Reverse Engineering...

[SchmidtMotorWorks]

The only difference, I understand we can now construct as many planes as we like from the captured data,
with as many points as we like on each plane,
And we can add more planes (even after digitizing has been completed) if we need more data for a complex curves.

Scanning more points than you need and projecting them onto a plane has been possible since the 70s.

BTW, the problem with RE is that you will get curves that are more complex than you want, rather than not complex enough. You want single segment splines with the fewest CVs possible (preferably 3 or 4). Where possible conics are prefered.

Ask them to post a finished port model.

[SchmidtMotorWorks]

By the way, Headman could use your STL file for his CAD/CAM shop-around.

That's a good idea, I'll agree to that if whoever wants to try will agree to return their best results in a neutral file format (IGES, STEP, Parasolid) surface model.

[headman]

OK, here's a challenge to any advocate of reverse engineering ports with automated reverse engineering software tools rather than parametric modeling of ports and machining them with an integrated CAM system.

We can start with a scanned STL model or I will provide a set of point data that is better than any scan you will ever accomplish on a port.
...

I have a couple of questions to clarify.

First you mentioned "automated RE software"
By that, are you refering to a cmm running a probing cycle to collect data, (like a downward spiral path through the port)?

After considering the pros/cons, I'm very interested in an articulate arm for data collection as opposed to a fixed cmm.


When you say "scanned STL model", what method do you advocate to collect the data? I do not know the acronym...
I've seen Stereo Lithography of a solid,
Even digital images of the actual internal port used,
Or do you just mean triangulation algorithms???


Please overlook the obvious fact that I lack background in design.
I think it is fundamental that I understand the method of scanning you are referring to so that I can grasp how you are even able to model the port.


It is starting to sound like there is less and less of a difference between your methods and what I am expecting from a solution.
The programs I am considering are parametric. The changes made are associative and there is no outside translator needed.

True, I will be waiting to see the final proof, as far as the CAM side is concerned.



Also I do intend to continue discovery to determine if Nx should be my solution.

[headman]

Ask them to post a finished port model.

Yes, I will, that is the plan, when I do have the opportunity.

In fact I expect to see simulation and a port actually machined at one of their customer's facility.

[SchmidtMotorWorks]

When you say "scanned STL model", what method do you advocate to collect the data?

In this case I had a foam model and used an optical scanner, only because I had quick free access to use it. The people that sell these machines claim -.001 accuracy but that is way exagurated especialy when you measure long distances.

What I mean by STL is a file format that represents a manifold solid with a large number od adjacent triangluar faces. Several people in the group I used to work in did their phd thesis on the modeling and optimization of those triangles. A good scanner attempts to record not only the corner points of the triangles (called facets) but also tries to determine the normal direction of the face at each point and compromise the data to get something between accurate and smooth. One of the biggest challenges for these problems is determining what the boundaries of the surface patches should be. When I model a port I spend the vast mojority planning what the boundaries and internal parameterization of the surface will be. I have lots of experience at this and every new port poses some new problems, to expect a software to do this is expecting magic.

If I have time I will make some simple examples that show how dificult the problem is to solve in a program.

Avoid methods that make lots of sections and then loft a surface through them. This method is even difficult to make smooth on airplanes where trhe sections are parralel, goinf around a curve is much more difficult.

You will find that editing the sections to make a new design is nearly impossible to do and get a smooth result in a reasonable time.

[SchmidtMotorWorks]

It is starting to sound like there is less and less of a difference between your methods and what I am expecting from a solution.

The main difference is that I would prefer to develop the port design in CAD and not with a die grinder. I would prefer to start with a scan of the cast port then model an idealized port relative to the cast port. That way you know exactly how much you are removing everywhere before you do anything to the head. You also know that your design is smooth without buldges and inflextions. Defects like this are much easier to see on screen with shape analysis tools than looking down a hole and feeling with your finger.

The critical hurdle is learning to model which I would say is probably less difficult than learning how to design and grind a good port with a die grinder. So yes in someways it is like starting all over and given that choice it would be wise to consider getting a medical degree instead.

[BritishTurbo]

Interesting thread this one... didn't see it till now though.. was too busy shopping over the weekend...

Below is a screen capture from an LS1 head that I am currently working on in Mastercam X.



Personally the surfaces look pretty good to me...
This file started out as many individual sections that were captured with out digitizing probe. Then the data was imported into Mastercam and manipulated from there.
Toolpath generation isn't really much of a problem... you just take your time to make sure it's gonna machine it how you want it to.

The main thing to realise is that I'd say 95% of the average head people will start with an already hand ported design that they know works, and then use that as the basis for the CNC model. Not many of these guys will start from scratch in designing the ports... This is my experience from dealing with people that do CNc head porting day in and day out.

Every CAD/CAM software package out there has it's advantages and disadvantages... and you can get most things done in most of the packages.

Just out of curiosity, what is the price of the NX4 software package?

[SchmidtMotorWorks]

Personally the surfaces look pretty good to me...

Hmmm, tastes differ, was the original port that lumpy?

If so, I think that supports my argument for modeling them.

If not, then I think that support my argument that scanned surfaces are crap.

[SchmidtMotorWorks]

Just out of curiosity, what is the price of the NX4 software package?

I will have to ask, I haven't checked in years. There are lots of optional bits so I'll find out what is needed to model ports and maybe CFD and CAM too.

[BritishTurbo]

Personally the surfaces look pretty good to me...

Hmmm, tastes differ, was the original port that lumpy?

If so, I think that supports my argument for modeling them.

If not, then I think that support my argument that scanned surfaces are crap.

Where is it lumpy? You can't make a judgement on if a port looks good or not without seeing the original port that was digitised... there are certain features that you want to keep.
The hand ported port was not bad to the eye, but as you pointed out previously I think, when you digitize that surface, you can sometimes see imperfections that weren't obvious to the eye. The port is hand ported and flow benched for a reason. You then digitize that hand ported port, and smooth out any imperfections from the die grinding process... and then they will all be as perfect as you want them from then on.

I don't see a problem with doing things that way... But I'm not a head porting guy who does this for a living... but they guys who do it for a living do it the same way... so who am I to argue...

I do see the benifits of using parametric modelling of the ports like you suggest... but that doesnt mean scanned surfaces are crap...

[BritishTurbo]

Just out of curiosity, what is the price of the NX4 software package?

I will have to ask, I haven't checked in years. There are lots of optional bits so I'll find out what is needed to model ports and maybe CFD and CAM too.

Please do, Mastercam X level 3 with multiaxis, what you need to do ports in 5 axis, lists for $17,050... for bang for the buck, I don't have a problem with it.

[SchmidtMotorWorks]

Where is it lumpy? You can't make a judgement on if a port looks good or not without seeing the original port that was digitised... there are certain features that you want to keep.

Send me the surface and I will show you with surface curvature analysis tools exactly where the lumps are.

To me that shape looks like what you would get if you cross-bred a port and a potato.

If you will send me the points, I'll model a port without lumps so you can see the difference.

[SchmidtMotorWorks]

But I'm not a head porting guy who does this for a living... but they guys who do it for a living do it the same way... so who am I to argue..

If the method used by pros is your criteria for determining the best method then you will have to join my side of the argument. The OEMs that make the best engines today, particularly Japan, threw the die grinder out years ago.

Just a couple of months ago I was at Nissan making example port models.
They were evaluating which CAD CAM PLM software to use for the future and chose UGS, it was announced just last week.

http://www.ugs.com/about_us/press/press.shtml?id=4378

A lot of racing teams are finally getting up to speed with advanced CAD modeling, Greg Anderson is an example that comes to mind (5 CNCs) using NX.

Cosworth uses NX, TRD uses a PMTC CAD product to model parametric ports. I worked at HPD and their ports weren't scanned even ten years ago.

[bobqzzi]

Thanks for all the great information gentleman.

Assume you are a small end user, and have a hand ground port that you know works well, and you'd like to make a run of heads for a customer. Am I right in thinking there are 2 basic approaches being sggested here?
u
1. Digitize the port with a probe, then have at it.

2. Get the basic dimensions of the port and model/design it it in the sofware.

Is the latter really an option for someone who will be having maybe 20 heads a year cut?

[headman]

When you say "scanned STL model", what method do you advocate to collect the data?

In this case I had a foam model and used an optical scanner...

That makes a lot more sense.

Maybe this is to simplistic, but early on (on other threads) I got the impression from your statements that you choose to build the port from scratch in your CAD program. I was unclear as to how you determined your geometry.

I am aware that data can be collected several ways,
but is this not still reverse engineering?