A critical challenge facing the integrity of many assets throughout the oil and gas industry is directly related to corrosion under insulation (CUI). Unfortunately, the lack of adequate inspection technologies adds to this well-known industrial challenge. Presented in this paper is an inspection tool enhanced using Artificial Intelligence (AI) that can provide field inspection engineers with a facility heat map of insulated asset integrity allowing inspection prioritization.
The approach used in this research, and presented here, was to enhance the output of already known and field approved thermographic technologies using a purpose built AI based on Machine Learning (ML). By examining the progression of thermal images, captured over time (<20 minutes), corrosion and factors that cause this degradation are predicted by extracting thermal anomaly features and correlating them with corrosion and irregularities in the structural integrity of assets verified visually during the initial learning phase of the ML algorithm. Additional benefits to this technique include enhanced safety through remote inspection and additional cost savings from monitoring assets online.
To develop and verify the CUI technology results from in-house laboratory tests followed by field validation outcomes will be presented. Laboratory trials were carried out using a series of insulated field assets with different levels of degradation and structural integrity set up to mimic the thermal behavior of in-process assets. This initial feasibility study allowed the definition of key parameters required to build an effective ML model. Following in-house trials a series of field tests and visual verification was performed on both hot and cold insulated assets to gather a sufficient amount of datasets to train the predictive algorithm. To enhance this learning process, synthetic data was created based on real field asset configurations and operating parameters. Finally, during the technology validation phase, again on field assets, the AI technique coupled with a commercial field approved thermographic camera returned a predictive accuracy in the range of 85 – 90%.
The work presented in this paper provides a solution for the current lack of technologies to monitor the presence of CUI by enabling and enhancing the output from already known and field approved technologies, such as thermography, using AI. Additional benefits of this approach include safety enhancement through non-contact online inspection and cost savings by reducing the complexity of asset preparation (scaffolding) and downtime.