Failure prognostics is an open research field with options of deeper investigations. Traditionally the applications are statistical methods which do not give accurate results. As the aeronautic systems have become more complex the accuracy of prognosis is already a fact that needs to be considered. Based on the necessity of methods that improve prediction failure time, based-model prognostic failure methods combined with data-model prognostic are receiving more attention than only data-model prognostic. This paper comes from this paradigm, which is based on mathematical modeling building-up a mathematical procedure which results in estimation of a system failure. In order to reach this goal a new definition is proposed, the definition of Failure Severity which estimates the fault evolution until that fault becomes unbearable. So, this paper gives a new methodology that predicts the time to failure before the systems flag it. The results are simulated with prevision for bench-test validation.