If you open the column:The new round of scientific and technological revolution and industrial transformation triggered by artificial intelligence represented by large models is developing in depth, and thousands of industries will face huge opportunities and subversive challenges. In the industrial field, artificial intelligence is a key variable to promote new industrialization, and the implementation of industrial models will provide a new path for manufacturing enterprises to explore new industrialization. Starting from this issue, China Electronics News has set up a column of "Artificial Intelligence Empowering New Industrialization", which goes into the front line, into factories and enterprises, and comprehensively reports on the vivid practice of artificial intelligence technology represented by large models in the industrial field.
China's FAW has introduced AI large-scale model technology into the automotive production chain.
As the engineer enters the demand instructions in the dialog box on the computer, after just a few tens of seconds, not only the required data report appears on the computer screen, but also the line analysis chart automatically generated by combining these data reports.
This is a scene that the reporter saw in the production workshop of China FAW Group. The data presented on the screen is the vehicle data delivered by the dealer to the consumer and the vehicle data delivered by the manufacturer to the dealer, and these data analysis is very important for the automobile manufacturing plant.
On January 22, the reporter of "China Electronics News" walked into the China FAW manufacturing plant in Changchun City, Jilin Province, and witnessed the application of China FAW and Alibaba Cloud based on artificial intelligence large model technology in the field of automobile manufacturing. Although this scenario is still in the preliminary exploration stage, the reporter smelled the huge trend of the automotive industry firmly grasping the key variable of artificial intelligence and reshaping manufacturing.
"The Ask the Large Model report is automatically generated".
With the acceleration of digitalization in the automotive industry, data has gradually changed from a "by-product" to a "core asset" and is regarded as the "soul" of car companies. How to make data flow freely like water between manufacturing, R&D and design, financial management, marketing and after-sales, travel services and other links, and open up the "artery" of automobile manufacturing information, has become a common issue faced by all automobile manufacturing enterprises.
FAW intelligent platform HIS display model.
The artificial intelligence model has opened a key breakthrough for the data "flow" in the entire automotive manufacturing industry chain. "I thought the big model was far away from us, but I didn't expect it to come to us. A staff member of China FAW told reporters. Business intelligence (GPT-BI) based on large models has now become an indispensable assistant in his daily work.
We have to do sales reports almost every day, such as sales, customer flow, dealer rankings, etc., in the past, this work alone would take a lot of time, but now as long as you pick up the mobile phone and ask the large model, the report will be automatically generated, and the results are clear at a glance. The above-mentioned staff member said.
It is understood that compared with the "fixed Q&A" of traditional business intelligence (BI), GPT-BI based on the large model of artificial intelligence can realize any combination of Q&A, and the data can be penetrated at any time, so that "Q&A is insight", and can achieve a high accuracy rate of nearly 90%.
Don't look at it as just a small dialog box, in fact, this is equivalent to breaking through the data barriers existing in each link of China's FAW production chain and ** chain, and realizing the automatic flow of industrial data throughout the life cycle. Put simply, it's the equivalent of having a "data analyst" for every employee who is always on call and ready to answer questions.
In the long run, the emergence of GPT-BI is equivalent to tailoring an "intelligent housekeeper" for China FAW, which can help it reduce costs and increase efficiency while precipitating industrial data to assist enterprise decision-making.
For example, when asked why the production of a certain model is lower than expected, GPT-BI compares the actual production with the expected production to determine the difference, and then conducts an in-depth analysis based on the data. The "analysis" here includes not only the analysis of explicit variables such as "production shutdown for 20 minutes due to abnormal equipment" and "abnormal quality of a certain model of accessories", but also the analysis of implicit variables such as "fluctuation of raw materials", "energy consumption" and "stability". Finally, based on the data checking, GPT-BI will help the questioner find out the reason for the most correlation and generate a visual report.
The real transformation is to transform the core of traditional industrial enterprises that rely on responsibilities and process operation into data-dependent, respond to user needs at high speed, and form business capabilities and development capabilities that are constantly iterating forward. Men Xin, vice president of the China FAW Hongqi Brand Operation Committee, told the reporter of China Electronics News. He said that China FAW has now put the core business data such as R&D and manufacturing on the workbench, and the next step is to redo all the business with a large model, using data as a production factor, and making artificial intelligence technology a new quality productivity.
FAW Hongqi Prosperity Factory ** workshop.
"The design can be written by the large model".
How hard is it to design a new car? A car from scratch is roughly divided into six stages: product planning, concept design, technical design, product prototyping, product testing and production preparation. Among them, the work content of engineers alone includes modeling feasibility analysis, main section definition and design, part sewing line setting, part 2D drawings and 3D data production, body process analysis (stamping, welding, painting, **, etc.). These are only the contents of the body development, and do not include the design of electronic components, autonomous driving, intelligent cockpit and other systems.
In today's ever-changing consumer needs and rapid product upgrading, how to make design and development more transparent and efficient has become a compulsory course for all car companies. However, at the level of incompetence or architecture, the complexity of automotive software is increasing, and the efficiency of development work is not keeping up at the same rate.
According to research by McKinsey, automotive software complexity has quadrupled in the last 10 years, while software development efficiency has only increased by 1 to 15 times. This problem is most acute in large modules that are becoming increasingly complex, such as infotainment systems and advanced driver assistance systems (ADAS). The efficiency of developing these modules is approximately 25 to 35 percent lower than that of traditional deeply embedded software.
In the most complex design process of automobile manufacturing, engineers with rich knowledge and experience need to find various combinations to meet the needs in more than 20,000 parts and hundreds of thousands of parameters, and then write documents and draw drawings. However, with the help of the large model, the engineer only needs to describe the requirements, and the large model can efficiently find the required combination information, automatically generate the first draft of the design document, and greatly reduce the vehicle development cycle and cost.
The reporter learned that China FAW is trying to use large models to lower the threshold for automotive product design and development and improve R&D efficiency. "Our development team wrote a total of 42.96 million lines in one year. Now that we have a large model, at least half of the ** can be handed over to the large model to write. Menhin said.
We used to make a model every 24 months, but now we can measure it in weeks. Meng Xiangyue, director of the enterprise operation department of the digital department of the FAW system, sighed. According to him, at present, China FAW has realized automatic design, automatic drawing, automatic generation, and continuous iteration of model-based system engineering.
Artificial intelligence is reconstructing the entire software system, all software-related systems will be reconstructed, artificial intelligence will drive the change of software development mode, software will achieve independent optimization and upgrading, and it will bring about the transformation of the entire manufacturing system. An Xiaopeng, vice president of Alibaba Cloud Intelligence and director of the Science and Technology Research Center, told the China Electronics News.
FAW Hongqi Automobile.
The emergence of a new model of "car building" will accelerate
According to public data, the software, manufacturing and service industries have a large number of patents for large-scale model innovation in China. Among them, the number of patents in the manufacturing industry has reached 340,000 pieces.
Looking at the domestic market, cross-border car manufacturing by Internet technology companies abounds, and cross-border artificial intelligence of automobile manufacturers is also a matter of course. From intelligent cockpit, intelligent driving, to automobile design, manufacturing, management, to sales and service, almost every link in the automobile production chain and the first chain can see the application of artificial intelligence. In particular, with the emergence of more and more new models and new formats such as large models "getting on the car", large models "making cars", and large models "selling cars", the development of China's automobile industry has run out of "acceleration".
Manufacturing is the main battlefield for the application of large models, and automobiles are an important field of manufacturing," An Xiaopeng said, "Although large models are still in the 'pre-Newtonian era', this does not affect the application of large models in automotive scenarios today." ”
In terms of autonomous driving and intelligent cockpits, large model technologies such as BEV (bird's-eye view), cognition, and NLP (natural language processing) are expected to boost vehicle intelligence to a new level. In the production and manufacturing process, large models can directly serve the R&D and innovation of smart cars, robots, chips and other products; In the production process, the natural language interaction capability based on large models bridges a large number of breakpoints in the process of enterprise data flow, and builds a new infrastructure for real-time and ubiquitous connections within manufacturing enterprises and between upstream and downstream industries.
Wang Ting, a distinguished researcher at the Shenyang Institute of Automation, Chinese Academy of Sciences, told the China Electronics News that the application of the general model in the field of industrial manufacturing will not only involve the whole production process from product design to process planning to production and after-sales, but also involve all links in the chain from raw materials to material processing, manufacturing and transportation, as well as ERP (enterprise resource planning), MES (production management system), PLC (programmable logic controller), etc., which is not an easy task.
We believe that the entry of large models into the core control systems of the production process, such as PLC, MES, SCADA (data acquisition and monitoring and control system), etc., to improve the intelligence of the process production process is the key symbol of the application of large models in the manufacturing industry. An Xiaopeng believes that from autonomous driving, vehicle-machine interaction, product design to chain optimization, intelligent marketing, vehicle use, content production, digital employees, etc., artificial intelligence technology will bring subversive changes to the automotive industry in the whole scene, the whole life cycle, and the whole industry chain, and this round of change has just begun.
In the future, generative AI technology represented by large models will open up more imagination space for the automobile manufacturing industry. In the spark of the passionate collision between emerging technologies and traditional manufacturing, the curtain of artificial intelligence empowering new industrialization is slowly opening.
Author丨Song Jing.
Editor丨Zhao Chen.
American editor丨Mary.
Producer丨Lian Xiaodong.