FDD, virtual sensors
Virtual sensors have previously been developed and demonstrated that can provide a low cost and relatively accurate estimation of the amount of refrigerant charge contained in packaged (rooftop) air conditioners. Â One particular virtual refrigerant charge sensor approach uses four surface-mounted temperature measurements to determine suction superheat, liquid-line subcooling and evaporator inlet quality that are inputs to an empirical model for charge. The empirical parameters of the model are determined using linear regression applied to laboratory data collected from the system. In previous studies, extensive psychrometric chamber testing was required at different refrigerant charge levels and ambient conditions to obtain sufficient data for the regression. This testing is expensive for equipment manufacturers and it can be difficult to find available test facilities.Â The current work describes the development of an automated open lab training kit for calibrating the virtual refrigerant charge level sensor in an open laboratory space. The developed automated training kit algorithm has the ability to modulate the condenser and evaporator fans to simulate the effects of different ambient conditions and automatically add different amounts of refrigerant. The charge level is automatically adjusted and monitored using solenoid valves and a digital weighing scale. This approach reduces the human involvement to a great extent and eliminates the need for psychrometric chambers. An optimal set of test conditions has been determined using optimal experimental design techniques and implemented as a Python application. An Arduino microcontroller is used to continuously send data from the sensors to a personal computer which is used to supervise the process, including determining when the system has reached steady-state. The training kit has been applied to several different rooftop units in an open lab space.Â A comparison of the virtual refrigerant charge sensor accuracy and time/cost for calibration determined using the automated system and using psychrometric chamber test facilities will be presented in the paper.