Applicability of the Statistical Pattern Recognition Paradigm for Structural Health Monitoring of Bridges

• Autor(a) principal: Pereira, João Tiago Martins Neves
• Data de Publicação: 2012
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.14/15656
• topic: Damage detection; Bridge failures; Statistical Pattern Recognition Paradigm; Structural health monitoring; Deteção de dano; Desastres de pontes; Paradigma de Reconhecimento de Padrões; Monitorização da integridade estrutural; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Civil

In the last decades, health monitoring systems have gained an increasing importance in our society. The main purpose of these systems is to support the engineers to get more insight into the behavior of structures under service conditions, so they can optimize and improve maintenance programs and, hopefully, to avoid structural failures or disasters. It is possible to integrate these systems in any type of civil or mechanical infrastructure. However, in this dissertation, the preferential targets are the civil infrastructures with major strategically importance in the social environment, such as bridges and viaducts. Therefore, the goal of this dissertation is (i) to review the most recent bridge collapses in order to unveil the main causes and challenges posed by those catastrophic events; (ii) to review the concept and need of Structural Health Monitoring (SHM) of bridges as well as its associated potential for significant life-safety and economic benefits; and (iii) to study the applicability of the SHM concepts. Due to recent promising research developments, the SHM process is posed in the context of the Statistical Pattern Recognition (SPR) paradigm, which tries to implement a damage identification strategy based on the comparison of different state conditions. The applicability of the SHM-SPR paradigm is studied by applying its concepts in two separate cases: firstly on data sets from a base-excited three-story frame structure, created and tested in a laboratory environment at Los Alamos National Laboratory; secondly, on data sets from a real-world bridge, namely the Z24 Bridge in Switzerland. The major contributions of this dissertation are the extension of previous results obtained by Figueiredo et al. from the three-story frame structure and the development and application of an algorithm that uses a Gaussian mixture model as a way of improving the feature classification performance under varying operational and environmental conditions.