AI Reduced Manufacturing Downtime by 90%

The V-Soft Digital team worked with a multi-national appliance and air conditioning manufacturer that produces thousands of different models of products at dozens of production plants in multiple continents. Despite using some of the world’s most sophisticated production robotics, die-presses, and process controls to meet growing product demand and ensure quality control, the organization’s production lines were challenged daily with the need for quick and accurate die plate changeovers.

Production changeovers demand die-plates be reconfigured constantly to meet the work-order requirements of a production run. The reconfiguration process is complex and time consuming.

With hundreds of different plates and dozens of permutations possible, ensuring the correct configuration is time consuming and prone to error. A misconfiguration leads to a cascade of production problems, starting with downtime to reconfigure, repairing crashed plates, and reduced production output.

Solution

The V-Soft Digital team deployed an artificial intelligence production changeover control system to improve workforce efficiency, time to competency, and inspection quality.

The workflows of the production teams required the solution to be portable and non-disruptive to the operation. This meant a traditional fixed camera wasn’t appropriate. The tens of thousands of configuration permutations, dynamic lighting conditions, and the need for near 100% accuracy 100% of the time, could not rely on traditional computer vision technology. The V-Soft Digital multilayered AI processing platform provided the flexibility needed to operate in the challenging environment while delivering accurate and precise results. Operators simply took a video with a tablet device of the die-plate configuration and were provided information on any misconfigurations and what corrective action was needed.

Technology Applied

• Image Preprocessing eliminated reflections and glare and accentuated product features such as holes and contours before misconfiguration identification occurred.
• Vector Space Multi-Frame Analysis matched objects from one frame of a video to another, providing hundreds of representations of a single shape from different angles, preventing the chance of obstruction, and increasing accuracy.
• Homography realigned images on the same planar surface for operators/production crews to take video without worrying about orientation or angle.
• Multilayered and Multi-cropping AI Computer Vision analyzed processed images or objects at different levels of granularity and with increased precision. First general shapes and die positions were viewed, then each shape for punches, holes, and screws, and subsequently each screw, punch, hole, and other reference objects.

Outcome

Tablet devices with edge-AI computer vision capabilities empowered production crews to check their configured dies and take immediate corrective action on misconfigurations instead of after it was too late.
This solution reduced die-press downtime by 90% and saved the organization $40,000/month for each of its production facilities (dozens globally).