Blue Room - Session 9
2:15 to 3:15 p.m. Wednesday April 22, 2015
A Practical Approach to Consider Several Usage Variables in Reliability Analyzes with an Example from Automotive Field Data
This presentation describes an iterative two-step methodology to analyze reliability data when multiple usage variables are assumed to influence the failure behavior. This situation is becoming more common in the context of technical reliability due to the fast development of complex new technology. In addition, reliability experts now have access to more data from various sources to work with (e.g., electronic control units in vehicles or machines).
The idea is to start off with an initial model which does not include any covariates. Then the influence of every possible covariate is measured so the model can be extended by an appropriate variable. Now the performance of the extended model is compared to the first model to indicate if the inclusion of the additional covariate led to a model improvement. Afterwards the two steps of identifying the influence of the remaining covariates and extending the model are repeated until the model's performance is satisfying. Based on an application to automotive field data it is explained how to identify the relevant variables, and several alternatives to model the influence of multiple usage variables are suggested. The methodology is a general approach and can be applied to all fields of reliability-related questions.
Key Words: Multivariate Reliability Modeling, Usage Based Reliability Analysis, Multiple Usage Variables, Warranty Data Analysis, Automotive Reliability Data
Thomas Köttermann, Andreas Jacobi and Stefan Bracke