DATA EXTRACTION
BIM data from federated models
Choosing what information to extract
In Navisworks, the federation brought together objects from many disciplines. To turn them into queryable data, the add-in needed to preserve the link between each property and the element it came from. Users could start extraction at the root of the set or limit it to a selection. A mode focused on properties related to areas, lengths, volumes, types, and categories was also prepared. I developed the add-in in C# to traverse the object tree and its property sets, organizing values for writing to Apache Parquet files. The operation made it possible to separate data preparation from the audit questions that would come later.
Two ways to structure properties
From the outset, I considered two formats. A WIDE output is useful when specific properties are needed as table columns, for example to compare particular codes across elements. A LONG output retains each property as a record linked to its object, allowing broader sets of attributes to be explored without deciding every report column in advance. In the LONG implementation, records preserve provenance, identification, and element-path data, property group and name, and text or numeric representations of the value. The schema has 25 fields and writes the information in batches of 20,000 rows. One piece may produce several rows if it has many properties; that volume represents attributes, not an equivalent number of objects.
Preparing data for different questions
Parquet made it easier to work with the output using DuckDB and Python. Quality-control rules were then applied there, and results were prepared for Power BI. The distinction matters: the add-in extracted properties; assessments of naming, coding, and consistency belonged to a separate process. This separation made it possible to introduce new checks without traversing the entire federation again for every question. The tool also generated diagnostic logs and a map of the properties encountered. These files helped investigate incidents, unexpected fields, or differences between extractions without requiring users to examine the log manually during each routine operation.
A source other teams could use
With initial support, BIM coordinators and project leads incorporated extraction into their weekly work. The output was stored locally and fed the audits they consulted afterward. Later, I developed an additional command to add custom properties to selected elements in Navisworks. That was another capability of the add-in, separate from generating reports or automatically comparing versions. The most useful outcome was having a structured source that could be queried repeatedly and through different tools, instead of having to redo an export for every review.