03 April 2014

Keep your BIM Model Real

Many of us only work in the world of our BIM models. The closest we get to something real is the drawings and schedules produced by our BIM software (even then we may never see them as physical objects, document issues are often via PDF or DWF).

Often we forget what we are doing is for the real world. That we are showing real people using real materials and real tools what we want built.

Another thing we tend to forget is that what we produce is going to be used by different people for different purposes.
An architect may only think his work is for showing what he has designed will look like, an engineer to explain the rationale of his solution. Sure, they might provide more information they think others will use, but not a lot of thought goes into it.

Conversely, those that use what has been produced don't understand why it doesn't explicitly cater for all their requirements.

None of these issues can be completely resolved. BIM software will never exactly match reality, authors will always privilege their own requirements, and a BIM model will continue to be used by multiple people for multiple purposes.

If it is always going to be a compromise what are the practical things we can do to address these issues?


An Example - WALLS

Within a BIM model walls have a number of purposes:

show what is to be built 

Walls are made from materials, with certain properties. There is an expectation there will be enough information to build them, that what walls are made from is identified, as is the order they are placed.

locate to be built 

Walls represent real objects that real people have to build. They need to know where it is located, with sufficient tolerance to make it humanly achievable.

boundary for analysis

Walls can be used as boundary elements for various software analysis. But each analysis will have different expectations. Not every wall may be required, the boundary may be at the edge of a wall or somewhere within it, particular information about what the wall is made from, or how it behaves, may be expected.

measure quantities 

Walls can be measured for costing and procurement purposes. Accurate area and volumetric information is expected.

construction schedule

Walls can be made up of different materials that are placed at different times. The expectation is that these materials can be separately allocated to different milestones.


I don't intend to go through every issue, instead I'll use accuracy as an example.
What are the expectations in terms of accuracy? It is not a two way street however, for example the cost estimator does not model walls, the architect does. So for the architect, what level of accuracy is reasonable?


TOLERANCES

We all understand a real person in the real world can not be as accurate a a computer. Let's put some numbers on that.

In our BIM software we can be precise up to 16 decimal places. This is often important to our software, but in the real (metric - millimetre) world that is accuracy to 0.1 femtometres, which is narrower than a proton (1.6 to 1.7 femtometres). If the unit is feet it becomes 30.48 femtometres, which is still smaller than the width of a hydrogen atom.


In the real world standards have been set to define what is reasonable. Some examples:

From the U.S. "Minimum Standard Detail Requirements for ALTA/ACSM Land Title Surveys 2011":
"The maximum allowable Relative Positional Precision for an ALTA/ACSM Land Title Survey is 2 cm (0.07 feet) plus 50 parts per million".
So the minimum degree of accuracy expected for land titles is 20mm.


From the Guide to Standards & Tolerances, published by the Australian Victorian Building Commission:
"Departures from documented external dimensions of buildings are defects if they exceed L/200 where L is the documented overall length of wall, or 5mm, whichever is greater."
"Departures from documented dimensions of service rooms and areas are defects if they exceed L/200 or 5mm, whichever is greater."
"Departures from documented dimensions of heights of building elements such as beams and posts are defects if they exceed L/200 or 5mm, whichever is greater."
i.e. 5mm per 1000mm (3/16" per 3' 3 1/2")
"Departures from documented dimensions of habitable rooms and areas are defects if they exceed L/100 or 5mm, whichever is greater."
i.e. 5mm per 500mm (3/16" per 1' 8")

Some allowable inaccuracies are quite large:
"Finished floor levels are defective where they depart from documented RL or FFL by more than 40mm" or depart from floors of the same level."
And there are more for specific materials:



From this it appears it is generally considered unrealistic to expect on-site measurements of less than to the nearest 5mm (3/16 inch).


MODELLING WALLS

Let's go back to our wall example.

A typical (metric) wall is two layers of plasterboard either side of a structural stud.
13 plasterboard + 92 steel stud + 13mm plasterboard, total thickness 118mm.


But is the built wall going to be exactly 118mm thick everywhere? Of course not.
For a start it will thicker wherever there is a joint.


It will also be thicker in corners (both internal and external), which like joints are also stopped up.



 And this is before we factor in the imprecision of on-site measuring.

If the architect is to achieve required minimum clearances then this needs to be taken into account.
You might rationalize that surely a code checker won't worry about a few millimetres. But no-one can take the risk. I have seen a building inspector insist walls tiles be removed from a toilet wall because the width he measured was slightly less than the required 900mm. There was also a court case here in Australia where the architect, building surveyor and contractor were found culpable for a balustrade 20mm less than code height (someone slid down the balustrade and fell, resulting in brain damage. Apparently, according to the court, if it was 20mm higher it wouldn't have happened).

What about rounding? We are all guilty of using dimensions that round to the nearest 5mm, or (gasp) 10mm.
The problem is Revit doesn't always round up. We always want to ensure minimum dimensions are met, we are not so worried about maximums being exceeded, so we need to round up.

Depending on the actual dimension, Revit may round up or down.


Also the 2 or 3mm rounding changes a dimension by may not be enough to account for construction tolerances. In-situ concrete requires quite large tolerances, 25mm (1 inch) is not uncommon.


Rounding generally is not good practice. Rounding errors mount up, and you can end up in the embarrassing situation where your short dimensions don't add up to longer dimensions.


The correct way is to locate things ACCURATELY. If the end walls are supposed to be 600 apart, model them 600 apart. Objects should be modelled so they are increments of 5mm apart, then actual dimensions you have in your model will be achievable in the real world.


I know it is not always possible, but just because it can't be done all the time is no excuse to never do it.

But I digress, back to our wall:

Therefore to ensure minimum dimensions are at least achievable on-site we need to increase the width of the wall. Based on the 5mm minimum measurable rule that means a multiple of 5mm, remembering that you don't want possible mis-measures to be minus 5mm.

So a 118mm wall becomes 120mm, (2mm tolerance) or to be safer 125mm (7mm tolerance).

Now, according to the Guide to Standards & Tolerances:
"Unless shown otherwise, dimensions shown on drawings for internal walls always refer to the structure's dimensions."
So we also want to rationalise the dimension of the structural steel stud.

The wall we make in our BIM software becomes:
17.5 plasterboard + 90 steel stud + 17.5mm plasterboard, total thickness 125mm.


Now it is true we could add another 'layer' into the wall for tolerance.
13 plasterboard + 4.5 tolerance + 90 steel stud + 4.5 tolerance + 13mm plasterboard.


But this adds another line into the graphic representation of the wall.
Not only is this confusing to anyone looking at the drawing (and a potential recurring RFI topic), it adds additional complexity to our BIM model. We already cope with large, slow files (in Revit that is), we don't need to make it worse.

So the architect is happy. But what about others?

When an estimator extracts quantities from the BIM model they are going to be wrong. But how wrong?

If we take volume of plasterboard, every 100sqM of 13mm plasterboard has a volume of 1.3cuM, for 17.5mm it is 1.75cuM, a difference of 35%. Not insignificant.

But something like plasterboard is measured by area, not volume. When measuring wall linings errors in area measurements will occur at corners and end walls.
Looking at internal corners (there are more internal corners than external corners or ends walls in a building), there is an under-measure if walls are thicker.
So 2 lots of 4.5mm will not be measured, and assuming 2.7m high walls, total area is 0.0243sqM.


Put another way there would need to be 41 internal corners to lose 1sqM of measured area, which is close to 10 rooms (4 corners per room). Say an average room is 3m x 3m, total wall area for 10 rooms is 324sqM. A loss of 1sqM equates to 0.3%. Which is insignificant.


So while it can be quite dangerous to rely on volumes out of a BIM model, areas are not such a problem. Of course this is not the case for every material. Mass materials like concrete are usually safe (tolerance in concrete placement is accounted for in attached linings). Structural steel may be OK if realistic structural members are used (which generally happens if the structural engineer is the modeller).


The take home point here is to not rely on thicknesses parameters to define what a material is. The above example has 13mm plasterboard, but when modelled at 17.5mm if its thickness is used to identify the type of plasterboard it could be mistaken for 16mm plasterboard, on the assumption a 1.5mm tolerance has been allowed for.

There is another reason thickness is not a good indicator of what a material is. Walls in BIM software have homogeneous layers. When materials overlap there is no way to use their real thickness in the thickness parameter.
The example below is a stud wall with insulation. The insulation is within the same 'layer' as the studs, but is not the same thickness as the studs. Therefore we have to nominate the stud layer thickness as 40mm, even though they are actually 92mm studs.


There are plenty of other parameters that can be used to define what a material is. It is not really necessary to use the thickness of a modelled material.


CONCLUSION

I've only provided limited examples here, but these lessons apply right across all things in BIM models. The specifics may be different, but the issues are the same.

Don't assume a BIM model has just been made for you and your purposes.

In particular no-one should presume, or ask for, absolute accuracy from a BIM model.
We should all expect, and actively include, tolerances in our BIM models.

And most of all, don't forget why we do the work we do. To build things in the real world.



2 comments:

  1. Antony great things to consider indeed. You nearly lost me when you used Engineer and rationale in the same sentence :)
    In MEP it has always interested me how the roughness parameter is derived for pipework. This value could be so subjective to quality control of manufacture. I guess we should always use a value "Large enough for worst possible example" to be safe.
    Cheers Pete



    ReplyDelete
  2. Antony Excellent article, as always. Know that in Spain you follow your articles for some time as very good reflections.
    In this particular article, I'll take that model can not please everyone. To me, in Spain I like to say, when in doubt, think as if was being on a construction site, and if themes of regulations and health and safety check things very well.
    Admittedly model taking into account all these factors is more tedious, but I hope the software gradually (or better as soon as possible) introduce these tolerances critoreios.
    I'll take your last sentence as a good summary: "don't forget why we do the work we do. To build things in the real world".

    ReplyDelete