e20-co326-MQTT-Data-Anomaly-Detection

Industrial Digital Twin and Cyber-Physical Security

Software-only edge-to-cloud Industrial IoT integration project for CO326 Computer Systems Engineering - Industrial Networks.

Team

Table of Contents

  1. Introduction
  2. Project Scope
  3. System Architecture
  4. Technology Stack
  5. How the Pipeline Works
  6. Main Topics and Services
  7. Project Links

Introduction

This project implements a simulated industrial digital twin for Motor-01 overheating monitoring with relay trip simulation. The system models one main process variable, Motor/WindingTemperature, and one actuator, a protective relay that can be controlled in AUTO, TRIP, and RESET modes.

The objective is to demonstrate a realistic Industrial IoT pipeline using a software-only setup that still follows the required edge-to-cloud architecture:

Project Scope

The project is centered on a single industrial use case:

This is a software simulation, not a hardware deployment. The edge device behavior and lightweight ML runtime are designed to represent an ESP32-class constrained environment using container resource limits and simple inference logic.

System Architecture

The pipeline consists of these major components:

  1. publisher Generates simulated motor temperature data and relay feedback state.
  2. mqtt Carries Sparkplug-style telemetry and relay command messages.
  3. node-red Bridges Sparkplug device data into a Unified Namespace and applies orchestration logic.
  4. ai-brain Runs a lightweight anomaly-scoring model using statistical thresholding.
  5. backend Consumes telemetry, calls the ML service, estimates Remaining Useful Life (RUL), writes historian records, and exposes APIs for the dashboard.
  6. influxdb Stores time-series telemetry and derived analytics.
  7. grafana Visualizes temperature, anomaly score, RUL, and other historian data.
  8. frontend Shows live digital twin state and allows relay commands.

For the detailed architecture diagrams and data-flow sequence, see docs/ARCHITECTURE.md.

Technology Stack

How the Pipeline Works

  1. The simulated publisher generates Motor-01 temperature and relay-state telemetry.
  2. Telemetry is published on a Sparkplug-style MQTT topic.
  3. Node-RED transforms and republishes the data into a Unified Namespace topic.
  4. The backend subscribes to the UNS telemetry stream.
  5. The backend sends motorTemperatureC to the lightweight ML service for anomaly scoring.
  6. The backend estimates RUL from temperature trend, anomaly score, and relay state.
  7. Processed records are stored in InfluxDB.
  8. Grafana reads historian data for trend visualization.
  9. The frontend shows live state and sends operator relay commands back through the backend and MQTT.

Main Topics and Services

MQTT Topics

Local Service Ports