Corrosion characterization of engine connecting rods using fringe projection profilometry and unsupervised machine learning

Abstract

The consequences associated with corrosion, a global industrial peril, cost an estimated $ 2.5 trillion annually to inspect, rectify, and prevent. In addition to significant economic losses, corrosion-induced failure of critical components in transport systems, like automobiles, may also lead to loss of human life. Hence, it is essential to eradicate corrosion in its early stages. The most vital automobile component is its engine, whose failure can cause fatal accidents. Regular quality inspection and maintenance by skilled personnel is essential to prevent this. Automating this task will address this domain's personnel shortage while mitigating the risk of human error. To enable the performance of this task without the need for human intervention, we determine the morphological parameters affected by corrosion in automotive engine components, namely connecting rods, using Fringe Projection Profilometry (FPP), a high-speed 3D profiling technique capable of achieving sub-millimeter accuracy. We then perform classification using k-means clustering into low, medium, and high corrosion bands, based on the parameters obtained from 3D imaging. The model was able to achieve a high accuracy of 88.57%. The accuracy was determined by considering the visual classification performed by a Material Science Expert as the ground truth.