This presentation outlines our vision on the D-Reader, the product our company develops. The product is focused on automated understanding of engineering drawings.

1. DII
Presentation 2
Digitising legacy assets

2. Digitising an asset requires two main tasks:
1. Creating the exact model based on the asset’s original design
2. Ensuring that model is identical to today’s condition of the asset
Two key steps

3. The first step is usually done by creating a 3-D model of the asset, or its
digital copy, combining all levels including forms, shapes, components,
materials and engineering systems (for complex objects). Usually all those
data must be kept in the original engineering documentation.
For legacy assets, that documentation is kept on paper or in ‘old’ 2-D CAD
files. Creating a single, 3-D digital model from them is mainly done manually,
though some operations are semi-automated.
That work requires engineering work hours, which immediately raises the cost
and time component of the initiative. For some large assets, this can become
a significant endeavour.
Digitising an original design

4. The original design is converted into digitial model

5. The second part is performed by applying the actual measurements to the
original model. This step is required because of the need to ensure that any
changes made during construction have been captured.
Normally, the design drawings had to be updated with ‘as-built’ drawings if
the changes were made during construction/fabrication of the asset.
However, in many cases the ‘as-built’ drawings were not created; hence they
must be created during asset digitising process.
Various measurement methods are used in this case. Mainly, there are three
options: manual measurements, comparing the model to photo/video data,
and laser scanning.
Ensuring it mirrors actual object

6. The designed model need to mirror actual object

7. “Laser scanning”, also known as “3D mapping” or “high definition surveying
(HDS)”, is the most quick but accurate way of providing the object’s shape. It
is being performed both indoors and outdoors with various types of
equipment available on the professional market. The result of laser scanning
is a 3-D drawing of the object performed with high accuracy. However, it does
not recognize any physical properties of the object apart from its shape. Laser
scanning can provide us with knowledge of the actual shape and, along with
analysing images of the object, can show if there are any missing parts or
changes.
The comparison process will include all data available, its alignment with the
original 3-D model, identifying the differences, and deciding on the correction.
In the construction industry, this process is called “Scan to BIM” or “point
cloud to BIM” and it is not automated.
Laser scanning and Scan-to-BIM

8. Laser scanning

9. Laser scanning

10. Laser scanning

11. Converting data from 3-D scanning into BIM
(“scan-to-BIM”)

12. Converting data from 3-D scanning into BIM
(“scan-to-BIM”)

13. Converting data from 3-D scanning into BIM
(“scan-to-BIM”)

14. Many actions required for creating digital models can be automated. Some
automation can be based on well-known methods and contemporary
software. Some can be developed using existing methods. However, some
methods have to be invented. Combining all of above can create a platform
able to automate the creation of the digital models.
Realistically, it is difficult to say at this point of time when that platform will be
able to provide complete automation, though manual input will still be
required at some point. However, there will be a substantial decrease of the
level of manual input, especially engineering hours, required to re-create the
model as the platform will dramatically simplify the digitisation process.
That simplification will allow massive numbers of assets to obtain a digital
form, which in turn will enable using various applications with it.
Automation

15. Our plan is to offer a high degree of automation and conduct constant work
and research and improve the process over time.
The platform will include both AI elements, specifically for computer vision
and analytical algorithms which, when combined, can enable the software to
“read” the drawings and build the models.
A specific case will be shown in the next presentation.
Automation

16. Turning old drawings into clean vector format

17. Some key methods, which have been developed and are available on the
market, will be used in the software:
- De-noising the original drawings
- Raster to vector conversions (the processing of scanned drawings and
defining symbols and lines).
- Recognising symbols
- Assigning the meaning to symbols
- Applying photo and video data to the drawing
- Applying 3-D scanning data to the drawing.
Key methods already developed

18. The methods can be developed based on existing algorithms and some
dataset training:
(when the software understands the role of the symbol on the drawing)
- Understanding key areas of the drawing
- Understanding the type of product and materials
- Determining the service graphics from the main lines of the object
- Determining the differences between drawings, photo, scanned images
Key methods can be developed

19. Actual object
BIM

20. Actual object
BIM

21. BIM
Actual object

22. Critical new algorithms required:
1. Combining all drawings into one 3-D model. Essentially converting a 2-D to
a 3-D model.
2. Assembling parts into complex components and components into overall
product
Its work is explained in the next presentation.
New algorithms

23. Creating 3-D model from the 2-D drawing

24. Creating 3-D model from the 2-D drawing

25. Creating complex assemblies of components

26. Thank you
This was Presentation 2 of 3 outlining
the DII business case.
Presentation 3 walks through the process of creating a 3-D
model from the 2-D drawing.
www.dii.ai