Enhanced Fault Characteristics Information Based Canonical Variate Analysis for Detecting the AHU's Faults

As an indispensable component of air-conditioning system, the air handling unit (AHU) always exhibits the issue of inconspicuous fault characteristics information and strong time-wise dynamics. To address these challenges, the enhanced fault characteristics information based canonical variate analysis (EFBCVA) monitoring strategy is developed in this paper. Firstly, to obtain the enhanced fault characteristics information, the normal dataset is deducted from the test dataset and multiple batches normal datasets are subtracted from each other to obtain multiple training differential datasets, which together compose the three-dimensional training differential dataset. Then, the three-dimensional training differential dataset is rearranged into a variable-wise two-dimensional differential matrix. Finally, to manage the time-wise dynamic attributes, the CVA is utilized to capture the slowly evolving driving forces during the AHU's operation. The experimental results on the ASHRAE RP-1312 datasets show significant enhancements achieved by our proposed EFBCVA based monitoring operation approach.