The problem the researchers set out to solve is semantic rather than geometric. Scan-to-BIM pipelines can already reconstruct three-dimensional models of existing buildings from laser scans with increasing automation, and tools can extract wall areas and element volumes from those models. What the models lack are the material identities, contamination diagnoses, and waste classifications that a legally valid audit requires. As the authors note, an automated system can identify a wall element, but it cannot reliably distinguish high-value recyclable gypsum from hazardous cement-asbestos boards without physical sampling and professional liability. Their answer is a strict human-in-the-loop division of labor: the software computes everything deterministically derivable from geometry, while the auditor alone assigns material identity, contamination status, and waste codes.

The framework rests on open, vendor-neutral standards. It ingests an IFC 4.3 model, either authored in a BIM tool or produced by an automated scan-to-BIM pipeline, and enriches it with property sets defined by the WASTEie dictionary, a publicly published dictionary on the buildingSMART Data Dictionary platform grounded in ISO 12006-3. Waste classification follows the European List of Wastes, restricted to chapter 17 for construction and demolition wastes. The auditor selects codes from a controlled list, so free-text variance is eliminated, and hazardousness is not an independent flag: mirror-entry codes marked with an asterisk automatically set the corresponding contamination properties, so the flag and the classification can never contradict each other. The output is an enriched IFC file whose semantics any third-party tool can resolve from the public dictionary, plus an Excel report structured according to the EU Construction and Demolition Waste Management Protocol.

The case study is the railway station in Mladá Boleslav, Czech Republic, a pre-1950 masonry building with load-bearing brick walls and semi-battened wooden beam ceilings. It was chosen partly because it typifies the older European building stock where as-built documentation is usually missing, and decisively because its demolition was already contracted, guaranteeing access to the contractor’s weighbridge consignment records. A terrestrial laser scan produced a raw point cloud of 504 million points, thinned to a minimum spacing of 0.01 meters into a final dataset of 273 million points, from which the Cloud2BIM pipeline generated the IFC geometry. The model contained 970 physical elements, none of which failed the volume computation, so the reported quantities cover the complete modeled inventory.

The audit quantified a total estimated volume of 2281.01 cubic meters across the modeled fabric. The dominant stream was 1841.07 cubic meters of masonry bricks suitable for recycled aggregate, followed by 237.75 cubic meters of concrete, 128.31 cubic meters of timber and wood, 14.29 cubic meters of ceramics and tiles, 12.83 cubic meters of metals, and smaller glass and mixed fractions. Where visual estimation would have been unreliable, the team supplemented the model with targeted physical interventions: a series of on-site test probes established an average internal plaster thickness of 25 millimeters, which was combined with wall surface areas extracted from the BIM model to calculate plaster volumes with a bounded margin of error.

Perhaps the operationally most consequential result concerned hazardous material. The auditors identified 32.03 cubic meters of chimney masonry as hazardous under waste code 17 01 06*, and the framework locked that classification to the spatial coordinates of the chimney elements within the digital model. The authors are careful about the evidentiary status of what followed: the demolition contractor confirmed the separate dismantling and disposal of the chimney masonry, and the consignment records contain no hazardous entry, while the primary brick stream was accepted for recovery without reclassification, which would have been precluded had contaminated masonry entered it. Direct documentary evidence, the hazardous-waste consignment note, was handled by the contractor and could not be obtained. The isolation is therefore reported as consistent with the audit’s intent rather than independently documented, though the mechanism itself, a spatially explicit hazard annotation a text-based report cannot provide, is demonstrated unambiguously.

The ex-post comparison is where the study breaks new ground. Converting the audit’s volumes to tonnages using bulk in-situ densities recommended by the EU CDW Protocol and PARADE guidance, the audit predicted roughly 4146 tonnes across the categories it covered, while the measured consignment records totaled 8418 tonnes. The apparent 51 percent gap is not a pipeline error. It decomposes into 2701 tonnes, 63 percent of the gap, of subsurface and external material invisible to interior laser scanning, including 2593 tonnes of soil and stones and 108 tonnes of bituminous mixtures; roughly 277 tonnes, 7 percent, of thin-layered and non-structural material the scan-to-BIM process did not resolve, such as insulation and bulky waste; and 1475 tonnes of mineral-fraction difference attributable to the contractor re-sorting predicted clean brick into mixed rubble on site, an economic decision rather than a calculation error. An auditor using this workflow on a comparable building must therefore expect the model-derived inventory to cover roughly half of the total outgoing mass.

For the streams the model did represent, the deviations were interpretable. Summed across the relevant mineral codes, the audit estimated about 3888 tonnes against a measured 5363 tonnes, a 28 percent underestimation. Wood fared worse: 77 tonnes predicted against 275 tonnes measured, a 72 percent shortfall traced to the digital model’s omission of non-structural elements such as floorboards, door and window frames, interior paneling, and roof carpentry. A sensitivity analysis that pushed every density simultaneously to the lower and upper bounds of its plausible range moved the aggregate prediction between 3720 and 4570 tonnes and the gap between 46 and 56 percent, but the sign and qualitative interpretation of every deviation remained invariant. No admissible density assignment can close either the mineral or the timber gap, confirming that the deviations originate in the scope of the input geometry and on-site sorting decisions, not in the volume-to-mass conversion.

The economic stakes of spatial hazard localization are considerable. Under the Czech Waste Act, the statutory landfill fee for hazardous waste is 2,000 CZK per tonne, roughly 80 euros, to which gate fees and a mandatory reclamation reserve are added, yielding realistic disposal costs of 140 to 200 euros per tonne. Had the contaminated chimney masonry been demolished together with the adjacent walls, the commingled 5363-tonne mineral stream would have been exposed to classification under the hazardous mirror entry, costing at least 0.4 million euros at the statutory fee alone and 0.75 to 1.1 million euros at realistic full disposal rates, against a disposal cost on the order of tens of thousands of euros for the separately handled 58-tonne hazardous fraction.

The team also verified that the deliverable is genuinely machine-readable. An Information Delivery Specification requiring every physical element to carry the WASTEie property set with populated contamination and recyclability properties, plus a six-digit EWC classification reference, was validated with the open-source ifctester toolkit: 2457 of 2457 checks passed. The enriched file was then loaded into BIMcollab ZOOM, a commercial model viewer unconnected to the toolchain, which correctly resolved the property sets and classifications and could filter components by their waste-related properties. The authors are candid about the limits: the validation is a single masonry case, the in-model classification carries only the dominant code for multi-material elements, and material diagnosis remains expert judgment that a formal inter-auditor repeatability study has yet to test. But the core finding stands: for the building fabric captured in the input model, standardized semantic enrichment adds no error of its own and yields a deliverable that third-party tools can consume without any bilateral agreement with the authoring tool.

Subject of Research: A digital framework for producing machine-readable, standards-based pre-demolition audits validated against measured demolition waste

Article Title: Standardized semantic enrichment of OpenBIM for machine-readable pre-demolition audits

Article References: Zbirovský, S., Šulc, S., Tetervov, V., & Nežerka, V. (2026). Standardized semantic enrichment of OpenBIM for machine-readable pre-demolition audits. Results in Engineering, 32, Article 113650. https://doi.org/10.1016/j.rineng.2026.113650

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113650

Keywords: pre-demolition audit, openBIM, IFC, construction and demolition waste, circular economy, bSDD dictionary, European Waste Catalog, scan-to-BIM, hazardous waste isolation, machine-readable data, Mladá Boleslav railway station, ex-post validation