Drillability of Rock Formations Assessment by Relevance Vector Machine Based on Particle Swarm Optimization

Abstract:

This paper presents a relevance vector machine based on particle swarm optimization method (PSO-RVM) for assessing the drillability of rock formations. Five parameters, including the depth of rock formations (H), acoustic travel time (AC), electrical resistivity (ρd), density of rock formations (ρ) and the shaliness of rock formations(Vsh), are selected as the basic parameters in the PSO-RVM model. The Du4 well drilling in an oil field is chosen as an example and the PSO-RVM method, multiple regression method and relevance vector machine (RVM) method are used to assess the rock formation drillability of the well. The results suggest that the predict results of PSO-RVM method accord well with the measured data and the prediction accuracy is significantly higher than that of multiple regression method and the RVM method. It is shown that PSO-RVM method can be applied in the prediction of rock formation drillability with its advantages and high accuracy.