Using MBSE for abstracting existing product solutions to generic solution matrices

[Jeremias Friesen]

Hi everyone,

for my master thesis I am tasked to look at existing product solutions for a certain product and based on that develop generic solution matrices which can be used to quickly develop new designs to show to potential customers.
For the past two weeks I got familiar with MBSE and SysML as the modeling language but I am kind of dumbfounded on how MBSE can be used to develop generic solutions. 
Since my supervisor for the thesis required me to use MBSE for that task, I just figured this was a common use case for MBSE but I have never seen it used like that.
Can any experienced user assess the feasibility of the utilization of MBSE in that context or is it not a good tool for that? And if it is, how would an implementation of MBSE look like roughly?

[Philip Oakley]

"... or is it not a good tool for that? " - As always, you end up with the aphorism "a fool with a tool, is still a fool" :wink:  Here the exercise is to learn about, and gain insights, into the value and limitations of 'MBSE and SysML' as tools [that support "Thinking"!].

As an example I would start with the almost ubiquitous SE model example of the car / automobile, and ask 'why is that the preferred example?'. Apparently Henry Ford said that if he'd asked the customer they'd have asked for 'faster horses'. So why don't the examples ask for 'Horseless carriage' models? or Personal transport vehicles (PTVs) [cf hovercraft, helicopter, speedboat, microlight aircraft, spaceship to Mars; are they all PVTs, so what is the top level PTV model?]

You didn't mention the discipline for your Masters, so taking it wide [engineering], reflect on the distinction between electrical bulk component modelling (resistors, capacitors, inductors, and perfect wires that join them) and the continuous distributed model of Maxwell's equations (no partitions, no blocks, etc).

If your Masters is computing related, reflect on the matched use of BNF (Backus Naur format) for programming language specification (C, Fortran, etc), as compared to free format languages such as FORTH (great language!), and which can be 'specified' in MBSE and SysM tools.

I hope I haven't answered the question.. but provided enough thoughts to allow you to investigate SE tools, their uses and abuses. Start by listing all the tools and their lead tutorial example...

PS. never tell 'management' it is a 'model' - its' always simply a diagram - it's all about priorities and communicating ideas [the 'never tell' anecdote is borrowed from Jon Holt who wrote one of the UML books]

[Jeremias Friesen]

Thanks for the elaborate answer!

I get what you are saying, the solution to a problem should ideally not start with the methodology to use. However in that case, using MBSE is a precondition as it is the main research focus of the master thesis. The problem of creating a solution matrix was provided by the company I work with and the methodology was proposed by the university that looks after the thesis. So I guess I have to "abuse" MBSE a bit, but I think I will find a way.
My idea is to model the system (I can´t name it because of company policy, but it is a subsystem of automobiles) on a concept and logical level and based on that create a generic solutions matrix. If the main insight of my research is to not use MBSE for that purpose, that´s still a valuable insight right? ;D

I´ve also heard that solution matrices were implemented in SysML-Diagrams, but I would have to research on that a bit more.
For now, thanks again for your input :)

Cheers!

[Walter Roscello]

Perhaps I'm not interpreting correctly, but it sounds like you may be describing reference architectures using a different terminology.  As previously stated, SysML/MBSE is the way to convey the model, but it does not give you methodology or product management on its own.  I suggest you look into Product Line Architecture, and modeling of variations, both in the standard MBSE references and the field in general.

Based on what you describe, you may be looking to model the functional and logical commonality among solutions in your domain, then you can do a specialization for each specific solution to add additional behavior and attributes, and/or allocate those reference objects to objects in your physical models for each solution.  In that case the key would be to analyze each logical component for what it does, rather than the form (and technology) that it takes.

[Jeremias Friesen]

Hi, I´ve just now looked into Product Line Architecture and found a really good article that describes how MBSE can be used to follow a PLA approach. This describes my problem really well, thank you for that! Funnily enough I just started modeling the coneptual and logical design/requirements for my problem, so I guess I already have a solid starting point. 
Thanks alot :)