sg-crest
A Singapore Government Agency Website
logo
Official website links end with .gov.sg
lock
Secure websites use HTTPS
Look for a lock () or https:// as an added precaution. Share sensitive information only on official, secure websites.

Challenge Statement

How might we reduce the time spent on the evaluation of Earth Control Measures submission drawings and accurately identify the lapses?

Challenge Owner

  • Catchment & Waterways Department

PUB receives more than 1,500 drawing submissions in PDF and CAD formats per year seeking approval for construction works requiring earth control measures (ECM). As part of the evaluation, PUB personnel needs to review the to-scale ECM drawings to ensure that the key design elements and key requirements are included and valid. The process requires a lengthy time for an experienced eye to manually inspect the drawings.

We are interested in solutions that use machine learning or artificial intelligence to help reduce the time spent evaluating 2D CAD drawings to pinpoint and identify lapses that require correcting. For example, the solution should be able to estimate the size and capacity of sump pits or tanks based on the drawing and determine if they are appropriately sized to the area of the worksite. Other examples include the ability of the solution to identify whether critical requirements for ECM are met based on annotations/texts within the drawing.

Previous attempts to use Building Information Modeling (BIM) methods to review ECM were put on hold as the industry is not ready to transition to BIM, which are much more complex 3D models of civil structures.

  • The solution must be operable from the stand-alone device or a private cloud, separated from the Internet.

  • The solution must be able to imbibe a range of file types, including PDF documents, scanned image files, 2D CAD drawing files, Microsoft Word Doc files, etc.

  • The solution must be able to decipher the different drawing formats adopted by different contractors.

  • The solution must neither increase the cost to the construction industry nor significantly complicate the process and level of detail required for submissions.

A solution that is able to decipher past submissions and identify more than 90% of the commonly spotted issues in submissions.

Challenge Owner

  • Catchment & Waterways Department

PUB receives more than 1,500 drawing submissions in PDF and CAD formats per year seeking approval for construction works requiring earth control measures (ECM). As part of the evaluation, PUB personnel needs to review the to-scale ECM drawings to ensure that the key design elements and key requirements are included and valid. The process requires a lengthy time for an experienced eye to manually inspect the drawings.

We are interested in solutions that use machine learning or artificial intelligence to help reduce the time spent evaluating 2D CAD drawings to pinpoint and identify lapses that require correcting. For example, the solution should be able to estimate the size and capacity of sump pits or tanks based on the drawing and determine if they are appropriately sized to the area of the worksite. Other examples include the ability of the solution to identify whether critical requirements for ECM are met based on annotations/texts within the drawing.

Previous attempts to use Building Information Modeling (BIM) methods to review ECM were put on hold as the industry is not ready to transition to BIM, which are much more complex 3D models of civil structures.

  • The solution must be operable from the stand-alone device or a private cloud, separated from the Internet.

  • The solution must be able to imbibe a range of file types, including PDF documents, scanned image files, 2D CAD drawing files, Microsoft Word Doc files, etc.

  • The solution must be able to decipher the different drawing formats adopted by different contractors.

  • The solution must neither increase the cost to the construction industry nor significantly complicate the process and level of detail required for submissions.

A solution that is able to decipher past submissions and identify more than 90% of the commonly spotted issues in submissions.

Info Session