SENTINEL-X

SENTINEL X: AI-powered Industrial Decision Intelligence Platform for Safety, Operations, and Predictive Risk Management.

Comment

Sentinel X Logo

🏭 Sentinel X — Industrial Safety AI Operating System

Autonomous IT/OT Convergence, Real-Time Compound Risk Intelligence, Multi-Agent AI Debate & Worker Biometrics for High-Risk Industrial Plants

Live Demo Demo Video MP4 Demo GIF API Docs ET AI Hackathon 2.0 License

Sentinel X Demo Showcase Video

📌 Executive Summary

Modern industrial facilities—petrochemical refineries, steel mills, and chemical processing units—operate complex machinery under hazardous conditions. Today, industrial safety systems are fragmented into isolated silos: SCADA handles process telemetry, physical permits are filed separately, gas detectors alert locally, and worker tracking is manual.

When multiple subtle anomalies happen simultaneously (e.g., a low-level toxic gas weep during confined space maintenance while an adjacent compressor exhibits abnormal vibration), traditional systems fail to correlate the compound hazard, leading to catastrophic industrial disasters.

Sentinel X bridges this gap by acting as an autonomous AI Operating System for Industrial Safety. It normalizes industrial IoT data into a Unified Namespace (UNS), continuously correlates multi-vector hazards through a 20-Rule Compound Risk Engine, streams real-time consensus via a 7-Agent AI Debate Engine, tracks worker physiological strain via smart biometrics, models atmospheric gas plume dispersion, and validates Permit-to-Work (PTW) compliance against live SCADA telemetry.

🚀 Key Innovation Highlights

🤖 1. Multi-Agent AI Debate Engine

When a high-risk safety incident occurs, Sentinel X triggers an autonomous debate between 6 specialized domain AI agents + 1 Executive Decision Agent:

  • 🔴 Safety Agent: Prioritizes life protection and mandatory evacuation thresholds.
  • 🟡 Production Agent: Evaluates operational downtime costs and partial isolation alternatives.
  • ⚖️ Compliance Agent: Enforces strict OISD & Factories Act statutory mandates.
  • 🔧 Maintenance Agent: Diagnoses root-cause mechanical failure signatures and repair timelines.
  • 💰 Finance Agent: Calculates ROI of proactive shutdown vs. catastrophic liability.
  • 🚨 Emergency Agent: Mobilizes response teams, exclusion zones, and protocol activation.
  • 📊 Executive Agent: Synthesizes all perspectives and renders an authoritative, decisive operational verdict.

⚡ 2. 20-Rule Compound Risk Correlation Engine

Goes beyond single-sensor threshold alarms by detecting multi-variable hazard interactions:

  • H₂S Buildup + Active Confined Space Permit
  • Hot Work Permit + Downwind Adjacent LEL Flammable Gas Spike
  • Lockout/Tagout (LOTO) Verification Failure + Active Maintenance
  • Worker Fatigue (>12h Shift) + High-Hazard Overcrowded Zone
  • Vibration Anomaly + Overdue Mechanical Calibration Cascade

🩺 3. Worker Wearable Biometric Intelligence

  • Computes Moran et al. Physiological Strain Index (PSI) (0–10 scale) using real-time heart rate and core/skin body temperature telemetry.
  • Evaluates WBGT Heat Index and calculates cumulative OSHA TWA/STEL gas dosimetry for H₂S and CO exposure per worker.

💨 4. Gaussian Hazard Plume Dispersion Modeling

  • Implements atmospheric Pasquill-Gifford dispersion models using Briggs rural coefficients.
  • Dynamically projects ERPG-2/3 & OISD hazard boundaries on SVG plant floor overlays, adjusting in real-time to wind direction, release rate, and stability classes.

🚶 5. Real-Time Location System (RTLS) & Exclusion Geofencing

  • Ultra-Wideband (UWB) worker coordinate tracking on interactive plant floor maps.
  • Automated 2-second evaluation loops checking confined space permit access and active machinery exclusion zones.

📋 6. SCADA-Integrated Digital PTW Validator

  • Calculates a 3-vector Vulnerability Index (VI): $\text{VI} = \text{Location Score} \times \text{Job Criticality} \times \text{Human Strain}$.
  • Validates isolation status and verifies training records against OISD-105, OISD-116, OISD-118, and Factories Act 1948 requirements.

🔑 7. Multi-Provider LLM & Custom Key Settings Interface

  • Built-in provider router supporting Google Gemini 2.0, OpenRouter, Anthropic Claude 3.5, and Local Ollama models.
  • Includes a dedicated Settings Page (/settings) allowing operators to configure custom API keys and test connections at runtime.

🏗️ System Architecture

flowchart TB
    subgraph OT_Layer["Industrial OT & Sensor Layer"]
        Sensors["📡 Gas, Temp, Press, Vib Sensors"]
        PLC["⚡ Modbus / OPC-UA / PLCs"]
        Biometrics["🩺 Worker Smart Vests (HR, Temp, Gas)"]
        RTLS["🚶 UWB Location Tags"]
        CCTV["📹 RTSP / HLS Camera Feeds"]
    end

    subgraph UNS_Layer["IT/OT Integration & UNS"]
        Normalizer["Unified Namespace (UNS) Normalizer"]
        MQTT["Mosquitto MQTT Broker (Sparkplug B)"]
    end

    subgraph Intelligence_Engine["Sentinel X Core AI Engine"]
        RiskEngine["⚡ 20-Rule Compound Risk Engine"]
        DebateEngine["🤖 7-Agent AI Debate Orchestrator"]
        PlumeEngine["💨 Gaussian Plume Dispersion Model"]
        PTWEngine["📋 Digital PTW & SCADA Validator"]
        LLMRouter["🧠 Multi-Provider LLM Router (Gemini/Claude/Ollama)"]
    end

    subgraph Presentation_Layer["User Interface & Emergency Response"]
        WebUI["💻 Next.js 14 Bento Dashboard (Vercel)"]
        FastAPI["🚀 FastAPI Backend Server (Render)"]
        Alerts["🚨 Emergency Alarms & WebSocket Feed"]
    end

    Sensors --> Normalizer
    PLC --> Normalizer
    Biometrics --> Normalizer
    RTLS --> Normalizer
    CCTV --> FastAPI

    Normalizer --> MQTT
    MQTT --> RiskEngine
    RiskEngine --> DebateEngine
    DebateEngine --> LLMRouter
    PlumeEngine --> WebUI
    PTWEngine --> WebUI

    FastAPI <--> WebUI
    FastAPI <--> Alerts

📱 Application Modules & Pages (17 Routes)

Page Route Module Name Core Features
/ Mission Control Live plant operations feed, real-time risk ring, sensor ticker, compound risk alerts.
/debate AI Debate Chamber Streaming multi-agent safety debate reasoning with interactive scenario controls.
/sensors Sensors Console Matrix of 10+ real-time plant sensors with sparklines and historical charts.
/cameras CCTV Vision Computer vision AI feed overlay with object detection and PPE monitoring.
/permits Digital PTW Automated permit verification engine with Vulnerability Index scoring.
/biometrics Worker Biometrics Fleet physiological strain index (PSI) gauges, heart rate, and gas dosimetry.
/dispersion Hazard Plume Real-time atmospheric toxic plume model with ERPG contour map overlays.
/workers RTLS Tracking UWB plant floor positioning map, geofence breaches, and machinery exclusion zones.
/compliance Regulatory Hub Real-time statutory compliance audit against OISD and Factories Act rules.
/simulator Scenario Simulator What-if risk simulation matrix for testing compound hazard escalation.
/executive Executive Briefing AI Copilot assistant and executive shift report generator.
/knowledge RAG Knowledge Base Vector semantic search over OISD regulations, SOPs, and incident archives.
/sites Multi-Site Overview Geographical map overview across multiple industrial refinery locations.
/connect Connectivity Manager Protocol connector settings for MQTT, OPC-UA, Modbus, and RTSP.
/handover Shift Handover Log Digital shift change log with pending permit tracking and AI summary.
/settings Settings & API Keys Custom LLM API Key manager (Gemini, OpenRouter, Claude, Ollama) with live test suite.

🛠️ Technology Stack

  • Frontend: Next.js 14 (App Router), TypeScript, React 18, Zustand, Recharts, Lucide Icons, Vanilla CSS (Claymorphism & Spatial Bento UI).
  • Backend: FastAPI, Python 3.11, Asyncio, Pydantic v2, Uvicorn, Starlette.
  • AI / LLM & ML: Google Gemini API, OpenRouter API, Anthropic API, Ollama, PyTorch, OpenCV, ChromaDB (Vector Search).
  • Industrial IoT Protocols: MQTT (Sparkplug B), OPC-UA, Modbus TCP, RTSP / HLS Video Streaming.
  • Infrastructure & Deployment: Docker, Docker Compose, Vercel (Frontend), Render (Backend).

⚖️ Supported Industrial Standards

  • OISD-105: Work Permit System & Gas Clearance Limits
  • OISD-116: Fire Protection Facilities for Petroleum Refineries
  • OISD-118: Inspection of Heavy Rotating Equipment
  • OISD-GDN-206: Safety Management System in Petroleum Industry
  • Factories Act 1948: Sections 36 (Confined Space), 37 (Explosive Gas), 41F (Hazardous Processes)
  • OSHA 1910.119: Process Safety Management (PSM) of Highly Hazardous Chemicals

⚡ Quickstart & Local Development

Prerequisites

  • Node.js 18+ & npm
  • Python 3.11+
  • Git

1. Clone Repository

git clone https://github.com/brovk2008/SENTINEL-X.git
cd SENTINEL-X

2. Frontend Setup

cd frontend
npm install
npm run dev

Open http://localhost:3000 in your browser.

3. Backend Setup

cd backend
python -m venv venv
# On Windows:
.\venv\Scripts\activate
# On Linux/macOS:
source venv/bin/activate

pip install -r requirements.txt
uvicorn main:app --reload --port 8000

API server will run at http://localhost:8000. Access Swagger UI at http://localhost:8000/docs.

4. Docker Compose Setup (Optional)

Run the entire stack including Postgres, Redis, Mosquitto MQTT, and API via Docker:

docker-compose up -d

🌐 Production Deployment

📄 License

Distributed under the MIT License. See LICENSE for details.

Built for the ET AI Hackathon 2.0 | Sentinel X — The factory has a brain.

Built With

Share this project:

Updates