Survival Estimation for Opportunistic Maintenance

The problem of rational maintenance of aircraft engines is studied with respect to the influence of random events. Aircraft engines can be more economically maintained and resources can be saved if the maintenance process is improved. The starting point is an optimization model suggesting what parts in the engine that should be replaced at each maintenance time. The input data is the age of the details in the engine. Statistical models are developed that estimates the remaining life of the components in the engine. The models work with different kinds of data. The first data set only contains times between repairs and is modeled with a non-stationary renewal process and a non-homogeneous Poisson process. With our data the non-stationary renewal process works better. Different repair stations affect the life of the components, which the non-stationary renewal process manages to model. This model also manages the aging component problem in an effective way. However, in this case no aging is present other than substantial degeneration after the first repair. The second dataset contains crack growth data. The remaining life is modeled with a empirical crack growth model. With data directly indicating the condition of the detail a more precise estimate of the reaming life can be made. In order to get an interface with the optimization model the distributions need to be discrete. Four methods to make discretizations are discussed and adapted to suit the model. The methods are compared and the choice concerning the number of points of support is discussed. Finally the consequence of using a narrow scenario tree is commented upon.