Back to Portfolio
Live in ProductionWater Infrastructure / Utilities
🔧

IoT Predictive Maintenance System — Water Pump Station Monitoring

An ESP32 vibration monitoring system deployed across 45 pump stations — using FFT and ML to predict bearing failures 3–4 weeks in advance, reducing emergency repair calls by 60% and slashing maintenance costs.

60%
Emergency Repairs↓
3–4 wks
Lead Time
$1.1M/yr
Cost Saved

The Problem

A water utility operating 45 pump stations was spending $1.8M/year on emergency pump repairs and emergency contractor call-outs. Pumps failed without warning, disrupting water supply to residential areas and triggering SLA breach penalties. Maintenance was calendar-based (every 6 months) regardless of actual condition — some pumps were overmaintained, others failed between maintenance cycles.

💡Our Solution

We installed ADXL355 high-resolution MEMS accelerometers on each pump's motor bearing housing, connected to ESP32 nodes. The ESP32 samples vibration at 3,200 Hz and performs on-device FFT, extracting 1×, 2×, 3×, and BPFO/BPFI bearing fault frequencies per ISO 10816. Data is published to InfluxDB every 5 minutes. A Python ML model (trained on 6 months of baseline data) calculates a bearing health score (0–100) per pump and projects failure probability for the next 30 days. When a pump reaches health score < 60, a work order is automatically created in ServiceNow CMMS with the predicted failure date and recommended action.

🔗System Architecture

ESP32 (FFT on-device) → MQTT → AWS IoT Core → InfluxDB → Python ML → React Dashboard + ServiceNow CMMS

Tech Stack

Hardware
  • ESP32 with MEMS accelerometer (ADXL355)
  • Current transformer (motor current)
  • Temperature (winding + bearing)
  • Vibration sensor (piezoelectric)
  • Industrial enclosures (ATEX-rated zones)
Communication
  • Wi-Fi / Ethernet (pump station network)
  • MQTT to AWS IoT Core
  • 4G where no site network available
Cloud
  • AWS IoT Core
  • InfluxDB (vibration + current history)
  • Python ML service (ISO 10816 FFT analysis)
  • ServiceNow CMMS integration (work orders)
Frontend
  • React pump health dashboard
  • FFT spectrum viewer per pump
  • Predicted failure timeline
  • CMMS work order tracking

Key Features

Vibration monitoring at 3,200Hz sample rate per pump
On-device FFT — bearing fault frequency extraction
ML bearing health score with 30-day failure prediction
Automated CMMS work order creation on predicted failure
Motor current monitoring (efficiency and overload detection)
Winding temperature trend analysis
ISO 10816 compliance vibration severity classification
Pump efficiency trend (flow vs. power)

Results Delivered

  • 60% reduction in emergency repair call-outs
  • Average failure prediction: 3–4 weeks advance notice
  • 45 pump stations fully instrumented
  • $1.1M maintenance cost reduction in first year
  • Water supply interruptions reduced from 18 to 3 in 12 months

Technologies

ESP32MEMS VibrationFFTPython MLMQTTReactCMMS

Who This Is For

Water utilities, oil & gas operators, HVAC companies, industrial pump operators

Need a similar solution?

We've built production IoT systems like this across 15+ countries. Let's talk about yours.

Let's Build Together

Got an IoT challenge?
We've shipped it.

Whether you need a fleet to track, a factory to monitor, or a farm to automate — our team has done it before and we'd love to build it with you. Typical response time: under 24 hours.

No upfront commitment99.9% uptime SLANDA on requestFixed-price options