振動傾向から推定した残存寿命

Vibration trending provides a reasonable estimate when the fault mechanism produces a consistent growth pattern. Key factors for reliability:

• More data points over a longer period improve accuracy
• Consistent measurement conditions (same point, direction, load, speed)
• Understanding the failure mode helps select the right growth model
• Bearing degradation typically follows exponential growth
• Gradual wear (erosion, buildup) tends to follow linear growth

Treat estimates as planning guidance, not precise predictions. Always err on the side of caution.

Linear growth: Vibration increases at a constant rate (e.g., +0.5 mm/s per month). The time to reach any level is easy to predict. Common for gradual wear processes.

Exponential growth: The rate of increase accelerates over time. Early stages show slow change; later stages show rapid deterioration. Common for bearing damage and fatigue crack propagation.

Why it matters: Exponential growth is more dangerous because it can appear stable for a long time, then deteriorate rapidly. The "hockey stick" curve means the window between detectable change and failure can be short.

General guidelines based on ISO 20816 / ISO 10816 philosophy:

• Plan action when vibration reaches Zone B/C boundary (alarm level) or has increased 50% above baseline
• Schedule maintenance when trend projection shows alarm will be reached within the next planned shutdown window
• Urgent action when vibration reaches Zone C/D boundary (danger level) or has doubled from baseline
• Immediate shutdown when danger level is exceeded or sudden step change occurs

The key principle: schedule corrective action while you still have time to plan, source parts, and execute safely.

Factors that reduce prediction accuracy:

• Inconsistent measurement: Different probe positions, mounting methods, or sensor types
• Variable operating conditions: Load, speed, or temperature changes between measurements
• Step changes: Sudden events (impacts, process upsets) that shift the baseline
• Wrong growth model: Using linear when actual growth is exponential, or vice versa
• Insufficient data: Fewer than 3–4 data points makes trend fitting unreliable
• Multiple fault mechanisms: If more than one fault is developing, the combined trend may not follow a simple model

Recommended measurement intervals based on machine criticality:

• Critical machines: Continuous online monitoring (24/7)
• Important machines: Weekly to monthly route-based measurements
• General machines: Monthly to quarterly

When a trend is rising:

• Zone A→B transition: Increase to weekly
• Approaching alarm: Increase to daily or install temporary online monitoring
• Above alarm: Continuous monitoring until maintenance

The measurement interval must be shorter than the P-F interval (time from detectable fault to functional failure) divided by at least 3, to ensure you capture the trend and have time to act.