Analysis of human risks due to dam-break floods—part 1: a new model based on Bayesian networks

Risk assessment of liquefaction-induced hazards using Bayesian network based on standard penetration test data

“…The former problem just needs to answer whether the soil is liquefied or not 11 after an earthquake, whereas the latter problem not only needs to predict whether liquefaction-induced 12 hazards occur or not after soil liquefaction but also needs to assess the severity of different hazards 13 induced by liquefaction. Prediction of liquefaction potential of foundation soils is only the first step of 14 assessment of liquefaction hazards, which was well studied in the recent decades, such as the simplified 15 methods Idriss 1971, 1982; Starks and Olsen 1995; Stokoe and Nazarian 1985) based on 16 standard penetration test (SPT), cone penetration test (CPT) and shear wave velocity measurement, 17 laboratory testing, numerical methods, and empirical liquefaction models (Goh 1994; Pal 2006; Toprak 18 et al 1999) based on historical data.…”

“…As for the empirical liquefaction 5 method, liquefaction potential index (LPI) has been used to characterize liquefaction-induced hazards 6 worldwide that is proposed by Iwasaki et al (1982). After that several approaches based on the LPI, 7 such as damage severity index (DSI) developed by Juang et al (2005) that allowed for evaluation of the 8 severity of liquefaction-induced ground damage at or near foundations, the Ishihara inspired LPIISH 9 extended by Maurer et al (2015) that was found to be consonant with observed surface effects and 10 showed improvement over LPI in mitigating false-positive predictions, and liquefaction severity 11 number (LSN) developed by Tonkin and Taylor (2013) that was developed following the liquefaction 12 damage observations from the Canterbury Earthquake Sequence to reflect the damaging effects of 13 shallow liquefaction on residential land and foundations. In addition, generalized analytical or empirical 14 techniques for estimating a single type of ground failures (e.g.…”

“…In addition, fines content can change the type of soil. Normally, purified clay 11 and silt cannot be liquefied, whereas poorly graded sand and silty sand are easily liquefied. The bigger 12 average particle size is, and the bigger SPT number is, the smaller the probability of soil liquefaction is.…”

“…Because of incomplete data, e.g. the proportion of missing data for D50 is about 15.2%, the proportion 10 of missing data for vertical effective stress is about 29.4%, and the proportion of missing data for the 11 thickness of soil layer is about 38.9%, expectation-maximization (EM) algorithm (Lauritzen 1995) was 12 used for training the 332 SPT borings data to obtain conditional probability table for the BN model due 13 to its more robustness compared with other algorithms and its suitableness for the data contained large 14 missing values. Briefly, the EM method is an iterative algorithm to find maximum likelihood estimation 15 or maximum a posterior estimation of parameters that repeatedly takes a Bayesian net and uses it to 16 obtain a better one by doing an expectation (E) step followed by a maximization (M) step until 17 convergence of the algorithm.…”

“…It can be seen that 5 (1) liquefaction doesn't have to induce hazards, but the occurrence of the liquefaction-induced hazards 6 is based on liquefaction, (2) LPI is not a good index for a description of the severity of (1) and slight liquefaction of the LPI occur in the severe SLH in Fig. 6 (4), but there is a rule that the 10 bigger the LPI is, the corresponding severity of the liquefaction-induced hazards is, (3) SB, S, GC, and 11 LS are macroscopical phenomena of hazards induced by liquefaction, and there is also a trend that the 12 bigger the values of these indexes are, the more severity of the SLH is, (4) the classifications for the 13 four different types of hazards in Fig. 6 almost accord with the descriptions of the field ground damage 14 status in Table 3.…”