🔍 Fault Detection and Diagnostics (FDD) for Building Systems

What Is FDD?

Fault Detection and Diagnostics (FDD) is the automated identification of abnormal conditions (faults) in building systems and their root causes (diagnostics). FDD systems continuously analyze data from the BMS — sensor readings, equipment run statuses, control signals, energy consumption — and compare them against expected behavior to identify when equipment is malfunctioning or operating inefficiently.

Without FDD, building faults often go undetected for months or years. Studies have found that over 50% of HVAC systems in commercial buildings have significant operational faults at any given time, causing 15–30% excess energy consumption. FDD makes these invisible inefficiencies visible and actionable.

Common Building Faults FDD Detects

Simultaneous heating and cooling — the most common and wasteful fault. A reheat coil adds heat to air that a cooling coil just cooled, wasting both cooling energy and heating energy. FDD detects this when the cooling valve and heating valve in the same air handler are simultaneously open.

Stuck dampers — VAV box dampers or AHU outside air dampers that are mechanically stuck open or closed. A stuck economizer damper open in summer causes massive overcooling loads; stuck closed eliminates free cooling opportunity.

Faulty sensors — temperature, humidity, and CO₂ sensors drift or fail over time. FDD can identify sensors reading impossible values (temperature sensor reading 150°F) or implausible values (zone temperature reading the same value for weeks, suggesting a failed sensor).

Chiller or AHU degraded performance — a chiller with dirty condenser tubes delivers less cooling per unit of electricity (reduced COP). FDD tracks the ratio of cooling output to power input and alerts when it degrades below a threshold.

Control hunting — PID loop instability causing valves or dampers to oscillate continuously, increasing wear and energy consumption.

FDD Approaches

Rule-based FDD — expert-written rules flag specific conditions: "If outside air damper position > 90% AND cooling coil valve position > 10% AND outside air temperature > 65°F, flag excessive economizer cooling fault." Simple to explain and implement, but requires manual rule development for each system type.

Model-based FDD — a physics or regression model predicts expected equipment performance. Deviations between predicted and actual performance trigger fault flags. More sophisticated and scalable than rule-based approaches.

Machine learning FDD — anomaly detection algorithms trained on historical normal operation data flag deviations from learned baselines. Requires no manual rule writing but needs sufficient training data and can produce false positives if baselines are not clean.

FDD Implementation

Commercial FDD platforms (Clockworks Analytics, SkySpark, BuildingIQ, CopperTree Analytics, Siemens Enlighted, Schneider EcoStruxure) connect to the BMS via BACnet, Modbus, or API and process historical and real-time data. Fault rules or models are configured for each piece of equipment. The platform provides a prioritized fault list — ranked by energy impact or criticality — that facility managers work through systematically.

FDD ROI typically pays back in 1–3 years from energy savings alone, plus additional value from avoided equipment failures and extended equipment life.