Identification of Eggshell Crack using BPNN and GANN in Dynamic Frequency Analysis

Abstract

Abstract The eggshell crack was detected using flexible piezoelectric film sensors on the basis of dynamic frequency response. The response signal was transformed from the time domain to the frequency domain using fast Fourier transform algorithm. In the normalized power spectrum, the first 10 and 20 features were extracted by interval frequency (IFM), maximum magnitude in turn (HVM), and frequencies by magnitude in turn (MVF). The discrimination between intact eggs and cracked eggs was performed using back-propagation neural network (BPNN) and combination of genetic algorithm and BPNN (GANN). The results were found that intact eggs and cracked eggs could be better identified from their dynamic frequency response using BPNN and GANN based on HVM and MVF. It is suggested that intact and cracked eggs can be distinguished using GANN based on 10 features in MVF.