Real-Time Pressure Monitoring System - iMining (Pty) Ltd
Production pressure monitoring system with 12 live measurement points. Replaced 4-hour manual checks with continuous real-time facility monitoring and automated anomaly detection. Deployed in a harsh industrial environment - 99.2% uptime over 6 months of operation.
12
Measurement points
99.2% over 6 months of continuous operation
Uptime
<5 seconds from out-of-spec reading to notification
Alert latency
Overview
Designed and deployed a complete real-time pressure monitoring system for iMining (Pty) Ltd. The system monitors 12 pressure sensors (4-20mA industrial-grade), processes data on a custom STM32 microcontroller board, and streams live readings to a browser-based dashboard via WebSockets. Automated alerts notify operators within 5 seconds of any out-of-spec reading. The system replaced manual 4-hour check cycles, providing continuous visibility into facility conditions and catching 3 anomalies that would have gone undetected under the previous regime.
The Challenge
Problem
Manual pressure checks every 4 hours created significant blind spots between readings. Facility anomalies could develop, escalate, and cause damage in the window between checks. There was no automated alerting - detection depended entirely on a technician being present at the right moment.
Environment
Harsh industrial facility with elevated dust levels, high ambient temperatures, and vibration from active equipment. Hardware needed to be ruggedized and firmware needed to be deterministic - a missed reading or software hang was not acceptable.
Goal
Continuous, real-time visibility into facility pressure conditions across 12 measurement points, with automated alerts for out-of-spec readings delivered within 5 seconds - eliminating the 4-hour blind spot entirely.
System Architecture
Sensors
12× 4-20mA industrial-grade pressure sensors - selected for operating range, ingress protection (IP65+), and linearity across the facility's expected pressure bands
Microcontroller
STM32 - chosen over ESP32 for deterministic real-time performance. Interrupt-driven ADC sampling ensures consistent sub-100ms acquisition latency regardless of system load
Custom PCB
Designed in Autodesk Eagle - includes precision current-to-voltage conversion circuitry, filtering for industrial noise rejection on 4-20mA lines, and STM32 integration headers
Communication
WebSocket-based real-time data streaming over local network - enables live dashboard updates without polling latency
Dashboard
Browser-based React frontend with real-time gauge displays, historical trend graphs, and configurable alert thresholds per sensor
Alerting
Firmware-level threshold detection with <5-second end-to-end latency from out-of-spec reading to operator notification
Technical Stack
- ›STM32 microcontroller (interrupt-driven architecture)
- ›12× 4-20mA industrial-grade pressure sensors
- ›Custom PCB - Autodesk Eagle (signal conditioning, noise filtering)
- ›Industrial enclosure - IP-rated, vibration-resistant mounting
- ›Power supply with industrial voltage regulation
- ›C firmware - interrupt-driven ADC sampling
- ›Real-time pressure data acquisition (<100ms latency)
- ›Threshold monitoring with immediate alert triggering
- ›WebSocket communication stack
- ›Non-volatile data buffering for network interruption tolerance
- ›React-based real-time monitoring dashboard
- ›WebSocket data ingestion and live display
- ›Historical data logging and trend visualisation
- ›Configurable alert thresholds per sensor channel
- ›Role-based dashboard access
Results
| Metric | Before | After |
|---|---|---|
| Detection speed | 4-hour manual check cycles | Real-time - continuous monitoring, instant detection |
| Alert latency | Hours (dependent on manual inspection schedule) | <5 seconds from out-of-spec reading to notification |
| System uptime | N/A - manual process, no uptime metric | 99.2% over 6 months of continuous operation |
| Anomalies detected | Would have been missed between manual checks | 3 anomalies caught and acted on before escalation |
| Alert accuracy | Human-dependent - subject to fatigue and gaps | Zero false positives in 6 months of operation |
| Measurement accuracy | Manual gauge reading - operator dependent | ±0.5% across all 12 channels |
Technical Insights & Lessons
STM32 was the right choice over ESP32 for this application. The deterministic interrupt-driven architecture of the STM32 guaranteed consistent ADC sampling intervals - critical when you need sub-100ms acquisition latency in a production environment where timing variation could mean a missed event.
4-20mA signal conditioning on the custom PCB required more careful noise filtering than anticipated. Industrial environments introduce significant electromagnetic interference on long sensor cable runs. Adding proper current-loop filtering on the PCB eliminated spurious readings that appeared during early bench testing.
WebSocket latency under concurrent browser connections required firmware-level buffering optimisation. The original implementation caused alert delays when multiple operators viewed the dashboard simultaneously. Restructuring the firmware transmit queue resolved this and confirmed <5-second end-to-end alert latency under load.
Project Photos
"Nduvho built exactly what we needed - a reliable monitoring system that works in our harsh environment. The system caught a pressure anomaly that our manual checks would have missed. Professional, technically solid, and delivered on time."
Ready to Add Real-Time Monitoring to Your Facility?
Tell us about your facility, environment, and what you need to measure. We'll scope a system that works - sensors, firmware, dashboard, and deployment.