M.Eng. Dongsheng Li

Structural health monitoring provides vital information for the safe operation of key civil structures and enables operational cost reduction by performing prognostic and preventative maintenance, and also shows great potential for disaster mitigation, for instance the catastrophic roof collapse of Terminal 2E at Paris Charles de Gaulle airport. The first issue in structural health monitoring is to deploy sensors, i.e. how to optimally place sensors in a large, spatially extended structure so that data acquired from those locations will result in the best identification of structural characteristics. The optimal sensor placement problem can be approached from several different directions, for instance, visual inspection method, modal kinetic energy method, the effective independence method, and many sensitivity-based methods. Because structural health monitoring is a relative immature research field, these investigations are unsystematic and in- adequate to guide emerging monitoring demands. Therefore, this research aims to develop a theory to seek as small as possible number of sensors that contain as much information as possible about the health state of a structure, and to establish corresponding criteria for evaluating the effectiveness of different sensor placement methodologies, and as well to provide practical solutions for engineering monitoring projects.

The objective of current work is to deepen the understanding of existing influential sensor placement methods and to develop an effective method to deploy accelerometers suitable for structure health monitoring. Furthermore, we aim to find a sufficient evaluation criterion to judge which topology configuration of accelerometers outperforms than others. Up to now, we have revealed the connections and interrelationship of the influential Modal Kinetic Energy method, the Effective Independence method, QR row decomposition method and the MinMAC method, etc. from a mathematical point of view.  An extended MinMAC algorithm is proposed to overcome the disadvantages of traditional MinMAC algorithm with the introduction of a forward- and backward combinational approach. Furthermore, we propose a loading dependent sensor placement method, specially tailored for structural health monitoring. The method is based on representative least squares method proposed by the authors. The theory of representative least squares method is examined, and compared with classical least squares method. Three computational approaches to find the solution for the representative least squares method are initiated.