Vibration Analysis

Analyzes wind turbine drivetrain vibration data (main bearing, gearbox, generator) from CMS trends, RMS/peak values, frequency spectrum, and SCADA alarms. Classifies severity (1-5) and recommends shutdown or monitoring actions.

Wind Turbine Drivetrain Vibration Analysis

Evaluates drivetrain vibration health across three subsystems: main bearing, gearbox, and generator.

When to Use

Load this skill when the user wants to:

• Assess drivetrain vibration health from CMS or SCADA data
• Interpret RMS, peak-to-peak, or spectral findings for main bearing, gearbox, or generator
• Correlate vibration alarms with operational events
• Decide whether to continue operating, increase monitoring, or shut down

Drivetrain Components

Component	Sensor Location	Key Frequencies
Main Bearing	Non-drive end, drive end	BPFO, BPFI, BSF, FTF
Gearbox LSS	Low speed shaft	Gear mesh (LSS x teeth), bearing defect freqs
Gearbox IMS	Intermediate shaft	IMS gear mesh harmonics
Gearbox HSS	High speed shaft	HSS gear mesh, bearing defect freqs
Generator NDE	Non-drive end bearing	Electrical harmonics, bearing defect freqs
Generator DE	Drive end bearing	Bearing defect freqs, rotor unbalance

Vibration Thresholds (ISO 10816 / CMS Reference)

Location	Normal	Warning	Critical
Main Bearing RMS (g)	< 0.3	0.3 - 0.8	> 0.8
Gearbox HSS RMS (g)	< 0.5	0.5 - 1.5	> 1.5
Gearbox LSS/IMS RMS (g)	< 0.3	0.3 - 1.0	> 1.0
Generator RMS (g)	< 0.5	0.5 - 1.2	> 1.2
Peak-to-peak step change	< 10%	10-30%	> 30%

Note: Always evaluate against site-specific baseline. A 20% rise from stable baseline is more significant than an absolute value alone.

Frequency Fault Signatures

Fault	Frequency Signature
Bearing outer race (BPFO)	(N/2) x (1 - d/D x cos a) x RPM
Bearing inner race (BPFI)	(N/2) x (1 + d/D x cos a) x RPM
Gear mesh	number of teeth x shaft RPM
Gear mesh sidebands	GMF +/- shaft frequency
Rotor unbalance	1x RPM dominant
Misalignment	2x RPM dominant, axial component
Looseness	Sub-harmonics (0.5x, 1.5x) or high harmonic content

Severity Scale

Severity	Label	Description	Action
1	Healthy	All values normal, stable trend	Continue normal operation
2	Early warning	1-2 parameters in warning zone, stable	Increase CMS polling frequency
3	Moderate	Multiple warning flags or single critical	Inspect within 2 weeks
4	Significant	Critical zone or rapid trend growth	Plan shutdown within 48-72 hours
5	Critical	Multiple critical flags, step-change	Immediate shutdown required

Procedure

1. Collect inputs: CMS trend (last 30-90 days), current RMS and peak-to-peak per component, frequency spectrum findings, SCADA alarms, operational context.
2. Evaluate RMS values against thresholds. Flag Warning or Critical zones.
3. Analyze trend:
   • Stable: value in warning zone but flat for >30 days = lower urgency
   • Gradual rise: value increasing steadily = schedule inspection
   • Step change: sudden jump >30% = treat as Critical regardless of absolute value
4. Interpret frequency spectrum if available:
   • Match dominant peaks to fault signatures table
   • Note sidebands around gear mesh frequencies
   • Note sub-harmonics or 1x/2x dominance
5. Correlate with SCADA alarms and operational events.
6. Assign severity per component, then determine drivetrain-level severity as highest.
7. Generate output report using the format below.

Output Format

=== DRIVETRAIN VIBRATION REPORT ===

ASSET : [Turbine ID] SITE : [Site name] DATA PERIOD : [Date range of CMS/SCADA data] MISSING DATA : [List any unavailable inputs]

MAIN BEARING: RMS : [value] g - [Normal / Warning / Critical] Trend : [Stable / Gradual rise / Step change] Spectrum : [Key findings or not available] SCADA Alarms : [Count and type] Severity : [1-5] - [Label]

GEARBOX (LSS / IMS / HSS): RMS : LSS [value] g / IMS [value] g / HSS [value] g Trend : [per shaft] Spectrum : [Key findings] SCADA Alarms : [Count and type] Severity : [1-5] - [Label]

GENERATOR (DE / NDE): RMS : DE [value] g / NDE [value] g Trend : [per bearing] Spectrum : [Key findings] SCADA Alarms : [Count and type] Severity : [1-5] - [Label]

DRIVETRAIN SEVERITY : [1-5] - [Label] SHUTDOWN : [Yes / No / Conditional]

FAULT HYPOTHESIS:

• [e.g., HSS bearing outer race defect - BPFO peak confirmed at X Hz]
• [e.g., Gear mesh sideband modulation - possible gear wear or load variation]

RECOMMENDED ACTIONS:

• [e.g., Increase CMS polling to daily for HSS channel]
• [e.g., Oil sample with ferrography within 72 hours]
• [e.g., Plan HSS bearing replacement at next scheduled outage]

ESCALATION TRIGGERS:

• [e.g., RMS exceeds 1.5 g on HSS - immediate shutdown]
• [e.g., Step change >30% on any channel - treat as critical]
• [e.g., New BPFO or BPFI peak confirmed in spectrum - escalate to Severity 4]

Cross-Skill Correlation

If gearbox visual data is available, load wind-turbine-gearbox skill and cross-correlate:

• High Fe ppm + rising HSS vibration = active wear confirmation
• Spalling in borescope + BPFO peak in spectrum = bearing failure progression
• Normal oil + rising vibration = early fault not yet generating debris (higher urgency)

If blade inspection data is available, check for rotor imbalance:

• 1x RPM dominant in main bearing spectrum + blade damage = aerodynamic imbalance
• Asymmetric blade damage across A/B/C = mass or aerodynamic imbalance source

Pitfalls

• Do not evaluate vibration in isolation. Cross-reference with oil analysis and visual inspection.
• A single high RMS reading during a storm or grid fault is not a fault indicator. Check operational context.
• Spectrum analysis requires RPM-normalized data. Raw frequency peaks are meaningless without shaft RPM.
• Generator electrical faults can appear as vibration. Check electrical data before attributing to mechanical cause.
• Stable high RMS is less urgent than rapidly rising moderate RMS. Trend rate matters more than absolute value.

Verification

After generating the report, confirm with the user:

• Does the severity match CMS system alerts or OEM recommendations?
• Is shaft RPM data available to normalize spectrum frequencies?
• Are there recent maintenance events that could explain vibration changes?
• Is SCADA power curve deviation consistent with vibration findings?