Bayesian prediction of peak resistance of a spudcan penetrating sand-over-clay

Abstract

Assessing the potential for a punch-through failure during spudcan installation in sand-over-clay is crucial for reducing risk in the operations of mobile jack-up platforms. Typically, in the offshore industry, the peak penetration resistance and the depth at which it occurs are determined deterministically without rigorously considering the uncertainties in the soil. This paper proposes a probabilistic approach to estimate the peak resistance and the corresponding depth, as well as a Bayesian method of incorporating installation data to update the predictions. Instead of a single value in the deterministic analysis, a range of the potential peak resistances and depths can be estimated by accounting for the uncertainties in the soil, the spudcan's geometry and in the calculation method itself, with a database of 66 geotechnical centrifuge tests characterising the model. This prior probability is then updated using the monitored data, allowing a real-time update of the probabilities associated with candidate values of peak resistance and depth during the installation. The advantage of such a probabilistic updating model is shown in a retrospective simulation of a mobile jack-up platform in sand-over-clay conditions in the Gulf of Mexico. The results show that the prior estimation can be effectively refined by incorporating the monitored data. The proposed method provides a powerful tool for assisting decision-making during the installation of jack-ups offshore.

Prediction of Pipe-Jacking Forces Using a Bayesian Updating Approach

“…A common implementation of the Bayesian updating approach is to generate predictions using a deterministic model where uncertainties due to (a) soil properties, (b) geometry, (c) measurement error and (d) imperfections and approximations on the calculation model are lumped together in a model 'bias factor' applied to the deterministic prediction (e.g. Li et al 2018). The prediction model adopted in this paper is defined as follows:…”

“…Lognormal distributions are an expedient option as they prevent unrealistic negative realisations of the non-negative geotechnical variables θ and have been widely adopted in geotechnical engineering e.g. Lumb (1966), Qi and Zhou (2017), Li et al (2016Li et al ( , 2018, Zheng et al (2018). Therefore, the uncertainty associated with the model parameters θ are quantified as a 'prior' distribution, which is assumed to be a multivariate normal distribution with μ′θ and σ′θ as the prior mean and uncertainty of the natural logarithm of θ respectively i.e.…”

Mitigating punch-through on sand-over-clay using skirted foundations

“…Recently, Li et al (2018) proposed a probabilistic approach to estimate the peak resistance and the corresponding depth. Li et al (2015Li et al ( , 2018 demonstrated the effect of the spatial variability of random soil on the failure mechanism and the ultimate bearing capacity of foundations and proposed a Bayesian method of incorporating real time installation data to update the predictions. However, previous prediction models did not consider skirted foundations.…”

Optimization for the Assessment of Spudcan Peak Resistance in Clay–Sand–Clay Deposits

“…The percentile curves or probabilistic contours of the spudcan load-penetration response can be predicted through Monte Carlo simulations by account-ing for the uncertainties and probabilistic distribution of the inputs in the predictive model [11,[13][14][15]. Bayes' theorem can then be adopted to update these predictions according to the measured load-penetration data [16,17]. The parameter optimization technique (POT) is another alternative option, which is also formulated based on Bayes' theorem, but the prediction is updated by optimizing the model parameters according to the measured data.…”