Browsing by Subject "Building automation systems"
Now showing 1 - 1 of 1
- Results Per Page
- Sort Options
Item Fault detection and diagnostics modeling for air-handling units on the main campus of The University of Texas at Austin(2019-02-06) McHugh, Megan Kellie; Nagy, Gyorgy ZoltanHeating, ventilation, and air conditioning accounts for about 44% of energy usage in commercial buildings. In HVAC systems, air handling units are used to condition air based on comfort for occupants or controlled environmental requirements. Faults in AHUs can occur due to failures in equipment, actuators, or sensors and feedback controllers. Leakage typically occurs due to faults in ducts, faults with valves that are stuck, broken, or in the incorrect operating position, faults in measurements of state variables, or faults with the controllers maintaining the setpoint from sensor feedback. The variety faults that can occur in AHUs can lead to increased energy consumption, especially when it remains undetected. AHU faults can also lead to uncomfortable conditions for building occupants or impact research and other special facilities as the campus building types include classroom/academic, hospital/clinic, housing, office/administrative, parking/garage, public assembly/multipurpose, and research laboratories. The building automation systems on the main campus of The University of Texas at Austin manage over 100 buildings each with multiple AHUs in different working conditions. In this paper, a methodology is proposed for the fault detection of AHU steam and chilled water valve leakage and for general fault detection and diagnoses of other common AHU faults on the UT campus. The approach is based on supervised machine learning classification models and compared to the ASHRAE fundamentals expert rule-set models. BAS data trended at 15-minute intervals for periods up to 400 days were used. Faults detected through these methods have been validated by UT Facilities Services upon inspection of the faulty AHUs. A dashboard web application was developed for the interactive use and visualization of the fault detection models by UTFS for continuous maintenance prioritization. A classification analysis allows for the prediction of leakage and provides UTFS a priority ranking of AHUs to address for maintenance in the future. The rule-set models provide a method for continuous tracking of AHU features for faults. Identifying and addressing valve leakage and other faults is expected to reduce energy usage and contribute to reduction in average annual energy use intensity in order to improve demand side energy efficiency while maintaining indoor environmental quality. This will contribute to reach the 2020 energy savings targets set in the 2012 UT Austin Campus Master Plan, which outlines a variety of initiatives for sustainable growth through 2030.