Anomaly MFL Signal Recovery based on Denoising Sparse Autoencoder

Abnormal signals in Magnetic Flux Leakage (MFL) signals seriously affect accurate assessment of pipeline health. To overcome this problem, this paper proposes a novel anomaly MFL signal recovery method based on denoise sparse autoencoder (DSAE). First, similar to the denoise autoencoder, the input of DSAE is the signals with anomalies, and the target signals are the complete signals. Second, relative entropy is added in the loss function as a penalty factor to improve the sparsity of the network. Finally, the proposed method is verified by comparison experiments. The results indicate that the method proposed in this paper is effective.