is a nonparametric probabilistic modeling approach. And the parameter determined from given data sets obeys Gaussian distribution. The capability of GPR for estimating s1 from given s2 and s3 is examined. Variance and 95 percent confidence interval of the predicted intact rock strength are obtained from the proposed estimated method using GPR. To validate the efficiency of the proposed estimated method using GPR modelling approach, comparative studies have been conducted with the predicted results from the four ISRM suggested three-dimensional failure criteria. The results show that the proposed estimated method using GPR modelling approach is a promising and reliable tool for estimating the strength of intact rocks.