BombaWatch: ₱800 pump listener that texts any phone before burnout

Problem Statement

Communal pumps and irrigation pumps die suddenly — dry-run, worn bearings, voltage sag — and repair costs triple after burnout. Caretakers check by ear once a day. Commercial SCADA costs more than the pump and needs cloud fees and stable internet the barangay doesn’t have.

Proposed Solution

A clamp-on ESP32 listens to vibration/current signatures, learns “normal” over two days, and flags anomalies locally. Alert goes as plain SMS to up to 3 caretaker numbers on any network, plus a loud buzzer at the pump. Log stays on the device; technician reads last 50 events over USB. Solar trickle keeps it alive through brownouts.

Target Market

• Primary users: Barangay water caretakers, irrigators’ associations, school pump custodians.
• Deployers: Electricians and youth tech volunteers; hardware shops as kit sellers.

How AI Is Used

• TinyML anomaly detection: Autoencoder on vibration spectra; runs on ESP32 with no cloud.
• Dry-run vs bearing classifier: Small supervised head trained on team-recorded pump sounds.
• Why AI: Fixed thresholds false-alarm on every voltage dip; learned hum is robust.

Independent Deployment Angle

• Carrier-agnostic SMS, no platform lock, no venue, no headset. Open firmware any TESDA grad can flash.
• Parts from Raon/online: ESP32, sensor, buzzer, solar cell.
• Repair stays local; no ticket to a distant NOC.

Demo Scenario (3 minutes for judges)

1. Show live “healthy hum” waveform on a laptop from a demo motor.
2. Pinch the hose to simulate dry-run; device buzzes and a caretaker phone receives SMS within seconds.
3. Show on-device event log with timestamps, no internet needed.
4. Swap to worn-bearing recording; different alert code appears.
5. Close: “Narinig bago masira.”

30-Hour Build Plan

Suggested stack: ESP32 + MEMS vibration sensor, TinyML autoencoder, any-SIM SMS module, solar trickle charger. Public data / APIs: Team-recorded pump audio; no external API.

Risks & Privacy Notes

• False alarms: Two-stage alert (watch vs urgent) and caretaker confirm button; no auto-shutoff of water in v1.
• Safety: Clamp-on only, no mains rewiring by students; licensed electrician sign-off noted.
• Privacy: No personal data collected; phone numbers stored only on the device with caretaker consent.

AI Use Disclosure (draft)

Uses TinyML anomaly detection trained on team-recorded sounds. AI coding assistants helped with firmware scaffolding. All demo faults were staged and labeled by the team.

Rubric Scoring