Skip to main content
IoTGPT for Boiler & Plant OEMs

Fault Screening for Boiler Plants

Plant data + OEM know-how, on-premises under data privacy — catch developing faults early and explain what to do next.

Lab test

costly fault types caught on a controlled national-lab dataset with injected faults

4

fault families tested: sensor drift, fouling, mis-tuned controls, pressure bias

15

standard plant readings used: flows, temperatures, pressures, pump power

The business story

  • Benchmarked on a US national-lab hot-water boiler dataset where faults were deliberately injected: drifting sensors, fouling, mis-tuned controls and pressure issues.
  • The costly fault types (temperature and flow drift) were caught almost every time on that controlled test bed — before they would show up as higher bills or comfort complaints.
  • For a boiler OEM: a commissioning-friendly screening layer — plant data finds the drift, your published docs explain the fix.

Why boilers are a fit

1. Site algorithms on this plant's loop: Purpose-built algorithms learn the plant's normal thermal pattern from this building's own control-system data and flag developing faults — on-site, no cloud.

2. OEM docs bundled at install — not other plants' data: Alert explanations draw on manuals, hydronic specs and drawings you ship with the product. Customer operating data never leaves the site and is not used to train other customers.

Sensors needed for this use case

Water flow · Supply and return temperatures · Pressure / differential pressure · Pump power

Standard readings already in the building or plant control system. Optional inputs: OEM manuals, specs, sketches.

How it rolls out

1

Connect

plant / BMS data plus OEM manuals, specs and drawings for the boiler loop.

2

Baseline

two weeks of normal-season operation establish the plant's fingerprint.

3

Screen

issues ranked by drifting readings, explained with site data + OEM docs.

4

Triage

service desk sees which loop — boiler, pumps or piping — drives each alert.

Validated facts

Evidence typeControlled lab test bed (injected faults)
Costly fault types on that bedCaught almost every time
Where it runsOn-site — data privacy preserved
Knowledge usedThis plant's data + OEM docs (not other fleets)
Extra hardware requiredNone

Honest scopeResults are from a controlled national-lab dataset with injected faults — not yet a multi-site field fleet. Subtle pressure faults stay hard to see in standard tags. Energy upside here was small (≤5%), so lead with fault screening, not savings. Validate per site in a pilot.

IoTGPT — Industrial AI for OEMsTest bed: US national-lab boiler-plant fault dataset, audited July 2026