Flexible, scalable, printed ultrasound sparse array for corrosion detection using machine learning

Corrosion detection is a huge industrial challenge, costing billions every year. Guided wave systems have been developed for long range inspection of pipelines however the low frequencies do not have the sensitivity needed to detect mm scale corrosion. Higher frequencies can be used for improved sensitivity but are much shorter range and are costly to install/scan over large areas of plant. This work demonstrates a low frequency, printed, flexible sparse array capable of detecting corrosion at 2 mm depth. The scalable nature of the probes allows for comprehensive monitoring over large areas. The sparse array was designed as a flexible mat with sensors positioned in a grid: a single, centrally located Tx probe and Rx probes located in a square formation around the Tx probe. Using the sparse array, data was captured for two 6mm thick, mild-steel plates: one with a 25 mm-squared area of induced corrosion (2mm depth), and the other being a clean, uncorroded plate. Implementing machine learning algorithms, the data from both the corroded and clean plates could be classified.