On the third day of the Lunar New Year, the workshop lights were already on. The old aerospace factory stabilized the production pace through its self-developed AI platform.

On the third day of the Lunar New Year, most people are still on holiday at home.

On an island by the Hengliang River in Songjiang, Shanghai, the lights in the workshop of Shanghai Aerospace 800 Institute are already on. Robotic arms precisely pick up parts, intelligent transport carts “didi” shuttle between machine tools, and the production line operates quietly and efficiently.

“We started work on the third day of the New Year because there are tasks waiting, and we can’t delay even a day,” said Liu Xiao, Chief Information Officer of Shanghai Aerospace 800 Institute. In recent years, Shanghai Aerospace 800 has been awarded the Ministry of Industry and Information Technology’s Demonstration Factory for Intelligent Manufacturing, a National Green Factory, and recognized as a “Gongfu Chain Leader” enterprise by Shanghai. This Spring Festival, instead of relying on manpower, Shanghai Aerospace 800 maintained production rhythm through solid digital means.

Compatible with “Universal” Equipment

At 8 a.m., beside the processing workshop’s operation area, a large electronic screen flickers with blue light, scrolling real-time task progress and equipment status. Yang Youcheng, current team leader of Tang Jianping’s team, is coordinating production schedules with colleagues.

While speaking, a new pop-up window titled “Tooling Complete Reminder” appears next to the date in the top right corner of the screen, indicating that all process equipment in the workshop is in place and in good condition, ready for operation. Everyone quickly begins the first shift of the Year of the Dragon.

Next to the work center, three skilled workers are gathered around a table discussing processing parameters for a certain part.

“On the first day of work after the New Year, just planning takes half a day,” Liu Xiao said. “Now, with a swipe card into the workshop, my phone automatically pushes what tasks to do and on which machine, clear and straightforward—no questions, no waiting.”

Behind the efficiency boost is Shanghai Aerospace 800’s self-developed digital scheduling system, which consolidates dispersed information from scheduling, warehouses, and processes into one cohesive system, halving the time needed for workshop coordination. “Basically, it makes information run, and fewer people need to run around,” Liu Xiao explained.

This system is the result of a “small team with makeshift tools” self-research project that Liu Xiao led starting in 2018. “We didn’t buy expensive software or hire big vendors; we relied on our core staff to develop a platform truly suited for our old factory.”

Shanghai Aerospace 800’s equipment is considered “world-class,” ranging from old machine tools from the 1960s to the latest domestic CNC equipment, with brands and interfaces that are diverse and incompatible.

“Other projects might use the same solution for ten identical machines. We might need ten different programs for ten devices,” Liu Xiao admitted. “But because of that, our self-developed system can truly be compatible with both old and new equipment.”

By integrating robots, visual recognition, and intelligent scheduling, this aerospace discrete collaborative manufacturing industrial internet platform has increased equipment utilization from 60% to over 80%, with a 100% pass rate for product inspections on the first try.

“The key is, with this digital scheduling system, machines can run continuously,” Liu Xiao said. “People can rest, but machines keep running, ensuring the schedule.”

Shanghai Aerospace 800’s skilled workers are focused, comparing drawings and operating CNC centers.

No Flaw in Weld Defects

In Shanghai Aerospace 800’s intelligent inspection laboratory, Liu Xiao taps the screen to display an image of a rocket fuel tank weld seam.

At the center of the image, a tiny dark spot almost blends into the background, difficult to detect with the naked eye. But the AI system has already marked it with a prominent red box—indicating a potential porosity defect.

“In the past, one rocket required tens of thousands of images,” Liu Xiao recalled. “Three rounds of manual review were standard, with inspectors sitting for hours, eyes getting dry and blurry.”

To solve this problem, Shanghai Aerospace 800 launched the “AI + Inspection” project. The team collected and organized over 100,000 images of historical welds, covering different materials, processes, welding parameters, and defect types, with each image annotated by inspection engineers, creating China’s first high-quality defect database tailored for aerospace high-reliability welding scenarios.

Based on this, Liu Xiao’s team independently developed a deep learning-based image recognition model specifically for automatically detecting typical weld defects such as cracks, porosity, and slag inclusions.

“Our approach is practical—preferably report too many than miss any,” Liu Xiao said. This human-machine collaboration not only significantly reduces inspectors’ workload but also achieves a 95.5% defect recognition accuracy and 100% recall rate for actual defects. “In other words, we don’t miss any real defects, and over 95% of the issues flagged are true defects.”

This system, called “Intelligent Nondestructive Testing Cloud Platform,” has deeply integrated into Shanghai Aerospace 800’s production process. More impressively, this technology, born from aerospace’s strict requirements, is accelerating into other fields, successfully providing AI inspection services for welding quality in automotive, nuclear power, and shipbuilding industries.

Liu Xiao said, “Because we are users ourselves, we truly understand the complexity of the现场. Variations in lighting, reflections, image noise, even the habitual viewing angles of veteran inspectors—these details are hard for outsiders to grasp but are critical for AI implementation.”

Key Production Lines Run Nonstop

In fact, the biggest challenge in transformation isn’t technology but people.

“At first, veteran workers refused to scan codes to report work. They said, ‘We used to just use paper and pen. Why do we need to scan codes or tap screens now?’” Liu Xiao recalled. When introducing this system, many thought it was unnecessary.

In front of large CNC machines, two skilled workers at Shanghai Aerospace 800 are carefully adjusting measurement devices to ensure machining accuracy.

Resistance stemmed from unfamiliarity. Some seasoned technicians, with decades of expertise, felt at ease with manual skills but were unfamiliar with smart equipment, even at a loss on how to operate.

What to do? Liu Xiao didn’t force the system on them but instead personally “squatted” in the workshop, digging into real pain points. She found the issue wasn’t the system itself but the user experience—so her team redesigned the interface by job type: machinists simply click “Start/Finish”; heat treatment workers can batch submit over a hundred parts; inspectors receive automatic task prompts on handheld terminals wherever they go.

“It’s not about making people adapt to the system but making the system adapt to people,” Liu Xiao said. “True digitalization is about making data serve people.”

Additionally, as an assembly unit, Shanghai Aerospace 800 understands that working alone won’t go far. Currently, the institute has exported mature modules like intelligent inspection, equipment networking, and structured processes to dozens of upstream and downstream companies, greatly improving cross-enterprise collaboration efficiency and building an efficient, interconnected industrial ecosystem.

From the third to the seventh day of the Lunar New Year, Shanghai Aerospace 800’s workshops will operate as usual every day, with key production lines running nonstop. What supports all this isn’t just responsibility but the digital foundation built over years and the collaborative manufacturing network developed with partners upstream and downstream, ensuring tasks stay connected, information flows smoothly, and production continues without interruption.

Over the past 60 years, Shanghai Aerospace 800 has developed and produced dozens of models, participated in the development of new-generation launch vehicles such as the Storm-1, Long March 2D, Long March 3, Long March 4, Long March 5, Long March 6, and contributed to manned spaceflight projects like the Shenzhou spacecraft…

“Sixty years ago, aerospace pioneers built rockets while living on this marshy land. Today, we produce and upgrade simultaneously,” Liu Xiao said, gazing at the still-operating robotic arms in the distance. “The old factory isn’t old because it’s static; it’s constantly changing. Using intelligent methods, we do what needs to be done. That’s how we celebrate the New Year in aerospace.”