Using Deep Learning and Computer Vision Techniques to Improve Facility Corrosion Risk Management Systems 2.0

Abstract

During fit for service or corrosion risk assessments of oil and gas facility systems, a key parameter required to design and implement an effective risk management strategy is visual inspection. This paper explains how using state of the art computer vision and deep learning techniques could address such challenges. We used majorly the python programming language, Tensorflow Application Programming Interface, Resnet deep learning architecture, GPU machines and cloud computing technologies to achieve this. Beyond the challenges of obtaining sufficient corrosion defects data, our final solution is a systematic method that would assist field personnel, facility engineers, service companies and management more accurately detect corrosion defect types and failure modes unbiasedly. This leads to more cost effective and quicker recommendation of preventive or corrective measures.