| --- |
| title: SteelAI Module2 EAF Intelligence Explorer |
| emoji: 🚀 |
| colorFrom: red |
| colorTo: green |
| sdk: docker |
| app_port: 8501 |
| tags: |
| - streamlit |
| pinned: false |
| short_description: Part of the TenderMatcher.Tech AI & Digital Intelligence |
| license: other |
| persistent_storage: 1Gi |
| --- |
| |
|
|
| # SteelAI Module #2 — Blast Furnace & EAF Data Intelligence Explorer |
|
|
| **Part of the TenderMatcher.Tech AI & Digital Intelligence for Metallurgy Suite** |
|
|
| Ready-to-deploy **Streamlit + SHAP application** for **energy and yield optimization** in |
| **Blast Furnace** and **Electric Arc Furnace (EAF)** operations. |
|
|
| --- |
|
|
| ## Objective |
|
|
| Predict and optimize key furnace variables such as: |
|
|
| - `furnace_temp` |
| - `tap_temp` |
| - `offgas_co`, `offgas_co2`, `o2_probe_pct` |
| - `arc_power`, `energy_efficiency`, `yield_ratio` |
|
|
| This module simulates a **complete furnace data intelligence environment** with ensemble modeling, SHAP explainability, and physics-informed feature engineering. |
|
|
| --- |
|
|
| ## Core Features |
|
|
| - Synthetic EAF dataset generator (3,000+ records) |
| - Derived physical proxies: |
| - `carbon_proxy`, `oxygen_utilization`, `slag_foaming_index` |
| - PCA and clustering for operating modes |
| - Ensemble regression (Linear, RF, GB, ExtraTrees) |
| - SHAP explainability for model transparency |
| - Business framing & annotated bibliography for metallurgy ML |
| - Fully local synthetic data generation (no external upload needed) |
|
|
| --- |
|
|
| ## Use Case Alignment (SteelAI Framework) |
|
|
| | # | Use Case | Alignment | Description | |
| |---|-----------|------------|--------------| |
| | **2** | Blast Furnace / EAF Data Intelligence | **Primary (100%)** | Furnace temperature, gas chemistry, and power-density modeling for yield & energy optimization | |
| | 7 | AI-Driven Alloy Design Tool | Partial | Shares compositional features (`chemical_C`, `chemical_Mn`, etc.) | |
| | 8 | Predictive Maintenance Framework | Partial | Includes rolling, lag, and vibration signals for maintenance AI | |
|
|
| --- |
|
|
| ## Example Targets |
|
|
| - Predict **`furnace_temp`** from operational data |
| - Analyze **feature importance** with SHAP plots |
| - Quantify business value: |
| - 5–8% yield improvement |
| - 3–5% energy cost reduction per ton |
| |
| --- |
| |
| ## How to Run Locally |
| |
| ```bash |
| pip install -r requirements.txt |
| streamlit run app.py |
| ``` |
| |
| Then open the local URL shown in your terminal (typically http://localhost:8501). |
| |
| --- |
| |
| ## Deploy on Hugging Face |
| |
| 1. Create a Space → choose **SDK: Streamlit** |
| 2. Upload: |
| - `app.py` |
| - `requirements.txt` |
| - `README.md` |
| 3. Hugging Face automatically installs dependencies and builds your Space. |
| |
| Your app will be live at: |
| |
| ``` |
| https://huggingface.co/spaces/singhn9/SteelAI_Module2_EAF_Intelligence_Explorer |
| ``` |
| |
| --- |
| |
| ## Business Value Snapshot |
| |
| | Dimension | Improvement | Impact | |
| |------------|-------------|---------| |
| | Productivity | +8–15% throughput | Higher process stability | |
| | Energy efficiency | 3–8% reduction | Lower cost per ton | |
| | Quality control | 10–15% better rejection precision | Fewer off-grade batches | |
| | R&D cycle | 25–35% faster property correlation | Shorter design-to-validation loop | |
| | Predictive reliability | 7–10% OEE gain | Reduced downtime | |
| |
| --- |
| |
| ## Generative AI & Industrial Innovation |
| |
| Beyond process modeling, **TenderMatcher.Tech** extends AI into Generative domains: |
| |
| - **Knowledge-Graph Assistant** — links research papers, alloy data & insights |
| - **Chat-based Technical Advisor** — LLM-powered metallurgical Q&A |
| - **Generative Report Builder** — auto-creates lab summaries & dashboards |
| |
| --- |
| |
| ## AI–ServiceNow Integration Mapping |
| |
| | # | AI Module | ServiceNow Integration | Business Value | |
| |---|------------|------------------------|----------------| |
| | 1 | Steel Property Prediction | QMS, Predictive Intelligence | QA traceability & ISO/BIS compliance | |
| | 2 | Blast Furnace Intelligence | OT IntegrationHub, EHS | Closed-loop efficiency alerts | |
| | 3 | Microstructure Classifier | KM, AI Search | Metallography knowledge base | |
| | 4 | Surface Defect Detection | FSM, Predictive Intelligence | Real-time defect case auto-routing | |
| | 5 | Corrosion/Fatigue Prediction | ORM, Asset Mgmt | Predictive asset health | |
| | 6 | Energy Optimizer | Sustainability Mgmt | ESG-linked savings reporting | |
| | 7 | Alloy Design Tool | Innovation Mgmt, KM | R&D portfolio tracking | |
| | 8 | Predictive Maintenance | AIOps, FSM, CMDB | 10–12% downtime reduction | |
| |
| --- |
| |
| ## About Naval Singh |
| |
| **Naval Singh** — Digital Transformation Advisor |
| Specializing in AI, analytics, and industrial systems. |
| Focused on production-grade decision systems for metallurgy, mining & manufacturing. |
| |
| **singhn9@gmail.com** |
| [LinkedIn](https://linkedin.com/in/navalsingh9) |
| **Rourkela, India** |
| |
| --- |
| |
| ## Why Rourkela Matters |
| |
| Rourkela — India’s steel and metallurgy hub — provides the perfect ecosystem for AI pilot collaborations among academia, consulting, and industry. |
| |
| --- |
| |
| © **TenderMatcher.Tech** — *AI & Digital Intelligence for Metallurgy* |
| **[Read More](https://tendermatcher.tech/ai-metallurgy)** |
| |
| # SteelAI — EAF Intelligence Explorer (MODEX) |
| |
| An industrial-grade AI demonstrator for Electric Arc Furnace (EAF) analytics at |
| **Steel Authority of India Limited (MODEX)** — combining synthetic metallurgical datasets, |
| automated feature engineering, and AutoML ensembles with SHAP explainability. |
| |
| ### Key Highlights |
| - Generates full synthetic EAF datasets (~3000 rows × 200+ features) |
| - Supports ensemble AutoML across RandomForest, XGBoost, LightGBM, CatBoost, etc. |
| - Performs Optuna-based hyperparameter tuning per family |
| - Uses meta-stacking (Ridge) and SHAP explainability |
| - Includes “Recommended Target Variables” and Business Impact framing |
| - Features annotated bibliography with direct research paper links |
| |
| ### Logging & Reproducibility |
| - All generated CSVs, JSONs, and logs are stored under `./logs/` |
| - Each session in Hugging Face ephemeral environment appends new timestamps |
| - Users can download artifacts directly via the **Download Saved Files** tab |
| |
| *Ephemeral note*: Data and models are cleared when the Space rebuilds, |
| but logs persist for the current runtime session. |
| |
| --- |
| |
| ### Credits |
| Developed as part of SteelAI MODEX initiative for AI-driven metallurgy R&D. |
