A hybrid piecewise FDD strategy for refrigeration charge fault of airborne vapor-compression cycle system

Abstract

Many researchers have carried out study on refrigerant charge fault for variable refrigerant flow (VRF) air-conditioning system, while few pay attention to refrigerant charge fault for vapor-compression refrigerant cycle system (VCS) of enhanced vapor injection (EVI) with a flash tank onboard. To develop a strategy for refrigerant charge fault of the above system, the optimization methods have been implemented to predict the refrigerant charge level (RCL) based on virtual refrigerant charge (VRC) models. Firstly, Pearson correlation coefficient analysis is employed to obtain the characteristic parameters. Secondly, a piecewise modified VRC model with the different characteristic parameters from VRC models is obtained for two cases of RCL < 75% and RCL > 75%. Meanwhile, combined with the principal component analysis (PCA) and the support vector machine (SVM) algorithm, a PCA-SVM model is developed to improve the prediction accuracy, and the average root mean squared error (RMSE) is only 7.8 compared to 14.9 of the modified VRC model when RCL > 75%, meanwhile, the correctly detected ratio is improved from 62.8% to 84.5%. Then, for solving the difficulty of fault detection and diagnosis (FDD) when RCL > 160%, the moving time window average (MA) algorithm is introduced and an online PCA-MA-SVM model is developed. Finally, a hybrid piecewise FDD strategy combined with three models for refrigeration charge fault in airborne VCS of EVI with a flash tank is obtained.