What is clash detection?

Clash detection (or “conflict detection”) corresponds to the process of superimposing two representations of the building, with the aim of identifying the “clashes” (differences) between the two entities. These two representations can be of different formats (BIM digital model or point cloud in particular), coming from different actors in the project, and representative of different timelines.

Thus, a clash detection process makes it possible to detect if two elements (networks, partitions, etc.) of a building interfere in an inappropriate or unexpected way.

We can identify two main types of conflicts in building geometry:

• The” Soft Clash”, which correspond to the lack of space left around equipment, preventing its proper functioning or maintenance. These conflicts are mostly related to design errors.
• The” Hard clash”: when 2 different objects are located in the same space, the two objects thus interfering.

Note: it's also common to hear about “workflow clash.” These don't refer to object conflicts, but rather to workflow and calendar conflicts.

‍

What are the benefits of clash detection?

By comparing two footprints of a building, project managers can identify and implement corrective actions based on the “conflicts” detected between the achievement objectives and the real one.

For example, a building point cloud generated in the EXE phase during a key phase of the construction site (e.g. before formwork) can be compared to the digital design model. Conflicts identified before it is too late will make it possible to implement corrective actions and reduce costs generated by unidentified errors during the construction phase.

The detection of clashes thus makes it possible to anticipate and limit site hazards and to improve the project's quality.

What causes clashes?

Several organizational and human factors can cause clashes:

• The synthesis of models produced by different teams for defined batches;
• Construction errors;
• Design mistakes
• The lack of updating of a digital model.