This paper presents a fully Bayesian approach for lifetime prediction of an
automotive fleet using real workshop-service data as input. Fleet lifetime prediction is an important issue for the decision-making process in the automotive industry. For instance, fleet lifetime prediction assists engineering teams in monitoring their products under real-life conditions in the field. Furthermore, it provides future cost-projections to the board which thereupon can guide financial support effectively to research and development projects.
Also, fleet lifetime prediction plays a key-role in planning spare part pre-production to enable early closing of current productions lines in favor of re-configuring them for new products.
The paper explores a problem instance containing more than 170\,000 individual vehicles driving in 100 different countries exhibiting a certain failure, pre-selected for this study. The shape of the corresponding lifetime distribution suggests the presence of at least two different mechanisms that lead to failure. This behavior has to be taken into account in the lifetime prediction model, i.e., simple models were not sufficient to describe the lifetime accurately,
but more complex mixture models had to be utilized.
The suggested lifetime estimator consists of a combination of two probabilistic models. For a given point in time in the future, the first model predicts the effective age of each individual of the fleet. The second model attempts to forecast the total number of failures that will arise for the entire fleet. Both probabilistic models are fully Bayesian, i.e., all parameters of the models are implemented as probability distributions and computations are solely performed on distributions rather than on summarizing statistics. Thus, the predictive output of a Bayesian model is not a point estimate but
again a probability distribution, the so-called posterior predictive distribution (PPD). As a consequence, uncertainty of the predictions is made accessible in a very natural way and can be taken into account in the decision-making process
explicitly. As will be shown in the paper, working with the PPD becomes especially important, or even indispensable, when dealing with decision-making cost-functions that are not symmetric.
How to Cite
Bayesian Lifetime Prediciton
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
The Prognostic and Health Management Society advocates open-access to scientific data and uses a Creative Commons license for publishing and distributing any papers. A Creative Commons license does not relinquish the author’s copyright; rather it allows them to share some of their rights with any member of the public under certain conditions whilst enjoying full legal protection. By submitting an article to the International Conference of the Prognostics and Health Management Society, the authors agree to be bound by the associated terms and conditions including the following:
As the author, you retain the copyright to your Work. By submitting your Work, you are granting anybody the right to copy, distribute and transmit your Work and to adapt your Work with proper attribution under the terms of the Creative Commons Attribution 3.0 United States license. You assign rights to the Prognostics and Health Management Society to publish and disseminate your Work through electronic and print media if it is accepted for publication. A license note citing the Creative Commons Attribution 3.0 United States License as shown below needs to be placed in the footnote on the first page of the article.
First Author et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 United States License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.