Written by Jijun Ma, Member of the EdgeX Foundry Governing Board and Industrial Internet Director at Wanxiang Group
The Industrial Internet Consortium® (IIC) announced the Optimizing Manufacturing Processes by Artificial Intelligence (OMPAI) testbed yesterday. The OMPAI testbed is led by IIC members Wanxiang Group (also a member of EdgeX Foundry) and Thingswise and supported by Dell EMC (a founding member of EdgeX Foundry), Xilinx, China Unicom, and China Academy of Information and Communication Technology (CAICT).
The OMPAI testbed, which is the first testbed based on EdgeX edge computing platform, explores the application of artificial intelligence (AI) and industrial internet technologies, deployed from the edge to the cloud, to optimize automotive manufacturing processes. It also seeks to create an ecosystem that will foster the exchange of IT/AI/OT domain knowledge and the co-development of smart manufacturing applications. For example, deep learning may be able to improve quality assurance of an automobile part to substantially increase the detection of defects and reduce the need for manual inspection.
Vincent Wang, Chief Innovation Officer of Wanxiang Holdings, said, “As a leading multinational corporation in automotive and renewable energy, with factories in Europe, North America and Asia, we believe that an industrial IoT platform will be a key enabler for our digital transformation and global synergy. We are glad to work with technology leaders to validate AI, edge-cloud collaborative computers, and high-speed cellular networks to optimize manufacturing productivity and quality. This is the first step toward an open, inclusive IIoT platform on which we will continue with further testbeds, incorporating new ideas, new data usage models and creating greater value add. We invite worldwide enterprises, innovators and entrepreneurs to enrich the ecosystem together.”
In the edge platform, AI models and edge applications are run for the local optimization of manufacturing processes. In the cloud platform, they are run to enable global and long-term optimization, e.g. across production lines and plants. The edge platform also supports connectivity to and data collection from the equipment while the cloud enables historical data accumulation and storage and supports AI model building.
The cloud computing platform also provides the capability for enabling industrial app DevOps processes supporting collaboration between AI/IT developers and plant engineers in creating, testing and running data/AI model-driven industrial applications. The following image shows the solution overview of this testbed.
Blow are the usage scenarios in our testbed.
Machine vision on-line quality assurance
The main theme of this scenario is to exploit the capability of deep learning in image pattern recognition to improve quality assurance effectiveness and efficiency by increasing defect detection accuracy, reducing dependence on manual inspection and at the end providing online feedback to the production process to reduce defect rate.
Battery Cell Welding Quality Control
In this scenario, it is going to use historical data to analyze the relationship between the welding process and environmental parameters and the product quality and use that to predict in real time quality product and provide recommendation for optimization.
Wheel Bearing Production Line Balance & Optimization
A high throughput discrete manufacturing line usually consists of many workstations involving with various equipment and processes. These workstations may have different production throughput that vary depending on their process parameters. Mismatched throughput between the workstations would impede the overall production line throughput, reducing overall equipment utilization and production capacity.
Big data analytics on data collected from the workstation equipment can be used to monitor production pace of each of the workstations and overall throughput, and to identify bottlenecks and recommend optimization solutions.
Predictive Maintenance of grinding machines
In a manufacturing environment, equipment failures interrupt production lines or cause product quality issues, aggravated by the prevailing condition that few or no spare parts are usually kept for key equipment, e.g., grinding machines and motors, resulting severe reduction of production capacity in the event of equipment failures.
The current solution of “preventive maintenance” relying on periodic manual inspection is ineffective, laborious and interruptive to production. Predictive Maintenance for the equipment enabled by machine learning will be experimented within the general framework to effectively address this common manufacturing issue.
This testbed is open to new innovative ideas and EdgeX Foundry members are welcomed to join us to widely use EdgeX for industrial internet solution.