The Solution Framework

The paper presents a solution which is a blend of various image processing and Machine Learning Algorithms. The flowchart for the complete solution is illustrated in Fig 3.

Fig 3: Flow chart of the Solution

Fig 3 describes two processes, viz. -

i. AI-assistance Digitization Process

a. Segregation of Non-Drawing Entities from the Document-

Engineering drawings typically can contain several views of the same component, and some additional scratch views can be added for a further understanding. The information specified in the drawing are mostly which could help in recreating these in digital format. Some other relevant information such as the dimensions, tolerances are usually only specified in one place on the drawing, for avoiding redundancy and the possibility of inconsistencies. Suitable tolerances are incorporated for critical dimensions to allow the component to be manufactured and functional.

Subsequently, a more detailed production drawing may be produced based on the information given in an engineering drawing. These drawings also have an information box or title block containing details of who drew it, who approved it, unit of dimensions, projection information, the title of the drawing and the component name, etc.

The primary objective in this step is to extract entities in the form of text, information box /tables, frame information and dimensions set which is so considered to be non-drawing part/region. The non-drawing entities are also significant from another point of view. The dimension set mostly consist of arrow heads, guidelines and extension lines. The dimension set are not necessarily required to be present in all drawings (mostly Mechanical Component Drawings). The text information is mostly available in tables, outside frame region and as dimension text and Instrument tagging in few scenarios. The dimension text contains geometrical information of drawings which represents length, radius, etc. These are also preserved separately for further validation purpose. The different components of 2D Engineering Drawing is descripted in Fig 4 (a).

Sometimes the text outside the frame is also used to locate the coordinate of dimension text of the drawings. The schematic diagram of different aspects is shown in Fig 4(b). The information box/Title contains information which are essential and useful during recreation of the drawings. Page Dimension (width X height), Page Specification type (A2, A3, etc.), Measurement units of the drawing are some of important features required. A typical title box and its useful information is shown in Fig 4(c).

Fig 4 (a): Different sections of a 2D Engineering Drawing of Mechanical Component.

Fig 4 (b): Content of 2D Engineering Drawings

Fig 4 (c): Information Containing in a Title Box.

b. After removal of most of the non-drawing entities it takes input as a “Image” containing mostly drawing entities, extracts out the features in the form of various geometrical shapes. These extracted features are passed into the CATIA tool for auto drawing. In CATIA tool various Macros are used to ingest these Geometrical entities and finally we get the “Partial Digitized Drawing Image” as output. This process is shown in Fig 5.

Fig 5 :Conversion of 2D Engineering to Drawing Image (left) to Partial Digitized Image (right) using AI-assist Digitization Tool. .

ii. AI-assistance Verification Process

It takes inputs as “2D Engineering Drawing Image” and “Partial Digitized Drawing Image”. After overlaying them, the verification tool highlights the missing components and displays this as an Image to the user. The user can then aid this image manually to draw these missed out components to get the “Final Completed Digitized Image”. This process is shown in Fig 6. The verification tool uses different colors to distinguish between the drawn and the missed-out components from the Digitization process. The missed-out components are highlighted in green color and the available component are highlighted in pink color.

Fig 6: Illustration of AI-assistance Verification Tool.

The scope of this current solution mainly covers extracting the geometrical entities and the “Verification Process” as illustrated in Fig 6, which will aid in reducing the Digitization time from Days/Hours to few minutes. The “Extract Feature Module” and “Verification Process” are combination of algorithms as illustrated in Fig 7(a) and Fig 7(b) respectively.

Fig 7(a): Different algorithms of “Extract Feature Module

Fig7(b): Different algorithms of “Verification process