BIM QUALITY CONTROL
BIM audits of models and drawings
Turning a property into an observation
The federated-model audit workflow began with information extracted from Navisworks and stored in Parquet. I used Python and DuckDB to apply rules to those data: checking the presence of fields, formats, identifiers, and relationships between properties. One example involved checking a classification code, a type mark, and the composite identifier that should correspond to them. For building services, other reviews examined consistency between marks, sizes, and components. The difficult part was not producing an endless list of conditions. A rule is only useful when applied to the right categories and disciplines. The analysis therefore needed to distinguish when a value should exist, when a combination might be inconsistent, and which elements should not be evaluated against the same criterion.
From data to the coordinator's work
Results were organized in Power BI so the team could filter by model, discipline, or issue type. Each record retained references to the elements involved. Coordinators could export lists and share them with contractors so they could examine the problem directly in their working models. When new versions arrived, extraction and analysis were repeated. The report displayed the current snapshot; tracking which issues had disappeared or remained open was handled through coordination work and the comparisons available, not through automatic correction of models.
Reviewing what happens on a sheet
A related development addressed as-built documentation. Using Dynamo and the Revit API, I prepared routines to query sheet information and identify the views placed on each sheet. To detect possible overlaps between text and dimensions, the routines used geometric data describing the space they occupied. The output recorded the sheet, view, and flagged elements, was exported to Excel, and was then reviewed in Power BI. This made it possible to communicate observations more specific than “check the annotations.” A coordinator could identify which sheet to open, which view to review, and which objects were involved. The team continued to assess which detections represented real graphic conflicts and request changes from those responsible. The two experiences share a logic, although they review different information. One checks the properties of objects in federated models; the other examines data and annotations in documentation. In both, the aim was to make reviews repeatable and provide specific references for discussing corrections, while keeping final acceptance of deliverables with the technical team.