Turbofan Predictive Maintenance

249

Engines in Fleet

FD004 DATASET

18

Failure Risk

FAILURE_SOON=1

67.4

Mean RUL [cycles]

MEAN PREDICTED

$1.2M

Preventive Cost

vs $8.4M FAILURE

RUL DISTRIBUTION BY REGIME

HI — BY REGIME

FLEET STATUS — CLICK ANY ROW TO INSPECT IN XAI

STATUS REGIME RUL ≤ RISK ≥% — engines

↗ CLICK ROW TO OPEN XAI ANALYSIS FOR THAT ENGINE

ENGINE ID	REGIME	LAST CYCLE	PRED. RUL	FAILURE PROB	HI	STATUS

ENGINE

REGIME

SENSOR

WINDOW 20

RAW + ROLLING MEAN

ROLLING STD (VARIABILITY)

SENSOR vs RUL

LIFECYCLE PHASES

REGIME

ENGINE

RUL CLIP 125

PREDICTED vs ACTUAL RUL

RISK vs RUL (MEAN FAILURE PROB)

12.3

MAE [cycles]

18.7

RMSE

2847

NASA Score

HEALTH INDEX DEGRADATION — REGIME + ENGINE

REGIME

INDEX 35

MODEL

SHAP GLOBAL — TOP FEATURES

SHAP LOCAL — WATERFALL (INDEX 35)

COUNTERFACTUAL — SENSOR RECOMMENDATIONS

CORRELATION: SENSOR vs RUL (PER REGIME)

ALARM THRESHOLD 70%

FAILURE COST [$K] 100K

SERVICE COST [$K] 10K

0

Engines Analyzed

0

Failures Avoided

0

Preventive Maint.

$0

Total Cost

POLICY CURVE — THRESHOLD vs TOTAL COST

ENGINE

REGIME

SPEED 300ms

—

Current Cycle

—

Pred. RUL

—

Failure Risk

PREDICTED RUL — STREAMING

FAILURE PROBABILITY — STREAMING

SERVICE DECISIONS — LAST 20 CYCLES

🏭

—

Overall Health Score

Fleet average · higher is better

⚠️

—

Machines Need Attention

out of 30 in fleet

💰

—

Savings This Month

vs. waiting for failures

📅

—

Days to Next Failure

predicted · if no action taken

🗺️ Fleet at a Glance — Click any machine to inspect

Machine —

—

Days of life left

—

Failure chance

—

Health score

Why is this happening?

✅ What To Do Today

💵 Money Impact — This Quarter

—

COST IF WE DO NOTHING

→

—

COST OF PLANNED MAINTENANCE

→

—

WE SAVE

—

Estimated savings this quarter by acting on AI recommendations

💡 How we calculate this: Each unplanned failure costs ~$100K in parts, labour and lost production. Planned maintenance costs ~$10K. The AI predicts which machines will fail and when — so you can act early.

📈 Fleet Health Trend — Last 8 Weeks

🔍 Top Degradation Causes

🤖 What the AI Found This Week

All predictions are generated by a machine learning model trained on NASA C-MAPSS FD004 data (249 engines, 6 operating conditions). Technical details available in other tabs. → Open Technical View