STRUCTURAL HEALTH MONITORING SYSTEM USING SUPPORT VECTOR MACHINE

摘要

A structural health monitoring system for a building structure utilizing Support Vector Machine (SVM) is proposed. The SVM is a new machine learning technique for pattern recognition. In our proposed system, modal frequencies of a structure are used for pattern recognition. As our system does not require-modal shapes, typically only two vibration sensors detecting an input signal and an output signal for a structural system are enough to define damage. Changes in moral frequencies normalized by original modal frequencies before suffering any damage are used as feature vectors for feeding into the SVMs. The proposed system is capable of identifying the damaged locations even when multiple stories suffered damages.