Nonparametric estimation of marginal failure distributions from dually censored automotive data

It is not uncommon that a component's reliability characteristics depend on two usage variables. As an example, for automotive components such variables are time in service and accumulated mileage. For certain airplane chassis components, these variables are accumulated flight hours and the number of landings. The problem of failure time distribution estimation from dually censored (at warranty limits) automotive data has been addressed by many authors. However, most of published work is focused on estimating marginal in time failure distributions or (less frequently) joint failure distributions. This paper focuses on nonparametric estimation of the marginal in mileage failure distributions, since for the majority of the automotive components mileage is considered a more relevant survival variable than time in service.