Artificial Intelligence China Uses AI for the Construction of Railway Tunnels

Related Vendors

Asia Waypoint

FAULHABER Drive Systems

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

For the first time, the drilling of a railway tunnel through challenging terrain has been entirely planned by an AI model. Chinese engineers and data scientists report on this world-first innovation.

The Yangcun Tunnel is part of a new route for the high-speed train "Gaotie," the Chinese equivalent of the Japanese Shinkansen. To enable trains to thunder through a region shaped by karst mountains, geological faults, and underground lakes at 217 mph in the near future, highly complex calculations must be made, and risky decisions must be taken.

A Nightmare for Tunnel Construction Engineers

The railway line runs straight through the Wuling Mountains, a range in China long known for its rugged beauty, but also for its remoteness and the poverty of the ethnic minorities living there. The government in Beijing is deliberately expanding the country's high-speed rail network in such areas to combat remaining "pockets" of poverty and provide residents with new economic opportunities.

The Yangcun Tunnel is being built in Wufeng County in Hubei Province, home to the Tujia people, among others. The terrain is a kind of nightmare for tunnel construction engineers. Due to soil erosion in the karst rock, underground rivers, and constantly changing rock layers, a single wrong decision by the engineers can have costly or even catastrophic consequences. Even seasoned tunnel construction veterans are not immune to such errors.

Deep Learning Model Develops Recommendations

That’s why artificial intelligence has been tasked with planning the excavation for the tunnel this time. A deep learning model was trained with a large dataset from previous tunnel projects in China and has developed recommendations on which segments of the new route would be best and safest for blasting or other excavation methods, such as benching or the CD method.

They had fed their AI model with the "design data of 4,100 lining segments from 248 tunnels," wrote the involved engineers and data scientists in an article published in the tunneling journal "Tielu Jianshe" in early November. The project was first introduced in the 45th issue of the Chinese tunneling journal. Their model achieved an accuracy of 89.4 percent for optimal construction predictions, which were subsequently verified by "human" experts, it states.

The Yangcun Tunnel, currently under construction, is the first high-speed rail tunnel in the world whose construction method has been primarily determined by artificial intelligence.

AI Construction Plan Reviewed And Approved By Engineers

The new AI model has proven itself, especially in risky drilling segments in this highly dynamic "underworld" in central China. It has outperformed traditional methods like Random Forest or SVM in accuracy of predictions in high-risk zones, the engineers write.

The construction plan developed by the AI has been reviewed and approved by senior engineers. It is already being implemented in practice as part of the Building Information Modeling (BIM) system.

This makes the Yangcun Tunnel, currently under construction, the first high-speed rail tunnel in the world whose construction method has been primarily determined by artificial intelligence. Chinese media call it a "milestone" for the industrial adoption of AI.

Data Treasure Thanks to Thousands of Railway Tunnels in China

The foundation for this technological breakthrough is the vast amount of data collected during the nationwide expansion of the Chinese Gaotie network. By the end of 2024, 18,997 railway tunnels will have been built in the People's Republic. High-speed trains now race through 4,917 of them.

This data treasure allowed them to feed sufficient experience in lining risky tunnel segments into several "Convolutional Neural Networks (CNN)," the experts wrote in their article. Using additional data processing methods, guidelines were developed whose accuracy exceeds the "subjective experiences" of even seasoned designers, the article states.

Worldwide, scientists are grappling with the question of how to balance the risks of inaccuracies or "hallucinations" in large AI models against their benefits to achieve a responsible use of artificial intelligence.

Use Artificial Intelligence Responsibly

This specific use of AI in industrial practice is not only technically and economically interesting but also has an ethical dimension. Scientists worldwide are grappling with the question of how to balance the risks of inaccuracies or "hallucinations" in large AI models against their benefits to achieve a responsible use of artificial intelligence.

If the combination of AI and human oversight by experienced tunnel construction experts now being used in the Yangcun Tunnel proves successful, AI could soon be used worldwide in the construction of dams, mines, or other infrastructure projects.

Henrik Bork, a long-time China correspondent for the Süddeutsche Zeitung and the Frankfurter Rundschau (Germany daily Newspaper), is Managing Director at Asia Waypoint, a consulting agency specializing in China and based in Beijing.

Fascination With Technology

Preventing Heart Diseases with AI