A random demodulation architecture for sub-sampling acoustic emission signals in structural health monitoring

Structural health monitoring (SHM) has received increasing attention due to its low cost and high performance in the field of non-destructive testing. However, the data acquisition step of SHM, especially in acoustic emission (AE) applications, often encounters a sampling rate barrier because of limited energy and storage resources. In this paper, we propose and evaluate a compressed sensing AE signal acquisition system to solve this problem. Our sampling framework is based on the existing random demodulation (RD) architecture, which is easy to implement in AE monitoring systems. Our sparse recovery algorithm is based on ℓ1-homotopy with a learned dictionary, which compared to alternative techniques/dictionaries is more accurate, fast, and easily-implemented for dynamic, non-stationary, streaming AE signals. Finally, we apply the proposed method to actual signals to verify its validity and efficiency. The results confirm that the proposed sampling model, dictionary, and algorithm can realize the goal of under-sampling and reconstructing AE signals with high accuracy and speed.