Where this conversation describes AI being used to build or deliver a product. Part of the AI in practice tracker.
Engineering & simulationSimulation accuracy improved to 100% by combining physics calculations (80%) with real data (20%)
The team used real-world physical trial data to correct simulation residuals, creating accurate models that predict how sheet metal deforms, allowing them to determine process parameters without expensive physical prototyping.
“We use physical calculations and physical equations to get, you know, 80% there, and the last 20%, we use actual physical trials and the data from real, uh, real world to, um, make the same very accurate.”
ManufacturingRoboCraftsman (AI-driven robotic system)Lead time reduced from 4 years to 2-3 days
The system uses online active control loops and machine learning to adjust robot movements in real-time based on sensor data, allowing it to form complex metal parts without specific tooling or molds.
“We have different computers and control systems that look at the sensor data and are adjusting what the robot does in every form millisecond. So it's an online active control loop. There is a software that comes with it that allows you to prepare. Uh, you know, your geometries, your parts for production, uh, and then analyze the final results.”
Quality & testRoboCraftsman scanning and analysis software
The robot scans the formed part after manufacturing, and the software analyzes the data to perform quality control and identify defects or deviations from the design.
“For example, the robot after it forms it, scans it, and it gives you, okay, what it did. So you can look at that data and do analysis, do QC.”
Engineering & simulationMachine learning models on RoboCraftsman
The system learns from the results of each formed part to improve its future performance and potentially learn new operations, continuously refining its process parameters over time.
“the system also uses that data to build models, uh, to improve what it does, right? So over time, as it does more and more, it also learns from the result that it achieved and constantly improves or potentially learns new operation.”