ABSTRACT: The general topic investigated in this thesis is “structural health monitoring”, with special care being devoted to damage detection and localization. The study is focused on methods which work in the space of the observed... more
ABSTRACT: The general topic investigated in this thesis is “structural health monitoring”, with special care being devoted to damage detection and localization. The study is focused on methods which work in the space of the observed variables, where the global information measures can be used to detect the damage. In this context damage detection is demanded to global measures such as entropy or Lyapunov exponents. The comparison of these measures computed in subspaces of the observed variables or in their expansions, allows one to perform a sort of sensitivity analysis on the role of the variables involved in the problem.
Such global measures are able to detect but not to directly localize damage. For this purpose, a new approach to detect and localize the damage is formulated and implemented. It exploits response surface techniques to approximate the relationship among the observed variables. Comparing the resulting models in different damaged and undamaged situations identifies the response differences and their causes.
The response surface procedure is described from its basic theoretical aspects up to the features of its numerical implementation. Additionally, the global measures introduced above drive the analyst in establishing whether or not the problem is well posed, i.e., whether or not the set of measured quantities are able to detect the specific damage to be localized.
The validation of the proposed methodology is first numerically pursued by applying it to the benchmark case set up by the Structural Health Monitoring Panel of the American Society of Civil Engineering (ASCE). The procedure is then tested on two different experimental situations. While the first study of a three-dimensional steel frame reproduces the features of the benchmark problem, the second investigation covers a monumental structure situation, where damage is represented by more or less extended cracks in the masonry. The potential expressed by the procedure also in this second application seems to promise interesting exploitation possibilities.
Such global measures are able to detect but not to directly localize damage. For this purpose, a new approach to detect and localize the damage is formulated and implemented. It exploits response surface techniques to approximate the relationship among the observed variables. Comparing the resulting models in different damaged and undamaged situations identifies the response differences and their causes.
The response surface procedure is described from its basic theoretical aspects up to the features of its numerical implementation. Additionally, the global measures introduced above drive the analyst in establishing whether or not the problem is well posed, i.e., whether or not the set of measured quantities are able to detect the specific damage to be localized.
The validation of the proposed methodology is first numerically pursued by applying it to the benchmark case set up by the Structural Health Monitoring Panel of the American Society of Civil Engineering (ASCE). The procedure is then tested on two different experimental situations. While the first study of a three-dimensional steel frame reproduces the features of the benchmark problem, the second investigation covers a monumental structure situation, where damage is represented by more or less extended cracks in the masonry. The potential expressed by the procedure also in this second application seems to promise interesting exploitation possibilities.
