Stationary Evaluation of CPTu Data in Brazilian Marine Clay

Resumo

Structural integrity is an important premise in oil well design. In this context, soil characterization helps the well
design team to assess soil resistance and design conductor and surface casing strings that supports all construction

and operation loads. Piezocone Penetration Test (CPTu) is commonly used as it is simple and provides comprehen-
sible soil data. Using CPTu dataset, it is possible to characterize geomechanical behavior using statistics presented

in the literature. Application of statistics procedures in parameter characterization must be done using samples that
attend stationarity criteria i.e. constant mean and variance of the fluctuating component. Present work applies a

modified Bartlett Statistics in experimental marine soil CPTu data to evaluate sample sizes that satisfy stationar-
ity criteria. Initally, trends in parameters are calculated using linear regression analysis and layer classification is

applied in detrended parameters. Subsequently, autocorrelation samples are evaluated and fit into models. These
autocorrelation models estimate the scale of fluctuation which measures high statistical correlation domain and
Bartlett profiling is then applied. Bartlett’s peak value is compared with critical modified Bartlett Test formula
evaluated by Phoon et al. [1] to assess if stationarity of the sample is within 5% confidence level. Results show
sample intervals that achieve stationarity.