The document describes D-Reader, a software tool being developed by DII to automate the conversion of engineering drawings from scanned paper formats into 3D CAD models using artificial intelligence techniques. D-Reader uses computer vision and complex algorithms to extract engineering data from drawings, contextualize the information, and recreate the designs as 3D CAD files. This addresses the challenge of digitizing legacy engineering assets still contained in paper formats. The tool will help accelerate the global transition to digital formats by automating what is currently a manual process.
2. DII has designed a software product which helps to automate the
conversion of engineering data from scanned paper drawings to a
CAD (computer-aided design) model.
It addresses one of the critical global challenges of the digital
transition: digitising of legacy assets.
We believe that newly available technologies, including AI, will enable
us to resolve this challenge successfully.
How to digitise world’s physical assets?
3. D-Reader [its working name] is a software package, which:
1. Extracts engineering data from drawings
2. It then puts it in a context
3. Finally, it re-creates the designs in a 3-D CAD file.
D-Reader uses a combination of AI tools and original complex
algorithmic methods. It takes a scanned drawing image as input data,
recognises symbols, contextualises them and connects to the object,
then constructs the object with a design that precisely matches the
original drawing.
D-Reader
4. Why it matters?
A common opinion is that soon, all important objects will have their
digital replicas. Multiple global tech trends of today already require
digital models of objects. The D-Reader idea stands on two strategic
premises:
1. All the world’s major assets will need to be digitised.
2. Most of them still have designs preserved in a paper (or digital
scanned) form.
The vast number of physical of assets are converting into digital form
every day. All that work is done manually. The practical implications of
the D-Reader are extensive. The D-Reader will contribute to a new
digital revolution as it will allow fast digitization of all physical assets
around us.
5. Key Tech trendsPhysical objects/assets
- Buildings
- Infrastructure
- Industrial facilities
- Industrial equipment
- Transportation facilities
- Tools, parts and components
- Digital modelling and
prototyping
- Internet of things (IoT)
- Digital twins
- Smart facilities and
infrastructure
- Smart cities
- Building information modelling
(BIM)
6. Legacy assets challenge
Obtaining digital models for pre-built ('legacy') assets presents a
fundamental problem. Most of the legacy assets' engineering
information is still contained in paper drawings or scanned files. If
anyone needs a digital model, then substantial, time-consuming and
labour-intensive manual work is required. That takes a massive
amount of time and effort. From a practical standpoint, it is hard to
imagine manually digitising all objects around us.
Automated conversion of paper drawing data to CAD models has
remained an unresolved issue for decades. It slows down the global
adoption of many technologies related to operating physical assets.
7. - Remodelling, renovation, restoration or conversion design.
- Industrial facility upgrades or automation projects. Smart
infrastructure and IoT projects at large facilities.
- Applying building information models (BIM) for pre-built assets.
- Smart city projects requiring digitization of all types of properties.
- Digital twin projects for buildings, equipment and cities.
- Engineering of decommissioning projects for large structures.
Practical examples
8. We approach the challenge from a perspective that traditionally has
been assigned only to human engineers – understanding the drawing
and recreating the object by analysing the meaning of its every
element.
Using computer vision tools, we read the data from the drawing,
define drawing’s areas and shapes of the figures and assemble all
elements in a 3-D projection.
Importantly, we provide these models with all characteristics (sizes,
dimensions, tolerances, materials) from the paper-based drawings.
The model then is saved in a CAD file readable by all conventional
software.
How do we do this?
9. Due to overall complexity of the end goal, we have started from the
most basic functions, so the first versions of the product will create
elementary drawings and models. The method is the core of the
solution. Once proven, we will increase its sophistication by adding
further detail.
We are currently working on the first critical version of the product – v.
0.1 with a completion date of December the 31st, 2020. The
functionality will be basic, as the main task is to prove the working
methodology.
Where we are now?
10. Who is behind the project?
The team is an amalgamation of traditional industry experts and machine
learning specialists with a strong record. The head of R&D has an
impressive academic and industrial background. He is a PhD graduate in
Computer Sciences from the Université Catholique de Louvain, Belgium. He
also is a co-founder of computer vision-based startup which participated in
SAP-Techstars’ collaboration for B2B AI integration. Together with
prominent young mathematicians from Ukrainian and Canadian academia,
a strong and talented research unit has been created. R&D is backed by the
technical team with 20 years of engineering experience in the heavy
industry, working in the most complex and challenging of environments for
several of the most prominent market players. This practical expertise,
when combined with innovative tools developed using innovative
computer-vision techniques, gives us a massive edge when it comes to
swiftly and effectively providing solutions to any of the major challenges we
are faced with. More on: www.dii.ai/team
11. What’s next?
After V0.1, we plan to raise seed funding and build up a commercial
functionality. The development of V.1 requires additional CAD
engineering, building libraries and AI training for learning various
symbols and meanings.
V.1 will process parts and simple assemblies. We believe it will take
one year to transition from V0.1 to V.1. The subsequent versions will
process more complex assemblies. We plan to issue a new version
each year adding more layers of engineering data with each new
release. Our ultimate goal is to be able to successfully process
equipment assemblies of any complexity, address electrical and P&I
diagrams and architectural drawings.
12. Business model
When D-Reader gets to an industrial version, it allows quick
automatic transition of intricate engineering designs into CAD models.
Some projects will still require manual output, but even 80% of
automation will cause a massive effect. It can be used by engineering
firms, which convert the paper drawings, dramatically increasing
speed and reducing effort.
Most of the software will market as a SaaS for a subscription. For
some complex projects, it can work on users servers for a license fee.
After the first commercial version, DII will become a development
partner of all major CAD packages.
13. Additional information
More information on our website: https://dii.ai/
Our Linked In page Concept White paper on Medium
YouTube Introduction video Demo video of the product (pre-release)
Presentations (on Slideshare):
1) This presentation
2) Annex 1 - Digital copies of the assets
3) Annex 2 - Digitising the assets
4) Annex - Our process explained