Because the substrate material of flexible antenna is easy to generate holes and cracks in the harsh airborne environment, Lamb wave test technology is proposed to detect and identify the damage in flexible polyvinyl chloride (PVC) substrate. The wavelet decomposition and extreme learning machine (ELM) pattern recognition combined to carry out wavelet threshold denoising and damage feature extraction on the echo signal, construct the feature vector of the echo signal from three levels of time domain, frequency domain and transform domain, train and test to classify the types of no damage, hole, and groove damage. The experimental platform for damage detection of flexible PVC substrate based on Lamb waves was built for verification. The results showed that the feature vector of no damage, hole, and groove damage echo signals on the flexible substrate was extracted by wavelet threshold denoising and decomposition. The characteristics of low-frequency slowly varying echo reconstruction signal were unmistakable, and the types of no damage, hole and groove damage were classified by training and testing. The classification accuracy was 92.56%, which was compared with BP Neural network classification method for comparison, ELM classification method has better classification performance, and it verified the effectiveness of a combination of wavelet decomposition and limit learning machine to identify the damage characteristics of flexible PVC substrate. This method has a good application prospect in the rapid detection of early damage of flexible electronic devices.

Key words: wavelet decomposition; extreme learning machine; flexible substrate; damage identification