Estimating and Forecasting Failure Rate Processes by Means of the Kalman Filter

摘要

A new nonparametric method is described for analyzing failure data. The approach used is to model the logarithm of the failure-rate process as a linear dynamic system with observations. This formulation permits the underlying failure-rate process to be corrupted by noise from various sources. In addition, the observations of the process are functions of simple nonparametric failure-rate estimates which are assumed to be noisy. The Kalman filter equations are used to provide the estimates and future forecasts. An example is provided. 5 figures. (ERA citation 02:023865)