xpertsystems/oil025-sample

OIL-025 — 合成泄漏检测数据集（样本）是一个免费的、模式一致的数据集预览版本，由XpertSystems.ai开发，专用于石油和天然气中游管道完整性领域。该数据集旨在支持机器学习任务，包括管道泄漏检测、破裂概率预测、声学泄漏特征分类、计算管道监测（CPM）异常检测、泄漏定位和管道完整性评估。样本数据涵盖300条管道、1,500个泄漏事件和180个破裂事件，涉及6种流体家族（如原油、天然气、精炼产品等），总共有140,230行数据，分布在12个CSV文件中（例如管道主表、泄漏事件表、压力波特征表等）。数据集强调物理耦合特征工程，已验证了八种近确定性特征耦合，包括托里切利孔口流动、斯特劳哈尔式声学共振、比尔定律压力波衰减以及API 579爆裂压力裕度逻辑耦合。数据集基于多个行业标准进行校准，如API 1130、API RP 1175、API 579等，并经过10项指标验证，总体评分为A+。它适用于多种用例，如泄漏率回归、泄漏检测分类、破裂概率回归等，并支持多表关系机器学习。数据集为合成数据，用于研究目的，不用于实际运营决策。OIL-025 — Synthetic Leak Detection Dataset (Sample) is a free, schema-consistent preview release developed by XpertSystems.ai, exclusively for the midstream oil and gas pipeline integrity domain. This dataset is designed to support machine learning tasks including pipeline leak detection, rupture probability prediction, acoustic leak feature classification, Computational Pipeline Monitoring (CPM) anomaly detection, leak localization, and pipeline integrity assessment. The sample data comprises 300 pipelines, 1,500 leak events, and 180 rupture events, covering 6 fluid families (e.g., crude oil, natural gas, refined products, etc.), with a total of 140,230 data rows stored across 12 CSV files (e.g., main pipeline table, leak event table, pressure wave feature table, etc.). The dataset emphasizes physics-coupled feature engineering, with eight near-deterministic feature couplings validated, including Torricelli orifice flow, Strouhal acoustic resonance, Beer's Law-based pressure wave attenuation, and API 579 burst pressure margin logical coupling. The dataset is calibrated against multiple industry standards such as API 1130, API RP 1175, API 579, etc., and validated via 10 metrics with an overall rating of A+. It supports multiple use cases including leak rate regression, leak detection classification, rupture probability regression, etc., and enables multi-table relational machine learning. The dataset is synthetic data intended for research purposes only and shall not be used for real-world operational decision-making.