TY - JOUR
T1 - LQD-RKHS-based distribution-to-distribution regression methodology for restoring the probability distributions of missing SHM data
AU - Chen, Zhicheng
AU - Bao, Yuequan
AU - Li, Hui
AU - Spencer, Billie F.
N1 - Funding Information:
This work was financially supported by the National Natural Science Foundation of China (Grant Nos. 51638007 U1711265 and 51678203 ), the Ministry of Science and Technology of China (Grant No. 2015DFG82080 ), and the China Scholarship Council (CSC).
Funding Information:
This work was financially supported by the National Natural Science Foundation of China (Grant Nos. 51638007 U1711265 and 51678203), the Ministry of Science and Technology of China (Grant No. 2015DFG82080), and the China Scholarship Council (CSC).
Publisher Copyright:
© 2018 Elsevier Ltd
PY - 2019/4/15
Y1 - 2019/4/15
N2 - Data loss is a critical problem in structural health monitoring (SHM). Probability distributions play a highly important role in many applications. Improving the quality of distribution estimations made using incomplete samples is highly important. Missing samples can be compensated for by applying conventional missing data restoration methods; however, ensuring that restored samples roughly follow underlying distributions of true missing data remains a challenge. Another strategy involves directly restoring the probability density function (PDF) for a sensor when samples are missing by leveraging distribution information from another sensor with complete data using distribution regression techniques; existing methods include the conventional distribution-to-distribution regression (DDR) and distribution-to-warping function regression (DWR) methods. Due to constraints on PDFs and warping functions, the regression functions of both methods are estimated from the Nadaraya-Watson kernel estimator (a local linear smoothing technique) with relatively low degrees of precision. This article proposes a new indirect distribution-to-distribution regression approach to restoring distributions of missing SHM data via functional data analysis. PDFs are represented by ordinary functions after applying log-quantile-density (LQD) transformation. The representation function of the missing distribution is first restored from a functional regression model constructed in Reproducing Kernel Hilbert Space (RKHS) by solving an optimization problem and subsequently mapping back to the density space through inverse log-quantile-density transformation. The performance of the proposed approach is evaluated through application to field monitoring data. Test results indicate that the new method significantly outperforms conventional methods in general cases; however, in extrapolation cases, the new method is inferior to the distribution-to-warping function regression method.
AB - Data loss is a critical problem in structural health monitoring (SHM). Probability distributions play a highly important role in many applications. Improving the quality of distribution estimations made using incomplete samples is highly important. Missing samples can be compensated for by applying conventional missing data restoration methods; however, ensuring that restored samples roughly follow underlying distributions of true missing data remains a challenge. Another strategy involves directly restoring the probability density function (PDF) for a sensor when samples are missing by leveraging distribution information from another sensor with complete data using distribution regression techniques; existing methods include the conventional distribution-to-distribution regression (DDR) and distribution-to-warping function regression (DWR) methods. Due to constraints on PDFs and warping functions, the regression functions of both methods are estimated from the Nadaraya-Watson kernel estimator (a local linear smoothing technique) with relatively low degrees of precision. This article proposes a new indirect distribution-to-distribution regression approach to restoring distributions of missing SHM data via functional data analysis. PDFs are represented by ordinary functions after applying log-quantile-density (LQD) transformation. The representation function of the missing distribution is first restored from a functional regression model constructed in Reproducing Kernel Hilbert Space (RKHS) by solving an optimization problem and subsequently mapping back to the density space through inverse log-quantile-density transformation. The performance of the proposed approach is evaluated through application to field monitoring data. Test results indicate that the new method significantly outperforms conventional methods in general cases; however, in extrapolation cases, the new method is inferior to the distribution-to-warping function regression method.
KW - Distribution-to-distribution regression
KW - Functional data analysis
KW - Log-quantile-density transformation
KW - Missing data
KW - Probability distribution
KW - Reproducing Kernel Hilbert Space
KW - Structural health monitoring
UR - http://www.scopus.com/inward/record.url?scp=85057782975&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85057782975&partnerID=8YFLogxK
U2 - 10.1016/j.ymssp.2018.11.052
DO - 10.1016/j.ymssp.2018.11.052
M3 - Article
AN - SCOPUS:85057782975
SN - 0888-3270
VL - 121
SP - 655
EP - 674
JO - Mechanical Systems and Signal Processing
JF - Mechanical Systems and Signal Processing
ER -