Model based fault diagnosis for drillstring washout using iterated unscented Kalman filter

Abstract

In this study, a model based fault diagnosis method for drillstring washout is proposed, which uses iterated unscented Kalman filter (IUKF) to detect the emergence of drillstring washout and to estimate washout depth and washout rate. To quantify the changes in pressure-loss and annulus outlet flow rate resulting from drillstring washout, pressure-loss factors and flow rate factor are introduced into the governing equations of circulating drilling fluid under normal drilling condition. Pressure-loss factors and flow rate factor are estimated by IUKF with updated measurements, and the emergence of drillstring washout is automatically detected by confirming the changes in pressure-loss factors and flow rate factor using generalized likelihood ratio test (GLRT). After drillstring washout is detected, continuously updated pressure and flow rate measurements are sent to the identification model of drillstring washout to estimate washout depth and washout rate by IUKF. In addition, the performances of fault diagnosis method for drillstring washout using IUKF and unscented Kalman filter (UKF) are contrasted.Numerical simulation indicates that IUKF and UKF have equivalent performance in drillstring washout detection, while IUKF shows a better performance in washout depth estimation. When using IUKF, the average relative errors of estimated washout depth and washout rate are 2.1% and 1.5% respectively. The errors of estimated washout depth and washout rate are 2.8% and 1.49% when using UKF. Robustness analysis indicates that IUKF is more robust than UKF to measurement noise. IUKF and UKF both have the ability to estimate washout depth and washout rate within the noise scope of 0%–0.8%, while IUKF is more accurate than UKF for washout depth estimation.