面向轨道交通设备运维的工业大模型高质量行为数据集建设与应用

本数据集由无锡金鑫集团与绿轴科技联合建设，依托自主研发的POHA（人类行为事实证明）底层架构，将轨道交通设备（卫生间系统、冷却系统等）运维中的现场人工操作转化为不可篡改、可审计的结构化高质量数据。数据覆盖从生产制造到产品30年全寿命周期维护的完整链路。系统通过六大并发引擎与区块链双链存证交叉核验，构建“采集即验证、验证即标注”的数据生产流水线，彻底解决传统维保记录易篡改、不可靠的盲区，为工业垂类大模型提供高保真、强对齐的基准事实数据底座。This dataset was jointly constructed by Wuxi Jinxin Group and Green Axis Technology. Relying on the independently developed POHA (Proof of Human Behavior Facts) underlying architecture, it transforms on-site manual operations during the maintenance of rail transit equipment (including restroom systems, cooling systems, etc.) into tamper-proof, auditable structured high-quality data. The data covers the complete lifecycle from manufacturing to 30-year maintenance of the products. The system utilizes six concurrent engines and cross-verification via dual-chain blockchain notarization to build a data production pipeline following the principle of "collection equals verification, verification equals annotation". This thoroughly resolves the blind spots of tamper-prone and unreliable traditional maintenance records, providing a high-fidelity, strongly aligned benchmark factual data foundation for industrial vertical large language models.

搜集汇总

背景与挑战

背景概述

该数据集由无锡金鑫集团与绿轴科技联合构建，专注于轨道交通设备运维场景，通过POHA系统将现场人工维保操作转化为不可篡改、可审计的结构化高质量数据。它覆盖设备全寿命周期，并利用多引擎与区块链技术进行验证，旨在为工业垂类大模型训练及智能决策应用提供高保真的基准事实数据底座。