A Fault Detection And Diagnosis Method Research For Control Valves Based On Support Vector Machine In Air Conditioning Water System

Air conditioning system has played an increasingly important role in people's living conditions. However, there are several defects in air-conditioning system, which increase energy consumption in architecture. A great many of buildings meet abnormalities in lifecycle due to all kinds of faults, which always lead to tremendous energy waste. Control valve is the key to flow control in air-conditioning system, sometimes even provide a fundamental protection to the chiller and heating source. When control valve fails, it may cause serious imbalance of chilled water system and put chiller, heating source and air conditioners into a risky condition, which can induce a reduction in service life. Such result not only jeopardize the safety of whole system and indoor thermal comfort, but also create unnecessary economic losses owing to the increase of maintenance fees and energy consumption. Thus, real-time control valve running status monitoring, and using reliable fault detection and diagnosis strategy to prevent the occurrence of failure and ensure the efficient operation of the control valve has vital practical significance.The fault detection and diagnosis method for control valves based on support vector machine(SVM) presented in this paper are consist of two parts: offline model training and online valve fault detection and diagnosis. This method use support vector machine theory to develop performance characteristics reference. The exponentially weighted moving average control(EWMA) chart is introduced to reduce the type ? errors in traditional t statistic method for a more accurate fault diagnosis. This paper takes differential pressure bypass control valve for example, from which 6 classical faults are chosen to exam FDD methods. After data processing, the SVM based model trained with input parameters, which are bypass flow and controller output, and fault characteristics, which is manipulate value, has been compared with simulation result calculated by Flowmaster.The new SVM based control valve fault detection and diagnosis method is respectively validated by simulation training set and test set data from Flowmaster model on a basis of real building project. The statistical results show that combined with EWMA control chart, the SVM based control valve FDD method is proved to obtain better accuracy and reliability.