What is fault detection and diagnostics (FDD) for rooftop units? A plain-English guide

Last updated: August 25, 2026.

Fault detection and diagnostics (FDD) is software that watches equipment data for the early signatures of developing problems and explains what it found. For rooftop HVAC units, that means reading the temperatures, runtimes, and staging a building already produces, comparing them against normal, and flagging refrigerant loss, failed stages, short cycling, and airflow problems days to weeks before comfort fails.

That is the whole idea. The rest of this guide explains how it works on rooftop units specifically, what kinds of problems it catches, what it cannot do, and how to tell marketing claims from real capability.

What does FDD actually do?

Every rooftop unit produces a continuous record of its own operation: supply and return air temperatures, which cooling stages are running, fan state, setpoints, runtime. Healthy units produce recognizable patterns in that data. A unit cooling properly drops the air temperature across its coil by a consistent amount. A correctly sized unit cycles a predictable number of times per day. Runtime tracks outdoor weather.

Failing units break those patterns, and they break them early. A refrigerant leak shows up as a slowly shrinking temperature drop across the coil long before the space gets warm. A failing compressor shows up as abnormal cycling or stages that engage but do not deliver cooling. A stuck damper, a blown fuse on one stage, a sensor drifting out of calibration: each one leaves a distinct signature in the operating data.

FDD software reads those signatures continuously. Detection is noticing that behavior has deviated from normal. Diagnostics is identifying which specific problem the deviation points to. Together they turn a rooftop unit from a black box that either works or does not into equipment that reports its own developing problems.

Why does this matter for rooftop units in particular?

Rooftop units are the workhorse of restaurant and retail cooling, and they occupy a blind spot. They live where nobody looks, they fail gradually rather than all at once, and the buildings they serve rarely have on-site engineering staff. A large office tower might have a building engineer who notices a chiller acting strange. A restaurant has a store manager who notices when the dining room is hot, which is the last stage of a failure, not the first.

Most rooftop unit failures follow a long arc: weeks or months of gradual decline, then a threshold, then a hot store and an emergency call. The decline phase is invisible from inside the building but obvious in the operating data. That gap between when a problem becomes detectable and when it becomes noticeable is where FDD lives, and on rooftop units that gap is typically measured in weeks.

What kinds of problems can FDD catch?

The failure modes that appear in operating data before they appear in the space include:

• Refrigerant loss. The temperature drop across the cooling coil declines gradually as charge leaks out. One of the most common and most catchable faults.
• Compressor degradation and failure. Struggling starts, abnormal staging behavior, stages that run without delivering cooling.
• Short cycling. A unit starting and stopping far more often than designed, which destroys compressors over time. Nearly impossible to observe on a site visit, obvious in continuous data.
• Electrical faults. A blown fuse or failed contactor on one stage can leave a unit limping at partial capacity while appearing to run normally.
• Economizer and damper problems. Dampers stuck open or closed waste energy for months without producing a comfort complaint.
• Sensor failures. A frozen or drifting temperature sensor quietly corrupts everything the unit's controls do.
• Schedule and setpoint drift. Units running all night, setpoints overridden and never restored. Not equipment faults, but among the most common findings and the fastest savings.

What can't FDD do?

An honest guide has to include this list, because vendor marketing often does not.

FDD does not fix anything. It observes and diagnoses. A qualified technician still does the repair. The value is that the technician arrives on a planned schedule, with the right diagnosis in hand, before the failure, instead of on an emergency dispatch after it.

FDD needs data to analyze. Rooftop units connected to an energy management system or connected thermostats already produce the necessary data. A unit controlled by a standalone manual thermostat produces nothing to analyze and would need connectivity added first.

FDD cannot see what the data does not contain. A unit reporting only space temperature supports weaker analysis than one also reporting supply air temperature and staging. More data points mean earlier and more specific detection. Good FDD is transparent about confidence: what it knows, what it suspects, and what it cannot see for a given unit.

Not every alert is a crisis. Mature FDD separates "this unit is drifting, watch it" from "this unit needs a technician." Systems that cannot make that distinction bury facilities teams in alerts until the alerts get ignored, which is worse than no system at all.

How is FDD different from an energy management system or a smart thermostat?

An energy management system controls equipment: schedules, setpoints, staging, sometimes demand response. A smart thermostat portal shows current conditions. Both answer the question "what is the temperature and what is the unit being told to do."

FDD answers a different question: "is this unit healthy, and if not, what is wrong with it?" A dashboard can show 74 degrees in the dining room while the unit holding that temperature runs twice its normal hours to do it. Control systems and FDD are complementary layers, and in practice the control system's data stream is usually what the FDD layer analyzes. Operators who already have an energy management system have already built the foundation.

Does FDD require new hardware?

Not necessarily, and this is the most important recent shift in the category. Traditional FDD deployments added sensors and monitoring hardware to each unit, which priced the technology for large campuses and out of reach for a 3,500 square foot restaurant. Software-only FDD instead connects to the systems a building already has and analyzes the data they already produce. Research from Lawrence Berkeley National Laboratory found software-based FDD to be the fastest-growing segment of the category, largely because removing the hardware removes most of the cost and all of the installation project.

The practical question for any operator is not "should we buy monitoring hardware" but "what data do our existing systems already produce, and is anyone analyzing it?"

How do you evaluate an FDD product?

Five questions separate real capability from a dashboard with alerts bolted on:

1. Does it diagnose, or just alert? "Unit 3 is abnormal" is detection. "Unit 3 is losing cooling capacity in a pattern consistent with refrigerant loss" is diagnosis. Only the second one tells a technician what to bring.
2. Does it learn each unit's normal? A kitchen unit and a dining room unit have different healthy baselines. Fixed one-size thresholds produce false alarms on one and missed faults on the other.
3. Does it account for weather? High runtime in a heat wave is normal. The same runtime on a mild day is a fault signal. Systems that ignore outdoor conditions cannot tell the difference.
4. Can it rank a whole fleet? For a multi-site operator the useful output is not 400 individual unit reports, it is "here are the five units across your portfolio that need attention this week, and why."
5. Are its catches verified? Ask for cases where the system flagged a problem and a technician then confirmed the specific fault. Verified catches are the only honest proof this category has.

See exactly how Elite Energy Management applies these layers in production on the methodology page.

The bottom line

FDD turns rooftop units from equipment you find out about into equipment that tells you. The technology is mature, the software-only version of it fits buildings and budgets that traditional monitoring never reached, and for any operator whose sites already have connected systems, the raw material is already flowing. The only question is whether anyone is reading it.

According to the U.S. Department of Energy, fault detection and diagnostics tools could deliver up to 50 percent HVAC and refrigeration energy savings if widely adopted.

Elite Energy Management applies AI-driven fault detection and diagnostics to rooftop units across restaurant and retail fleets in 28 states, layered on the energy management systems operators already have. No new hardware required.

For a concrete example of FDD in action, see how refrigerant loss shows up in monitoring data. Related reading: software-only vs hardware-bundled HVAC monitoring.