Guest editorial by Alexander Yampolsky
It is strange, there is much talk about digitalization, but nothing about how does it end? How do big data, virtual reality, generative design finish? As a rule, they result in the decision that a person takes using a model built in his head.
Apparently, creating models in a human head is not considered as a big problem, otherwise building universities would have a special course introducing students into the basics of semiotics and linguistics. What will it take? Approximately, learning languages at schools achieves the similar result - elimination of illiteracy. Future constructors will understand that there are no other tools allowing reach the foreign brain other than linguistic ones and will learn to express their thoughts in professional language of constructors - in drawing language.
Interaction between models
Let keep in mind three types of models.Subjective models are models in the human head.Digital models are models in computer memory.Linguistic models - notional drawings (including texts) on any flat surface – are transitional bridges between subjective and digital models.Consider options for models interaction.
Person to Person. The owner of a subjective model creates a linguistic model as an integral, structured, multi-page document - a set of drawings. This process may be compared to writing software using a high-level language. The main purpose of this software is to create and spread understanding. To understand, how a complex object is arranged, it is sequentially split into simple blocks. The upper splitting levels are intended for a general understanding; the lower ones are for the detailed one. It is clear that this is not a literal split like engine dismantling at a car repair shop. This is a conceptual analysis using step-by-step object detailing method.
The recipient interprets (compiles) the linguistic model as a whole. During compilation, the reverse process takes place - the object conceptual synthesis. As a result, an understanding - a subjective model in the recipient mind - emerges.
Thus, drawings by their nature are an algorithm for constructing a subjective model, i.e. algorithm of understanding.
Person to Computer. The owner of a subjective model sequentially enters commands as separate linguistic sentences using the command line, menu or toolbar. The computer interprets each command at the time of its entry. As it is known, interpreter-programs operate according to this algorithm. As a result, a digital model is formed in the computer memory.
The main purpose of the digital model is to verify the subjective model, perform digital experiments, and generate new data.
Computer to Person. Similarly to the option "a person to a person", the computer should automatically generate a drawing set. The problem is that this task is absolutely irrealistic.
The computer cannot independently choose an understanding strategy, assign steps of detailing, and determine the abstraction degree at each step.
There are attempts to solve the problem by embedding linguistic model elements (views being stored, annotations, drawings, etc.) into a digital model. The result is a hybrid model, or BIM-model - a kind of digital models.
It is believed that a hybrid model can be used as a unified environment, a reliable support for making the right decisions by all construction participants. For the design phase, this is a new idea, a new application area for digital modeling. Let check feasibility of this idea for a real episode.
A model (digital twin) of a one-storey industrial building was created (see Fig. 1). Checks were carried out, clashes were eliminated, and specifications were calculated.
The next step is an expert meeting to assess the project quality. It is a form of collaboration which should be based on the data got from the model according to the new technology.Expert opinions were different.Those, for whom the presented model was the only source of truth, assessed the project as excellent. However, some experts arrived at another conclusion - an absolutely amateurish project. All manipulations with its data were a waste of time.
What is the reason that the “source of truth” has turned out into a source of false?
In my opinion, the reason is the dubious worth of the collaboration based on a single model. Because of a defective model, collaboration on its basis leads to a dismal result.
Being inside the model, it is impossible to test its correctness. But having left the digital reservation, we will enter the world of normative documents, reference books, textbooks, and scientific articles. It turns out that collaboration in this world is based only on the knowledge (concepts) while the talk is carried out using only the language of drawings. The skill to play with toy blocks is not enough; it is necessary to “write” and “read.”
The ability to “write” allows a person to abstract his/her understanding of the building structure to the conceptual level - simple schemes. For the building shown in Fig. 1 the ordinary frame scheme will look like the one shown in Fig. 2.
The ability to “read” allows making a realistic image of an object based on simple schemes (like ones shown in Fig. 3).
Without such schemes (Fig. 2 and Fig. 3), it is impossible to make the right design solution.
Designing is collecting of the knowledge acquired in different ways from different sources. Knowledge, as we know, is abstract and fragmentary. Each new knowledge may devaluate (sometimes, in an unobvious way) the previous one due to the incompatibility problem. To eliminate this issue at the final design stage, a single digital model of building should be made and verified.
If conceptual design solutions are found and properly formalized, creation a single model on their basis does not require high qualification and can be automated. This issue is already being solved.
Computer to Computer. Everything is simple here. If two models can interact with each other without human intervention, they may be considered as a distributed single model. With regard to a gradual decrease of the human role and responsibility in the designing process, I do not see such a trend. As 40 years ago, "disclaimers" is a mandatory paragraph of any license agreement for the software purchase.
Extracting drawings from a hybrid model
BIM software developers promise to automatically receive drawings of excellent quality. The following statement would be more realistic: the higher automation degree of obtaining drawings, the lower their quality. Fig. 4 illustrates the technology drawing obtained from the hybrid model.
It is an overloaded ("dirty"), hard-to-perceived drawing resulting from too textual copying of the model. It is not clear, for whom this drawing is intended, and for what purpose.
For example, a designer needs to obtain from a technologist a formalized assignment, which indicates loads and apertures for the equipment transportation. Instead of Fig. 4, I would prefer to obtain something similar to the one shown in Fig. 5.
Difference between a linguistic model and a digital one
Any tool other than its primary purpose has also to “teach” a user how to work properly. This means that the tool should be based on an ideology, which dictates the correct way of using and developing this tool.
A drawing board was the first generation of 2D editors.
Everything was perfect with the drawing board ideology. It took into account specificity of visual perception, understood indispensability of the paper carrier, set the clarity and definiteness of the intention above the graphical accuracy. In other words, the drawing board targeted a designer to create just drawings - linguistic models.
"Autocad" was the second generation of 2D editors. The history of this generation is an example of how an erroneous ideology leads a tool into a deadlock. Wrong orientation, misunderstanding of the drawing essence was the main mistake. Instead of a tool intended for the linguistic presentation of the knowledge, "autocad" gradually turned into a digital modeling tool, first in 2D and then in 3D. As a result, it lost face in a foreign field and is considered now as an anachronism, which exists only due to habit of using it.
What did the second generation editor get and what did he refuse?
Fig. 6 illustrates a digital model of a ship at an anchorage, Fig. 7 – linguistic model of the same ship. Let try to find differences.
- The highest accuracy is the motto of a digital model. Correspondingly, objects in Fig. 6 are shown with a plenty of details. (Objects in Fig. 7 are shown notionally.)
- When measuring the distance from the sea floor to water surface using a “ruler” in the digital model, we get the value 20. It is possible to extract additional parameters directly from the model, such as horizontal projection of the anchor chain.
All measurements are meaningless in the linguistic model since they will give unpredictable results. It is possible only to read what is written in it, in particular, that the anchoring depth is 20 m.
- The filled polygon in a digital model (for example, an anchor) can turn out to be a set of merged lines.
In the linguistic model the polygon will remain a polygon when any approaching. Details of the anchor design may be viewed on a separate sheet depicting the anchor in a large scale.
- If the image in Fig. 6 is a view of a three-dimensional model, any change in it will automatically result in changes in Fig. 6.
The linguistic model is completely independent of external changes. It does not reflect something; it declares what should be. Fig. 7 is a fragmentary knowledge, concept, standard solution of anchorage. No doubt, concepts should be linked to each other, but this goal is not achieved automatically. Any change in the concept may be made only after a profound analysis of the causes and consequences for the proposed change.
At first glance, an accurate “data-rich” digital model is much better than a linguistic one. However, everything has a downside.
- You have to pay for the accuracy. We simply correct the value 20 to the desired one in the linguistic model when the anchorage depth increases; and this ends the editing.
In the digital model, it is necessary to raise the sea level, shift the ship, move the anchor, and edit the chain sagging curve.
- Data abundance of digital models (for example, of buildings) ends up at the level of rebars, embedded parts, bolted and welded joints.
Linguistic models do not have boundaries. The explanation is simple. Designer's handbooks, typical projects, national standards are usual linguistic models; references to them are a natural continuation of a drawing. If desired, we can trace the structure of the object (for example, the anchor in Fig. 7) up to the crystal lattice of the steel used for object producing.
- It is easy to determine in linguistic models, what the author of the model ensures exactly and is responsible for.
In a digital model, the boundaries of responsibility are not so clear. We are proposed to recover data from the model by ourselves, for example, using measurements. Those persons, who offer doing so, seem to be confident that:
- The user is so proficient in the modeling tool, as the author of the model;
- Aa measurement operation is so simple and clear that it is impossible to make a mistake when performing it;
- The user will understand himself/herself, which elements of the model may be measured, and which ones should not be measured since the accuracy of their modeling is not guaranteed;
- And so on.
4. Fig. 6 is a view of the model and, at the same time, imitates a drawing explaining the concept of anchorage. Such duality is dangerous. Automatic changes are required for the view, but they are not allowed for the concept.The structures of the digital model (assemblies, sub-assemblies, components) and design documentation (sections, sheets, schemes on sheets) are interconnected by the “many-to-many” relationship. One sheet of the documentation can describe many elements in a digital model, and one element of the model may be described in many sheets. The puzzling task to harmonize two structures is assigned to a user. Hypercomplexity, confusion and substitution of notions are inherent for hybrid models, and it is so-called a systemic propensity for errors. It was necessary to rely on hybridization because of despair, since drawings did not want to "die". Apparently, the variant of automatically model creation using drawings was not considered.
I foresee the question: where is the new editor description?
All the issues covered above are an attempt to clarify the common situation, to identify fundamental problems.
In my view, the principal task is to draw a clear boundary between digital and linguistic models. It is also necessary to ensure problem-free transition of this boundary at any place, at any moment in time. I see solution in the wide use of machine interpretation of drawings.
From the software engineering viewpoint, drawings are the universal independent format for exchanging information, the same as IFC, for example.
IFC language is more formalized but, as a result, it is clear only to programmers. The language of drawings includes pictures; it is less formalized and, as a result, is clear to any engineer.
As you can see, the task is to update the language of drawings, to add the missing feature - machine interpretability with simultaneous preserving its main advantage – human perception.
This is the main challenge of the next, third-generation of 2D editors.