A probabilistic Bayesian methodology for the strain-rate correction of dynamic CPTu data

Abstract

Dynamic Cone Penetration Test (CPTu) profile offshore sediments by impact penetration. To exploit their results in full, the measured data are converted to obtain a quasi-static equivalent profile. Dynamic CPTu conversion requires calibrated correction models. Calibration is currently done by using paired (i.e., very close) quasi-static and dynamic tests. It is shown here that paired test data, which may be inconvenient to acquire offshore, are not strictly necessary to convert dynamic CPTu data. A new probabilistic methodology is proposed to call upon quasi-static results from a much wider area in the conversion procedure. Those results feed the prior distribution of a converted profile, within a Bayesian updating scheme where strain-rate coefficient and correction model error are also described by updated stochastic variables. The updating scheme is solved numerically using the Transitional Markov Chain Monte Carlo sampling algorithm. To avoid undue influence of local profile heterogeneity, the statistic treatment of the quasi-static CPTu data takes place in the frequency domain, using a discrete cosine transform. The new procedure is applied to a CPTu campaign offshore Nice (France): dynamic tests are converted with equal precision using quasi-static data acquired at distances orders of magnitude larger than what was previously employed.