Health Score Calculation Logic
The ESP-GUARDX Health Score is an AI-based predictive maintenance index for the complete ESP system. It combines surface power behavior, VSD stress, transformer and cable condition, motor diagnostics, seal behavior, pump hydraulics, intake/gas behavior, and fluid/reservoir trends into one live value from 0 to 100.
Unlike a normal alarm counter, the score is not only asking whether one threshold has tripped. It evaluates how serious each developing condition is, how confident the AI diagnosis is, how long the condition has persisted, and whether physical exclusion logic rules out competing diagnoses.
Core Equation
S_health = 100 - D_total
D_total = sum(W_d * R_d)
| Term | Meaning |
|---|---|
S_health |
Final ESP Health Score |
D_total |
Total weighted degradation |
W_d |
Domain weighting factor |
R_d |
Risk score of each domain |
Domain Risk Equation
R_d = sum(S_i * C_i * P_i * E_i)
| Term | Meaning |
|---|---|
S_i |
Severity of the fault or abnormal condition |
C_i |
AI confidence level for that diagnosis |
P_i |
Persistence factor |
E_i |
Exclusion logic factor |
Calculation Elements
Severity measures how serious the abnormality is right now. It is derived from the magnitude of deviation from the healthy operating baseline.
Confidence measures how certain the AI engine is about the diagnosis. It is produced from multi-sensor confirmation, historical fingerprint matching, and differential diagnosis logic.
Persistence measures how long the abnormal condition has continued. This prevents short startup transients or one-off disturbances from unfairly reducing the Health Score.
P_i = 1 - e^(-t / tau)
Where t is abnormal duration and tau is the characteristic degradation time.
Exclusion logic removes impossible diagnoses. For example, if intake pressure is fluctuating, the pump is still rotating and interacting with fluid/gas, so a broken-shaft diagnosis can be excluded.
Recommended Domain Weighting
| Domain | Weight |
|---|---|
| Surface Equipment / VSD | 10% |
| Power Quality | 10% |
| Cable / Surface Connection | 10% |
| Motor | 15% |
| Seal Section | 15% |
| Pump | 20% |
| Intake / Gas Handling | 10% |
| Fluid / Tubing / Reservoir | 10% |
Domain Coverage
| Domain | Signals and fault modes considered |
|---|---|
| Surface Equipment / VSD | VSD switching stress, inverter ringing, DC bus instability |
| Power Quality | Voltage THD, current THD, harmonic heating, PCC instability |
| Cable / Surface Connection | Cable splice degradation, partial discharge, leakage current |
| Motor | Rotor condition, slip behavior, torque oscillation, mechanical vibration |
| Seal Section | Thrust-bearing wear, seal temperature, water ingress, oil contamination |
| Pump | Efficiency degradation, BEP deviation, cavitation, scaling, rubbing |
| Intake / Gas Handling | Gas lock, intake-pressure oscillation, slugging behavior |
| Fluid / Tubing / Reservoir | Water cut, GOR trend, sand production, paraffin/asphaltene buildup |
Score Bands
| Health Score | Status | Interpretation |
|---|---|---|
| 90-100 | Excellent | System operating normally |
| 75-89 | Good | Minor degradation detected |
| 60-74 | Warning | Early failure indicators developing |
| 40-59 | Critical | Significant degradation detected |
| < 40 | High Risk | Immediate intervention recommended |
Operational Use
The Health Score should be read as a predictive-maintenance index, not as a single trip alarm. A falling score means degradation patterns are accumulating across the ESP system. Operators should open the contributing screens, review the AI diagnostic calls, and confirm whether the same pattern is visible in power quality, motor behavior, pump curves, well behavior, and event history.
See the worked 65/100 Health Score example for a complete multi-domain calculation.