'An army of robots' and zero human workers will build a dam in China
China is using artificial intelligence (AI) to effectively turn a dam project on the Tibetan Plateau into the world's largest 3D printer, according to scientists involved in the project.
The 180 metre (590 feet) high Yangqu hydropower plant will be built slice by slice - using unmanned excavators, trucks, bulldozers, pavers and rollers, all controlled by AI - in the same additive manufacturing process used in 3D printing.
When completed in 2024, the Yangqu dam will send nearly five billion kilowatt-hours of electricity each year from the upper reaches of the Yellow River to Henan, the cradle of Chinese civilisation and home to 100 million people.
The power will travel via a 1,500km (932 miles) high voltage line built exclusively for green energy transmission.
According to the project's lead scientist Liu Tianyun, in a paper published in the peer-reviewed Journal of Tsinghua University (Science and Technology), dam construction and 3D printing are "identical by nature".
After years of development testing, 3D print technology for large, filled infrastructure had matured enough for mass applications and would "free humans from heavy-duty, repetitive and dangerous work", he said.
Liu, an associate researcher with the state key laboratory of hydroscience and engineering at Tsinghua University, and his team came up with the idea of "printing" large-scale building projects about ten years ago.
They thought an entire construction site could be turned into a giant printer, with a large number of automated machines working seamlessly together as different components.
The 3D printer was initially developed as a less wasteful way to manufacture components from precious materials. Printing - or adding - materials produces less waste than cutting and grinding.
Since then, some architects have started to apply the technology to buildings, although projects have so far been small. The first 3D-printed office building, the Dubai Future Foundation headquarters, stands at just six metres (20 feet).
Chinese civil engineers are no strangers to AI, which was used to build Baihetan, the world's second-largest dam, in just four years. But until now, it has mainly played a coordinating role in projects.
Testing of the technology in previous construction projects suggested smart machines could do a better job than humans, ", especially in some harsh and dangerous environments", said Liu and his colleagues.
Liu did not immediately respond to questions about Yangqu dam's progress, but according to state media reports work started at the end of last year in Hainan Tibetan autonomous prefecture, Qinghai province.
After "slicing" a computer model of the dam into layers, the AI at the heart of the project would assign a team of robots to add one layer at a time, the paper said.
Unmanned excavators will be able to identify and load materials from a stockpile yard into a fleet of automated trucks, some powered by electricity.
Following an optimised route calculated by the central AI, the trucks will deliver the right materials to the right locations, at the right time, to be located by robotic bulldozers and pavers and turned into a layer of the dam structure.
Automated rollers press the added layer until it is tight and firm, but they are also equipped with sensors. The central AI uses these to monitor build quality by analysing ground vibration and other data.
Breakthroughs in AI technology, including deep reinforcement learning, mean the machines can now recognise nearly all objects on site, deal with uncertainties in a changing environment, and perform various tasks flexibly, according to the paper.
They also do not make human errors. Liu said truck drivers often delivered materials to the wrong location, while shocks and strong vibration prevented roller operators from maintaining a perfectly straight path. And most workers were unable to read the technical design papers correctly, he added.
But where the machines shine is their ability to work in a life-threatening environment, without getting headaches from a lack of oxygen or exhausted after working continuously for 24 hours, according to the researchers.
Not all jobs in the dam's construction will be handled by machines. The team said the mining of fill-rocks from nearby mountains would be done manually because of the task's complexity.
Liu's team said the technology could also be used in other infrastructure projects, such as airport and road construction.
"AI-based on knowledge, information and data is a new tool … that will shape our future," they said.
A Nanjing-based civil engineering scientist, who asked not to be named because of his role in the technical evaluation of some major infrastructure projects, said there were limits to 3D print technology but it would find more uses in the future.
"It cannot print a structure consisting of different materials, such as reinforced concrete made of steel and cement," the scientist said.
"An army of construction robots can offset the sharp decline of manual labour caused by low birth rates," he added.
This article was first published in South China Morning Post.