How AI-Powered Business Processes Are Revolutionizing U.S. Manufacturing

Delta frequently adjusts and optimizes its production processes in response to the components and products that their customers require. It is not easy to accomplish this on an extremely complicated, fast-moving electronics production line for power supply. Their team joins machines in the production line, then uses mockups and static images to detect and eliminate physical collisions, problems, and faults. However, this method uses position switches to identify trays and objects on the production line. Delta's production lines required additional position switches as the number of robots and machines expanded. This was neither scalable or sufficient to meet their demand for flexible manufacturing, as reconfiguring each position switch required more than 30 minutes, resulting in significant production downtime whenever the layout needed to be modified.

To overcome these issues, Delta created a physically realistic digital doppelganger of their production lines.

This enabled them to iterate considerably faster on reconfiguration and run numerous "what-if" simulations to discover the best line designs. Another significant benefit of a digital twin was the ability to swiftly generate synthetic data at scale in order to test and train extremely accurate computer vision models for detecting trays and products in the production line. This innovative technique enabled the team to significantly accelerate the production planning process by eliminating the need to reconfigure every position move. Delta Electronics Production Line Delta Electronics A unified asset pipeline for creating digital twins. The process of developing a digital twin starts with accumulating and visualizing 3D assets from the surroundings. Delta models and simulates its production line using numerous 3D tools such as Autodesk 3ds Max, FlexSim, and Visual Components. Historically, bringing data from these modeling and simulation tools into one environment was practically impossible due to time-consuming data transfer, model decimation, and interoperability issues. Even after the data has been aggregated, if any modifications are made to the models or processes in the source 3D applications, the aggregation process is restarted to update any changes. Delta may use Omniverse to connect their multiple applications and data into a unified asset pipeline, allowing their teams to visualize and work on the full output in a single environment.

This is all made possible by OpenUSD and other connectors and extensions that enable third party technologies to send live data into a USD stage via a Nucleus server.

Once all of the components are included in the digital twin, the Delta team can conduct simulations to detect flaws early in the design and reconfiguration process, before making changes to their actual production lines. This is critical in avoiding costly downtime and change-order requests. "When apps connect to Omniverse, all of our files and data are synced simultaneously, allowing team members to see updates in the USD Stage and collaborate in real time," stated Ares Chen, PSBG General Manager at Delta Electronics. "USD enables seamless collaboration with each team member, so we can design a production line faster and more efficiently than before." Delta use computer vision for automated inspection of the final assembly, looking for faults like as missing components or misplaced screws. AI-assisted Automated Optical Inspection (AOI) greatly speeds up the inspection process, allowing Delta to discover errors earlier and reduce the need for manual intervention. Delta encountered difficulties when training AI models because manually collecting and classifying data takes up to two days for 1000 photos. To speed things up and save money, they switched to using synthetic data with Omniverse Replicator. They can now produce the same number of annotated photos in 10 minutes that it would take two days to do manually.

The team also obtains 90% accuracy on synthetic data, which is similar to real data.

The great win is that they achieved this level of precision in a mere fraction of the time it normally takes to collect real data now done in one-hundredth of the time. This not only speeds up AI training, but also reduces costs and improves efficiency for computer vision applications. Delta also employs NVIDIA Isaac Sim, a scalable robotics simulation application, to precisely replicate model performance and detect the position of a tray on the manufacturing line. Delta can reduce downtime and hazards by generating digital twins in Omniverse and reoptimizing production processes. Powered by artificial intelligence (AI), the application of AI algorithms in manufacturing has altered firm operations, resulting in significant increases in productivity, quality, and cost savings. The fragmented deployments we see today will eventually give way to the agility and operational intelligence that this new level of data management provides. ADVERTISEMENT AI in manufacturing nowadays. AI is rapidly gaining popularity, and new applications are altering the industry's landscape. According to Fortune Business Insights, AI's value in the global manufacturing sector was $8.14 billion in 2019 and is expected to reach $695.16 billion by 2032. This is part of a larger trend known as Industry 4.0, in which connectivity and advanced analytics pave the way for more agile and productive manufacturing that occurs on the go. Several examples follow.