Transfer Learning for Tonal Noise Prediction in VRF Units Using Thermodynamic and Vibration Signals
Researchers have developed an unsupervised transfer learning method, Domain-invariant Partial Least Squares (Di-PLS), to predict tonal noise in VRF units. This approach utilizes thermodynamic and vibration signals to forecast noise levels under varying conditions. The study found that vibration signals provided more accurate predictions than thermodynamic signals, with prediction errors within 3 dB. AI
IMPACT This research could lead to more accurate noise prediction systems in HVAC and other machinery, improving product design and user experience.