← Back to portfolio

Industrial AI: APM & Reliability (Remaining Useful Life)

Config-driven data science pipeline for high-fidelity industrial domains.

Industrial AI: APM & Reliability (Remaining Useful Life) Public dataset

At a glance

Task: regression  |  Headline: r2 = 0.9207

rmse
414.9999
mae
275.7019
median_ae
172.4507
mape
23.2257
r2
0.9207
adj_r2
0.9176
explained_variance
0.9289
mean_residual
-133.1397

Summary

Project

Config-driven data science pipeline for high-fidelity industrial domains.

What was investigated

The engagement modelled the target 'rul_hours' as a regression problem in the Industrial AI: APM & Reliability (Remaining Useful Life) domain, using 2000 records across 10 columns (data source: Synthetic sandbox). The pipeline audited data quality, engineered features, split the data honestly into train/test, and compared several models by cross-validation.

Outcome

The selected model was ExtraTrees, with r2 = 0.9207 on the held-out test set. Supporting metrics: rmse=415.000, mae=275.702, median_ae=172.451, mape=23.226, adj_r2=0.918, explained_variance=0.929, mean_residual=-133.140. Full results, the deployment gates and the audit trail are in the run's report.

Figures

02_distributions.png
02 distributions
03_correlation.png
03 correlation
04_vif.png
04 vif
05_target.png
05 target
10_pred_vs_actual.png
10 pred vs actual
11_residuals.png
11 residuals
12_leaderboard.png
12 leaderboard
13_feature_importance.png
13 feature importance
20_degradation_envelope.png
20 degradation envelope
21_weibull_reliability.png
21 weibull reliability
22_rul_vs_health.png
22 rul vs health

Artifacts

Data source: Synthetic sandbox

Download report (.docx)

Audit SHA-256: 7b501e38a129f1335385ada6f9efcd14dc1f37a4aca98490becbb4ad9c10da88