NSF EAGER, 2016-2017
Lead PI: Ravi Srinivasan with PIs from UNC Chapel Hill & G2 Inc.
We employ an unsupervised learning approach to model and encode the regular pattern in the acoustic time-series data, and to discover if a running HVAC system has deviated from its regular pattern of operation at any point in time. At first, the audio stream is converted to a stream of 50ms frames and passed through a frame-level feature extraction stage where Mel-Frequency Cepstral Coefficients (MFCC) are computed for each frame. Then k-means algorithm is used to cluster similar audio frames. Cluster assignment is used as an encoding for each frame. This step maps acoustic frames to k clusters. Finally, we compute the transition probabilities between each pair of clusters. Once the transition probabilities are in steady state, a sequence of unlikely transitions would mean that the HVACs behavior is unusual with respect to the currently learned model.
This work will radically reduce energy waste of centralized HVAC systems. In the U.S., centralized HVAC systems are used in 88.7% of all commercial buildings. In terms of energy, over 42.8% of commercial building energy usage is due to HVAC systems. In future, the proposed research can be applied to industrial systems such as commercial washing machines in laundromats re the total annual repair cost for all multi-load washers in the U.S. exceeds $2B.