A response to blog “Got Brake NVH Problems?” by Dr Mo Esgandari:
On an educational visit to India, I was talking to a well-known company about upskilling of staff. The main concern expressed by the chief engineer was the company could readily employ young engineers who could “drive” the CAE systems but that they did not understand the background theory. That understanding was regarded as necessary so the company could “tune” the models to meet the company needs.
I often wonder whether we need to rename CAE as CDE – Computer Dependent Engineering. We have forgotten that the “A” means “aided” and as a result there is now a tendency to use computational techniques, rather than think about the task in hand. Worse still, there is a tendency for young engineers to believe the results. When did the philosophy “use appropriate technology” be discarded from the toolbox of our engineers?
As an educationalist (with a strong interest in the education of engineers to meet the demands of the 21st century), and a practic
ing design engineer, my biggest “initial” concern with computer aided design/drawing was that the student engineer ceased to think how detail parts were made. They were more impressed with colours, shadowing, and the ability to view a component in 3D. Up to that point the engineer could look at a drawing and visualise its 3D shape by reading the 2D drawing – that was part of their training.
One of the necessities not mentioned in the article was that all CAE models need to be verified.
With aerodynamic analysis (CFD) the car (or a scaled model) is placed on a 6-point balance machine and subjected to a variety of wind conditions. The laminar flow lines may be replicated using smoke trails and as a result separation readily observed. Even the wind effects around wing mirrors are analysed and the computational model optimised to actual events.
Similarly, FEA has the advantage of access to endless historical experiments and material data that may even indicate the magnitude of stress raisers. Standards have been produced to advise the radii at the root of k
eyways and gear teeth to minimise such concentrations of stress. That element of detail design is now lost in the ether of CAE. Theories of elastic failure cease to be understood and if questioned many engineers are unaware what a “Mises” stress means.
As such, the models are used in these activities with some degree of confidence and are readily validated and optimised. The packages provide valuable results which are readily accepted - and in general, without question. It is not clear whether “an analysis in order of magnitude” accompanies the use of FEA – indeed it is easier to accept the results without challenge.
The use of CFD, FEA and dynamic modelling is now commonplace because it may be used with confidence. This is not the case with brake NVH modelling.
One of the most serious problems with any NVH modelling is how to validate the results. In addition, there is insufficient reliable information available for effective use within the model and most explanations begin with “assuming”. Brake noise is fugitive in nature. Identical brakes may generate noise on differe
nt corner assemblies, under apparently identical conditions, but then not happen again under the same conditions.
In essence we lack a basic understanding of brake NVH and place too much faith in the modellers who have even less knowledge of brake NVH.
You will all recall the manufacturing director who ordered CNC machines and then had to use them for trivia jobs and indeed contract work simply to keep them employed – to justify that decision. Similarly, it is regrettable, but I believe to be true, that if our senior design engineers invest a significant amount of their budget in NVH models then (you can be assured) the junior engineers will have to use them, regardless. It is claimed that it is cheaper to use models than testing on NVH rigs and so testing ceases, and the source of realistic information dries-up. Is it cheaper to use CAE and get dubious results or test and get real results? Because the modelling is tending to replace rig testing, the younger engineers are removed from the basic thinking process of the basic causes of NVH (the DOE stage), leading to lazy thinking about why NV
H is such a persistent issue.
NVH modelling is not appropriate for brake systems – we simply don’t understand the problem sufficiently to use such models with confidence. The models cannot be verified for anything other than basic information. That is why we still have NVH issues.
When an engineer is faced with a system of connected parts, there are two options to predict its performance: 1) tell it what you want to know (modelling) or 2) ask it what it must give (testing). Use appropriate technology for brake NVH to accumulate sensible and repeatable data and only then move progressively towards modelling.
Professor John Fieldhouse is currently an advisor to industry and provider of short courses regarding braking and NVH issues. He
is a National Teaching Fellow and has for many years been a visiting professor at The University of Leeds - teaching chassis systems and vehicle performance. He holds a BSc from the University of Leeds and gained a PhD at the University of Huddersfield where he was awarded professorial status. John is a member of and an instructor in Brake Academy.