Brent Dreher, NVH and durability manager at Faraday Future, discusses the importance of virtual data correlation with physical results for fruitful simulation.
Uncorrelated CAE simulations are like your morning weather report: you’ll get numbers, such as the day’s expected highs/lows, but on the face of it, do you have any idea how grounded those numbers are in reality? Despite this, we generally trust objective data at face value. Today’s object lesson in trusting numbers comes from the ongoing refinement of Faraday Future’s soon-to-be-released electric vehicle, the ultimate intelligent tech-luxury FF 91. This 1,050hp EV is loaded with advanced e-mobility technology such as sophisticated voice-control algorithms.
For all of this technology to function properly, and for maximum occupant comfort to be achieved, the cabin of the vehicle should be quiet and well isolated from all sources of noise and vibration. A source of noise and vibration in every EV is the electrically driven air conditioning compressor. In the FF 91 preproduction vehicles, the A/C compressor assembly is mounted to the adjacent structure using three rubber isolators. During early development, a nominal stiffness for these isolators was selected based on a quick-and-dirty (uncorrelated) CAE simulation. This simulation provided a target baseline stiffness that would minimize rigid body excitation and maximize attenuation.
Unfortunately, initial development testing of the A/C system revealed subpar NVH performance. The subjective issues with vibration and noise were also corroborated by objective measurements: the vibration and noise responses at the compressor first order and harmonics were objectively recorded using accelerometers and in-cabin microphones.
We knew we had to fix this problem right away but we didn’t yet know whether the baseline isolators were too soft or too stiff. To investigate further, we ordered tuning sample isolators in ±15% and ±30% stiffness to cover both possibilities. This simple development test was expected to zero in on the optimal configuration. But that’s when a funny thing happened – we measured the NVH performance of the various sample isolators, but surprisingly observed no improvement from any of the stiffness variants.
So, what was going on here? Was our isolator stiffness already optimized as per our CAE simulation and no further improvement was possible? Maybe, but to be absolutely certain, we expanded our experimentation to cover the fullest range of stiffnesses: at one extreme an isolator was entirely removed from the system, and at the other extreme the rubber isolators were replaced with solid aluminum pucks. That’s when another funny thing happened – both of these extreme configurations performed better than any of our fine-tuning samples, in the order of 10-15dB additional attenuation. The removal of one isolator (simulating an extremely soft mounting condition) resulted in some compressor mid-RPM NVH performance deterioration but a huge performance gain in the higher RPM range. The solid aluminum pucks (simulating an extremely stiff mounting condition) showed NVH performance gains across the entire compressor RPM range compared to the performance with rubber isolators. What a fascinating result!
Here’s the debrief: we thought we had a good starting point based on CAE simulation but development testing showed that we had in fact selected the worst baseline stiffness. Our initial development testing was narrowly focused on fine-tuning the isolators when actually we should have been considering a far more dramatic range of the stiffness parameter. Therefore, the lesson of the day is to always develop a proper design of experiments covering the widest practical range of parameter values, especially when one’s starting point is based on uncorrelated CAE simulations. We are now taking lessons like this to refine the tech-luxury FF 91’s NVH performance to astounding levels of quietness and comfort.
Going back to the morning weather report analogy, it seems a curious vulnerability of our minds that we sometimes accept objective data without fully vetting the methodology used to generate it. And I’m certain that when CAE started coming to the forefront of the product development process, many engineers felt that it would largely replace development testing. However, we see now that real-world testing is as crucial as ever for refinement of vehicles such as Faraday Future’s tech-luxury FF 91.