🤖 How AI Nameplate Scanning Is Changing HVAC Field Work

Rooftop units bake in the sun for decades. By the time a technician needs the serial number or refrigerant charge weight off a 15-year-old nameplate, half the text is faded, the rest is corroded, and the label is often mounted in a spot that's impossible to see clearly.

AI nameplate scanning changes that. One photo — even a bad one — is often enough to extract the data you need.

The Technology Behind It

Equipment Tracker Pro uses Google's Gemini multimodal AI to analyze nameplate photographs. Unlike basic OCR (optical character recognition) which just reads text, Gemini understands context. It knows that a 4-digit code next to "MFG DT" is a manufacture date in YYMM format, not a serial number. It knows that the number in the Model box that starts with "FD" is part of a Bosch production code that belongs appended to the model number, not treated as a separate field.

The system extracts up to 40+ specification fields from a single image: manufacturer, model number, serial number, BTU capacity, voltage, phase, full load amps, MCA, MOP, horsepower, refrigerant type, SEER/EER rating, factory refrigerant charge weight, static pressure, filter size, belt size, manufacture date, and more.

SEER Ratings and Refrigerant Charge

Two fields that were recently added to the extraction model are the SEER/EER efficiency rating and the factory refrigerant charge weight. These are critical numbers for any HVAC technician servicing a unit.

The SEER rating tells you what the unit was designed to deliver and is essential context when diagnosing efficiency complaints. The factory refrigerant charge weight is the single most important number when recovering and recharging a system — it tells you exactly how many ounces or pounds the system calls for, which is stamped on the nameplate by the manufacturer. Having both of these in your digital equipment record means you don't have to look them up every time.

Strict Data Integrity Rules

One problem with early AI extraction tools was that they were too eager — pulling filter sizes off nearby sticker labels and jamming them into the equipment record whether they belonged there or not. Equipment Tracker Pro uses strict rules to prevent this.

Filter sizes and belt sizes are only extracted if they appear explicitly labeled on the main equipment nameplate. Dates written on filter media are captured as filter install dates, not manufacture dates. Job numbers, work order numbers, and contractor tracking codes are never mistaken for serial numbers. The result is clean, reliable data that you can actually trust.

What This Means for Your Workflow

Before AI scanning, building a complete equipment record meant writing down specifications by hand, typing them in later, and hoping you got them right. A single unit might take 10-15 minutes to fully document. With AI scanning, the same unit is documented in under a minute — often with more fields filled in than a technician would have manually recorded.

For a company maintaining a portfolio of 200 rooftop units, that difference compounds into dozens of hours saved per year, and a database that's actually complete enough to be useful.

Equipment Tracker Team

Product · Equipment Tracker Pro