Over the past decades, noise sources such as wind noise or engine noise have been significantly reduced leveraging improvements of both the overall vehicle designs and of sound packages. Consequently, noise sources originating from HVAC systems (Heat Ventilation and Air Conditioning), fans or exhaust systems are now becoming Tier-1 problems affecting quality and passenger comfort. Furthermore, existing experimental techniques are not adapted to internal flows and fail at identifying the location of noise sources, as well as corresponding design changes to reduce noise. This study focuses on HVAC systems and discusses a Flow-Induced Noise Detection Contributions (FIND Contributions) numerical method enabling the identification of the flow-induced noise sources inside HVAC systems. Moreover, this method provides the contribution of each source at the passenger’s ear locations considering the propagation of the noise through the system. This methodology is based on the post-processing of unsteady flow results obtained using Lattice Boltzmann based Method (LBM) CFD simulations (Computational Fluid Dynamics) combined with LBM-simulated Transfer Functions (TF) between the position of the sources inside the system and the passenger’s ears. By identifying and quantifying sources, this method guides engineers at treating the main sources in confined systems, usually a daunting task experimentally. In a first part, the accuracy of this numerical approach is proven by comparing the predicted aeroacoustics results to measured data at various operating conditions such as defrost, fresh air ventilation or recirculation mode. In a second part, the Flow-Induced Noise Detection Contributions method is used to highlight the location and intensity of the various noise sources for the different operating conditions. In the third and last part, limited modifications are made to the HVAC system geometry within design constraints , and LBM simulations are performed on the modified design to evaluate the achieved reduction of the noise levels.