Data-driven based reliability evaluation for measurements of sensors in a vapor compression system

Sensors play essential roles in the refrigeration and air conditioning systems. The faults of sensors may result in the decrease of system performance and waste of energy. It is not easy to discover the sensor bias, since its occurrence is always random and unpredictable. The data-driven based evaluation logic is proposed to assess the measurement reliability of sensors in the refrigeration and air conditioning systems. The subtractive clustering is presented to classify and recognize the various operation conditions adaptively. The principal component analysis models constructed upon the known conditions are developed to detect the measuring faults of sensors. Two statistics of T2 and SPE are combined to evaluate the measurement reliability of variables, which are divided into monitoring-type and controlling-type according to their attributes in the control loops. Ten fault cases, which include the fixed and drifting biases of various temperature and pressure sensors, are tested in a real vapor compression system.