AEGIS · Predictive Maintenance

Project 7 — AI4I 2020 Machine Failure Risk Scoring

Model: Random Forest

Enter current sensor readings

F1: 0.8413 Precision: 0.9138 Recall: 0.7794 MCC: 0.839 ROC-AUC: 0.9766

Fields start pre-filled with typical values for normal operation (median readings from non-failure machines) — use the field's up/down arrows or type over them with this machine's actual readings.

API

POST JSON to /api/predict with the six feature keys shown above (Type as "L"/"M"/"H", the rest as numbers) to integrate this into a monitoring pipeline instead of the form.

This is a decision-support tool, not an automatic shutdown trigger — the RED band means "send someone to look," not "the machine will fail." The underlying model is trained on the AI4I 2020 synthetic dataset; before relying on this for real equipment, retrain on your own machines' sensor history.