Abstract: A slope reliability analysis method based on limited shear-strength parameter data is proposed by adopting the Bootstrap sampling method. The traditional method of slope reliability analysis is first introduced briefly, and then the Bootstrap method is adopted to model the statistical uncertainty in probability distributions of shear strength parameters. An example of the infinite slope is given, the influence of the distribution parameters of shear strength and the uncertainty of distribution type on the reliability of the slope is studied. The results indicate that the sample mean, sample standard deviation and AIC values derived from the limited data of shear strength parameters show a large variability. This variability further aggravates the statistical uncertainty in the probability distributions of shear strength parameters. By considering the statistical uncertainty in the probability distributions of shear strength parameters, the reliability index of a slope is presented as a confidence interval instead of a single, fixed index derived from a classical reliability analysis. The range of this confidence interval increases with increasing slope safety factor. Meanwhile, when both the uncertainties in distribution parameters and distribution types of shear strength parameters are taken into account, the reliability index has a higher variability and thus a wider range of confidence interval. The Bootstrap method provides a practical access to modeling the uncertain probability distributions of shear strength parameters and to determining the associated slope reliability with limited data of shear strength parameters.

Key words: slope, reliability, shear strength parameters, probability distribution function, statistical uncertainty, Bootstrap method