Number one is boring, and that's exactly why it pays back first. I ranked ten AI use cases for factories by payback. Digital twins, cameras and sensors came last, because they need data most plants don't have yet. #1...

یه اکسل شلوغ تولید دادم به چت‌جی‌پی‌تی؛ داشبورد ساخت، بعد گلوگاه رو اشتباه گفت. شلوغ‌ترین صف ≠ گلوگاه. عددها مثال فرضی‌اند. I gave ChatGPT a messy production Excel; it built a dashboard, then named the wrong bottleneck. Busiest queue...

۲۴ سال در تولید کار کرده‌ام. یه اکسل شلوغ و ساختگی دادم به چت‌جی‌پی‌تی؛ داشبورد ساخت، بعد گلوگاه رو اشتباه گفت و تعریف دو تا شاخص رو عوض کرد. توی این ویدیو قدم‌به‌قدم می‌بینید چی بهش دادم، چی...

AI won on accuracy in a spare-parts forecast test. Then something surprised me. Illustrative synthetic demo: moving average ~58% WAPE, Croston-SBA ~53%, gradient boosting with the overhaul plan ~44%. Under a simple reorder rule, better accuracy did not automatically...

Your most frequent breakdown is probably not your most expensive one. On an illustrative press with six months of maintenance log, electrical faults happened 14 times and hydraulic leaks only 9. But sort by minutes down, and hydraulic...

Can AI forecast spare parts better than a moving average? In this video I race three methods on a fully synthetic data set: a 12-week moving average, Croston-SBA, and gradient boosting that also sees the planned overhaul calendar. In...