ACTIVE RESEARCH · IEEE IES · OPEN BENCHMARKS

Industrial AI Research Lab IEEE IES Initiative

Open benchmarking initiative for industrial AI, aligned with IEEE Industrial Electronics Society publication tracks — TIE, TII, JESTPE, and IECON. Reproducible baselines. Real datasets. Honest numbers.

Explore Projects → View on GitHub
4
Repositories
15+
Model Impls
10+
Datasets
benchmark_runner.py
$ python benchmarks/run_benchmarks.py --dataset FD001 --all-models
# NASA CMAPSS FD001 · seed=42 · window=30
Train: (17631, 30, 14) Test: (100, 30, 14)

Running LSTM ...
RMSE: 13.42 · NASA↓: 214.3
Running Transformer ...
RMSE: 12.89 · NASA↓: 198.7
Running TCN ...
RMSE: 13.15 · NASA↓: 207.1

✓ Results → benchmarks/results/results_FD001.json
$
// 01 — REPOSITORIES

Four Research Repositories

Each repository targets a specific IEEE IES publication track with real implementations, runnable benchmarks, and honest results.

🔧
TIE / TII

Industrial Predictive Maintenance

RUL prediction and fault detection for turbofan engines and rolling element bearings. End-to-end pipeline from raw sensor data to deployment-ready inference.

CMAPSSCWRUIMS PaderbornLSTMTransformerTCN
13.42
RMSE↓
198.7
NASA↓
4
Models
View Repository →
Active
📈
TII

Industrial Time-Series AI

SOTA time-series models for industrial sensor streams — multivariate forecasting, anomaly detection, and feature engineering for ICS/SCADA systems.

ETTh1/h2SWaTPSM PatchTSTDLinearPoint-Adjust
0.9999
ROC-AUC
1.000
F1-PA
5
Models
View Repository →
Active
⚡
JESTPE / TIE

AI Power Electronics Diagnostics

Fault detection in 3-phase VSI inverters and PMSM motor drives. Physics-informed synthetic data — runs out of the box, no download needed.

InverterIGBT FaultMotor Drive 1D CNNBiLSTMSTFT
77.9%
Acc
0.777
Macro F1
5
Models
View Repository →
Active
🏭
TII

Smart Manufacturing AI

Vision-based defect detection and quality inspection for industrial manufacturing. Focused on anomaly detection benchmarks aligned to IEEE IES TII standards.

MVTec ADDefect Detection Visual InspectionAnomaly Detection
15
Categories
CNN
Backbone
🔧
In Progress
View Repository →
In Progress
// 02 — RESULTS

Benchmark Results

Reproducible results. Seed 42. Official train/test splits. Real numbers — no ranges, no estimates.

NASA CMAPSS FD001 · Window: 30 · Max RUL: 125 · Seed: 42 · Official NASA split
ModelRMSE ↓MAE ↓NASA Score ↓Params
Transformer12.899.71198.7412K
TCN13.1510.05207.1287K
LSTM + Attention13.4210.28214.3523K
Synthetic Inverter · 9 classes · Quick mode: 5 epochs, 50 samples/class · Seed: 42 · 70/15/15 split
ModelAccuracyMacro F1ParamsMode
BiLSTM + Attention77.94%0.7772.2Mquick 5ep
1D CNN Waveform73.53%0.6874.0Mquick 5ep
Spectrogram CNN36.76%0.31511.2Mquick 5ep
Transformer32.35%0.2510.8Mquick 5ep
Synthetic Motor Drive · 5 classes · Quick mode: 5 epochs, 50 samples/class · Seed: 42
ModelAccuracyMacro F1ParamsMode
BiLSTM + Attention63.16%0.5942.2Mquick 5ep
Spectrogram CNN50.00%0.41711.2Mquick 5ep
1D CNN Waveform44.74%0.4094.0Mquick 5ep
Transformer21.05%0.0700.8Mquick 5ep
Synthetic SWaT · 51 features · 3 epochs · Seed: 42 · Unsupervised
ModelROC-AUCF1F1-PAParams
LSTM Autoencoder0.99990.99811.0000108K
// 03 — DATA

Benchmark Datasets

10+ industrial datasets across all four repositories. Access requirements and known gotchas documented for each.

NASA CMAPSS
Free
Turbofan engine run-to-failure. FD001–FD004. Auto-downloads via loader.
Sensors: 21Cycles: ~70K
⚠ CMAPSS ≠ N-CMAPSS — different datasets.
N-CMAPSS (2021)
Free
Updated NASA turbofan used in recent TII papers. Roadmap item.
Year: 2021Sensors: 18
⚠ Most post-2021 papers use this version.
CWRU Bearing
Free
Case Western Reserve bearing vibration. Four fault types and sizes.
Loads: 0–3 HPFaults: 4 types
⚠ Specify motor load when comparing baselines.
Paderborn Bearing
Free
32 operating conditions. Artificial and real damage. Registration required.
Conditions: 32Size: ~32 GB
⚠ Use official Paderborn split protocol.
ETTh1 / ETTm1
Free
Electricity transformer temperature. Standard forecasting benchmark.
Features: 7Variants: 4
SWaT / WADI
Request
Water treatment ICS datasets. Formal access request to iTrust Singapore.
SWaT: 51 featWADI: 123 feat
⚠ Apply at itrust.sutd.edu.sg
PSM / SMAP / MSL
Free
Server metrics and NASA telemetry anomaly detection benchmarks.
PSM: 25 feat
MVTec AD
License
15 industrial categories for visual defect detection. Non-commercial only.
Categories: 15Images: ~5K
⚠ Cannot be redistributed. mvtec.com only.
Synthetic Signals
Built-in
Physics-informed inverter and motor drive signals. No download needed.
Inverter: 9 faultsMotor: 5 faults
// 04 — PIPELINE

Shared Pipeline Architecture

All four repos follow the same end-to-end pipeline from raw industrial sensor data to deployment-ready inference.

STEP 01
📥
Data Ingestion
  • Dataset loaders
  • Auto-download scripts
  • Format normalization
  • Train/test splits
STEP 02
⚙️
Preprocessing
  • Signal cleaning
  • Min-max normalization
  • Sliding windows
  • STFT / FFT
STEP 03
🧠
Model Training
  • LSTM / Transformer / TCN
  • 1D CNN / BiLSTM
  • Autoencoder anomaly
  • Early stopping + ckpts
STEP 04
📊
Evaluation
  • RMSE / MAE / NASA score
  • F1 / ROC-AUC / F1-PA
  • CSV output + JSON
  • Markdown result tables
// 05 — TEAM

Authors

The people behind the IEEE IES Industrial AI Lab initiative.

IEEE Industrial Electronics Society
IEEE IES Volunteer Initiative

This lab is an independent volunteer initiative created to combine all industrial electronics AI projects into one open research lab. Built by IEEE E-JUST Student Branch active members who are volunteers in the IEEE Industrial Electronics Society. It is not an official IEEE IES chapter or branch — it is created independently in the spirit of IEEE's mission: advancing technology for humanity, aligned with IES publication tracks (TIE, TII, JESTPE, IECON).

🌍 IEEE IES Society Volunteers 🎓 IEEE E-JUST Student Branch 📖 Open Benchmarks · MIT License
Assem
Assem ElQersh
Main Author & Lead Developer
Ex Treasurer — IEEE CS E-JUST SBC Graduate CSE · E-JUST 2025 IEEE IES Volunteer

Lead developer of all four repositories. Designed the benchmark architecture, implemented core models, and established the IEEE IES alignment framework for this open research initiative.

GitHub LinkedIn ✉ IEEE Email
Abdelrahman
Abdelrahman Wael
Co-developer & Contributor
Chairman — IEEE EMBS E-JUST SBC Undergrad BME · E-JUST 2028 IEEE IES Volunteer

Co-developer and contributor. Chairman of the IEEE EMBS E-JUST Student Branch Chapter. Second-year undergraduate in Biomedical Engineering, bringing an interdisciplinary perspective to industrial AI research.

GitHub LinkedIn ✉ IEEE Email
// 06 — CITE

Citation

If you use any repository in your research, please cite using the BibTeX entry below or the CITATION.cff file in each repo.

BibTeX · IEEE IES Industrial AI Lab
@misc{ieee_ies_industrial_ai_lab_2026,
  author       = {El-Qersh, Assem and Wael, Abdelrahman},
  title        = {{IEEE IES Industrial AI Lab: Open Benchmark Initiative}},
  year         = {2026},
  organization = {IEEE Industrial Electronics Society},
  url          = {https://github.com/IEEE-IES-Industrial-AI-Lab},
  note         = {Volunteer initiative · IEEE E-JUST Student Branch}
}
// 07 — CONTRIBUTE

Contributing

All contributions that improve reproducibility and benchmark coverage are welcome. Open an issue before submitting a pull request.

📦
Dataset Loaders
Add loaders for N-CMAPSS (2021), FEMTO bearing, PHM 2012, or real SWaT/WADI after obtaining access.
🔬
Model Implementations
Implement models from recent IEEE TIE/TII papers. TimesNet, iTransformer, Mamba-based architectures welcome.
📊
Benchmark Results
Run full benchmarks and contribute honest, reproducible results. Seed, split, and library versions must be documented.