Investigations on the inversion model of membrane resistance-residual life for sulphur-bearing shale gas pipelines

In order to address the 6-9 month lag in traditional time-corrosion model warnings and the frequent perforation accidents in Sichuan-Chongqing shale gas gathering and transmission pipelines in CO₂-H₂S-Cl⁻-SRB multi factor coupling environment, a real-time residual life inversion model based on membrane resistance R was established. At a shale gas site, 82 sets of L245NS steel hangers were engaged to note the measured perforation time and pre-perforation R. Following cleaning, R outliers were identified using autoencoder neural network combined with the box plot approach (IQR＞1.5). A corrosion environment similar to the field conditions was created using a Parr 4575 high-pressure reactor and dynamic immersion was performed for 1, 4, and 7 days. Corrosion rate, R, and morphological evolution were obtained using weight loss technique, in-situ EIS, and SEM-EDS. Early warning accuracy and error analysis were carried out on 20 blind samples. The data from field steel hangers indicate that R drops in a three-stage pattern over time "gradual decline-platform-steep decline" Dynamic immersion acceleration test show that the corrosion rate rose from 0.184 mm/a to 0.863 mm/a exponentially. The remaining operating time was fitted to yield t_R=9.94e^0.065R （R≤2.0 kΩ·cm²） with linear compensation for high-resistance zone. Verification on site show that MAPE is 3.8% and the greatest error is 18.2%. Setting off a red warning when R is no higher than 0.20 kΩ·cm² produced an accuracy rate of 75.0% and a false alarm rate of 0%. Sichuan-Chongqing shale gas gathering and transportation pipeline has a quantitative basis for the joint emergency response of "corrosion inhibitor + bactericide" thanks to the development of an engineering deployable "R-remaining life" index inversion model. This model can also be extended carbon steel pipelines in other high sulphur gas fields.