# KNOWLEDGE BASE: PREDICTIVE MAINTENANCE

## 🔧 Maintenance Paradigms Contrast
1. **Reactive Maintenance**: Fix it when it breaks. High repair costs, catastrophic asset damage, massive unplanned downtime.
2. **Preventive Maintenance**: Fix it on a calendar schedule. Leads to over-maintenance and premature replacement of healthy components.
3. **Orbit Predictive AI**: Continuous telemetry monitoring (acoustic, vibration, thermal, current signatures) to diagnose real-world asset health.

---

## 📈 Remaining Useful Life (RUL) Modeling
Orbit AI Maintenance maps structural decay curves by analyzing multidimensional variables. We model the time-to-failure envelope:
- **LSTM Networks**: Trained on historical temperature, pressure, and accelerometer logs.
- **Dynamic Regressions**: Projects current sensor trends against established failure baselines.
- **RUL Prognosis**: Auto-compiles remaining run-hours (e.g. *960 hours remaining before Bearing #4 cage failure*), letting engineers schedule repairs during natural shift intervals.

---

## 🔊 Fast Fourier Transform (FFT) Vibration Scans
Acoustic and vibration accelerometers capture complex physical frequencies. The Orbit Maintenance solver runs 1024-point FFT calculations to convert raw wave data into spectral plots:
- **1x RPM Frequencies**: Rotor dynamic unbalance.
- **2x RPM Frequencies**: Structural misalignment.
- **5x - 20x RPM Frequencies**: Bearing inner/outer race micro-cracks (localized surface defects).
- **High-Frequency Jitter**: Lubrication starvation or cavitation.