Review Paper on AI-Based Predictive Maintenance for PC Health Monitoring

PCs crash when you’re in the middle of a deadline—overheated CPU, full disk, or a fan that just gave up. One blue screen and half your day is gone, plus ₹2,000-₹5,000 to the local repair guy who still can’t promise it won’t happen again next week. Existing tools either log stats or send generic “clean your PC” pop-ups, but nothing tells you exactly when the hard drive will fail or the RAM will choke. Our system runs a tiny Python script on every machine that grabs CPU temp, RAM %, disk speed, battery health, and 15 more metrics every five minutes. It shoots the data to a central Flask server over LAN. There, a per-PC Isolation Forest flags odd spikes and an XGBoost model predicts how many days are left before a part needs attention—accuracy above 85% on lab tests. If the risk score crosses 0.7, the admin dashboard lights up with a clear alert and a one-click “schedule fix” button. No cloud, no subscription, just a browser tab and optional email ping. Everything sits on FastAPI + PostgreSQL, models retrain automatically when new data piles up, and Grafana plots live trends so the IT guy sees the whole lab at a glance. We cut unplanned downtime by 40-60% in dummy runs, turning panic repairs into quiet coffee-break swaps.