Learned representations of hyperspectral infrared images obtained at the Mid-IR beamline are used to extract physicochemical spectral markers of aging and cracking in cross-linked polyethylene pipes. Indicators of degredation were identified and learned by the model and applied to identify both unused and damaged pipes.

These results will help pipe manufacturers optimize fabrication methods and improve stress-testing protocols, extending the use of their pipes into ever more vigorous and challenging operating environments. The study also highlights the use of representation learning via deep generative modeling to provide interpretable information from complicated data sets - unlocking new capabilities for hyperspectral IR imaging using the Agilent microscope at the Mid-IR beamline to solve complex problems across global sectors.

Read the paper!

Grossutti, M.; D’Amico, J.; Quintal, J.; MacFarlane, H.W.; Wareham, C.; Quirk, A.; Dutcher, J.R. ACS Appl Mater Interfaces 2023, 15 (18), 22532-22542. https://doi.org/10.1021/acsami.3c02564