A new framework for remaining useful life estimation using Support Vector Machine classifier

In this paper a framework for remaining useful life estimation is presented. Remaining useful life of a system or an equipment is the time period between the current time instant and the time instant when the system stops operating within its predefined specifications. It is an important part required for condition based maintenance, which increases the safety, quality, reliability and reduces the operating costs of a process. The framework consists of two parts, which are health feature creation and remaining useful life estimation. Therefore, a new health feature creation approach is proposed using binary Support Vector Machine classifier, which is also used to obtain fault detection as an additional feature. As degradation of the health feature, a Weibull distribution is assumed, which is common for performance degradation of equipment due to aging. The remaining useful life is then calculated using an identified Weibull function, where a weighted least squares algorithm is employed for the identification of the Weibull parameters.