HVAC Fault Detection and Diagnostics: What to Require in a Pilot

An HVAC fault detection and diagnostics pilot should demonstrate that your building's data can produce findings a technician can verify, an owner can prioritize, and a maintenance team can resolve. Before buying a portfolio-wide subscription, define the equipment, data, responsibilities, and evidence that will make the pilot worth expanding.

The most useful deliverable is a traceable record: what the software detected, what investigation found, what action was approved, and whether the condition improved afterward.

Controls and equipment data provide the foundation for a carefully scoped fault-detection pilot.

What does HVAC fault detection and diagnostics do?

Fault detection and diagnostics, commonly called FDD, analyzes operating data to identify abnormal equipment or system behavior. Some tools also suggest likely causes. DOE notes that diagnosing faults and prioritizing large numbers of findings remain important challenges; an alert should therefore be treated as an investigation lead rather than a confirmed repair instruction. See DOE's FDD research overview.

A building automation system supplies control functions and often the data used by analytics. FDD can support a commissioning or maintenance process by reviewing that data repeatedly. Its usefulness depends on the equipment, available points, configured rules, and people responding to findings.

For the broader existing-building investigation process, see Tustin's commercial retro-commissioning guide. Here, the purchasing decision is narrower: can a defined analytics service improve the way your team identifies and closes operational problems?

Start with a specific operating question

Choose a pilot around a question the facility already needs answered. Examples include whether air handlers run outside approved schedules, whether heating and cooling operate together unnecessarily, or whether sensor readings make control sequences unreliable.

Select equipment that is representative of the intended rollout and accessible for field investigation. A carefully chosen group of air handlers or rooftop units can reveal more about deployment effort than a large equipment count with incomplete records.

Record the operating conditions the pilot must observe. Cooling, heating, economizer operation, occupancy changes, and process loads may not all occur during the same window. If a needed condition never occurs, classify that test as deferred or unobserved. Avoid calling an untested rule successful.

Require a data-readiness check before judging results

Have the provider deliver a point map that connects each analytics input to the correct equipment and physical meaning. A useful checklist includes:

Point name, unit, equipment relationship, and data source
Trend interval, timestamps, time zone, and retention period
Missing values, repeated values, communication gaps, and obvious outliers
Sensor condition and any required field checks
Commands versus actual equipment status or measured response
Current schedules, setpoints, overrides, and sequences of operation
Which diagnostic rules cannot operate with the available data

Ask the bidder to distinguish existing integration from additional gateways, licenses, sensors, and engineering work. Establish who resolves mapping errors and who maintains the mapping after equipment changes.

Include the owner's IT and controls teams in connection approval. Define permissions, remote support, data ownership, export formats, and access termination before connecting systems. If automatic control changes are proposed, give them a separately reviewed scope and approval process.

Define what counts as an actionable finding

A useful finding should identify the affected asset, observed behavior, supporting trend, applicable operating condition, and reason it deserves attention. The report should separate the suspected cause from the verified cause.

For each finding, require:

Evidence that can be reviewed without relying solely on a dashboard score
A priority tied to operational impact and confidence
A named person responsible for investigation
A proposed next action and its approval requirements
A work-order reference when corrective work is authorized
A verification record showing the outcome under relevant conditions

An apparent simultaneous-heating-and-cooling fault, for example, needs the design sequence and actual operating context before it becomes a recommendation. A rule that ignores intended humidity control or a critical process requirement can create avoidable work.

Put the response process in the proposal

Software licensing and ongoing engineering support are separate purchasing questions. DOE's EMIS procurement resources include technology specifications, ongoing-service specifications, and evaluation criteria. That structure is useful when comparing offers with different service boundaries.

Ask each bidder to explain who reviews findings, how often the review occurs, who investigates in the field, and how unresolved issues escalate. Clarify whether repairs, controls programming, retesting, operator training, and seasonal review are included or separately priced.

Request a sample closeout record. It should show more than an alert disappearing: a rule can stop firing because data stopped arriving, an operating mode changed, or someone disabled it. The owner needs enough context to distinguish those outcomes from a successful correction.

Set pilot acceptance criteria before launch

Agree on an evaluation method rather than accepting a promised number of alerts or a generic savings percentage. Use these buyer questions:

Is the required data available and trustworthy?
Can the team reproduce and explain representative findings?
Are duplicate, low-value, and false findings being identified and tuned?
Can the maintenance team handle the investigation workload?
Are approved corrections verified and documented?
Are excluded equipment, untested modes, and remaining integration costs clear?
Can the owner export the equipment map, findings, and history?

If energy savings are an objective, define the baseline, measurement boundary, and adjustments for weather, occupancy, and operations before implementation. Keep estimated opportunities separate from measured results. DOE's EMIS planning guidance emphasizes goals, staffing, scope, and measurement considerations.

Frequently asked questions

Does FDD replace the building automation system?

Usually it uses data from controls and other sources. Confirm the proposed architecture and integration requirements for the existing system.

Does an alert prove a component has failed?

No. Data quality, operating context, and field investigation determine whether the finding supports repair, programming changes, or no action.

How long should a pilot run?

Long enough to observe the operating conditions needed to evaluate its agreed questions. Define seasonal gaps and deferred checks in the scope.

Should we expand after finding many faults?

Expansion should depend on verified usefulness, manageable workload, completed corrections, data access, and a clear operating budget. Alert volume alone is insufficient.

Discuss a building-analytics pilot with Tustin

The Tustin Group's energy solutions include monitoring, custom dashboards, systems integration, commissioning, and measurement and verification. Contact Tustin to review your controls and operating priorities, confirm available analytics and integration services, and define responsibilities for any potential FDD pilot.