Probabilistic fault detection and diagnostics for packaged air-conditioner outdoor-air economizers

Andrew L Hjortland, Purdue University

Abstract

Poor economizer control, economizer damper failure, and excess outdoor-air contribute to these performance degradations. In order to promote optimal rooftop air-conditioner (RTU) performance and reduce operating costs, an automated fault detection and diagnostics (AFDD) tool has been designed for RTUs with integrated economizers. Based on previously proposed methods, the proposed method advances the economizer fault detection and diagnosis components by using statistical classifiers in order to provide more robust, probabilistic fault outputs. A set of air-side virtual sensors has also been added to the method in order to expand the applicable range of conditions fault detection and diagnostics can be applied. The operational performance of the outdoor-air damper was characterized using a series of laboratory tests in order to model the expected outdoor-air fraction at different damper actuator control signals and ambient conditions. Two temperature correction models were developed in order to minimize the sensor error caused by stratification. The first correction was to the outdoor-air temperature sensor. This sensor was influenced by return-air that was recirculated back into the outdoor-air stream, an effect of economizer hood design. The second temperature correction modeled was for the single-point mixed-air temperature. At the mixed-air temperature sensor location, significant thermal stratification and non-uniform flow is present due to ineffective mixing in the RTU mixing box. Finally, the temperature rise across the indoor fan was modeled, along with the expected mass-air flow rate and power consumption of the indoor fan. Using these models of normal performance, deviations from normal are detected using a fault detection classifier. Using a Bayesian classifier a comparison of expected and actual performance is made when the RTU operates at steady-state. Outdoor-air damper position faults and temperature sensor faults, including faults in the outdoor-air, return-air, mixed-air, or supply-air temperature measurements, are considered by the AFDD tool. After a fault has been detected, an active economizer diagnostic procedure is performed by sweeping the outdoor-air damper from the fully-closed to fully-open position. When the damper is at these positions, redundant system measurements can be compared and a set a fault diagnosis residuals can be calculated. These residuals yield unique responses to different faults when they are present in the system. Using this as a guide, faults are isolated using a statistical fault diagnosis classifier. Experimentally collected data were used to test the effectiveness of the AFDD method under different normal and faulty conditions. The false alarm rate of the fault detection method was approximately 1.0 %. The misdiagnoses rate of the diagnosis classifier for normal data was approximately 4.9 %. When taken together, the overall false alarm rate of the AFDD tool was approximately 0.05 %. This low false alarm rate can be attributed to the accuracy of the temperature sensor correction and outdoor-air fraction models that can be attained when using experimentally obtained training data for an individual RTU. This also shows of the advantage of embedding diagnostics into the equipment over a tool that is applied retroactively. The diagnosis tool was also able to correctly identify greater than 90 % of the different faults studied. The most significant faults studied, stuck outdoor-air damper faults, were correctly diagnosed in 93.2 % of the fault cases. As a first step towards determining optimal FDD thresholds, several performance tests were conducted in the laboratory in order to observe the affects of a stuck damper fault on system performance. Tests with warm, humid outdoor-air temperatures were considered. Different damper positions were tested and their impact on the outdoor-air fraction entering the system were examined. The damper faults were shown to increase system capacity and efficiency due to the higher evaporation temperature caused by the higher fraction of warm outdoor-air at the evaporator air inlet. However, a negative impact on required RTU run-time was also determined, yielding increases in required energy consumption in order to meet equivalent conditioned space loads. The cause of this increased run-time was the increased ventilation load component introduced by the opened damper. These conditions lead to a reduction in available cooling capacity to meet the space load. (Abstract shortened by UMI.)

Degree

M.S.M.E.

Advisors

Braun, Purdue University.

Subject Area

Mechanical engineering|Operations research

COinS