Construction’s Misguided ‘Model Based’ Ideology

Paraphrasing Country singer Johnny Paycheck, “Take this Model and Shove It.”

Yes. You heard me correctly. But just to be sure, let me rephrase it another way….”F_ck the Model“.

There. That should do it.

That may seem a bit of an odd thing to say. Especially for someone like me. After all, I’ve spent most of my career working with technology in Manufacturing and Construction. It’s been good to my family and I. But like many things, it can go too far. And it has. It’s time someone stands up and says something.

Why Do You Oppose 3d Models?

Why am I opposed to 3d models? Simply put, I’m not.

I’m opposed to idealistic visions of a utopia where everything can be solved with a 3d model. That may be the case one day, but it won’t be in my lifetime. Technology moves fast. But we’re also a long way off from promises made even a mere decade ago. The value of creating many 3d models does not overcome the cost to generate them (or maintain them).

There’s a lot of reasons NOT to model things. I won’t elaborate too deeply here. You’ll either get it or you won’t. But to summarize, here’s some of the major reasons to NOT 3d model something…

  • No time / schedule
  • No resources / staff
  • Lacking tools / technology
  • Less efficient / takes longer
  • No value / creates waste

If you disagree and think we should 3d model everything…like right now today, consider you might be part of the problem.

The 3d Model Vision

There’s a lot of folks who tell you how things “should be”. Digital Twin this and that. Everything will have a perfectly pristine working digital clone. Identical in every detail.

For the most part, I agree with them. However there’s an economic ecosystem at play. It’s beyond the control of a mere few wishful thinkers and prognosticators. It takes a while to turn around an industrial complex the size of constructon. So until modeling everything adds value offsetting it’s cost, some things (many things) should never be modeled. Ever.

Some will say this is because the industry full of laggards. Those reluctant to change. But look around. The skyscrapers, the roads and bridges, the dams and monuments. Do they look like they were constructed by laggards? Next time you’re in a big city, walk up to a construction worker and call them a laggard. See how that goes for you.

It’s also important to keep in mind how things “are” didn’t just happen in an ignorance vacuum. Things evolved for a reason. And until those reasons go away, we’re left with what “is” not “what should be”. And if things haven’t changed as you expected, consider the problem might by more complex than you give it credit. Those ‘laggards’ as they’re called, just might know something you don’t.

Resistance to Overboard

3d Modeling in manufacturing preceded common use in construction by about two decades. Same with concepts like PLM which is manufacturing’s ‘BIM’. Lean? Started in manufacturing (Just like me).

I pushed hard for 3d modeling back in one of my old companies. Endured the eye rolls and comments about how “3d is for boxes, not complex things like the warped surfaces” which they were doing in 2d AutoCAD. I also promoted Revit for a manufacturer of construction materials before Autodesk acquired the technology.

In each of those cases, when they finally saw what I was trying to convey, they over reacted and went too far. Complete abandonment of all 2d even where it made sense and 3d had no added value. They even attempted to do manufactured piece part modeling in Revit where Inventor or Solidworks was better suited. What was need was nuance and a blended approach. Instead the result was a binary shift. Classic throwing the baby out with the bath water. Lessons learned the hard way.

Sometimes when you apply too much force to a seemingly immovable or stuck object, it’ll eventually break free and go beyond it’s acceptable tolerance. If you’ve ever busted a knuckle trying to remove a stuck bolt while working on your car, you know what I mean.

3d Modeling has went that same way. Gone too far. If you ask me, as an industry are suffering from bruised, broken and bleeding knuckles.

The Problem With 3d Models

So what’s wrong with the current 3d modeling approach? Absolutely Everything.

No, not THAT ‘Everything’. There’s a lot of 3d modeling that creates incredible value and eliminates waste. It greatly exposes risk before cost is incured. In those cases we should do more not less.

The ‘Everything’ I’m talking about is that we’ve somehow moved into this unattainable, low intellect thinking that 3d modeling is the answer for….well, just about EVERYTHING. Revit projects have become nothing more than an erotic orgy of data vomit.

Oh look…corn. I don’t remember having corn. Is that a piece of baked potato? A date dump from a 3d model often results in forensics to even determine what the hell it is.

  • Need data? Add it to the model.
  • Nothing modeled? Model it.
  • Data changed? Open the Model and edit it.

Somehow the 3d Model turned into not only the single-source of truth but the ONLY source of truth. Merit for what to put in a model is typically controlled by somone with limited perspective. Can the computer can open it, navigate efficenetly and display/print sheets in an asthetically pleasing way? That’s the benchmark used today. There’s not a shred of data lifecycle cost or maintenance perspective let alone the cost and impact of those decisions up and down stream from the model.

This is absolutely flawed thinking. When you look at the data required for the built environment, there literally is no commercially available applications on the market today to help you manage this data. Hardly at the project level, and certainly not at the enterprise level. None. They’re all focused on the project model at it’s core….not project data. And certainly not enterprise data.

Yes, there are applications that help you manage portions of data or small genres of data. But nothing exists today which allows a company to manage their AEC ‘Data’ and leverage it across the enterprise. Once again I repeat, NOTHING.

Baseless Claim or Reality?

I’ve been thinking along these lines for some time. Years perhaps. I’ve been in User Groups and Meetups. Vendor Webinars. Industry Conferences. Hallway conversations. Far too many times I’ve heard “Model it” or ‘Add it to the Model”. No debate. No discussion. Just an accepted fact without any hint of value focused thought.

There’s a lot of examples in other industries that suggest alternate thinking. One is Manufacturing which typically leads trends in constriction by a couple decades. Manufacturing has Data Management systems and PLM (Product Lifecycle Management) systems. These are database systems that manage not only models but the data about the models. The key is that the data is associated to the model from an external database. They typically contain a lot of data that isn’t modeled. Anyone model Grease or Paint? A well implemented PLM system will tell you exactly how much you need and where to procure it all without a model of it.

Another notable example is GIS (Geographic Information Systems). A database is at it’s core, not CAD data. CAD data is merely just a small subset of data in a GIS system. Imagine a home changes owners. How does someone at the County get their GIS system updated? They update a database. What aren’t they doing? Opening CAD…Editing a graphical polygon with the owner information and saving the CAD file.

In both of these cases, external databases are linked to and help manage graphical models. The models don’t contain all of the data. And the databases usually contain MORE than just data for the models themselves. Unlike construction where we try to embed every conceivable piece of information right into the model itself.

I’m Not Alone…Anymore

As I said earlier, I’ve been thinking about this for a while. I frequently question my own sanity. Perhaps not as much as those who know me. But I do none the less. So imgine my surprise when I recently came across a couple articles that reaffirmed my thoughts.

This first is this article from AEC Magazine. ‘BIM is Bust’.

https://aecmag.com/opinion/bim-is-bust-how-should-aec-data-work-hok/

The next is the ‘Data Centric Manifesto’. You can read that here…

http://datacentricmanifesto.org/need/

To summarize, enterprise data is locked inside applications (and models). It’s created there and those applications serve as the gate keeper. Yes there’s Cloud services with API’s. But the data they host and control is still application/model centric. It doesn’t integrate with your data. If you want the data, you have to integrate with it. As such, it’s application/model centric. Not enterprise centric.

How Should We Use Models?

By examining how modeling tools came to be, we can understand how those tools should be used,

That start was as a drafting and documentation tool. At it’s core, creating drawings is the act of working out design decisions and documenting them. That documentation is then used to physically build whatever it is you were designing.

As tools progressed, they offered some natural enhancements. 3d helped us visualize not just during the design process but for others who weren’t skilled at visualizing in 3d from a 2d document. Parametric functionality also helped us build smart objects which helped us quickly make derivative designs. These functions also helped increase efficiency in the design process. We lessened the need to calculate small details and merely used smart objects to help build a design. They’re really ‘Digital Pre-Fabrication’.

Here’s one example: The following image is a conveyor support. It’s top and bottom width are variable. So is it’s height and the number of cross bracing panels.

Now look at the holes on the half circular mounting plate. Look also at the holes where the cross-bracing meets. It’s not in the center. What’s the length of that steel angle anyway?

Imagine trying to calculate all of this information from purely numerical data and no geometry. It’s no small task. The number of right-triangles and trigonometry required is quite complex.

If you need to build a few of these, 3d modeling is the perfect fit. The smart objects have made it a digital measuring tool. But what if you fabricated thousands of these? All different. Copying design files and making derivatives with new sizes gets to be time consuming. And the same 5 or 6 inputs drive all of them. And eventually, some of the parts will likely be duplicate as parts from one design just “happen” to match another. And what if the design needs to change? All those models as opposed to just regenerating them.

If you wanted to build a manufacturing system around this, you need a configurator and move data OUTSIDE the model. In this case, that’s what was done.

The assembly is 100% parametrically controlled from a spreadsheet. ALL data driving the model is driven from Excel from 6 inputs. From the length of the angle and position of holes. Even the size and shape of the structural steel shapes is driven from Excel. Here’s the formula for just the length of the selected cross-brace…

So what type of software tools do we have to manage this data at the enterprise level? Sure there’s product configurators but what about you parts library? Historical models and drawings? Common reference data defining shapes and configurations? Vendors and pricing? That’s where you implement data management or PLM systems in manufacturing.

What do we do with that data in Construction? Embed it in a model. Editable only by a Revit/AutoCAD/Microstation user. Locked up and application/model centric.

If you ask me, the solution was obvious. Or so I thought.

The Solution Should be Obvious

One of our technology vendors was pushed hard from many customers like us. The ask was to handle non-model based workflows and data. Their solution? A tool to model data in the field. They built some cool tech that will be helpful in cases, it’s not what was needed or asked. The data wasn’t modeled for a reason. So we didn’t need another tool to model it.

The real solution is to build a tool that can help us manufacture and build from data. No model required. Cutting of linear materials requires very little data as an example. Other operations like purchasing or assembly can be accomplished without a model too. Which means we can procure, manage pricing and labor, and even manufacture vast amounts of parts and data all without a model. None.

If we had tools that let us build WITHOUT a model, then where a model IS required it’s very simple to generate a model and/or link a model to that data. The manufacturing and construction process would have 100% coverage because it’s based on data, not a model.

When you use a model based system, you have partial coverage. You miss all non-modeled work and introduce a second workflow and processes. Also lost are analytics to view your entire operation.

One company I know literally publishes purchased items like buckets of adhesive as generic models. This is done so they can have all work go through their “model based” system and link to an ERP. Is that really what you want? It’s yet another workaround to compensate for lacking tools.

To summarize…

If you build tools that work only with models, they’ll only work with a model.

If you build tools that work with only data, they can also use models because a model is a shortut to data.

Now all we need is someone to start building “Data” centric tools.

Shop Math 101: 1/100″ = 1/4″

Does your fabrication shop lack confidence in your drafting/detailing department?

Have you struggled to get buy-in when trying to roll out new processes, technology or deliverables?

Did you wonder why? More importantly, do you KNOW why?

Getting to the Root of Trust Issues

My entire career spanning manufacturing to construction, fabrication shops have had trouble trusting the information they’re given. And there’s good reason. It takes time to master a domain and learn the work. And production staff are busy building and fabricating. They don’t have time to run into the office every time something is wrong. As a result, office staff take longer to train and often persist with producing lower quality work.

But there’s also a lot of reasons that are not good. Downright bad in fact because they’re simple to resolve. Things you’re NOT doing wrong but are causing problems. These trust issues are easily corrected if they’re understood properly. One such issue is fabrication tolerances that I categorize as Shop Math.

The Dynamic Between Tolerance and Rounding

I’ve had trouble explaining this verbally so I figured a more graphic (yet generic) representation would be in order. In basic terms, you need to use a rounding factor 1/2 the amount of your fabrication tolerance to achieve the desired result. As the title of this post suggests, a value of 1/100″ of an inch, can result in a deviation of 1/4″. That’s real Shop Math in practice.

My example uses both fractional inch and decimal to more clearly illustrate the point. You don’t want to get me started in why everyone should use decimal, that’s another post. But decimal also has the same dynamic, it’s just less hidden and more easily fixed. You get a lot less pushback in a shop by adding an extra decimal than by changing the denominator of a fraction to a number the shop says they don’t fabricate to.

Let’s take the following example..,

In this example, lets assume our tolerance is 1/8″ (construction field tolerances, that’s fairly common).

The TOP RED dimensions are all ROUNDED to the nearest 1/8″ to match our target Tolerance. A fairly common practice. The BOTTOM GREEN dimensions are all ROUNDED to the nearest 1/16″ which is half our target Tolerance.

Each graphic shows a line 6″ long. Half that is 3″. We’re going to make a gap and dimension from each end to that gap. Maybe it’s a weld joint in pipe or perhaps a mortar joint in concrete block. Doesn’t matter what to illustrate the problem.

For the graphics on the LEFT (Quadrants 2 & 3), the gap is just shy of 3/16″ of each side of middle to forced the dimension to round UP. For the graphics on the RIGHT (Quadrants 1 & 4), our gap just heavy of 3/16″ of middle to force the dimensions to round DOWN.

Rounding = Tolerance = Confusion = Mistrust

Let’s focus on the TOP RED in the following illustration…

If you take the two parts and add them, they vary by 1/4″. Adding 3/8″ (our rounded gap size) to either of them does not equal 6″ either.

Quadrant 1: 2 3/4″ + 2 3/4″ + 3/8″ = 5 7/8″
Quadrant 2: 2 7/8″ + 2 7/8″ + 3/8″ = 6 1/8

Between these 2 examples, a mere 1/100″ difference in our gap results in a 1/4″ difference and neither adds up to the 6″ of the total length. This is a 1/4″ TOLERANCE because we set ROUNDING to 1/8″ to match our fabrication accuracy.

If you ever wondered why your shop doesn’t trust your drafting/detailing, this is one reason. The Shop Math just doesn’t add up. They see the sum doesn’t add up to the whole and leads them to question the accuracy of your drawing and your staff.

Rounding = 1/2 Tolerance = Trust

Now lets focus on the BOTTOM GREEN portion of our illustration…

Between the left and right (Quadrants 3 & 4) we’re still making the gap just shy and just heavy of 3/16″ from the center. As you recall, we said our fabrication tolerance target was 1/8″ but here, the dimensions are ROUNDED to 1/16″. This is HALF of our target fabrication Tolerance.

Here, it doesn’t matter of the dimensions round UP or DOWN due to the slight variation in the gap, the dimension are the same. Furthermore, if you add the parts, you get 6″.

Quadrant 3: 2 13/16″ + 2 13/16″ + 3/8″ = 6″
Quadrant 4: 2 13/16″ + 2 13/16″ + 3/8″ = 6″

Same Geometry – Different Clarity

As you can see, we had different results between rounding UP and DOWN when our ROUNDING value equals the Tolerance we’re trying to achieve. When we round to HALF the Tolerance, those small variations are masked and all our numbers add up.

So if you’re dimensioning for the shop, it’s important to realize this little change can mean questioning or trusting your data and staff. Additionally, if you take the time on the shop floor to explain WHY they see these differences, they quickly realize that the information that leads them to question the data (and your people), is also the very same data that’s most likely to be wrong. What IS accurate and didn’t change, is the geometry itself. This is an extremely critical point to highlight if you’re trying to get your shop to use automation and drive machine tools from CAD/BIM geometry.

The model/geometry, is the MOST right data we have. It’s just not human readable and what we provide as human readable is prone to errors such as these. This is one reason you’re seeing terms like “Model Based Enterprise” starting to float around in the Manufacturing space. It’s also a reason you’re seeing more shop go paperless, eliminating dimensions when possible by leveraging automation.

These efforts can be challenging and often require a leap of faith. But if everybody understands dynamics like this, it can be extremely helpful in moving all of your staff to more digital workflows. Because they trust the geometry and you eliminate what’s confusing them.

Intolerance of Tolerances

In a recent LinkedIn post, the topic of Tolerance Stacking was brought up. I’m not a machine designer, but I’ve spent a lot of my past life in Manufacturing. In that world, the term was used frequently. If the term was used in Construction, it certainly wasn’t when I was listening.

Tolerance Stacking can be described (in my mind) as the accumulation of allowable tolerances to a point where the design is no longer suitable for it’s intended purpose. Errors resulting from Tolerance Stacking are caused by a few things…

  • Lack of tolerance awareness
  • Poor annotation and documentation of tolerances
  • Both of the above

Tolerance Stacking Explained

The best way to understand Tolerance Stacking is from a few examples. In our first example, we see a part 10 Units long with 9 holes, equally spaced 1 Unit apart. Take note of the RED dimension on the right.

10 Unit Long Part with 9 Holes Spaced 1 Unit apart

You may have seen parts dimensioned like this. Looks pretty normal. Now lets consider this same part and assume the dimensions have a tolerance of +/- 0.0625 (1/16 Inch). Now lets also assume that all the dimensions are in the negative -0.0625. The following graphic illustrates this condition. Again, notice the RED dimension on the right.

Tolerance Stacking using the an allowed -0.0625 on each dimension.

Is the overall length really have enough tolerance to compensate for the accumulation of those tolerances?

Now lets look at the same part, same tolerances but annotated/documented differently. It’s not as “pretty” and takes up a lot more real estate on your drawing.

Same part as before but dimensioned differently.

But lets look at that same -0.0625 extreme case tolerance in this scenario. Once again, keep an eye on that RED dimension to the right.

Using an alternate annotation approach solves the Tolerance Stacking problem.

This latest example solves the Tolerance Stacking issue by clearly outlining where the tolerances are allowed. In fact, in construction, we’re already doing this. We just don’t call it Tolerance Stacking.

In construction, one of the ways we eliminate Tolerance Stacking is by dimensioning to gridlines and columns. Dimensions in relation to known fixed points minimized Tolerance Stacking.

Are you old enough to remember when rafters were layed out by hand on the job site and cut individually? You would cut one and use it as a template and use that to mark the others. You never installed the template and used the next cut as your template for another. This minimized Tolerance Stacking as well.

Geometric Dimensioning and Tolerancing – GDT

What’s less familiar, is another concept used heavily in automotive and other precision manufacturing. It’s called Geometric Dimensioning & Tolerancing or “GDT” for short.

Traditional linear tolerances have flaws. GDT on the other hand more accurately describes “features” and allowable deviation from the desired location using a more complex form of graphics and symbols.

Once again, the best way to explain this is with some illustrations. The following example shows a square part with a hole in the middle. Pay close attention to the RED dimensions.

Top Left – Perfect Part (not real world)
Top Right – Hole moved 0.0625 to the right
Btm Left – Hole moved 0.0625 up
Btm Right – Hole moved 0.0625 in both directions

In this example, you see when the hole is moved to the maximum tolerance in both directions, it’s actually further away from it’s desired position than 0.0625.

This is where GDT comes in. In this last example, GDT is used to “Describe” the allowable deviation from it’s ideal position.

GDT can more accurately describe tolerances.

There’s actually an ASME Standard for GDT (Y14.5.2) and a full explination of GDT is not only beyond the scope of this blog but my knowledge, There’s a lot of courses out there specifically for this but a good “101” description can be found here.

Given trends in Prefab, Modularization, and Construction becoming more like Manufacturing….Makes you wonder….should there be a “GDT” style of documentation for construction?

Tag Opposite End of Fitting

In CADmep, using the Size command, you can tag the size of a fitting. But on a fitting like a transition, what if you wanted to tag the size of the opposite end?

This can be easily done but the sequence is a little nuanced. Type “SIZE” from AutoCAD’s command prompt or select the “Size” tool button on the CADmep toolbar.  

When prompted to Select Objects, select the fitting. Once the fitting is selected, instead of pressing <Enter> to end the selection like you normally would, press and hold the <TAB> key while you press the <Enter> on the keyboard at the same time. Your tag will display the size of the opposite end.

Depending on your AutoCAD and Mouse settings, right-clicking to end the select objects prompt may not work and instead bring up a right-click menu. For this reason, it’s recommended you use the <Enter> key on your keyboard while pressing <Tab>

The following video shows to transitions of the same size side by side. The left transition has it’s size tagged like you normally would. The fitting on the right, the <Tab> key his being held down when the <Enter> key is pressed which results in the tag displaying the size of the opposite end of the fitting.