FA-PI

A study proposed converting the Fourier amplitude spectrum into a two-dimensional frequency-amplitude persistence image (FA-PI) for impact echo defect detection.

FA-PI is constructed by mapping each retained Fourier component into a two-dimensional space, with one coordinate being the frequency bin position and the other being the amplitude, and then applying a Gaussian kernel function for smoothing to generate the persistence image.

The research team evaluated FA-PI on an experimental backfilled void dataset and a public bridge deck impact-echo dataset.

In the summary of noise conditions across the two datasets, FA-PI consistently outperformed the raw Fourier amplitude spectrum and Gaussian-smoothed Fourier amplitude spectrum baselines.

Representing Fourier spectral components as structures in a two-dimensional frequency-amplitude space can improve impact-echo defect classification, and this improvement goes beyond the effect of merely smoothing along the frequency axis.

The construction of FA-PI relies on exact multiparameter persistent homology with fixed parameterization, which means that parameter selection may affect the final performance.

The current evaluation of FA-PI is limited to linear classifiers, and whether more complex classifiers can further amplify the advantages of FA-PI remains to be explored.

The computational complexity, real-time performance, and difficulty of integration with existing impact-echo inspection equipment are all issues that require further consideration for FA-PI.