Data is vital for tunnel boring machines (TBMs), as intelligent equipment needs professional assistance when it functions underground.
However, the industry has long been facing four major challenges: data is scattered across diverse projects and systems; data annotation requires massive manual work, incurring high costs; data collection under special working conditions is difficult; and it is hard to translate on‑site construction experience into reusable expertise.
To tackle these challenges, China Railway 11th Bureau Group Corporation Limited, a subsidiary of China Railway Construction Corporation Limited (CRCC), launched a dataset‑building initiative in 2022 and developed China’s first dataset covering TBM’s full‑life‑cycle scenarios, with a total volume exceeding 20 terabytes. The data were collected from over 300 crucial railway and metro projects in more than 50 cities nationwide. The dataset spans diverse TBM types ranging from 2 to 18 meters in diameter and incorporates all-embracing information including geological conditions, equipment parameters, construction records and monitoring results.
The company has developed an annotation system featuring “AI‑assisted pre‑annotation plus three‑tier human review”. This mechanism improves the automation rate of data processing to over 70 percent and cuts annotation costs by an order of magnitude.
The dataset, supported by CRCC’s AI computing center, can play a supportive role in eight major business scenarios, including intelligent TBM type selection, assisted driving, driverless transportation and intelligent fault diagnosis. Previously, TBM type selection relied on engineers’ experience and extensive document reviews, but now AI can rapidly offer the best option from the data of massive cases. In the past, cutter wear had to be assessed through human review, but now the system can provide early warnings.
Success in practice speaks for itself. At the Jintang Subsea Tunnel of the Ningbo‑Zhoushan Railway, the world’s longest subsea high‑speed railway tunnel, an intelligent material transportation system has, for the first time, successfully accomplished unmanned transportation of super‑large components, both inside and outside the tunnel.
Looking ahead, the company will continue to upgrade the dataset into industry‑wide public data infrastructure, enabling high‑quality data to serve as the “engine” for new quality productive forces driving intelligent construction.
(Executive editor: Zuo Shihan)
