Impact identification in structures using a sensor network: The system identification approach

This research in this dissertation is motivated by the need for reliable monitoring technologies for the detection of impact events on structures. A system that automatically detects and reports impacts would ensure structures' integrity while reducing inspections only to investigation of possible damage at significant impact points. In this research, a new method and an implementation procedure for identifying impact events is presented based on a system identification technique using built-in sensors. The method identifies the impact location and force time history of an impact event on structures without the need for information about actual mechanical properties, geometries and boundary conditions.; The proposed technique does not require constructing a sophisticated mathematical model to simulate the behavior of a structure or building a specific neural network with exhaustive training; instead, the technique uses the inverse of system transfer functions to reconstruct the loading history from sensor measurements. In this proposed approach, the system transfer functions for the entire structure are constructed by two sequential approaches: (1) impact tests at selected points to generate sensor data for establishing the system transfer functions between the sensor and the impact points of the structure; (2) interpretation functions based on a traditional finite element approach to approximate the system transfer functions everywhere on the structures.; A specific implementation procedure based on the proposed technique is proposed for practical application of the technique in structures with a built-in sensory system. Extensive tests on various structures were performed to verify the proposed method and accuracy of load and position predictions.