Cross-signal anomaly detection across production, OEE, downtime, and BOM variance — surfacing the runs, machines, and materials that need attention this period.
AI analyzed 95 production runs across 17 machines and 18 finished goods, covering 2,175 h of shift time. 5 signals need attention this period — roughly OMR 6,390 of material cost is at risk.
Unspecified downtime accounts for 551 hours across 48 events on 14 machines. The next largest reasons are far smaller: No Plan 12 h, Dye Change 9 h, Heat Check 6 h.
When downtime is not classified, teams cannot target the real failure modes, so repeat losses keep coming back.
Shift A produced 24,763 units, while Shift B shows 0 units against 10 docs. Total plant output came mostly from BOTH-shift runs at 36,812 units.
A zero-output shift usually signals a scheduling, staffing, or reporting breakdown that can hide lost capacity and delay recovery actions.
Production run DPE2026011764 for FG7210 used 1,163.347 kg vs 120 kg expected, a variance of 1,043.347 kg. That is 8.69x above plan.
Extreme input variance points to scrap, rework, or master-data errors and can quickly erode margin on the run.
The plant produced 61,575 units across 95 production runs, but 525.5 down hours cut availability to 75.8%. That is 3.5 h lost for every 10 h scheduled.
Low availability directly reduces output and increases the cost per unit by spreading fixed labor and overhead across fewer good units.
Unspecified downtime accounts for 551 hours across 48 events on 14 machines. The next largest reasons are far smaller: No Plan 12 h, Dye Change 9 h, Heat Check 6 h.
When downtime is not classified, teams cannot target the real failure modes, so repeat losses keep coming back.
Shift A produced 24,763 units, while Shift B shows 0 units against 10 docs. Total plant output came mostly from BOTH-shift runs at 36,812 units.
A zero-output shift usually signals a scheduling, staffing, or reporting breakdown that can hide lost capacity and delay recovery actions.
Production run DPE2026011764 for FG7210 used 1,163.347 kg vs 120 kg expected, a variance of 1,043.347 kg. That is 8.69x above plan.
Extreme input variance points to scrap, rework, or master-data errors and can quickly erode margin on the run.
LSP100 logged 435 run hours and only 21 down hours, giving 95.4% availability and 0.954 OEE across 19 docs. SJ-60/36-20 is also strong at 92.6% availability.
High-performing machines are the best candidates for protecting schedule stability and replicating their operating practices elsewhere.
APEX-7 ran 58 hours and was down 58 hours, resulting in just 50% availability and 0.50 OEE across 6 docs. APEX - 1 is also weak at 44.4% OEE.
Low-OEE bottleneck machines constrain total plant throughput and often create downstream waiting, changeover, and rescheduling losses.