BIM and machine learning: predicting the impact of cost changes

Abstract image to illustrate BIM machine learning
Image: Xtau |

A research team has proved that BIM data can be combined with machine learning to predict the impact of design changes.

The team of researchers (from Northumbria University, Alexandria University in Egypt and the UCL Bartlett School of Sustainable Construction, plus Dr Marzia Bolpagni, head of BIM international and associate director at Mace) set out to design a conceptual model that demonstrated that machine learning (ML) software can analyse BIM data and thus predict the impact of design changes.

In a paper published in the November issue of Automation in Construction, the researchers noted: “Even while ML is beginning to be applied in construction projects based on BIM, it is still not being deployed productively when it comes to predicting the consequences that are caused by changes in the design. Therefore, the process of predicting implications in general, particularly concerning cost and time, occurs later on in the timeline of the project.

“As a result, this tardiness does nothing to assist in executing the project on schedule and within budget, and it may even lead to disputes. ML and BIM should and could play a role in these predictions.”

The researchers’ conceptual model draws on the numerous characteristics that each design change features to create a design changes dataset. The researchers said: “These characteristics include a nearly comprehensive description of the change as well as an evaluation of the change’s impact on the project’s budget and schedule.”

The characteristics include geometry, position, specification and dependencies between components as well as the documented impact to cost, schedule and quality.

Training machine learning on BIM data

The researchers’ concept then required the ML software to be trained on historic data to develop the algorithm necessary to analyse fresh data.

The researchers also created a use case (a single-floor building with architectural and structural components) to test the model on.

They concluded: “The implementation of the model on the use case delineates that ML can predict the design change impacts based on generating and structuring data via BIM, an approach that many practitioners in the industry, currently, do not seem to apply.

“The implementation of the BIM-ML model can significantly enhance decision-making regarding design changes in construction projects with a higher level of confidence.”

They added that the model could be used for cross-industry collaboration: “The study sets a new path to utilise BIM data in an ML context, wherein design change attributes are registered in a way that enhances data structuring, allowing ML algorithms to use them as training data. This can be achieved through intra- or inter-organisational collaborations on similar projects, as they would share similarities that ML can recognise as patterns.”

For further (real-world) development of the BIM-ML model, the researchers state that:

  • the model needs to be tested on real-world data;
  • data in BIM software needs to be structured so that it can be used effectively by ML algorithms; and
  • a user-friendly interface is required.

Don’t miss out on BIM and digital construction news: sign up to receive the BIMplus newsletter.

Story for BIM+? Get in touch via email: [email protected]

Latest articles in News