Realtime e-motor noise prediction for automotive attribute development in driving simulators

Find out more about this research below.

Introduction

Modern electric powertrains offer the opportunity for greener transport but come with their own unique challenges in meeting other vehicle performance attributes such as noise, vibration and harshness (NVH). Several decades of improving the NVH of internal combustion powertrains have led to unforgiving customers and expectations of luxury design, who are new to the highly tonal noise that is emitted by electric motors and resonances from inverter housings.

Optimisation of the design of the motor itself, its mounts and the way it interacts with the rest of the vehicle cannot be an afterthought and has to be considered from the start. To this end, predicting the noise emitted by electric motors at an early development stage is crucial for addressing NVH issues.

Working on the APC-funded, £39m Virtual Vehicle Integration and Development project (ViVID) in collaboration with Ford, IPG carmaker and Horiba-MIRA, academics and researchers at the Aeronautical and Automotive Engineering Department at Loughborough University, developed a real-time multi-physics model to predict the NVH performance of electric motors.

The model can predict motor noise as it is generated and can allow for quick what-if scenario to be tested by modifying model parameters and operating conditions, for example the motor load, its mounting method, the geometry of the motor casing or even the modulation scheme implemented in the motor inverter. 

This is a significant leap compared to current practice which largely relies on artificial noise, sampled noise from physical testing of existing motors or from costly and lengthy simulations that are performed in isolation and cannot account for changes in a wide range of parameters or operating conditions.

Method

Functional diagram of the e-motor NVH simulation process
Figure 1: functional diagram of the e-motor NVH simulation process: design data to simplified motor model / electromagnetic model / FE structural model to torque / modal superposition to noise vibration.

Tackling the problem of real time e-motor noise prediction required the adoption of a multi-domain/multi-fidelity modelling approach. A lumped parameter, systems-level electrical model of the motor allows for the calculation of the torque produced by the motor, which, in combination with a drivetrain/vehicle model, provides the angular acceleration and speed of the rotor shaft, implemented in real time Simulink. A separate, semi-analytical model predicts the electromagnetic forcing as a function of rotor position and the current through the stator windings. Finally, a reduced structural model of the motor casing is created using a subset of vibration modes from a high-fidelity structural finite element model of the motor casing. The response of the motor casing to the electromagnetic excitation is then solved in the time domain by modal superposition and finally the sound transmission is simulated using a simplified acoustic distribution of monopole sources with the correct amplitude and phase.

Motor casing illustration that contains the outer shell and inner magnets.
Figure 2: Motor casing.
FE-generated mode shapes.
Figure 3: FE-generated mode shapes.

The model was tested and validated at different stages using experimental data. For this purpose, a scale test rig was commissioned comprising two motors connected via their rotors, one being the driving unit and the other offering controlled resistance. The driving motor was powered using a custom-built motor drive that allowed for variable pulse-width modulation frequencies, while simultaneously controlling the resistance applied by the second motor.

Sensors were used to measure the speed of the motor as well as the voltage across and the current through the windings. A finite element model of the driving motor casing was created in Abaqus and validated by performing modal testing on the actual casing of the test motor. Following validation, the real-time model was scaled up to provide torque levels representative of a full-size powertrain and was integrated with a driving simulator to provide instant sound cues to the driver. 

Experimental setup for e-motor noise prediction
Figure 4: The experimental setup showing the laser tachometer, main motor, ancillary motor, power supply, signal generator, motor controller and data acquisition.

References

Gao B, O'Boy DJ, Mavros, G. Real-Time Sound and Vibration Modelling for Electric Motor, SAE Technical Papers. 31 Aug 2021.

Gao B, O'Boy DJ, Mavros G. High frequency modelling of electric motor vibration in the presence of adhesive bonded components, Internoise 2022 - 51st International Congress and Exposition on Noise Control Engineering, Jan 2022.