A multiple-point monitoring model for concrete dam displacements based on correlated multiple-output support vector regression. (November 2022)

• Record Type: Journal Article
• Authors: Ren, Qiubing; Li, Mingchao; Bai, Shuo; Shen, Yang
• Is Part Of: Structural health monitoring. Volume 21:Number 6(2022)
• Page Start: 2768
• Page End: 2785
• Publication Date: 2022-11
• Subjects: Dam health monitoring -- displacement behavior -- multiple-measurement points -- spatiotemporal correlation -- improved support vector regression; Structural health monitoring -- Periodicals; Structural stability -- Periodicals; Strength of materials -- Periodicals; Nondestructive testing -- Periodicals; Constructions -- Stabilité -- Périodiques; Résistance des matériaux -- Périodiques; Contrôle non destructif -- Périodiques; Electronic journals; 624.17
• DOI: 10.1177/14759217211069639
• Languages: English
• ISSNs: 1475-9217

Abstract

Displacements reflect the overall behavior of a concrete dam; thus, it is of vital importance to evaluate the overall structural health status by displacement-based mathematical monitoring models. However, most of the existing monitoring models focus on point-by-point displacement modeling, ignoring the correlations among displacements at different measurement points. This study therefore proposes a model for dam multiple-point displacement monitoring based on the support vector regression (SVR) algorithm. The improved SVR-based model with multiple-output formulation is a new development based on the statistical learning theory, which can simultaneously analyze and predict displacements at multiple-measurement points. Furthermore, by introducing the weight vectors that separate the common and individual information, the potential correlations among multiple-point displacements can be fully exploited by the multiple-output SVR. Combining the above two improvements, a multiple-point monitoring model for dam displacements considering spatiotemporal correlations, referred to as correlated multiple-output SVR (CMOSVR), is constructed. The proposed model is verified using in-situ monitoring from a full-scale concrete gravity dam. The accuracy, robustness, and efficiency of the CMOSVR-based model are compared with those of conventional single-point monitoring models, such as classical hydrostatic-seasonal-time model and standard SVR-based model. Empirical results show that in both real and simulated noisy scenarios, the CMOSVR-based multiple-point model can achieve a better monitoring performance with less modeling time cost. Moreover, the superior performance of CMOSVR-based model does not require a very strong correlation among multiple-point displacements, which considerably improves the adaptability of the monitoring model to various possible scenarios. The novel multiple-point model will provide an effective technical support tool for ensuring the safe operation of dams.