Open Access

ARTICLE

Interpretable Damage State Identification of Buried Pipelines under Rotary Tiller Loading Using a PSO–CatBoost Framework

Liqiong Chen¹, Haoyu Jia¹, Mailun Liu², Kai Zhang^1,*, Song Yang¹, Zongjun Jiang¹
1 Petroleum Engineering School, Southwest Petroleum University, Chengdu, China
2 National Petroleum and Natural Gas Co., Ltd. Oil and Gas Regulation Center, Beijing, China
* Corresponding Author: Kai Zhang. Email:
(This article belongs to the Special Issue: Greening the Pipes: Achieving Sustainability in Pipeline Engineering)

Structural Durability & Health Monitoring https://doi.org/10.32604/sdhm.2026.077675

Received 15 December 2025; Accepted 08 April 2026; Published online 12 May 2026

Abstract

Buried natural gas pipelines are critical components of energy infrastructure, and their durability and safe operation depend on effective structural health monitoring and the early identification of damage states. In farmland environments, rotary tillage imposes repeated and often concealed mechanical loads on buried pipelines, resulting in stress accumulation, progressive deterioration, and potentially structural failure. However, predictive and interpretable health monitoring approaches that explicitly incorporate rotary tiller-induced damage mechanisms remain scarce. In this study, a physics-informed and interpretable hybrid framework is proposed for the structural health monitoring of buried pipelines subjected to rotary tiller loading. A three-dimensional multiphysics-coupled finite element model of the rotary tiller-pipeline-soil system was developed to simulate the mechanical response and damage evolution of pipelines under varying wall thickness, internal pressure, blade number, operating speed, and soil density. Based on the simulation results, pipeline conditions were classified into three damage states, namely elastic deformation, plastic deformation, and failure, with the first two regarded as warning states. A multi-class CatBoost model optimized using Particle Swarm Optimization (PSO) was subsequently established for damage-state identification. On the test set, the model achieved an accuracy of 0.94, and the AUC values for all three classes reached 0.93. SHapley Additive exPlanations (SHAP) were further employed to interpret the model outputs and quantify the contribution of individual parameters. The results revealed critical risk thresholds associated with the transition from warning states to failure under the shallow-cover rotary tiller disturbance scenario considered in this study. In particular, the risk of failure increased markedly when the blade number exceeded eight and the internal pressure was greater than 8 MPa. These findings indicate that wall thickness and internal pressure govern the baseline structural resistance of pressurized pipelines, while the identified thresholds can support the screening of high-risk conditions and the operational control of shallow-cover farmland sections.

---

Keywords

Structural health monitoring; buried pipelines; third-party interference; damage state identification; finite element analysis; explainable machine learning

---