Plant Asset Health Index and Remaining Useful Life (RUL) Predictor
Comprehensive asset health management programs require combining multi-sensor vibration, temperature, and maintenance work orders to accurately forecast Remaining Useful Life (RUL). You will build an asset health index engine using PyTorch or LightGBM to process multi-sensor telemetry and forecast asset RUL. The model must achieve a Root Mean Squared Error (RMSE) below 15 operating hours on RUL predictions.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop an end-to-end asset health score and RUL prediction engine using LightGBM and PyTorch to prevent unplanned catastrophic equipment failures.
How work is evaluated
Evaluates prediction accuracy (RMSE) of forecasted Remaining Useful Life (RUL) in operating hours.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
health_index_bounds_check
Verifies Asset Health Index strictly remains within 0.0 to 100.0 score bounds.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
rul_rmse_hours
Root Mean Squared Error of predicted RUL vs ground truth • target: 12.4 • range: 0-50
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Extract temporal features from raw multi-axis vibration RMS and bearing temperature streams
Train LightGBM / PyTorch models predicting time-to-failure (Remaining Useful Life)
Formulate standardized 0-100 Plant Asset Health Index (AHI) metrics incorporating operating history
Evaluate RUL prediction error across run-to-failure degradation datasets (e.g., NASA C-MAPSS style)
Reference links and supporting material
Run-to-failure multi-sensor vibration and temperature degradation profiles for industrial bearings and gearboxes.
Challenge setup
Evaluates prediction accuracy (RMSE) of forecasted Remaining Useful Life (RUL) in operating hours.
Challenge input
JSON containing vibration_rms_window array, peak_temp_c_window array, and total_operating_hours
LightGBM
Standard high-performance ML library for tabular prognotics.
PyTorch
Standard deep learning framework.
Evaluated output
JSON containing asset_health_index_0_100, predicted_rul_hours, and failure_risk_category ("LOW" / "MEDIUM" / "HIGH")
- Verifies Asset Health Index strictly remains within 0.0 to 100.0 score bounds.
- Root Mean Squared Error of predicted RUL vs ground truth • target: 12.4 • range: 0-50
- Benchmark: Industrial Asset RUL Prognotics Benchmark
- Rul Rmse Hours target: 12.4
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- PyTorch Lightning · Required
- Captum (PyTorch) · Optional
Evaluation contract
- health_index_bounds_check · Weight 1
- rul_rmse_hours · Weight 1
Metrics
- rul_rmse_hours · Target 12.4
Benchmarks
- Industrial Asset RUL Prognotics Benchmark
- robotics
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing vibration_rms_window array, peak_temp_c_window array, and total_operating_hours
Output: JSON containing asset_health_index_0_100, predicted_rul_hours, and failure_risk_category ("LOW" / "MEDIUM" / "HIGH")
Execution harness: Python
Evaluation tools
- LightGBM · Gradient Boosted RUL Model
- PyTorch · Neural Feature Engine
Recipe freeze marker
No recipe freeze time is recorded.