Comparison study of feature extraction methods in structural damage pattern recognition
This paper compares the performance of various feature extraction methods applied to structural sensor measurements acquired in-situ, from a decommissioned bridge under realistic damage scenarios. Three feature extraction methods are applied to sensor data to generate feature vectors for normal and damaged structure data patterns. The investigated feature extraction methods include identification of both time domain methods as well as frequency domain methods. The evaluation of the feature extraction methods is performed by examining distance values among different patterns, distance values among feature vectors in the same pattern, and pattern recognition success rate. The test data used in the comparison study are from the System Identification to Monitor Civil Engineering Structures (SIMCES) Z24 Bridge damage detection tests, a rigorous instrumentation campaign that recorded the dynamic performance of a concrete box-girder bridge under progressively increasing damage scenarios. A number of progressive damage test case data sets, including undamaged cases and pier settlement cases (different depths), are used to test the separation of feature vectors among different patterns and the pattern recognition success rate for different feature extraction methods is reported. © 2011 Copyright Society of Photo-Optical Instrumentation Engineers (SPIE).
Proceedings of SPIE - The International Society for Optical Engineering
Comparison study of feature extraction methods in structural damage pattern recognition.
Proceedings of SPIE - The International Society for Optical Engineering,
Retrieved from: https://digitalcommons.mtu.edu/michigantech-p/12157