At the 2026 China Marine Economy Expo — known as "China's No. 1 marine industry show" — held in Shenzhen from October 22 to 24, Shenzhen-based robotics company Lide (Lide Technology) unveiled what it calls the world's first shipbuilding "world model," named "Dingfengbo" (literally "Calming the Storm"). Alongside the model, the company showed a family of intelligent work terminals, including an autonomous welding terminal, an autonomous spraying terminal, and an autonomous surface-roughening (gouging) terminal.
Letting AI "Read" an Entire Ship
On the only national-level marine economy exhibition platform in China, shipbuilding — a traditional heavy industry — now has, for the first time, a world-model system capable of "understanding a whole ship," converting artificial intelligence into on-site delivery capability in the shipyard.
The backdrop is a structural shift in China's shipbuilding industry. Ministry of Industry and Information Technology data show that in the first half of 2026, China accounted for 62.2% of global completed shipbuilding output, 82.3% of newly received orders, and 71.2% of the order backlog — all three indicators leading the world. Behind the full order book and busy docks, the industry's main contradiction is shifting from production capacity to delivery capability, and the core bottleneck on delivery is people.
Shipbuilding remains a labor-intensive heavy industry with a punishing work environment: workers endure high temperatures in enclosed hull compartments, flying welding sparks, dust, and toxic paint fumes. Young workers are increasingly unwilling to take on welding and coating jobs, while the craft experience of older skilled workers is difficult to pass on in standardized form through traditional apprenticeship. The gap between labor shortage and delivery pressure is a common headache for shipyards. On top of that, a single ship involves tens of thousands of structural components, and no two steel-plate joints are exactly alike — one-off custom designs and non-standard sections are the industry norm, and the biggest obstacle to intelligent upgrading.
Traditional industrial robots have hit a wall in this environment. For decades they have relied on "teach-and-repeat" operation: engineers program each point, the machine repeats the path, and any change in section structure means reprogramming from scratch. Faced with constantly varying section dimensions, groove shapes, and station conditions, this "teach once, remember one seam" approach simply does not scale.
The breakthrough lies in making the machine "understand actively." Lide's Dingfengbo world model provides a new technical path: design drawings, process documents, and bills of materials are fed into the model as unified inputs; the model parses the engineering logic among them, decomposes the entire ship into blocks, sub-assemblies, assemblies, and parts, and reconstructs an understandable, plannable 3D structure. After that, robots no longer depend on manual teaching — they perceive the site on their own, break down tasks autonomously, and decide how to work by themselves.
This capability was built on engineering data accumulated from more than a hundred ships, plus field judgment repeatedly refined in simulation. The more the system is used, the deeper its understanding of process and working conditions becomes.
Three Smart Terminals Free Workers From High-Risk Scenes
Strong engineering cognition still has to land in the physical world. The three terminals shown at the expo correspond to three key processes of hull construction, all sharing Dingfengbo as the same "brain" while each adapting to the contact mode, mobility, and working posture of its own process. The autonomous spraying terminal takes over anti-corrosion coating, replacing workers in enclosed, toxic compartments with paint-film thickness controlled to the millimeter. The autonomous gouging terminal handles surface preparation, fixing the old problem of uneven manual grinding that shortens coating life. A handheld laser welding tool puts flexible welding capability back into workers' hands, enabling human-robot collaboration across complex shipyard stations. The system dispatches tasks; the terminals execute the process in real working conditions — closing the loop of "physical AI" from knowing how to do the work to actually doing it on site.
The results have been verified on a live production line. In September, Lide's autonomous welding terminal entered the sub-assembly welding line of Wuchuan Heavy Industry (Wuchuan Shipbuilding). From arrival to producing the first qualified weldment took only 7 days. On its first batch of tasks — blocks B101 through B107 of a 25,900-DWT chemical tanker — the terminal completed a cumulative 4,358 meters of weld seams, with per-shift capacity above 420 meters; weld forming and fillet consistency passed both the customer's quality inspection and crew acceptance.
The change on the shop floor is tangible: welding stations that used to rotate three to five workers per shift are now covered by one or two terminals working in succession. The veteran welders have not disappeared — they have moved up to deployment, commissioning, and quality-control roles. High-risk, repetitive, physically punishing work goes to the terminals; judgment, verification, and coordination stay with people.
That is precisely the significance of Dingfengbo landing in real shipyards: using physical AI to release production capacity, fill the industry's labor gap, and free people from high-temperature, high-humidity, dusty, and otherwise hazardous environments.