Modeling Rate of Occurrence of Failures with Log-Gaussian Process Models: A Case Study for Prognosis and Health Management of a Fleet of Vehicles

摘要

Gaussian Process Regression (GPR) is a flexible non-parametric regression technique that provides an alternative solution to the model selection problem commonly seen in parametric models. GPR models are able to easily model complex non-linear relationships that are often present in rate of occurrence of failure data. The predictive distributions that result from the regression models are then used to provide valuable insights into the behavior of the data. An example of the application of this approach has been demonstrated for modeling the rate of occurrence of failure of a fleet of vehicles on a monthly basis. The Log-GPR model applied in this context is useful for detecting significant reliability problems that may occur over time. In order to assess the generalization capability of the model to accurately represent a test data set, a goodness of fit test is also presented.