Bearing Degradation Prediction Based on Support Vector Regression
(a)Sutawanir Darwis, (b)Nusar Hajarisman, (c)Suliadi, (d)Achmad Widodo
(a),b),(c)Department of Statistics, Bandung Islamic University
(d)Department of Mechanical Engineering, University of Diponegoro Semarang
(a)std.darwis[at]gmail.com, (b)nusarhajarisman[at]yahoo.com, (c)suliadi[at]gmail.com, (d)awidodo2010[at]gmail.com
Abstract
A life table gives probabilities based in failure per thousands in a given year and used to help determine premiums. The construction of life table is based on time to failure of the unit which is converted to a duration vector and used to calculate the estimate of life table using moment method. Bearings are one most present in turbine power generation. To assess turbine degradation, risk accident prognosis of bearings is needed. Predicting bearing degradation before reaching the state of risk of accident is one important issues in power generation insurance. This paper proposes a method based on support vector regression to achieve the goal. The results will be used to predict the remaining useful life of the bearings. The method is applied on PRONOSTIA dataset which is an experimental platform dedicated to test methods related to bearing health assessment. The results shows that the method can effectively model the evolution of the bearing degradation.
Keywords: bearing degradation, remaining useful life, support vector regression, power generation insurance, PRONOSTIA
Topic: Statistics