Abstract
Signal features extraction is a significant issue during the damage assessment in structural health monitoring (SHM). Multiple sensor arrays would generate data redundancy in long-term monitoring. It is a challenge to extract signal features and reduce data size. For extracting signal features with low storage consumption, this research proposed a monitoring data processing method based on signal musicalization. First, the time domain signal is sampled by music spectrum to complete the mapping from imperceptible signal to audible audio. The damage signal feature is encoded into musical indexes suitable for serialized storage. Then, the damage modal identification based on musicalized indexes is completed using Long Short-Term Memory (LSTM) network. Further, a data recovery mechanism for failure sensor based on Kalman algorithm is presented. Experiments show that multiple sensor signal features could be extracted by this method. Moreover, it has advantages in damage identification and data compression.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call