Acquire
Read paired air sensors, roof temperature, water contact, fan rotation and fan-branch power about once per second.
TEACHER / PANEL PRESENTATION
From a sensor reading
to a decision you can explain.
Follow the block diagram. Inspect the labelled shelter. Click any component to learn what it does, why it is there, and what still needs verification.
Two scales, one clearly marked explanation. GPIOs and control behavior below describe the implemented tabletop firmware. The human-scale CAD shows proposed placement. No physical assembly or cooling result is claimed; this is not a full-size wiring approval.
01 / SENSE → DECIDE → ACT → VERIFY
A proper IoT block diagram.
Green arrows carry measurements and commands; the blue return arrow is fan-rotation feedback; amber arrows carry power. Each linked hardware block opens its CAD explanation below. On a phone, scroll the diagram horizontally or open the standalone version.
Open / download full-resolution diagram ↗ · Detailed bench wiring schedule
Where the data goes
1 · ESP32USB serial JSON · 115200 baud
Sequence, session and acquisition time
Sequence, session and acquisition time
2 · Laptop bridgeSerial frame validation
Authenticated local HTTP POST
Authenticated local HTTP POST
3 · Local FastAPI + SQLiteStore source, configuration and calibration metadata
4 · Dashboard / CoolProofMonitor measurements; withhold invalid comparisons
Optional read-only AI explanation
Optional read-only AI explanation
Public Vercel website = static demonstration. It does not receive live ESP32 data or control the fan. Current firmware uses USB serial; Wi-Fi/MQTT/cloud telemetry is a future integration, not an implemented feature. A network or laptop failure must not erase the local assistance latch.
02 / EVERY SOLID IS SELECTABLE
What is it? Why is it there?
Start with Cooling view to see the reference worker, fan blades, sensors and intended airflow. Change temperature, humidity, roof temperature or fan feedback beneath the model to explain the bench control response. These are synthetic teaching inputs, not cooling predictions. Use the 16-stop tour for your presentation, the searchable list for all 143 instances, or click a labelled component. Selecting an internal board reveals the enclosure and zooms to the actual CAD location. The information panel retains original dimensions and coordinates even in exploded view.
03 / IMPLEMENTED BENCH LOGIC
Explain one complete working cycle.
Validate & decide
Reject missing or invalid readings. Manual mode commands 70% duty. Automatic mode is disabled by default; its optional 32°C bench trigger is not a safety threshold.
Act locally
Drive the fan PWM interface, update the OLED and latch local assistance. Tach feedback can flag a rotation fault after startup grace.
Record & compare
Send timestamped serial records to the laptop. CoolProof requires valid reference, timing, calibration and coverage before allowing a comparison conclusion.
Pin map — classic ESP32 bench only
| Connection | GPIO / bus | Reason / electrical condition |
|---|---|---|
| Reference + seated SHT31, OLED, INA219 | SDA21 / SCL22 | I²C addresses 0x44, 0x45, 0x3C, 0x40. Short bench bus, 3.3 V-compatible pull-ups. Human-scale cable lengths are not validated. |
| Fan PWM / tach | 25 / 34 | NPN open-collector signal driver; separate fan supply. Tach needs external 10 kΩ pull-up to 3.3 V. Never feed 12 V to an ESP32 input. |
| Roof probe | 4 | Powered 1-Wire DS18B20 with 4.7 kΩ pull-up. Surface contact and calibration required. |
| Water contact | 35 | External 10 kΩ pull-up. Closed to ground = available; open wire = low/check. Water does not control the fan. |
| Mode / help / acknowledge | 32 / 33 / 27 | Active-low buttons with internal pull-ups. Help persists in NVS; explicit acknowledgement clears it. |
| Status LED / assistance buzzer | 26 / 14 | LED series resistor; transistor buzzer driver. Local alerts only, no remote emergency dispatch. |
| Controller / fan power | USB / separate 12 V | Common signal ground. Fan branch fused and switched; INA219 excludes USB-side energy. Do not simultaneously back-feed the board 5 V pin. |
What happens when something fails?
| Condition | Actual intended behavior / limitation |
|---|---|
| Sensor disconnected | Mark invalid readings and show a sensor fault; automatic operation is inhibited. Manual control remains available. |
| Fan stopped | A commanded fan without expected tach feedback raises a fault after startup grace. RPM does not prove useful airflow. |
| Low water | Show low/check. No automated evaporation or pump exists. |
| Laptop/network lost | Local powered controls continue. Missing telemetry intervals are not invented. |
| All power lost | No powered sensing, fan or active alert; shade remains. A persisted assistance latch restores after reboot. |
04 / CURRENT GAP REVIEW
Clear explanations. Honest unfinished work.
Presentation gaps addressed
The original architecture was hidden in a collapsed section. Selecting ESP32 showed only its size, while the board stayed hidden inside the enclosure. The new visible block diagram, component rationale, cutaway, search and tour connect the explanation to the CAD.
Electronics still to detail
The discrete PWM/buzzer drivers, fuse, supply, connectors, full wire harness and standalone power conversion are not detailed CAD parts. They are explicitly shown as electrical requirements in the diagram. Full-size cable interfaces and outdoor actuator selection remain open.
Sensor placement still to validate
The “reference” sensor is near the roof overhang and is not yet a valid external control. Shields are simplified envelopes. The roof probe needs a thermal contact fixture. Placement rationale is a design hypothesis, not a measurement result.
Manufacturing & cost still open
Foundation, structural connections, roof weatherproofing, enclosure sealing and supplier fit need professional review. Current full-size allowance is about ₹94,000–₹1.71 lakh, unquoted; the below-installed-AC target is not demonstrated.
Verified artifacts: source code, automated software checks, valid CAD solids and browser interactions. Not verified: physical hardware, occupied performance, thermal safety, production readiness, novelty or manufacturer endorsement.
05 / SPEAKING NOTES
Questions your panel may ask.
Why two air sensors?
One samples the occupied zone, the other is intended for comparison. Both need matched shielding, calibration and appropriate placement. Temperature alone cannot demonstrate fan benefit; useful air speed needs a separate instrument.
Why put the controller inside, but air sensors outside the enclosure?
The enclosure protects electronics and supports service access. Its heat and trapped air would bias environmental readings, so the air sensors sit in dedicated shields. The metal enclosure also means wireless performance cannot be assumed.
Why a double roof and louvres?
They test shade and ventilation mechanisms inspired by climate-responsive architecture. Their effectiveness depends on sun direction, wind, surface properties and construction. We will measure those effects rather than promise a temperature drop.
Why not claim the fan proves cooling?
Tachometer rotation is an actuator-health signal. Useful air speed, radiant conditions, humidity and human responses are different measurements. No occupational safety claim is made.
Where is the AI?
An optional local model can explain finished experiment records. It is read-only, disabled by default, and not connected to the public demo. Deterministic firmware handles control; AI cannot declare a person safe.
Is this a connected IoT product already?
The implemented path is ESP32 → USB serial → laptop bridge → local API/database → dashboard. Internet hosting currently publishes the demonstration. Wi-Fi/cloud device telemetry is a future integration.
What is novel?
The proposal combines a repairable shelter with traceable comparison records and explicit invalid-result handling. Novelty remains a research question; this work does not establish a world-first invention or patentability.
What is the next physical experiment?
Assemble the unoccupied tabletop bench, calibrate sensors, test disconnections and fan faults, then run repeated paired trials with swapped positions. Record airflow, roof response, energy, uncertainty and exclusions. Miniature results do not directly predict full-size results.