Skip to Main content Skip to Navigation

Surveillance de santé structurale des ouvrages d'art incluant les systèmes de positionnement par satellites

Abstract : Structural health monitoring (SHM) aims to characterize and monitor the behaviour and performance of civil engineering structures over time. SHM relies on the definition of health or performance indicators in order to detect, and if possible locate, quantify or predict damage on the structure. Nowadays, SHM still makes little use of the absolute displacements of a structure, as the majority of the deployed sensors are limited to local or relative measurements. The Global Positioning System (GPS) has been fully available publicly since 2000. It has since been joined by other systems with similar characteristics, forming the Global Navigation Satellite Systems (GNSS), which can achieve absolute positioning with an accuracy up to a few millimeters. This PhD thesis takes place in a context where GNSS solutions appear as good complementary tools to traditional SHM instrumentations. Results of recent studies using lowcost GNSS solutions (small antennas, single frequency receivers) have motivated this research work to focus on similar hardware. The decrease of hardware performance is offset by using some specific processing parameters, and by the possibility to deploy more sensors on a structure for a similar cost. This work focuses on large structures such as bridges, dams, or high-rise buildings, whose monitoring is complex due to their dimensions, and which are more likely to have displacements that can be detected with GNSS stations. The purpose of this work is to answer the following question: is it possible, and how, to use low-cost GNSS stations to detect abnormalities or changes in a structure's behaviour? The first part of this thesis is dedicated to the experimental assessment of various low-cost GNSS solutions. Multiple combinations of antennas and receivers have been evaluated using relative phase positioning in order to select an efficient station. In parallel, several processing parameters have been evaluated in fixed and dynamic experimental scenarios, and an optimal GNSS solution, including hardware and processing, was selected. This solution is able to achieve sub centimeter accuracy. In the second part, datasets from two civil engineering structures equipped with smart GNSS sensor networks are studied. The components of the deployed sensors, the Geocube, are similar to those evaluated in the first part. Time series analysis through comparison and correlation with temperature data validated that the on-site low-cost GNSS stations are able to monitor the effect of environment on structural response. This study also highlights the redundancy of observations with various levels of correlations between the GNSS time series. The last part of this research is devoted to the use of predictive models of on-site GNSS time series for novelty detection. In the absence of complete thermal data or mechanical models available for the investigated structures, an approach using machine learning models with data from multiple sensors of a GNSS network was chosen. The comparison of several regression tools led to select recurrent neural networks (RNN) as the optimal tool. Finally, an automated novelty detection strategy, exploiting several predictive models for each sensor, was proposed and tested on both real and simulated anomalies.
Document type :
Complete list of metadata
Contributor : Ifsttar Cadic <>
Submitted on : Wednesday, February 17, 2021 - 9:01:31 AM
Last modification on : Monday, May 10, 2021 - 9:57:05 AM


Files produced by the author(s)


  • HAL Id : tel-03143785, version 1



Nicolas Manzini. Surveillance de santé structurale des ouvrages d'art incluant les systèmes de positionnement par satellites. Structures. Université Paris-Est, 2020. Français. ⟨tel-03143785⟩



Record views


Files downloads