LAUSR

Abstract To increase structural reliability of complex mechanical components under harsh working conditions, a vibration-based structural health monitoring (SHM) method is a promising tool. However, due to non-linear dynamic characteristics and interfering contents in the measured signals, applying a reliable SHM system to practical project is far from an easy work. On the basis of recent studies, an application-oriented SHM method based on a Dual-Tree Complex Wavelet enhanced Convolutional Long Short-Term Memory neural network (DTCWT-CLSTM) has been designed. The method combines three advanced SHM technologies. In its application process, DTCWT is firstly used to obtain multiscale characteristics information of measured signals. A DCNN model is then employed to automatically extract useful damage features. Subsequently, a LSTM model is utilized to predict damage values. The proposed DTCWT-CLSTM is applied to damage prediction of an automotive suspension component under real operation conditions. The experimental results confirm the superiority of our method when compared with several state-of-the-art baseline methods.