Title: Facility health maintenance through SVR-driven degradation prediction

Authors: Xiangang Cao, Pingyu Jiang, Guanghui Zhou

Addresses: State Key Laboratory for Manufacturing Systems Engineering, School of Mechanical Engineering, Xi'an Jiaotong, University, Xi'an, Shaanxi 710049, China. ' State Key Laboratory for Manufacturing Systems Engineering, School of Mechanical Engineering, Xi'an Jiaotong, University, Xi'an, Shaanxi 710049, China. ' State Key Laboratory for Manufacturing Systems Engineering, School of Mechanical Engineering, Xi'an Jiaotong, University, Xi'an, Shaanxi 710049, China

Abstract: In order to realise the health monitoring and maintenance of complex facilities with multiple degradation parameters, a facility synthetic failure probability model to map between inputs and probability of failure is established through adopting the logistic regression to synthesise each degradation parameter. Then, a SVR-driven degradation trend prediction and estimate of Remaining Useful Life (RUL) method is put forward. Last, based on Monte-Carlo method, a multi-parameters equipment emulator according with Weibull distribution is established to test the model. The results show that these methods are practicable.

Keywords: health monitoring; health maintenance; synthetic failure probability models; logistic regression; support vector regression; SVR; degradation trends; facility degradation; fault diagnosis.

DOI: 10.1504/IJMPT.2008.019781

International Journal of Materials and Product Technology, 2008 Vol.33 No.1/2, pp.185 - 193

Published online: 30 Jul 2008 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article