Detection and Classification of Corrosion-related Damage Using Solitary Waves

Abstract

ABSTRACT This paper presents an inspection technique based on highly nonlinear solitary waves, wireless transducers, and machine learning. The technique was demonstrated on a plate subjected to accelerated corrosion while monitored with wired and wireless transducers, designed and assembled in laboratory. The tethered device consisted of a chain of spheres surmounted by a solenoid wired to and driven by a data acquisition system to control the first particle of the chain in order to induce the impact between the particle and the chain needed to generate the stress wave. The chain contained a piezoelectric wafer disk, also wired to the same data acquisition system, to sense the waves. The wireless transducers were identical to their wired counterparts, but the data acquisition system was replaced by a wireless node that communicated with a tablet via Bluetooth. Four wired and four wireless transducers were used to monitor the plate for nearly a week to detect the onset and progression of electrochemical corrosion. A few features were extracted from the time waveforms and then fed to a machine learning algorithm to classify damage. The results showed the effectiveness of the proposed approach at labeling defects close to the transducers.