Application of 5G+ Industrial Internet
By the end of 2020, China Mobile will have 350,000 5G base stations and provide commercial 5G services in cities above the prefecture-level. In order to accelerate the application of 5G in vertical industries, China Mobile plans to launch 100 group-level key projects and 1000 province-level feature projects, accelerate the building of high-quality 5G demo projects, and promote the deployment of 5G private network products.
Jiangsu Mobile has actively explored the application of "5G+ industrial Internet" in the industrial manufacturing field. The manufacturing cluster in Jiangsu Province is developed but also has the difficulty in improving production efficiency, guaranteeing production security, and an insufficient level of digitalization. Taking 5G network as the cornerstone and industrial intelligence application as the entry point, Jiangsu Mobile has created 13 key group-level demo projects and 140 province-level demo projects. It has carried out the "5G+ industrial Internet" exploration jointly with 46 industrial customers, including analyzing the industrial informatization scenarios, sorting out network requirements and preparing pilot schemes, and among them, 12 industrial customers have started the "5G+ industrial Internet" transformation project.
5G+AI Empowering Quality Inspection
Changzhou Branch of Jiangsu Mobile, together with ZTE and MICRO-Intelligence, has successfully implemented an end-to-end solution for the 5G+AI quality inspection demo workshop at Jiangsu GIAN Technology Co., Ltd (hereinafter referred to as GIAN).
As a professional metal injection molding (MIM) product manufacturer and solution provider, GIAN provides high-complexity, high-precision, and high-strength customized MIM core components for consumer electronics and the automotive industry. GIAN attaches great importance to the R&D of product technology, and has won the second prize in the 2019 National Technological Invention Award for the near-net shape manufacturing technology and applications of high-performance special powder materials. The company's products have been applied to well-known consumer electronics and automotive brands.
In the actual production of GIAN, the quality inspection of 3C product parts requires a large amount of manpower. In particular, overseas customers impose strict quality requirements. Each part requires the use of industrial electron microscope and takes 30 seconds to 1 minute to complete the quality inspection. Therefore, GIAN invests nearly 3000 workers in the quality inspection workshop, accounting for 50 percent of the total number of workers in the factory. The labor cost per month exceeds 25 million yuan. However, quality inspection in manufacturing has always faced the pain points of unstable manual test quality, difficulty in recruiting, keeping employees, difficulty in training, and high costs.
Changzhou Branch of Jiangsu Mobile, together with ZTE and its ecological partner, MICRO-Intelligence, provides intelligent solutions for industrial quality inspection of GIAN (Fig. 1). As a company focusing on the industrial visual inspection and industrial big data platform, MICRO-Intelligence has a leading position in AI-powered quality inspection in China. The AI quality checker customized for GIAN integrates industrial cameras, robotic arms, programmable logic controllers (PLCs), and other components. The AI-powered quality inspection machine takes several consecutive UHD photos of the inspected items, and transmits them to the AI computing power platform for visual inspection and detection. Based on the image recognition technology, the AI computing power platform simulates the surface detection procedure of production line workers, and performs machine learning through samples to achieve accurate defect detection.
The quality control machine sends multiple HD photos to the AI computing power platform. Depending on the number of photos, the uplink network rate required by the network is 150 to 300 Mbps. The conventional 2.6 GHz new radio network with a smaller number of timeslots allocated to uplink cannot meet the machine's requirements. According to the service requirements, ZTE has designed the 5G+MEC private network solution. China Mobile has 100 MHz in the 4.9 GHz band, which can be used for private network deployment without affecting the public network. On the other hand, the industrial private network usually requires a much higher uplink rate than the downlink rate. The general 2.5 ms two-period frame structure has a low proportion of uplink timeslots and cannot meet the service requirements. ZTE has developed the 4.9 GHz base station with the 2.5 ms single-period 3U1D frame structure, which enhances the uplink transmission rate. The single-user peak rate reaches 700 Mbps, meeting the uplink rate requirements of the quality inspection machine.
For the AI computing power platform, ZTE provides the enhanced integrated MEC edge cloud solution (Fig. 2). Based on the three-layer NFV architecture, the edge cloud system uses a full-stack integration architecture consisting of the basic platform layer, core capability layer, and service application layer with diversified hardware, heterogeneous openness and lightweight management. It coordinates with the cloud to provide edge computing services. The edge network cloud provides comprehensive cloud computing services such as computing, network, storage, acceleration and security at the edge, and provides a stable basic environment for the deployment, scheduling and operation of the NEs and Internet/IT applications. It can reduce the response latency, pressure on the cloud end, bandwidth cost, and meet the diversified edge application scenarios. The data (quality inspection photos) is forwarded locally without traversing the 5G core network of the region. On the other hand, MEC provides powerful computation capability of CPU and GPU to support the operation of AI computing power platform.
After the first phase of the 5G+AI quality inspection system is implemented, its commercial value gradually becomes apparent. First, a single quality inspection machine is 10 times more efficient than manual work, with qualitative improvements in both accuracy and stability. Second, investment in the entire system will be recovered in about one year, and will continue to create value in the future. In addition, based on the big data analysis, the mold dimension tolerance in the injection molding process can be precisely adjusted to improve the precision of the mold and reduce the loss, and optimization suggestions and analysis can be made on the sintering process parameters such as the temperature and pressure to improve the yield rate of the product. After the phase-2 and phase-3 projects are implemented in the future, the company will complete the intelligent transformation of the quality inspection production line to bring greater value to its customers.