upFront.eZine #1,118: Which Comes First: Models or Drawings?
Guest editorial by Alexander Yampolsky
Inside the Business of CAD | 17 January 2022
This is the fifth in the series of thought-provoking pieces written by Alexander Yampolsky on the importance of 2D drawings for 3D construction:
Operating System for Structural Design (ebook; 2005)
Interpreting 3D Models from Formalized 2D Drawings (upFront.eZine; 2014)
Reflection: What are Drawings? (WorldCAD Access; 2018)
Why BIM Does Not Work (WorldCAD Access; 2021)
- - -
There is a way to determine people’s ability to structure designs. Ask them to arrange virtual furniture in a room:
When they draw rectangles and they annotate them with words like “wardrobe,” “bed,” and “table,” then place dimensions and distances, you are looking at a designer. The designer is a planner solving the “What can and should be done” problem.
When they create perspective views with details like armrests and cushions, it seems that they are “constructors” who would be better trying their hand at mechanical engineering. (See figure below.) Constructors solve the “How to make” problem using original, non-standard solutions. This is welcome in mechanical engineering, which has a pre-production testing stage, which does not exist in the design of buildings where non-standard solutions are risky.
“What is a model for, if you already have drawings?” This is the question I am asked during discussions on automatically generating 3D models from 2D drawings. People have no idea that it’s possible to make drawings without first constructing 3D models. For instance, this is the weighty opinion of a developer of one 3D modeling program: “The model is the only source of information for drawings. Drawings are nothing more than views of the model.”
And what do standards and textbooks say? When we consult them, we read terms like “projections,” “sections,” and “cross-sections,” such as illustrated by the figure below.
When we read further, we learn that a floor plan is a horizontal section at the level of the window openings, or 1/3 of the story’s height. We might think that this confirms the priority of models over drawings. An example from a real project, however, casts doubt on this conclusion.
Depicted in the figure below is a shop. This is practical, clear, but how can it be called a “horizontal section” when it shows a variety of elevations? Some are at -2.500; there is a rail track at floor level; some platforms at +1.000 and +4.000; many columns and vertical bracings (above the floor, and below the crane rails); walls at level of windows; bridge cranes and a crane landing platform (just under the roof).
This exactly is a plan, a schematic representation of what should be built. It is an extremely simplified representation, with all plan objects refined and detailed elsewhere. By the way, sections are the same as plans, except that they describe builds from different aspects.
PROJECT NEWPLEX v2.0
Project Newplex 2.0 is the advanced SHX font designed for technical documents; compatible with every DWG-based CAD program that supports Unicode SHX files.
Advanced character forms for readability & clarity
Designed for drafting
Includes the most-requested technical characters
Extensive international character set
For more info or to purchase, go to tcfonts.com/projectnewplex.html.
Parametric Modeling as Linguistic Description
The synthesis of spatial objects from their projections is called “the inverse problem of descriptive geometry,” and is an unsolved technical problem. Nevertheless, for hundreds of years, designers and builders have been synthesizing mental versions of 3D building models from 2D drawings. Then, builders transform the mental objects into real ones — how do they do it?
It is impossible to build models using projections, but it is possible using conditional images that have parameters tied to them. When we draw a circle (even if it is not perfect), and then write just two words, such as “Sphere d1800,” we create the parametric description of a sphere as a spatial object.
A parametric description consists entirely of conventional signs, and as a result is a linguistic description. Parametric modeling is essentially linguistic modeling. As a rule, parametric descriptions include the following elements:
Object type designations, formatted as text (e.g., “sphere”)
Conditional graphical representations of objects (e.g., drawing of a circle)
Parametric designations as text (e.g., “d1800”)
Conditional auxiliary images (e.g., dimensions and extension lines)
There are cases where the object type designator is not explicitly presented. In such cases, it can be reconstructed from the drawing context. For instance, the context may be provided by the drawing’s name, such as “column layout scheme.”
Any parametric description assumes the existence of an associated execution procedure, such as a mathematical formula, a computer program, or a detailing algorithm linked to detail drawings. From a programming point of view, a parametric description is a function call with parameters passing to it. When executed, the function transforms conventional signs and images into mental or digital models.
We can come to the conclusion that the primary method for analyzing drawings is to establish parametric descriptions for each building element. In drawings, the same building elements are usually described in several different contexts, such as being present in different views, on different drawings, in different drawing sets, and so on.
As an example, let’s try to restore the parametric description of the column located at A-17 in the figure below. The type of object (column) is easily determined from the context: column sections in this drawing are rectangles. There is, however, no other information tied to the object.
So we apply this well-known rule:
If, in a group of objects that are identical in appearance and purpose, there is only one object provided with parameters (dimensions, marking, designations), then all other objects in the group inherit these parameters.
In the group of columns along the A-axis, we see column A-7 is associated with detail #4. (See figure below.) If we were to refer to detail 4, we would see a refined column section with its dimensions, and snaps of the section to the co-ordination axes on the plan.
Using the details of section 1-1, we can determine the column’s top and bottom elevations. We apply the parametric description we obtained from A-7 to all columns (except for the corner ones) of row A, including the column at A-17.
Step-by-step Detailing of Objects
So. We managed to determine the definition of a simplified architectural column. A short parametric description cannot completely define complex structures, as real columns are. Nevermind: during the early stages of planning, our knowledge of objects is very approximate and vague.
During the planning process, structural drawings appear following the architectural ones, from which the columns obtain designations (that will be marked). A separate drawing of an abstract column representative of the columns of this type will appear. The structural elements of this abstract column — the formwork, rebars, embedded items — will, in turn, be detailed.
The essence of step-by-step detailing can be described this way:
Any project detail is itself a project
Any project is itself a project detail
During step-by-step detailing, a hierarchical structure is built by which the types and parameters of the upper-level objects are the context for lower-level objects.
Knowledge and Data
Parametric descriptions and their associated execution procedures are knowledge of what and how the work should be done. As a result of the execution of the parametric description, we get an object’s 3D model, which is the object’s database. Let’s take a closer look at the process.
In the figure below, there is a drawing fragment explaining the principle of reinforcing the floor slab.
The recipient of the drawing extracts a parametric description from it, applies the execution procedure, and, as a result, gets a mental model of the slab reinforcement. When the recipient is a builder, he will transform his mental model into real reinforcement in the field.
The process of creating a model in a 3D modeling program is, in principle, no different from a traditional drawing. By some means or other, the reinforcement contour is defined; the diameter, steel class, spacing, and possible overlap of rebars are specified.
The end result is, however, different. The execution procedure creates a computer model of a set of rebars with attributes attached to every bar. This model is saved. The figure below represents a top view of the reinforced slab as displayed by a computer model.
As mental models are built solely on the basis of visual perception, let's compare the two figures:
Drawing — shows a single abstract bar, with information attached. The bar defines the size, location, and attributes of all other bars in the planned reinforcement slab. We do not see individual bars, but we get a very good idea of the slab reinforcement.
Model — shows lots of bars. Annotations need to be added to all bars, and many dimensions need to be placed. When the slab is large and contains a lot of rebar, we get a messy, poorly readable picture. The result is a poor-quality mental model and, accordingly, a poor understanding.
- - -
Collaboration begins with understanding. Like it or not, we have to do what BIM calls "automatic" extraction of drawings from models.
A comparison begs. We cook borscht according to a recipe, throw away the recipe, and then, using complex algorithms, try to extract the recipe (the drawings) from the finished borscht (the models). This is the wrong way around of doing things.
The parametric drawing style is based on the assumption that recipients of the drawings have a professional set of algorithms for drawing analyses and for building mental models based on the analyses’ results.
In practice, this means that drawings should be made so that for each structural element, its parametric description can be restored. The mental object is modeled by calling the appropriate execution procedure associated with the parametric description.
Parametrics Determine Drawing Precision
In many cases, deviations from the image’s scale and accuracy contribute to the readability of drawings and the ease with which they can be adjusted. Consistent adherence to accuracy prevents this, surprisingly enough. The use of parametrics makes the meticulous precision of graphic images meaningless, and so it makes no sense to measure things on drawings. Indeed, the Russian standard prohibits such measurements.
The geometric accuracy of parametric drawings is determined by the accuracy of the values of dimensional parameters. In most cases, editing is limited to adjusting parameters, without affecting the graphic images. (It should be noted that parametric drawings are difficult to read when the proportions of objects are grossly distorted.)
As we see, this contradicts the autocad-style of drawing with its associative dimensions. The autocad style is based on drawing graphic images with tools that ensure great accuracy; the primary method for adjusting drawings is by editing the graphic images as required.
Machine Interpretation of Drawings
When professional algorithms are transferred to a computer, we get the ability to carry out automatic drawing analyses and building a computer model based on the results of the analyses.
A set of drawings can be considered a program written in a professional language — the equivalent of a high-level programming language. The program is compiled as a whole; a machine interpreter of drawings is in fact a compiler. As a result of the compilation, an overall execution procedure is formed, such as an IFC text file. In the end, a computer model is created. Note that the primary method of programming drawings is the step-by-step detailing of objects, better known as top-down design.
How is this different from the traditional way of creating models with 3D modeling software? All of them use a command interpreter as the input interface for 3D modeling. Complex objects are created by assembling pre-prepared components, whether primitives or parts. This technique implements the bottom-up principle; it simulates the construction process and so has nothing to do with planning.
It is clear that for many, assembling a building from cubes is easier and clearer than creating a linguistic description of a building with help of drawings. Unfortunately, except for the model in the computer’s memory, what’s actually needed are models in our heads. A set of drawings is an algorithm of understanding. It is impossible to replace the reading of this algorithm with something else, such as wandering across a computer model.
Crucially, the delineation of legal responsibility within the framework of this algorithm is also not a problem.
Machine Interpretation of Sketches
The parametric style of drawing allowed engineers with blunt pencils on broken-down Kuhlmann drafting machines to create absolutely accurate (not in the autocad sense) drawings. The independence instrumental to parametrics creates the principled opportunity of interpreting freehand sketches and drawings.
The figure below illustrates an example of such an interpretation. The problem is to turn a freehand line (possibly the sketch of a beam) into an accurate two-dimensional model.
Here are the steps involved:
Step 1. Bridge the gap, created by careless sketching. The extent of the gap is insignificant and is not numerically specified.
Step 2. Straighten the curve. The bend radii and the curve deviations from the straight line passing through the curve endpoints are insignificant and not numerically indicated. From this, we deduce that the curve is a straight line segment.
Step 3. Refine the x coordinates of the line’s endpoints. Horizontal deviations of the line’s endpoints from the A-axis and the 6000-dimension extension line are insignificant and so are not numerically specified. The x-coordinate at the line’s left endpoint is equal to xA (xA is the x-coordinate of axis "A"); the x-coordinate of the right endpoint is xA + 6000.
Step 4. Refine the z coordinates of the line’s endpoints. Vertical deviations of the line’s endpoints from the 3.000 elevation’s extension line are insignificant and not numerically specified, as is the inclination angle of the line. The z coordinates of the line’s endpoints are the same, and both are equal to 3000.
We completely restored the parametric description of the object.
Standards for Preparing Drawings
Current rulemaking in the area of preparing drawings stopped trying to combine the Gaspard Monge legacy with the drafting practice of the 1980s. [Mr Monge invented descriptive geometry, which became the basis for technical drawings.] In my opinion, construction drawings in Russia reached their peak in the eighties. From this, progress regressed, as drawings were made with 3D models and autocad styles of drawing.
Arrested development results in degradation, as shown by the figure below comparing Russian state standards from 1980 (top) and 2018 (below).
The Grammar of Drawings
Due to the appearance of 3D modeling programs, a question has arisen about teaching descriptive geometry in technical colleges: Is it still relevant? The crisis in descriptive geometry is a consequence of the same misconceptions about drawings as in engineering practice.
Back in the 19th century, drawings were called “the language of technique,” and descriptive geometry was the grammar of this language. In my opinion, the reasons for misunderstanding were these:
Incomplete awareness that drawings have all the properties of the language in a direct, literal sense
The consequences of not fully thinking this through. For instance, if we speak of drawings as text, then it is not clear what accurate projections have to do with the text
Descriptive geometry can and should develop in the direction of machine analysis of drawings: from classical descriptive geometry to parametric (linguistic) descriptive geometry, a kind of analytic geometry.
[Alexander Yampolsky has designed residential, public and industrial buildings, mainly as a structural analyst. He was an early participant (1982) in BIM at the Minuralsibstroy construction ministry of the Russian government, and developed technology for machine interpretation of drawings. Mr Yampolsky is a graduate of Tula State University.]
“The CDC said we can leave the Christmas lights up til January.”
- Aubrey Strobel
Thank You, Readers
Thank you to readers who donate towards the operation of upFront.eZine. To support upFront.eZine through PayPal.me, then I suggest the following amounts:
$25 for individuals > paypal.me/upfrontezine/25
$150 for small companies > paypal.me/upfrontezine/150
$750 for large companies > paypal.me/upfrontezine/750
Should Paypal.me not operate in your country, then use www.paypal.com to send funds to the account of email@example.com.
Or ask firstname.lastname@example.org about making a direct bank transfer through Wise (Transferwise).
Or mail a cheque (US$ or CDN$ only, please) to upFront.eZine Publishing, Ltd., 34486 Donlyn Avenue, Abbotsford BC, V2S 4W7, Canada.
upFront.eZine is published most Mondays. This newsletter is read by 4,600 subscribers in 70 countries. Read our back issues at www.upfrontezine.com.
Editor: Ralph Grabowski
Copy editor: Heather MacKenzie
Letter the editor are welcome at email@example.com. All letters sent to the editor are subject to publication, and may be edited for clarity and brevity.
To subscribe, click this link to sign up with with our mailer, Substack.
To change your address, send both your old and new email addresses to me at firstname.lastname@example.org.
To unsubscribe, click the Unsubscribe link at the end of this newsletter.
Copyright © 2022 by upFront.eZine Publishing, Ltd. All rights reserved.
Legal. All trademarks belong to their respective holders. “upFront.eZine,” “Inside the Business of CAD,” “WorldCAD Access,” and “eBooks.onLine” are trademarks of upFront.eZine Publishing, Ltd. Translations and opinions expressed are not necessarily shared by upFront.eZine Publishing, Ltd. By accessing this newsletter in any manner, you agree to settle disputes within ten days of publication date by arbitration within the city limits of Abbotsford, British Columbia, Canada with the arbitrator selected by an agent acting on behalf of upFront.eZine Publishing, Ltd.
Our mailing address:
upFront.eZine Publishing, Ltd.
34486 Donlyn Avenue