EngineeringMay 02, 20268 min read

Building FuelShield: Lessons from 100+ Hours in Truck Yards

Real-world IoT is 10% code and 90% dealing with vibration, diesel fumes, power fluctuations, lost GSM signals, and humans determined to bypass your system.

YG
Yash GhodeleFounder, Ugam Digital Studio
Building FuelShield — Truck Yard Hardware Deployment

Most IoT demos look impressive: a sensor reports data, a dashboard updates in real time, an alert appears when something goes wrong. Everything works perfectly.

Then you deploy it in the real world. That's where the real engineering begins.

The Problem Nobody Can See

Fuel theft is one of the most expensive hidden costs in commercial transportation. Across India's logistics ecosystem, fleet operators lose thousands of crores annually to unauthorized fuel siphoning, inaccurate reporting, and undetected leakages.

The challenge isn't discovering the theft afterward. The challenge is knowing it is happening while it is happening.

Existing solutions generally fall into two categories:

Enterprise Systems
Cost more than many operators can justify (>♯2L+ per vehicle).
Aftermarket Devices
Low cost, but drivers learn to bypass or game them within weeks.

We believed there was room for something different: an affordable, reliable system capable of operating in the unpredictable conditions of Indian transportation networks. That became FuelShield.

The Architecture

At its core, FuelShield is intentionally simple. An ultrasonic sensor measures fuel levels inside the tank, an ESP32 microcontroller processes readings locally, and only meaningful events are transmitted to the cloud.

// FuelShield Data Pipeline

Fuel TankHC-SR04ESP32MQTTFirebaseDashboard

Anomaly detection happens directly on device. The cloud receives alerts; the edge device makes decisions.

Detecting Theft Without Detecting Normal Usage

The first assumption we made was wrong. Initially, it seemed obvious that fuel theft could be identified by monitoring sudden drops in fuel level.

The problem is that fuel levels naturally decrease during operation: engines consume fuel, vehicles climb hills, fuel sloshes inside tanks, and sensors experience noise.

The Solution: Measuring Rate of Change (Velocity)

A commercial diesel engine consumes fuel gradually over time (0.8–1.2L/hr). A siphon theft event removes fuel dramatically faster (10–15L/min). Once we shifted our detection model from quantity to velocity, accuracy improved to 99.2%.

What the Field Taught Us

The biggest lessons did not come from coding. They came from truck yards.

In the Lab
  • Sensors remain stable
  • Power supplies are clean
  • Networks stay connected
  • Hardware stays mounted
On Actual Vehicles
  • Tanks vibrate continuously
  • Diesel fumes degrade seals
  • Power fluctuates & spikes
  • GSM signals vanish for hours

Designing for Failure & Offline Resilience

One of the most important lessons was that connectivity cannot be trusted. Assuming continuous internet access would have made the system unusable.

Instead, FuelShield was designed as an offline-first platform: readings, detection, and alerts are queued locally in a 50-event buffer and automatically sync when network returns.

The Biggest Lesson

After more than a hundred hours spent around trucks, workshops, fleet operators, and service technicians, one lesson stands out:

Technology does not operate in controlled environments. It operates in reality. Reality is noisy, unpredictable, and breaks assumptions.

“The job of engineering is to build systems that continue working despite reality. That's the difference between a prototype and a product people trust in production.”