Oil condition monitoring of gears onboard ships using a regression approach for multivariate T2 control charts. (October 2016)

• Record Type: Journal Article
• Authors: Henneberg, Morten; Jørgensen, Bent; Eriksen, René L.
• Is Part Of: Journal of process control. Volume 46(2016:Oct.)
• Page Start: 1
• Page End: 10
• Publication Date: 2016-10
• Subjects: 70-05; Condition monitoring -- Gear oil -- Mean value regression -- Quasi-stationary -- T2 control chart; Process control -- Periodicals; Fabrication -- Contrôle -- Périodiques; Process control; Periodicals; Electronic journals; 660.281
• DOI: 10.1016/j.jprocont.2016.07.001
• Languages: English
• ISSNs: 0959-1524

Abstract

Abstract : Highlights: We model condition of equipment in a lubricating oil system from sensor data measuring oil parameters and wear debris. We use a model that takes into account ambient conditions by mean value regression for T 2 charts. A selection of phase I period for T 2 modelling is established from run-in data and historical data. The model is applied on data from four ship gear and used to determine equipment conditions. A warning model for equipment progression and trending is presented. Abstract: In this paper, we present an oil condition and wear debris evaluation method for ship thruster gears using T 2 statistics to form control charts from a multi-sensor platform. The proposed method takes into account the different ambient conditions by multiple linear regression on the mean value as substitution from the normal empirical mean value. This regression approach accounts for the bias imposed on the empirical mean value due to different geographical and seasonal differences on the multi-sensor inputs. Data from a gearbox are used to evaluate the length of the run-in period in order to ensure only quasi-stationary data are included in phase I of the T 2 statistics. Data from two thruster gears onboard two different ships are presented and analyzed, and the selection of the phase I data size is discussed. A graphic overview for quick localization of T 2 signaling is also demonstrated using spider plots. Finally, progression and trending of the T 2 statistics are investigated using orthogonal polynomials for a fix-sized data window.