• The Journal of Engineering Geology
• /
• v.22 no.1
• /
• pp.15-25
• /
• 2012

Probabilistic design methods have been used as a design standard in Korea and abroad for achieving reasonable design by considering the statistical uncertainties of soil properties. In this study, the following techniques for reflecting geotechnical uncertainty are analyzed: quantification of the uncertainties of geotechnical random variables, and consideration of economic feasibility in design by minimizing the uncertainties related to the number of samples. To quantify the uncertainties, the techniques were applied to soil properties obtained from samples collected and tested in the field. The results showed an underestimation of the standard deviation by the 3-sigma approach in comparison with calculations using data from the samples. This finding indicates that economical design is possible in terms of probability. However, when compared with the Bayesian approach, which does not consider the number of samples, variability in the 3-sigma approach is underestimated for some variables. This finding also indicates a safety issue, whereas the number of samples based on the Bayesian approach showed the lowest variance. The variance of the probability density function showed a marked decrease with increasing number of samples, to converge at a certain level when the number exceeds 25. Of note, the estimation of values is more reliable for random variables having low variability, such as soil unit weight, and can be obtained with a small number of samples.

• The Bulletin of The Korean Astronomical Society
• /
• v.46 no.1
• /
• pp.30.3-30.3
• /
• 2021

In order to generate the 3D structure of the 21-cm signal during the reionization, semi-numerical simulations based on Excursion set formalism are broadly used. However, semi-numerical simulations in the realization of the 3D structure are known to be the ionizing photon non-conserving by the structure of the Excursion set approach. Recently, explicit photon conserving algorithms for semi-numerical simulations introduced, but they are still too slow when forward modelling the 21-cm signal with high-dimensional parameter spaces. Here, we introduce a new method for approximately correcting photon non-conservation, which can be applied on-the-fly. This method is tailored towards the efficient simulation and Bayesian inference with high-dimensional parameter space. Then, we investigate how large an impact that photon non-conservation has on astrophysical parameter inference by performing an MCMC analysis. We find that the ionizing escape parameter is deviated from the fiducial value by 2 sigma when we infer astrophysical parameters without this correction.