Sensors mainly play monitoring and controlling roles in air-conditioning systems and affect their normal operation

thereby causing adverse effects such as increased energy consumption if there are faults in sensors. In this study

an improved principal component analysis method combining wavelet transform data optimization and neural network-based fault diagnosis optimization is proposed for the sensor fault detection and diagnosis in an air-conditioning system. By comparing the results of principal component analysis before the data optimization and the results of the principal component analysis after the data optimization

it was found that in the principle of the same 0.8500 cumulative contribution rate after the wavelet transform used to remove noise

the number of principal components was reduced by two; the detection effect was improved by 0.0207

0.0208

and 00415respectively; and the effect of the airflow sensor fixed deviation failure detection was improved by 0.1606. To find the source of the fault

the principal component analysis was used as the input of the neural network to test five sensor fixed deviation faultsbased on the wavelet transform and principal component analysis. The fault diagnosis results were 0.7667

0.8667

0.9000

1.0000

and 1.0000

respectively.