Damage detection and health monitoring of structures using dynamic response and neural network techniques

The goal of this thesis research is to achieve a damage detection and health monitoring system for advanced structures like aircraft, helicopters, and spacecraft. Three novel contributions toward the research goal have been accomplished in the thesis research.;1. Improved modeling of the dynamics of a delaminated composite structure. The model includes the effects of shear, rotary inertia, the coupling of longitudinal and transverse deformations at the delamination boundaries and the longitudinal inertia. Nonlinear interaction represented by a piecewise linear spring model between the delaminated sublaminates is also included. To verify analytical results, I have designed, constructed and carried out experiments on delaminated beams with properly attached sensor and actuator.;2. Novel damage detection scheme based on experimental data. A rigorous relation between the damage and the structural response has been derived. A damage detection algorithm using experimental data only without empirical judgment has been established. The damage detection algorithm enables one to detect the damage location and the corresponding damage severity simultaneously.;3. High performance network method. The neural network technique is employed in signal processing and decision making. Due to its learning, memorizing and predicting abilities, neural network technique enables one to apply damage detection algorithm to health monitoring. A high performance training algorithm has been designed. A dynamic learning rate training method based on the global error and error change rate information to dynamically change the network learning rate is considered first. Then the training performance can further be optimized by introducing the fuzzy concepts into the learning rate control process. The training method has been demonstrated to be highly effective and efficient. The method has also shown the training robustness at the initial learning rate from benchmark and practical problems.;4. Integrated health monitoring. A health monitoring system has been designed to detect the damage related to delamination and stiffness loss of the composite structures. Structural dynamic response is used as network input and neural network techniques are used as real-time signal processor and delamination pattern recognizer. Subnetworks have been designed to simplify the network topology and training complexity.