On September 10th, the "2024 BRICS Industrial Innovation Competition" hosted by the Ministry of Industry and Information Technology and the BRICS New Industrial Revolution Partnership Innovation Base came to an end in Xiamen. The competition lasted for 4 months, with 1357 participating projects covering tracks such as "Artificial Intelligence Big Model", "Low altitude Industry General Navigation", and "Green Industry Low carbon Technology". After preliminary selection, final defense, expert deliberation, and approval by the organizing committee, nearly 200 outstanding projects were ultimately selected. With its unique algorithm, model training technology, products and industry solutions, SIE Information "PCB industry big model and application based on Shanmou GPT" has stood out in more than a thousand projects, won the "AI big model" track excellent project award in the 2024 BRICS Industrial Innovation Competition, and was also included in the "BRICS industrial innovation cooperation excellent project collection".

Award certificate

In recent years, with the continuous breakthrough of new generation information technology, new electronic products such as intelligent cars and VR devices have been constantly developing. The emerging electronic product market represented by in car ADAS, in car radar, wearable devices, AR/VR metaverse devices, etc. has risen rapidly, driving the rapid growth of demand for mid to high end PCB products. At the same time, the rapid development of artificial intelligence technology represented by ChatGPT is expected to bring about a major explosion of AI servers and products in the field of artificial intelligence, bringing new opportunities to the PCB industry. According to data from the China Academy of Commerce Industry Research, the size of China's PCB market will reach 307.816 billion yuan in 2022 and approximately 309.663 billion yuan in 2023. It is expected that the size of China's PCB market will further grow to 346.902 billion yuan in 2024.

Benefiting from the continuous strengthening of downstream demand such as AI servers, 5G, cloud computing, and automobiles, the PCB (printed circuit board) industry is accelerating its recovery, showing characteristics such as a "steady off-season" and a positive industry development trend. At the same time, based on Yimeike's over 15 years of digital service experience and customer accumulation in the PCB manufacturing industry, the rapid development of the PCB industry and the increasing demand for high-end products, as well as China's important position in the global PCB market, the "PCB industry big model and application based on Shanmou GPT" has high feasibility for large-scale implementation and application. In addition, the low concentration of China's PCB industry provides opportunities for the application and market penetration of new technologies.

Trend Chart of China PCB Market Size Forecast from 2019 to 2024

However, with the increasing miniaturization and functionalization of electronic products, downstream customers' demands for PCB accuracy, density, performance, and reliability continue to rise, as well as the growing demand for customized solutions, leading PCB manufacturing enterprises to face multiple challenges in technological innovation, cost-effectiveness, quality control, delivery speed, and customer service. Especially with the advancement of technology and the continuous changes in market demand, customers have increasingly high requirements for quotation speed and accuracy. Under the current service mode of PCB production and manufacturing enterprises, they need to rely entirely on manual interpretation of various file formats (such as Gerber, PDF, Word, Excel, TXT) and multilingual business descriptions of order data, extract key engineering parameters from them, and compare material cost prices to form the final quotation. This places extremely high demands on process engineers, and the process is extremely time-consuming and error prone.

Pain points in the PCB order quotation process at the final defense venue

In order to solve the above industry pain points, SIE Information, together with its holding subsidiary Yimeike, has built the industry's first PCB industry big model based on Shanmou GPT, relying on the self-developed AI middle platform - Shanmou GPT, and based on universal big model technologies such as natural language processing (NLP), computer vision (CV), machine vision (MV), multi-mode (text, image, audio, video). Through the knowledge reasoning, knowledge understanding, and content generation capabilities of the big model, it provides the PCB/FPC manufacturing industry with two major supporting applications of automated MI parameter extraction and intelligent quotation:

Automated MI parameter extraction: Utilizing advanced image recognition and natural language processing technologies to automatically and accurately identify and parse key parameters such as layer number, board thickness, and material in engineering drawings, and convert these unstructured data into structured data that can be used in the manufacturing process, greatly reducing the need for manual input, improving data accuracy and processing efficiency, and accelerating the preparation time of the entire PCB manufacturing process.

Intelligent quotation system: By analyzing the MI parameters and design complexity in the drawings through a large model, combined with factors such as material cost, manufacturing process, and market demand, it automatically generates accurate quotations, greatly improving the speed and accuracy of engineering quotations, helping enterprises optimize cost control, and enhance market competitiveness.

PCB industry big model architecture based on Shanmou GPT

Based on the construction of two core applications, automatic recognition, reading, and understanding of engineering drawings and documents have been achieved. According to Guo Taoshan, General Manager of SIE Yimeike, currently, the "PCB Industry Large Model and Application Based on Shanmou GPT" can accurately identify and extract key engineering parameters in PCB engineering drawings, such as size, material, layer number, etc., shortening the parameter extraction work that originally required 4-6 hours to a few minutes, while ensuring an accuracy rate of over 95%. This innovation not only greatly improves the efficiency from order receipt to production preparation, but also significantly reduces quality issues caused by human errors. Its core competencies are as follows:

Automated drawing recognition: Using advanced image recognition and artificial intelligence technology, it automatically interprets drawing files in different formats such as Gerber, PDF, Word, Excel, TXT, etc., reducing the drawing interpretation time to minutes.

Automated drawing interpretation: capable of handling multiple file formats and languages, simplifying the operation process for engineers, and reducing the difficulty and time cost of interpretation caused by format diversity.

Automated parameter extraction and verification: Using machine learning algorithms to automatically identify key parameters such as components, circuits, and layers in PCB drawings, and introducing an automatic verification mechanism to reduce manual error rates and improve accuracy.

Real time material cost update and matching: Establish a real-time updated material cost database, which is linked with the automatic quotation system to ensure the accuracy and updating of cost data.

Final defense venue - PCB industry big model capability

The BRICS Industrial Innovation Competition aims to explore and collect a group of projects with outstanding innovation capabilities, excellent application effects, and promising market prospects from BRICS countries, and promote pragmatic cooperation in the field of the new industrial revolution. SIE won the honor in this competition, which not only demonstrated the recognition of SIE products and solutions in the industry, but also demonstrated the high innovation and excellent application value of the PCB industry model. SIE will continue to adhere to the concept of innovation driven development, constantly break through technical bottlenecks, optimize product performance, and commit to turning more innovative achievements into a powerful driving force for social progress and industrial upgrading.