Modélisation de la viscosité des fluides de gisement : Apport de la PLS bootstrap et des réseaux de neurones

Partial Least Squares linear regression (PLS) with bootstrap sampling, and Multi Layer Perceptron (MLP) non-linear regression were used to elaborate new viscosity models for estimating crude oil viscosity using fluid composition, temperature, and pressure. Statistical models obtained concern values from 0.21 cP to 10000 cP.