Industrial Machine Monitor
Retrofit vibration and temperature monitoring for legacy CNC machines with an AI-powered anomaly dashboard.
What we delivered (snapshot)
- Retrofit sensor module (vibration + temperature) with edge preprocessing on ESP32-S3.
- MQTT-based firmware pipeline delivering pre-processed feature vectors to the cloud.
- Web dashboard with anomaly detection (Isolation Forest), trend charts, and maintenance alerts.
The Challenge
A precision engineering SME was experiencing unexpected CNC machine downtime costing several hundred pounds per hour. They needed a cost-effective way to monitor machine health across 12 machines without replacing existing equipment or integrating with proprietary machine controllers.
The solution had to retrofit onto existing machines (no drill-and-tap), stream vibration and temperature data, and flag anomalies automatically — reducing reactive maintenance and giving the maintenance team early warning.
Constraints
Timeline
12 weeks to a 3-machine pilot
Budget Band
£20k–£35k (firmware + AI + dashboard)
Technical Constraints
- No modification to machine controllers — sensor-only approach
- Wi-Fi connectivity (factory network access provided)
- Anomaly detection without labelled failure data
- Dashboard accessible on tablets from the factory floor
Our Approach
- 1
Fit check & constraints clarification
Site visit to review machines, confirmed mounting approach (neodymium mag base), and Wi-Fi coverage mapping.
- 2
Requirements & acceptance criteria for V1
Defined anomaly alert latency target (< 2 min), false positive rate acceptance, and dashboard access requirements.
- 3
Architecture & risk register
MQTT broker architecture, Isolation Forest baseline training approach, risk of Wi-Fi dead spots mitigated with fallback local storage.
- 4
Design & build
Firmware for ESP32-S3 (sensor sampling + MQTT), anomaly model training pipeline, and React dashboard with Recharts.
- 5
Integration & validation
3-machine pilot installation, 2-week baseline data collection, model training, and alert threshold tuning.
- 6
Handover package
Firmware repo, dashboard source, model training notebook, deployment runbook, and handover call.
What We Delivered
- BOM for sensor module (COTS components + ESP32-S3 DevKit)
- Wiring and mounting diagram
- Assembly notes for field installation
- ESP32-S3 firmware source (sensor sampling + feature extraction + MQTT)
- Flash guide and over-the-air update module
- Diagnostics API documentation
- Magnetic mount bracket CAD (FDM-printable)
- Assembly instruction card
- React web dashboard source
- Python anomaly detection pipeline (Isolation Forest + MQTT consumer)
- Model training notebook and retraining guide
- Dashboard deployment guide (Docker Compose)
Outcome
"Pilot ran across 3 machines for 6 weeks. The system flagged 4 anomalies — 2 were confirmed as early bearing wear by the maintenance team. False positive rate was within agreed tolerance. SME approved roll-out to all 12 machines."
Handover
- Firmware and dashboard repos with deployment scripts
- AI model files and retraining notebook
- Installation and maintenance runbook
- Handover call and recorded session
Related Services
Start Your Project
Ready to build your prototype?
Tell us about your idea and we'll help you plan the fastest path to a working prototype or V1.