The journey of SOURCE
30 October 2020
Maarten van der Seijs, expert on dynamic subsctructuring & co-founder of VIBES, worked many years on innovating NVH engineering. Now he designs software that allows engineers to really understand TPA and to perform reliable FRF measurements in a user-friendly way. Read more about the journey of SOURCE:
My first encounter with Transfer Path Analysis was during my PhD, co-organized by TU Delft, TU München and the BMW Group. I had always held a firm interest in the dynamics of structures, and to me the relevance was in the vibration modes, resonances and transfer functions (FRFs). Up until then, I never cared much for the excitation forces or vibration paths: I figured, the magic happens in the passive dynamics.
As part of the project I was told to find so-called “blocked forces” on a steering system, to be measured on a test bench, that would cause a similar response on the inside of a car. Standing under that car for the first time, I was wondering where I should put all these shakers to apply the blocked forces, and when that time would come, if I would have the slightest clue what I’d be doing. Coming from the school of Daniel Rixen and Dennis de Klerk, I was quite familiar with the principles of Dynamic Substructuring, which seemed to have an explanation for just about anything that is structure-borne and linear. So, it did occur to me that blocked forces are just a special case of it, and that all these shakers are really not needed if you can just use FRF functions to predict the same response.
However, both explanations didn’t seem to resonate well in the round-table discussions about “TPA”, which apparently I had become part of due to the nature of our activities. Clearly, that world of TPA was much bigger and needed diving into.
A unified approach
In the years to follow, we have explored numerous methods that loosely relate or refer to TPA. I’ve always felt that understanding Substructuring principles helps to see things more clearly, or at least more generic and structured than the TPA variants often got presented in their seminal papers. Seen through “substructuring glasses”, most of the methods would come down to combinations of an active and passive subsystem, possibly some mounts and a well-chosen set of sensors or “degrees of freedom” with particular roles.
At the same time, we experimented with several test bench concepts to get to better blocked forces: rigid, flexible, with and without force sensors. We showed that key to all of this was my Virtual Point Transformation technique derived from Dennis’ EMPC technique some years before. Indeed, it establishes the strict condition of a fully blocked interface, but also to actually have a common interface to transfer force descriptions between different assemblies. In a nice turn of events we ended up at an alternative derivation of a blocked force method better known as the in-situ characterization method by Andy Moorhouse et al. It turned out that source characterization was again another term for similar techniques.
We decided to unify these methods into a general framework, including as many methods as we could find. We chose to categorize them into the families of classic, component-based and transmissibility-based TPA, depending on the way the source was described.
Since the start of VIBES technology in 2016, we got to work on many more large-scale TPA projects, on anything ranging from a coolant compressor to full-vehicle tire noise and often applying several methods from different families. Quite commonly we would use OTPA as a first quick-scan after measurement, blocked forces concepts for independent source characterization and classic-TPA interface forces to express path contributions. Using DIRAC for reliable FRF measurement in conjunction with our VIBES MATLAB Toolbox for the TPA calculations, a process arose which lent itself for re-application in other projects.
SOURCE: the software that speaks TPA
With the development of SOURCE we’ve taken this a step further: SOURCE not only provides the tools to process source characterization and our common TPA framework efficiently, we wanted to make the software such that it really understands TPA! To reach this, we have created software abstractions of TPA concepts, using the same semantics as they appear in the TPA framework.
Now, SOURCE understands how combinations of subsystems and measurement channels lead to certain types of characterization, how active and passive-side vibrations need to be interpreted and how a noise floor measurement can be used to optimize the computation of blocked forces. On top of that, we wanted to create an application that tackles some of the common obstacles of current-day analysis tools. SOURCE is optimized to handle large amounts of data efficiently, whether they are stored locally or on a network or cloud location. The user interface is streamlined, but not too restrictive, such that it also allows the user to try and apply new methods.
SOURCE appeals to the engineer by being clear about what the datasets in terms of channels, length and representation. We have taken inspiration and combined principles from application we love, not only from NVH software but also MATLAB and Excel, audio & video editing workstations and big data analysis. With the release of SOURCE we are proud to offer you this complete experience to TPA and are thrilled to continue building new tools to innovate NVH engineering!
Interested to find out more about SOURCE? Please get in touch and we’ll be happy to tell and show you more!