NVIDIA founder and CEO Jensen Huang recently appeared in Taipei, China. He announced at Hon Hai's "Tech Day" that he would support manufacturing giant Foxconn to build an "artificial intelligence factory."
Foxconn is the world's largest manufacturing company. Now, Foxconn hopes to use NVIDIA's artificial intelligence software and hardware combination platform to produce smart cars. Previously, Tesla was also exposed to be using digital design and casting sand 3D printing technology to improve production efficiency.
Dassault Systèmes, one of the world's leading industrial software vendors, owns the digital simulation software platform 3DExperience. "Using our digital simulation platform, Fuyao Glass has accelerated the research and development of more than 52,000 glass materials, which is very impressive to me. We are also exploring how digital twins can be used to enable innovative engineering of glass materials to ensure digital continuity breakthroughs in manufacturing. Xu Shansong, executive vice president of Dassault Systèmes Asia Pacific, recently told First Financial Reporter.
"1 aircraft can be produced in 9 days"
In the field of aviation manufacturing, companies represented by Boeing and Airbus are embracing AI intelligent manufacturing and 3D additive technology (3D printing). Boeing is already building a "digital twin" of the plane and simulating the production systems that make the plane.
Boeing's Gigafactory, located in Everett, Washington, USA, is known for its high efficiency and high profits, and some industry insiders estimate that the plant can produce one aircraft in 9 days.
Airbus is using augmented reality AR technology to revolutionize its quality control process. Their team used drones equipped with lidar sensors (LIDAR) sensors to conduct flight inspections, which then transmitted data to human inspectors who examined the information using tablets and AR glasses.
Electric car giant Tesla is also using digital simulation to validate its "gigacast" process, which is expected to help Tesla slash production costs. If successful, it will become another innovative breakthrough for Tesla in the field of automobile manufacturing. The innovation will make it possible to die-cast almost all of the complex underbody of an electric vehicle into a single whole, rather than hundreds of parts as is the case with conventional cars.
If Tesla can actually cast the underbody of an electric car, it will be able to develop a car from scratch in 18 to 24 months, which will further disrupt the way cars are designed and built, because most competitors currently take three to four years, and will also help push Tesla to launch a lower-priced electric car.
In the case of a large metal test mold, a machining tweak during the design process can cost $100,000 at a time, and if a complete redo of the mold needs to cost, it can cost $1.5 million. On average, the entire design process for a large metal mold typically costs about $4 million.
To minimize costs, Tesla wants to use cast sand 3D printing technology, which is currently the lowest cost of the design verification process for metal mold prototypes, which means that Tesla can adjust the prototype as many times as needed. Through 3D printing technology, a new mold prototype can be reprinted in a few hours. In terms of time, the design verification cycle using sand casting takes only two to three months, while metal mold prototypes generally take six months to one year.
The "accelerator" to turn data into profit
In Huang's view, the "AI factory" will become an "accelerator" for new manufacturing companies to turn data into profits. Large-scale deployment of NVIDIA GPU systems, including HGX, GH200 super chips, etc., in manufacturing plants can power a variety of industrial applications.
According to reports, through these computing chips, unstructured data can be taken from a large number of IoT sensors deployed in manufacturing plants and turned into useful "products", which may be a computer or a car.
Two years ago, NVIDIA launched the "enterprise metaverse" platform Omniverse, a tool for visualizing machine learning and sensor data that manufacturing companies can use to create digital twins of factory floors or warehouses. NVIDIA has partnered with Siemens to leverage the Omniverse platform to bring "digital twins" to the mainstream of manufacturing.
Huang believes that the real value behind the Omniverse concept is simulation, which is based on real-world scenarios. "End-to-end simulation of the entire robotics and automation process will provide Foxconn with a way to improve operational efficiency and save time and costs," NVIDIA said. ”
Siemens was one of the first companies in industry to propose the concept of a "digital twin". "Digital twins" are closely related to simulation technology. Every day, Siemens helps OEMs develop thousands of new models using simulation.
Last year, Siemens partnered with NVIDIA to bring NVIDIA's Omniverse to Siemens' manufacturing plants, making its "digital twin" technology even cooler.
Wang Haibin, executive vice president of Siemens (China) and general manager of Siemens Greater China Digital Industry Group, told First Financial Reporter: "Simulation technology is a computer environment and software environment that can see exactly the same scene as the real world. Industry is using this technology to accelerate innovation cycles, so that products, machinery, production lines, plants, can be simulated in a software environment. ”
For example, when the layout of the factory floor changes, the digital twin system can be used to predict the change of workers' movement routes and judge production safety. At the same time, according to the data of the sensors in the system, the source of the machine fault can be accurately found through the virtual model display. This digital twin technology also enables geographically diverse teams to collaborate and solve problems virtually.
Building a "model factory" for Chinese automobiles to go overseas
At a recent industry-oriented digital intelligence innovation summit, Xu Shansong told the first financial reporter: "In the high-tech field, many companies are reconstructing their technology and architecture on servers to improve computing power, and fast computing power will bring competitive advantage." ”
The wave of generative AI has also brought new development opportunities to software vendors. "We hope to integrate generative AI technology into manufacturing solutions to make the entire production process and system layout more intelligent." Xu Shansong told the first financial reporter, "We also integrate the simulation platform with the operation side data through AI technology, and make full use of the operation side data for subsequent machine learning." Because machine learning can help us refine and optimize the entire production system. Generative AI technology is important because it is embedded in almost the intellectual property (IP) of the entire company. ”
Chinese new energy vehicle manufacturers are actively embracing software technology to accelerate the cycle from R&D design to production. In August this year, Nezha Motors and Dassault Systèmes announced a cooperation to fully deploy DELMIAAPRISO, a solution based on Dassault Systèmes' 3D experience platform, from R&D design to manufacturing.
"We will invest in data, collaboration and efficiency to improve the efficiency of plant operations with both data-driven and model-algorithm-driven." Dai Dali, chief technology officer of Nezha Automobile, said in an interview with the media at the summit of Dassault Systèmes.
Nezha Motors has announced that it will produce 1 million electric vehicles by 2026. The company has an automotive production line in Tongxiang, Jiaxing, Zhejiang Province, and is expanding its production capacity.
As China's new energy vehicle overseas plans become more frequent, some domestic manufacturers' intelligent production plants can also be copied abroad through digital simulation technology. In March, Nezha Motors built a technology factory in Thailand.
Xu Shansong told the first financial reporter: "Chinese automakers can first build benchmark projects locally, create some best cases, such as involving production modules or production systems, and then replicate them in other countries through virtual twin technology as a model to expand their influence." ”