Open AccessParameter adaptive LS-SVR based on multi-stages division and cluster sampling for remaining useful life prediction
Author(s) -
Jianwen Yan,
Xiao-hu Zhong,
San-min Guo,
Yu Fan,
Qiang Zhang
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/806/1/012041
Subject(s) - prognostics , support vector machine , robustness (evolution) , computer science , data mining , division (mathematics) , sampling (signal processing) , cluster (spacecraft) , test data , field (mathematics) , reliability engineering , artificial intelligence , engineering , mathematics , biochemistry , chemistry , arithmetic , filter (signal processing) , computer vision , gene , programming language , pure mathematics
Remaining useful life (RUL) prediction is an advanced technology to manage life cycles of equipment and to reduce maintenance cost. Support vector regressive (SVR) is one of the frequently-used data driven method in the field of prognostics, and is much suitable for analyzing small samples and multi-dimensional data. However, there is a limitation of traditional SVR model that the linearity will increase as the training data increase. In order to overcome it, this paper proposes the parameter adaptive least square support vector regressive (LS-SVR) based on multi-stages division and cluster sampling. It divides the degradation process of the equipment into several stages according to the value of degradation data and selects the model parameters stage by stage. The optimal model parameters of each stage are calculated through the novel testing data, which are obtained through appending the cluster sampling data to the original testing data. Therefore, the optimal model parameters can be selected adaptively online as the degradation evolves. Finally, the proposed method is verified systematically by a simulated dataset of fatigue crack growth and a real-world degradation dataset of GaAs-based lasers. It is shown that the proposed method is of better effectiveness, reliability and robustness.