Bio-inspired Computing Algorithms for Adaptive Structural Health Monitoring

This paper studies a bio-inspired pattern recognition algorithm for the structural damage detection and classification in changing environments. Biological system such as natural immune system is a remarkable distributed information processing system. It uses feature extraction, signaling, learning, and memory mechanisms to solve pattern recognition and classification problems. Artificial-immune-system-based damage pattern recognition algorithms detect damage patterns by examining the deviations of real-time sensor data from a normal pattern model (a set of representative feature vectors for the normal pattern). To enhance the adaptability of the presented algorithm, the normal pattern model is updated when the temperature changes. The performance of the updating of the normal pattern model is evaluated using environmental monitoring data collected by the SIMCES (System Identification to Monitor Civil Engineering Structures) project. The validation result shows that the updating of the normal pattern model increases the success rates of the damage detection.