Digital Transformation and Business Process Management in the U.S.
Delta's production lines are often changed and improved based on the parts and goods they need to make for their customers. It's not easy to do this on a very complicated and fast-moving electronics production line for putting together power supplies. After connecting the machines in the production line, their team uses mockups and static pictures to find and fix any problems, collisions, or mistakes that happen in real life.
Position switches are used to find trays and other things on the production line, though
Delta had to add more position switches when the number of robots and tools in their production lines grew. Because it took more than 30 minutes to reconfigure each position switch, this wasn't scalable or flexible enough to meet their needs. Whenever the plan had to be changed, there was expensive production downtime. To deal with these problems, Delta created a digital twin of their production lines that was exact in every way. This made it possible for them to change things around much more quickly and run many "what-if" models to find the best line designs.A digital twin also had the important benefit of being able to quickly and large-scale create fake data that could be used to test and train very accurate computer vision models for finding boxes and products in the production line. It took a lot less time for the team to plan production with this new method because they didn't have to rearrange every job switch.Line for Making Delta ElectronicsFrom Delta ElectronicsA single stream of assets for creating digital twinsPutting together and seeing 3D assets from the world is the first step in making a digital twin. Delta models and simulates their production line with a number of 3D programs, such as Autodesk 3ds Max, FlexSim, and Visual Components. If you wanted to bring data from these modeling and simulation tools into one setting in the past, you would have had to go through time-consuming data transfer, model decimation, and interoperability problems. If any changes are made to models or processes in the source 3D applications after the data has been gathered, the aggregation process starts all over again to reflect those changes.Delta can use Omniverse to connect all of their different applications and data using a unified asset flow. This lets their teams see and work together on the whole output in a single space. OpenUSD and other add-ons and
links make this all possible by letting third-party tools send live data to a USD stage through a Nucleus server
Once all the assets are put together in the digital twin, the Delta team can use simulations to find problems early on in the planning and reconfiguration process, before they make any changes to their real production lines. This is very important if you want to avoid expensive downtime and calls for changes.Delta Electronics' PSBG General Manager, Ares Chen, said, "When apps connect to Omniverse, all of our files and data can be synced at the same time. This way, team members can see updates in the USD Stage and work together in real time." "USD enables seamless collaboration with each team member, so we can design a production line faster and more efficiently than before."The use of Isaac Sim and Omniverse Replicator has greatly improved our ability to create synthetic data, making it over 100 times better than with past training methods. This makes things much more efficient and speeds up the process of designing new production lines.Ares Chen is the general manager of PSBG and Delta Electronics.
Automated Optical Inspection with AI Delta Electronics
Using fake data to speed up industrial inspection.
Delta uses computer vision to check the final assembly automatically for problems like missing parts or screws that aren't lined up right. Automated Optical Inspection (AOI) with AI speeds up the inspection process by a large amount. This lets Delta find problems early and reduces the need for human intervention.They moved to using fake data with Omniverse Replicator to get things done faster and save money. Now, it only takes 10 minutes to make as many labeled pictures as it would take two days to do by hand.The team also gets 90% accuracy with fake data, which is the same as real data. The big win is that they got to this level of accuracy in a hundredth of the time it usually takes to collect real data. This is good for computer vision uses because it speeds up AI training and also makes it cheaper and more effective.Delta also uses NVIDIA Isaac Sim, a scalable robotics simulation app, to correctly simulate how well the model works and find where a tray is in the production line.Delta can re-optimize production lines and cut down on downtime and risks by making digital twins in Omniverse.
Comments
Post a Comment