Blue Room - Session 13

2:15 to 3:15 p.m. Thursday April 23, 2015

Bayesian Monte-Carlo Methods to Consider Dynamic Failure Modes in Early Field Prognosis

As part of the monitoring process inside the use phase, consistent analysis and forecasting methods are used to extrapolate the expected failure amount of technical components. A specific challenge is the precise forecasting shortly after start of production (SOP) of a new product, due to the given technical boundaries like short usage time or minor operational demands. Besides the fact that most failure modes are not totally known at an early stage of the use phase, an additional impact is the skewed field data which can be the result of strategic failure part requests.

This presentation shows additional methods that can be implemented into existing forecasting models to enhance the results of early field data prognosis. Thus, the focus is the mathematical implementation of failure mode distributions into existing forecasting models and the consideration of dynamically changing failure modes under usage of Bayesian Monte-Carlo methods.

Key Words: Field Reliability Prognosis, Bayesian Monte-Carlo-Simulation, Automotive Engineering, Case Study

Jens Michalski and Andreas Braasch

Institute for Quality and Reliability Management GmbH