GitHub Repo for the Project

https://github.com/HajjSalad/Embedded-Predictive-Maintenance

Overview

• A multi-node embedded predictive maintenance system that monitors industrial machine sensors, detects anomalies in real-time, and streams telemetry to the cloud via a gateway node.

• Built on three sensor nodes - STM32F446RE, nRF52840, and ESP32-WROOM-32 - running identical Zephyr RTOS firmware, each communicating with an ESP32 gateway over UART (STM32 + ESP32 nodes) or BLE (nRF52 node).

  [Image here]

The system is built across these components:

• Zephyr Setup Guide: development environment and toolchain configuration
• Machine and Sensor Modelling: C++ OOP with Factory Method and Object Pool patterns
• Multithreaded Pipeline: five-thread producer-consumer architecture with priority scheduling
• Synchronization Primitives: mutex, message queue, and semaphore-based coordination
• Anomaly Detection: statistical threshold detection with event-driven alerting
• UART Transport: binary protocol with CRC16 for STM32 and ESP32 sensor nodes
• BLE Transport: GATT-based wireless telemetry for nRF52 sensor node
• CI/CD Pipeline: automated multi-board builds with hardware-in-the-loop testing

Components

1. Main Entry Point

File: main.c

• Entry point of the whole system
• One-time initialization:
  • Run C++ demo
  • Create machine and register sensors
  • Spawn application threads