A Reduced Order Model for Prediction of the Noise Radiated by a High-Speed EV Transmission using Statistical Energy Analysis

Abstract

<div class="section abstract"><div class="htmlview paragraph">The transmission is an integral part of the driveline in an automotive vehicle. Global vehicle pass-by noise regulations are becoming more stringent and transmissions are expected to be very quiet. Typically for an automotive system, engine is the most dominant noise source and transmissions have been considered a secondary noise source but as the trend is shifting towards more electric vehicles where engine noise is absent and overall vehicle is becoming quieter, the transmission can be more of a significant noise contributor. Gear whine is the major concern for sound radiation from the transmission. The gear whine simulation and acoustic radiation analysis of the transmission using traditional methods (FEM and BEM) is a crucial but very time-consuming part of the product development cycle. On top of that, electric vehicle transmissions operate at higher RPM which in turn increases the excitation frequency arising from the gear whine phenomenon. Hence present work focuses on the development of system level reduced order model using Statistical Energy Analysis (SEA) which could take fraction of computational time compared to FEM and BEM and can provide quick design solutions such as changes in ribbing pattern, enclosure thickness etc and hence making entire transmission product development process leaner and more efficient. The entire geometry of the enclosure is divided into SEA subsystems, such as flat plates, curved plates and beams. The gear whine force is provided as excitation to the SEA model. This work includes the sensitivity analysis of all the parameters influencing the SPL. The results from the SEA method are compared with actual test data for final validation. The obtained results are within the limits of +/- 3 dB with respect to test data. On top of that, computational time taken by SEA is 1500 times lesser than deterministic methods (BEM).</div></div>