Sound Quality Evaluation of the Interior Noise of a Tractor HVAC System Based on Prediction Model

This thesis investigates a comprehensive evaluation of sound quality within the interior noise of a tractor HVAC (Heating, Ventilation and Air Conditioning) system using a predictive modelling approach. The research follows a systematic methodology that includes literature review, signal acquisition tests, psychoacoustic parameter analysis, subjective evaluations, and the development of a sound quality prediction model. During the literature review phase, relevant papers on psychoacoustic parameters, subjective noise ratings and noise prediction models were tabulated. Cabin noise was recorded at the operator's ear level using a 19-channel microphone, an artificial head with binaural headsets, and an omnidirectional microphone. Psychoacoustic parameters such as loudness, A-weighted sound pressure level (A-SPL), sharpness and roughness were calculated. The subjective test involved the rating of recorded sounds using two methods: a 1-10 rating scale measuring annoyance and a semantic differential table with paired bipolar adjectives related to loudness, A-SPL, roughness and sharpness. The sounds were presented to the subjects in two ways: binaural listening with a computer and spatialised audio listening (3rd order ambisonics) with a VR headset simulating the tractor cab. Multiple linear regression was used for the prediction model. The 1-10 rating scale showed better results in predicting noise annoyance with an R-squared of 0.96, while the Semantic Differential Method (SDM) showed an R-squared of 0.82. Despite the lower R-square for noise annoyance, the Semantic Differential Method (SDM) demonstrates its utility in predicting psychoacoustic parameters, achieving strong R-squares for roughness (0.94) and loudness (0.87). However, sharpness could not be reliably predicted due to significant errors and discrepancies in subjective ratings.