Study Of Fault Detection And Diagnosis Of Air Conditioner Based On Inverse Modelling Of Air Conditioner Faults And Machine Learning

Due to the limitation of design and manufacturing,there are various faults happening in air conditioners.The energy wasted by these faults is vast and common.Many researchers have paid attention to the fault detection and diagnosis(FDD)of faults in air conditioners.Due to the cost of experiments,the data generated by experiments is limited and FDD tools based on machine learning can easily overfit the training and testing data.This research builds an air conditioner simulation tool according to the inverse modelling of the air conditioner.Besides the FDD protocol based on machine learning,this research also proposes a two-step validation process of the FDD development.Based on the idea of inverse modeling,this research builds the simulation models of evaporator,condenser,liquid line,restriction valve,compressor,suction and dissipation line by using moving boundary method and partialwet-partial-dry model.This research establishes the object functions for regressing the parameters in each component model.The accuracy of the models are guaranteed by data filtering and improved by experiment data clustering.PSO is used to regress parameters.After building different component models,the air conditioner simulation tool is established.The experiment results and simulation results show good match.At last,this research comes up with the idea that the normal FDD d a new fault detection protocol based on support vector machine and fault diagnostic protocol based on PSO-BPNN.By using Pearson Coefficient Analysis,each fault diagnose is designed to have specialized inputs which are sensitive to each fault.This research also proposes a two-step validation process of the FDD development.The FDD protocol proposed in this research is tested by the fault data of evaporator fouling,compressor valve leaking and liquid line restriction increasing generated by air conditioner fault simulation model.By this way,the adaptivity of this tool is validated under different working conditions.