A reliability analysis framework coupled with statistical uncertainty characterization for geotechnical engineering

Reliability analysis plays an important role in the risk management of geotechnical engineering. For the random field-based method, it is expected that the uncertainty characterization of geo-material parameters and the realization of random field can be integrated effectively. Moreover, as the increase in measured data size is generally difficult in the field investigation of geotechnical engineering due to limitation of budget and time etc., the statistical uncertainty resulting from sparse data should be paid great attention. Therefore, taking the determination of hyper-parameters for Bayesian-based conditional random field as the breakthrough, this study proposed a reliability analysis framework to achieve the expectation above. In this proposed reliability analysis framework, the present characterization method of statistical uncertainty is improved by setting the lognormal distribution as the prior distribution of scale of fluctuation (SOF). Subsequently, the performance of statistical uncertainty characterization method is tested by a set of unconfined compressive strength (UCS) database about rocks. Then, a case study about the stability analysis of slope is employed to demonstrate the beneficial effect of the proposed reliability analysis framework. It is found that the uncertainty in both the realization of random field and the reliability analysis results can be significantly mitigated by the proposed reliability analysis framework.