To address the multi-fault concurrency problem

which may occur in the actual operation of variable refrigerant flow air-conditioning systems

a multi-fault diagnosis strategy combining linear discriminant analysis (LDA) and random forest (RF) algorithms was proposed. After the completion of fault type identification

the best detailed diagnosis model was adaptively selected according to the fault type to further diagnose the fault level or determine the cause of the fault in detail. First

the original data set containing normal operating conditions

four-way valve failure

electronic expansion valve failure

and refrigerant charge failure was divided into a training set and test set with a ratio of 7:3. The training set was used to establish an RF algorithm-based fault type identification model. Then

the LDA method was used to reduce the dimensions of the three types of faults in the training set. The training set after the dimension reduction was used to establish a fault refinement diagnostic model. Finally

after the sample data in the test set were identified by the fault type

the test sample could adaptively input different fault refinement diagnostic models according to the recognition results. The results showed that the accuracy rate of the fault type identification model on the test set reached 99.99%

while the refinement diagnostic accuracy rates of the three types of faults were 96.12%

100%

and 97.44%

respectively. The results indicated that the strategy proposed in this paper could better complete multiple types of fault diagnosis tasks for different refrigerant flow systems.