← All case studies
SES-005 Manufacturing · legacy line IoT retrofit · predictive maintenance · ESP32
1996 production line with predictive maintenance — no digital interfaces and no machine replacement.
6
machines
78
measurement points
Edge FFT
ESP32 · MEMS
18 h
avg. warning time
Situation
A manufacturer wanted to connect an older line (built 1996) to modern predictive maintenance. Six machines, 8–15 measurement points each — analog only (4–20 mA, 0–10 V, thermocouples, PT100), no digital interface.
Unplanned downtime drove maintenance: failures were detected only after a stop. Market alternative: replace the entire line; vendor SCADA retrofit for these models was not available.
Goal: parallel data capture and early warning — machines keep running without control retrofit or line replacement.
Approach
1
1 Analysis
Signal and failure analysis
- Mapped 78 analog measurement points across six machines
- Identified twelve common failure patterns from maintenance logs (5 years)
- Correlations: which signals indicate which state and which precursors
2
2 PCB
Custom IoT nodes (hardware)
- Per machine type: ESP32 data logger, 16-channel 24-bit ADC
- Inputs 4–20 mA, 0–10 V, thermocouples, PT100; MEMS accelerometer
- IP65 enclosure, 24 V DC from cabinet, Wi‑Fi + MQTT
3
3 Embedded
Firmware: FFT and on-device early warning
- ESP-IDF; ADC sampling up to 10 kHz for vibration
- FFT with ESP-DSP — 0–500 Hz spectral analysis at bearing points
- Rule-based early warning on-device, telemetry raw + KPIs + alarms, OTA
4
4 Deployment
Installation without machine retrofit
- Tap existing measurement points — no change to machine control
- About four hours installation per machine, no planned production stop
- Six nodes live, first time series within days
5
5 Backend
Cloud pipeline and dashboard
- MQTT over Wi‑Fi into broker/adapter and Node.js pipeline
- TimescaleDB for all points, InfluxDB for high-frequency vibration
- FastAPI + React: FFT spectrum, trends, predictive score, escalated alarms
6
6 6 months
Results in operation
- Nine of twelve failure patterns caught by early warning
- Unplanned downtime sharply reduced (reported: 32 h/month → 4.5 h/month)
- First complete digital machine history; spare parts orderable weeks ahead
- Energy use measurably reduced through optimization (reported: about −12%)
How the solution works
6× IoT nodes (ESP32 · ADC · MEMS · IP65)
↓
MQTT over Wi‑Fi (telemetry · alarms)
↓
Broker / pipeline (AMQP/MQTT adapter · Node.js)
↓
TimescaleDB + InfluxDB (time series · HF vibration)
↓
React dashboard (FFT · trends · predictive score)
Results
Predictive maintenance on a legacy line with no digital interfaces — without machine replacement
78 analog points brought into one IoT network
Edge FFT and early warning on ESP32 — cloud for history and dashboard
Installation parallel to running production, typically a few hours per machine
Maintenance shifted from reactive to proactive — patterns and warning time documented
Technical path scalable to more machine types with the same signal landscape
Technologies
IoT retrofitESP32ESP-IDFESP-DSP / FFTCustom PCB24-bit ADCMQTTTimescaleDBInfluxDBNode.jsFastAPIReactPredictive maintenance
Analog plant, high downtime, no data interface?
In 30 minutes we clarify whether an IoT retrofit with edge analytics and a dashboard is realistic for your legacy line — without replacing machines.
Request a free assessment Reply within 24 hours No commitment Confidential