Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
In developed countries, the management of existing civil infrastructure is challenging due to evolving functional requirements, aging and climate change. Due to conservative approaches in construction design and practice, infrastructure often has hidden reserve capacity and this has potential to improve decisions related to asset management. For example, improved knowledge of behavior of load capacity through bridge measurement can be exploited to extend lifetimes and optimize retrofit designs. The assessment of bridge reserve capacity requires predictions of structural behaviour under actions. This behaviour is influenced by several parameters that are difficult to estimate, such as material properties. Field measurements, collected through load testing, may help in the identification of unknown parameter values and this process is called structural identification. Then, the reserve capacity is assessed using the updated behaviour model. The design of measurement systems is usually carried out by engineers using only qualitative rules of thumb. However, finding the optimal design is difficult due to redundancies in information gained from sensors. Suboptimal sensor configurations are often selected by engineers, reducing the information collected during monitoring. This thesis proposes a range of quantitative methodologies for measurement-system design in order to improve structural-identification performance. Measurement-system design for structural identification depends on the actions applied during measurements. This aspect is often neglected in sensor-placement algorithms. A methodology is presented to maximize the information gain of measurement systems when multiple static load tests are involved. The optimal design of measurement systems depends on multiple conflicting performance criteria such as information gain and cost of monitoring. A multi-objective approach for measurement system design thus leads to more informed decision making when several performance criteria are important. A framework is proposed that quantitatively accounts for multiple objective functions to recommend measurement systems based on asset-manager preferences. Monitoring is justified only when asset-manager decisions regarding bridge safety can be influenced by the information gain during load testing. A methodology is introduced to quantitatively assess whether monitoring information has the potential to influence reserve-capacity assessments and the value of information is assessed probabilistically. These proposals have been tested and verified with three case studies of full-scale bridges and one case study of a retaining wall in an excavation. Proposed methodologies for measurement-system design outperform other approaches and qualitative engineering rules of thumb. Sensor-performance predictions have then been corroborated using field measurements. Furthermore, the excavation case study demonstrates that the selection of the most informative data sets is crucial when sensor information is used to update complex models. This thesis demonstrates that rational methodologies to design measurement systems enhance the performance of structural identification, leading to better asset-management decisions. More generally, the measurement-point-selection methodologies proposed in this thesis have the potential to help maximize the information extracted from large data sets collected by the Internet of Things at city scale.
,