Dynamic Maintenance of Underground Pipelines via a Systematic Approach for Conservative Estimation of Pipeline Defect Probability Density Under Data Scarcity

The scarcity of the defect data may lead to the underestimation of defects, resulting in maintenance plans with inspection intervals that may not guarantee timely repairs. To address the low reliability of defect distribution models developed from insufficient data, we propose a systematic approach for deriving conservative probability distributions of pipeline defects. Based on the formal definition of conservative probability distributions, we present methods for modeling such distributions for pipeline defects, with the flexibility to adjust the degree of conservativeness. Furthermore, by incorporating Bayesian inference, we introduce a method for dynamic maintenance planning. The method enables effective utilization of the limited defect data samples obtained during pipeline inspection to assess overall pipeline conditions and dynamically determine subsequent maintenance intervals. The simulation results demonstrate that the proposed method can achieve cost-effective and safety-assured pipeline maintenance plans by quantitatively adjusting the degree of conservativeness, making it broadly applicable to various types of pipeline defects.