Intelligent diagnosis of structural cracks with optimized sensing network

Abstract

This paper considers intelligent diagnosis of structural cracks emanating from rows of rivet holes in thin metallic plates using active sensing network. Lamb waves are generated using actuators and propagate across the plates and received by sensors. We extract an effective feature called energy ratio change from time domain signals using wavelet transform. Then we develop neural networks using this feature to diagnose health condition. The sensing network is optimized by developing a mixed integer programming model. The results show that our method can effectively detect cracks and determine their locations, and the number of sensors of the sensing network can be significantly reduced while keeping high diagnostic accuracy. Important insights are also obtained such as in which area the sensing network has the weakest diagnostic capability.