Practical Use of Accelerated Test Data for the Prognostics Methods

Dawn An, Joo-Ho Choi, and Nam-Ho Kim
Submission Type: 
Full Paper
phmc_13_032.pdf105.08 KBSeptember 10, 2013 - 1:12pm
phmc_13_032.pdf694.04 KBSeptember 11, 2013 - 9:26pm

Prognostics is to predict future damage/degradation of in-service systems and the remaining useful life based on the damage data obtained at previous usage. The damage data is of great importance regardless of prognostics methods used, while it is very expensive to obtain the data because of time and cost. Instead, companies frequently use accelerated test data for the purpose of design, which is obtained under much severe operating conditions. This paper presents a method of utilizing accelerated test data for the purpose of prognostics. The uncertainty caused by mapping between nominal and accelerated operating conditions is taken into account using the Bayesian framework. As an example, crack growth data are synthetically generated under over-loaded conditions, which are utilized for both of data-driven and physics-based approaches under different conditions. Using accelerated test data increases prediction accuracy in early stage of physics-based prognostics as well as it covers insufficient data problem of data-driven prognostics.

Publication Year: 
Publication Volume: 
Publication Control Number: 
Submission Keywords: 
accelerated testing
data driven prognostics
Physics based prognostics
neural network
particle filter
Crack Growth
Submission Topic Areas: 
Data-driven methods for fault detection, diagnosis, and prognosis
Submitted by: 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed