Application of Dynamic Time Warping for Corrosion Detection in Downhole Casing Inspection

Well casing integrity management is key to safe operation of oil and gas wells. High-resolution Magnetic-flux-leakage (MFL) inspection tool measures high-resolution MFL data (logs) from downhole casing for risk evaluation. The MFL tool used in this study [Vogtsberger et. al. 2005] measures the MFL data against time and require further analysis to infer depth of corrosion defects, casing joint collars and other features identified from these data. We demonstrate results from an artificial intelligence (AI) approach that automatically infer physical depth of features identified in MFL inspection data. This allows to automatically process and compare historical inspection data, enabling consistent and accurate estimation of corrosion progression rate in well casing over time. The high-resolution MFL casing inspection tool used in this study comprises of upper and lower inspection modules, each housing 48 tri-axial flux leakage (FL) and discriminator (DIS) sensors, covering 100% of the casing wall circumference from two different measuring points. These sensors in each inspection module measure flux perturbations caused by defects on casing wall. Due to the physical distance between the sensors on the upper and lower modules, their MFL measurements are misaligned and require further analysis to align them before they can be used for corrosion risk evaluation. Experts utilize traditional methods with semi-automated workflows to align and estimate depth of the MFL data for evaluating casing integrity, by locating and identifying corrosion defect, casing collar, and other defect signatures in MFL data. However, these traditional methods are time consuming, require human interventions, and are prune to inconsistent results. To mitigate this issue, we developed an AI-based algorithm that utilizes dynamic time warping (DTW) and genetic algorithm (GA) to automatically estimate alignment and depth of MFL data from casing inspection. We used historical casing inspections logs, reports and gamma-ray and neutron correlation logs from a vertical well in North America (Well 1) to demonstrate results from our algorithm. Accuracy of the depth alignment of the high-resolution MFL data was evaluated using the average distance (mm) between the casing joint collars in memory, and the previous inspections reports. MFL casing inspection logs are noisy due to the vertical and rotational movement of the inspection tool inside the well casing. MFL sensor data are also misaligned due to the spatial different between the sensors in two separate inspection modules inside the tool. Without correcting these issues in the MFL sensor data, it is not possible to analyze the data for corrosion risk evaluation. Moreover, the casing inspection system used in this study logs high-resolution MFL data against time, instead of depth. This requires further analysis to associated depth to the MFL data for corrosion analysis. Traditional MFL inspection logs processing tools mostly rely on human experts to spend significant time to manually correct these misalignment issues in data, as well as associate depth from casing tally or gamma ray and neutron logs, which often lead to inconsistent results. Our proposed algorithm aims to fully automate the depth alignment process of the MFL inspection logs, enabling higher accuracy and consistency in evaluating corrosion progression in well casing over time.