ESP32-Based Machine Vibration Monitor with FFT Analysis and Predictive-Maintenance Alerts

The FFT needs real compute — 1024 points every window plus Wi-Fi telemetry. The ESP32's dual-core processor and built-in Wi-Fi handle both, while an Uno-class board would struggle with the FFT at these window sizes.

It converts the vibration waveform into a spectrum: the height of each peak shows how much energy exists at that frequency. The 1× peak sits at the motor's running speed; a growing 2× peak hints at misalignment and rising high-frequency tones hint at bearing wear — the classic pattern the dashboard marks.

The firmware ships with configurable defaults (6.3 / 11.0 mm/s, adapted from ISO 10816 teaching practice). The included calibration procedure has the buyer measure their own motor's healthy baseline first, then set the thresholds relative to it — so the alert numbers in the report are the student's own measurements.

It detects the vibration signatures that precede failures — rising overall RMS and growing fault harmonics — which is exactly what predictive-maintenance monitoring does at teaching scale. The report is careful: the system flags vibration conditions, it does not diagnose specific faults without the student's own analysis.

No — the node and dashboard communicate over the local Wi-Fi network. No cloud account, no SMS and no GPS are part of the build, which keeps the demo working in a lab without internet.

Yes — for IoT, Embedded Systems and Electronics programs. It demonstrates sensor interfacing, real-time signal processing (FFT), threshold-based alert logic, Wi-Fi telemetry and a live dashboard — a complete hardware-to-software loop. Suitable for B.E./B.Tech final-year projects in IoT, Embedded Systems and Electronics engineering. problem: Bearings, imbalance and misalignment are among the earliest precursors of rotating-machine failure, and rising vibration is their most accessible symptom. Industrial condition monitoring uses expensive analyzers and software that students never get to operate, so final-year students learn FFT spectra, aliasing and ISO 10816 zones only as theory. This project gives them the full chain to build and demonstrate themselves: a low-cost ESP32 + MEMS accelerometer node that samples vibration, computes the spectrum on-device, and raises threshold-based predictive-maintenance alerts streamed live to a dashboard. concept: An ADXL345 accelerometer mounted on the machine body feeds 3-axis vibration data to an ESP32 over I2C at a 200 Hz design target. Each window is Hanning-windowed and transformed with a 1024-point FFT on the device, producing a 0–100 Hz spectrum from which the firmware extracts RMS velocity, the dominant frequency, and the 1×/2× running-speed harmonics. These metrics are compared against configurable warning and critical thresholds with persistence filtering; confirmed alerts update the on-node OLED, sound the buzzer on critical, and push timestamped events over Wi-Fi (MQTT/HTTP) to a companion dashboard that renders the live waveform, marked FFT spectrum, KPI cards and an alert event log. A buyer-run calibration procedure sets thresholds from the student's own baseline measurements. features:

• ADXL345 3-axis MEMS accelerometer sampling over I2C at 200 Hz (design target)
• 1024-point on-device FFT with Hanning window; 0–100 Hz spectrum
• 1× running-speed and 2× harmonic identification and dashboard marking
• Configurable warning (6.3 mm/s) and critical (11.0 mm/s) RMS velocity thresholds
• Persistence filtering so single-window spikes do not raise alerts
• Companion live dashboard: waveform, FFT spectrum, KPI cards, alert event log
• On-node 0.96-inch OLED status display (NORMAL / WARNING / CRITICAL)
• Local buzzer alarm on confirmed critical alert
• Wi-Fi telemetry (MQTT/HTTP) to dashboard every ~5 s (design target)
• Buyer-run baseline and threshold-calibration procedure requirements:
• ESP32 development board with Wi-Fi
• ADXL345 3-axis accelerometer breakout (I2C)
• 0.96-inch OLED display (I2C)
• Active buzzer module, jumper wires, breadboard or base
• Small DC motor test rig with optional imbalance weight (demo)
• Computer with Arduino IDE for firmware; any modern browser for the dashboard
• Local Wi-Fi network for node-to-dashboard telemetry (no internet required) deliverables:
• Assembled sensor-node prototype (ESP32 + ADXL345 + OLED + buzzer)
• ESP32 firmware source with comments (sampling, FFT, thresholds, telemetry, OLED UI)
• Companion live dashboard (single-file HTML/CSS/JS)
• Wiring diagram and accelerometer mounting guidance
• Buyer-run baseline and threshold-calibration procedure
• Abstract + project report PDF (theory, methodology, student's own readings)
• PPT presentation for final review
• Viva Q&A preparation document limitations:
• 0–100 Hz spectrum only (200 Hz sampling → Nyquist limit); high-frequency bearing tones above ~100 Hz out of scope
• Demo rig is a small DC motor — thresholds are teaching values adapted from ISO 10816 practice, not industrial certification levels
• RMS velocity is integrated from acceleration in firmware; buyer measures their own noise floor per the procedure
• Wi-Fi telemetry is local-network only; no cloud, SMS or GPS in the build
• Alert quality depends on student-calibrated thresholds; no claim of diagnosing faults not demonstrated on the student's rig
• Teaching prototype, not certified condition-monitoring equipment timeline:
• Typical built-to-order delivery: hardware assembly and wiring in the first week, firmware (sampling, FFT, threshold logic) in the second, dashboard and report documentation in the third, with the calibration procedure run and verified on the finished build before handover. Exact schedule is confirmed at order time. support:
• Explanation support covers the full chain: ADXL345 interfacing, FFT and windowing choices, RMS integration, threshold and persistence logic, and the dashboard telemetry flow — so the student can answer viva questions on every part. Report, PPT and viva Q&A are included; code walkthrough provided on request.