Abstract
Oil extraction in deep waters implies the drilling of a salt layer, whose thickness ranges from 1000 m to 2000 m. Unscheduled shutdowns with interruption of production can result from the deposition of lime scale, which can block valves, tubing and flowlines of the oil wells plants. Thus, the prediction of lime scale is fundamental in order to control the performance of valves and actuators, which are expected to operate for a certain period of time under required operating conditions. To this purpose, laboratory tests have been conducted on different materials (25-Cr, 13-Cr, In718, WC) used in off-shore drilling in order to study lime scale deposition under controlled environmental and working conditions (Moura et al., 2011). The data collected in these tests are analysed in this work, to study the influence of the experiment inputs (e.g., material, roughness, sample initial weight, test temperature, etc.) on the outputs (i.e, rate of scale deposition and lime scale final weight and thickness). The idea is to identify the inputs which have the greatest influence on lime scale thickness variations to eventually use them to build an empirical model of the degradation of the oil wells valves and actuators. A sensitivity analysis is performed by resorting to a classification tree, a tree-like graph in which each laboratory test occupies one of the leaves of the tree. The key point of the method is the choice of the selection rules to classify the data. The input variances in the subclasses which are generated at every branching step provide useful information about the most significant inputs. Results are confirmed by different correlation coefficients (Linear, Spearman Rank Order and Kendall’s Tau).
