Not all 3D Models are Created Equal

In light of my article 7 reasons to not get your whole facility remodeled, I wanted to do some more in-depth articles on a couple of the points I raised. In these articles, I will look at one point at a time and give the reasons behind my logic. These articles will be aimed more at project managers and people in management roles than 3D modelers so will be kept fairly general and (hopefully) clear.  Hopefully this will give some insight into the best ways to work with 3D data and save some folks some time, money and a lot of frustration in the process.

The first topic I wanted to cover is file formats. In my original article I skimmed over the fact that there are a lot of different types of files and that there were a lot of considerations when creating the files as to how they will be used. Here we will take a deeper look at what I was talking about and how or why this should effect your decision making process if you are looking to do a lot of remodelling form scan data.

It is worth noting here that 3D modelling is an incredibly complex subject due to the huge number of software packages out there that have 3D modelling capabilities.  These articles will be written with relation to doing and working with 3D models based off 3D laser scan data, also known as point clouds. As such, there may be some caveats or exceptions to some of these comments if talking about them with relation to different workflows or very specific software. But for the most part they hold true across the board.

Understanding the basics of how 3D modelling programs work.

It is critical to understand when talking 3D modelling programs and the models they produce that there are 2 basic kinds of 3D model. There is a Parametric model and what is referred to in modelling circles as a “dumb” model or “dumb solid”. The primary difference is that the parametric model is driven by parameters in the model like dimensions and quantities where the dumb solid is purely a 3D representation of whatever you want to display. The Parametric models are easily modified where dumb solids are often not able to be modified at all or are very difficult to modify.

The other big distinguishing factor is that Parametric models can have a lot more information in them like material types and properties, surface finishes and colours, attributes like descriptions, part numbers, stock numbers and costs just to name a few. Most if not all of which can be easily called up in drawings or exported to spreadsheets for use in things like quantity take offs and other reporting. Like wise it is possible for some programs to link to spreadsheets like design tables or calculation sheets so you can drive your models from the spreadsheet based on the design criteria and numbers it generates.

When it comes to 3D models Parametric is king and is the way most dedicated 3D design and drafting programs work.

Most, if not all, parametric 3D modelling programs can import dumb solids and work with them. Most of them can even import files from other parametric programs and work with them. But here is a big catch, almost no program out there can import a file from another program and work with it as though it had been created natively with that program. Some programs can sometimes get some of the information across but this process is unreliable at best.

The bottom line here is, 3D files created in one program are often not compatible with other 3D programs unless you only want or need a dumb solid. Meaning if you get a model generated of a point cloud, there is a better than average chance that it will be a dumb solid.

The other basic thing to understand about 3D modelling is that there are typically 2 ways the programs handle the 3D objects that they are displaying. They either do everything in one model/file or they have one file for parts and another file for assemblies of those parts. Typically the part files display one 3D object and the assembly shows how those objects are arranged.

Again this is a massive simplification of a complex matter, but the gist of the matter is that depending on the program you are using you either end up with one file that can be displaying anything from a single 3D object to hundreds of thousands of 3D objects or, 1 file for each of the 3D objects you want to display and then 1 or more assembly files to display how all the parts fit together.

The end result is either one huge file that is difficult to work with because even the most up to date computers will struggle with the file. Or a huge number of files that can be not only difficult to manage by your CAD and IT team, but fraught with dangers if you get that management wrong.

File Formats

As everyone knows, file format is mainly governed by the program you are using regardless of the type of software you are talking about. This is not entirely true when it comes to design and drafting software though. Most people know that a DWG file is a bit of an “industry standard”. Most programs that are used in design and manufacturing processes can open or work with a DWG. It is also true that most of them can output a DWG even if it is not a native file format for that program. Anyone that has opened a DWG file will know that even in 2D they can be incredibly complex things with layers, line types, layer colours, annotation styles and scales as well as model spaces and paper space.

While a lot of parametric modelling software can bring across a lot of the layer information, when it comes to exporting 3D models in DWG format they are ALL dumb solids with all the 3D objects in one file. Parametric modelling programs typically don’t/can’t work well with this kind of data. Especially if a DWG was created by a non-autodesk software and then converted to your parametric program’s format. There is a lot that can and does frequently go wrong during conversion processes. Especially of large complex data sets like production facilities. Typically nothing can be done about any of it.

Now that we have a couple of “ground rules” established lets have a look at how models are created of 3D scan data.

Typical workflow

Before we go any further it is necessary to look at a typical workflow of remodeling from scan data. Normally it goes something like this;

  • 3D scan the site
  • register the scans in registration software. Clean up of the point cloud for use with 3D modelling can typically be done in the registration software.
  • The workflows can go one of two ways here, they can either be worked with in the registration software to create the required 3D solids or the point cloud will need to be exported to work in a different program.
  • If the point cloud was exported to work in another program then it will need to be imported to that program.
  • Create 3D solids
  • Export to “generic” 3D format (Typically DWG)
  • Import into program to be used to do design/drafting.

Creating solids off 3D scan data.

I went into how this works in some detail in my previous article, so I won’t cover that again here, what it is important to know is this;

Most of the programs I have personally seen go about this one of two ways. You can get the program to automatically extract things like pipes, planes, cylinders (tanks) and structural steel. While this sounds like the ideal thing to do these programs get a LOT of false hits. Things like assigning pipes to corrugated roof/wall sheets and making planes out of things that shouldn’t be planes. They are getting better fast, but they still have a long way to go yet. The problem with this is that it can take a long time to clean up and check to make sure the things it got right are actually right. By default this is often expensive.

The other option is to extract the information manually. In this process the CAD operator needs to manually pick points from the point cloud and tell the program what it is meant to be. The program then does a best fit to the selection and surrounding points based on algorithms and user inputs to produce a 3D object. While this is a far more accurate way of working it is very time consuming even for experienced operators. By default this can also make the process expensive.

The really important thing to understand here is that ALL of these programs produce dumb solids typically in a DWG format. Meaning the parametric programs you are using to do your production facilities layouts/models won’t like the output data much.

If you are lucky when you get your DWG it will be broken down into basic groups like equipment, structural features and pipes. This will be done by either putting the different elements onto different layers or by X-References of individual DWG files with each of the elements grouped into them. Neither of these approaches is ideal for downstream workflows either. The main reason for this is that typically someone who knows the facility and your workflows well has to put in considerable effort into breaking the DWG(s) into smaller more manageable files that make more sense. The file(s) will be then be imported as one or a large number of dumb solids into your native design program. If the designer/drafter needs to manipulate the supplied 3D data in any way during their processes it can be slow, painful and ultimately expensive.

What about the generic 3D formats?

So by now anyone who has been around 3D modelling for a while may be thinking, “But i know we can convert files into our software using other formats, we have done it a lot”. This is true and there are a lot of “generic” formats that can be used like IGES, STL, STEP, and SAT just to name a few. This is where the whole conversion process can get really complex and we run the risk of getting to far into the technical side of things. So to keep it brief I will just say this;

This will be heavily dependent on your individual software and work flow, but some programs do work much better with some generic formats of 3D model than others. The biggest issue here is that all generic 3D formats are dumb solids and even if your program imports them really well, you can almost never manipulate them in any large scale meaningful way without a lot of hard work. So if you are going to get a facility modeled off scan data a DWG is often the best deliverable as you can at least use almost any program that will allow you to open a DWG and see/manipulate the layers and 3D objects to work with your data to then take into your native program. Most programs that work with DWGs will also allow you to export to a number of generic 3D file formats.

What about companies that say they can deliver in “Native format X”?

There are two things to consider here.

  • They are likely using automated extraction programs and then just doing the dumb solid to “Native format X” conversion for you in order to keep the processing time and cost to a minimum. Meaning you are still only getting a dumb solid, it will just be in your native format instead of a generic format like a DWG or STL.
  • If the company claiming to do this conversion is actually using your native modelling program to do the remodeling off the point cloud, there are so many issues that could arise with this that I could write a novel on it and still likely not cover every potential problem. When talking whole facilities though, the biggest thing to consider with this method is time. While it is possible and cost effective to extract a handful of details from the point cloud, to do a whole facility like this would take an extraordinary amount of time in most design and drafting software. As we all know time = money, so it will also likely cost a fortune. Then there are all the other issues, so it really isn’t worth the price.

To summarize.

  • There are 2 types of models, Parametric and dumb.
  • Parametric is what you will mainly be working with in 3D design and drafting software.
  • When remodeling from scan data, the resulting models are almost always dumb solids unless.
  • These solids, while useable, are often painful to work with, which makes them slow to work with, which means increased downstream costs in both time and money.
  • No matter how you have your model done, it will likely still be a dumb solid, or cost prohibitive to be a parametric model.

Any way you slice this it is not easy or cheap to get a whole facility into 3D.

What’s the best approach?

If you want to, or have to, for some reason (I still can’t think of a good one), go down this path of 3D scan to full 3D model, get the company doing it for you to deliver a DWG file. In that DWG ensure that the model is broken down so that each major group of components (discipline) is placed onto it’s own layer(s). The more specifically you can break up the model the better, but all mechanical on one layer and all structural on another is better than nothing.

However;

I have said it before and I will say it again. I will continue to say it until I see a change in the way things are getting done too.

Just work straight with the scan data (point cloud).

Most design and drafting software can import it and work with the point cloud(s). It is just a case of knowing to ask your scan service provider for the format you need to work with. (If they are any good they will already know.)

Some programs won’t show the point clouds in the 2D drawing environment at this stage. However, it is actually really easy to remodel the handful of bits you need to see in the 2D drawings yourself in your native format. The added bonus to this is that if it is done properly you can slowly add your projects together and build a complete parametric model of the facility. But this is a topic for a future article, so stay tuned for that one.

Share this with Friends and Colleagues:

Leave a Reply

Your email address will not be published. Required fields are marked *