Adaptive Fault Detection

Over $9 billion is wasted because of undetected faults in building sub-systems. Equipment, such as air handling unit fans, pumps, roof top units, and controls, can be in a fault condition but still provide appropriate comfort and therefore not recognized or fixed.

Typical alarms can alert users that a unit is off or outside a certain parameter, but they are not capable of defining optimal performance to alert a user of inefficiencies.

There are many different types of methods for detecting building system faults:

  • quantitative model based
  • qualitative model-based
  • process history based
And, the research team at the UNM ISES Lab is exploring existing and new, innovative techniques for discovering faults. Specifically, the research team is applying Adaptive Resonance Theory and Lateral Priming Adaptive Resonance Theory artificial neural network algorithms to identify these tough to finding faults.

UNM Research Team
Building Fault Example

The figure above shows a typical thermal cooling load (in thermal kW) at the UNM Mechanical Engineering Building. The red line describes actual measured data, which alone looks normal. But, through advanced analysis using Adaptive Fault Detection techniques it is clear that the system did not perform as expected. The mixed air dampers to one of the air handling units malfunctioned and too much outside air was entering the unit, and required more power to cool the building.