Аннотация
This study proposes an integrative methodology for residual life assessment of trunk gas pipelines in Kazakhstan, combining laboratory investigations of pipeline steel with digital analytics and machine-learning techniques. Most of Kazakhstan’s pipelines, built during the Soviet era, have exceeded their design life, creating risks of corrosion, embrittlement, and failure under aggressive operating conditions. The experimental program involved tensile, bend, hardness, metallographic, and chemical tests of two pipe specimens made of 17G1S-K52 steel. Results confirmed compliance with GOST 31447-2012 but revealed local anomalies: elevated carbon content (0.29 wt%) in one specimen, ferrite–pearlite banding, and pitting corrosion with FeS surface films. These features indicate potential risks of brittle fracture and stress-corrosion cracking. To complement testing, a dedicated software platform, GasPipelineInsight, was developed to process more than 20,000 in-line inspection records. The system integrates ETL operations, physics-informed feature engineering, and a Random-Forest classifier with 93% accuracy. Defect classification showed 67% of anomalies in category C (borderline), 1.5% in category A (safe), and none in category D (critical). Visualisation tools, including heat maps, histograms, and burst-pressure curves, supported decision-making and repair prioritisation. The findings demonstrate that integrating laboratory data with machine-learning analytics enables objective integrity ranking, early risk detection, and risk-based maintenance planning. The approach enhances the reliability and safety of gas-transport infrastructure and offers scalability for industrial application. Its adaptability provides potential for integration into national standards and for broader use in pipeline integrity management worldwide. Keywords: residual life; trunk gas pipelines; in line inspection; machine learning; corrosion; burst pressure.