A Framework for the Development of Semantically Enriched 3D Retrofit Models for Existing Assets

Abstract

While the use of Building Information Modelling (BIM) in new build projects has gained a lot of momentum, its use in existing buildings has been hampered by the challenges surrounding the limitations of available technologies used for generating retrofit models. Different data collection methods, such as image-based (e.g. Photogrammetry and Videogrammetry) and range-based (e.g. Laser Scanning), are used to collect data from an existing building in the form of images and three-dimensional point measurements, also known as Point Cloud Data (PCD). In current practice, the raw data is analysed and processed manually to generate BIM models utilising commercial BIM-driven platforms. Accordingly, several studies have been undertaken, proposing semi-automated approaches for generating parametric models by using PCD as the primary geometrical data source. Although some progress has been made in generating building geometries using PCD, an appropriate, full-blown, semantically rich 3D model is still not achieved. An appropriate 3D model that is fit for purpose for a BIM-based process of design, construction, as well as operation and maintenance (O&M) of assets should incorporate geometrical and non-geometrical data. While the geometrical data can be extracted from the collected data, non-geometrical data may need to be appended to this for generating a genuinely semantically rich BIM model.

The design of this research focuses on the challenges and limitations involved in the generation of parametric models, and the exchange of information embedded in models are first addressed by conducting an extensive literature review based on scientific and technical contributions in parametric modelling and information interoperability within the Architecture, engineering and Construction (AEC) industry. In addition to the literature review undertaken in this research, the Historic Environment Scotland (HES) BIM project is utilised as a case study for identifying the challenges of generating BIM models for retrofit assets, historical building in particular.

Research objectives are then defined based on the identified challenges and limitations. Subsequently, a framework for generating semantically rich parametric models for existing assets is proposed. The proposed framework is composed of three main processes, viz. 1) Data Collection, 2) Data Processing, and 3) BIM Generation. The Data Collection process involves geometrical and non-geometrical data that can be extracted and retrieved from distributed data sources, including PCD, offline and online sources. Comma Separated Values (CSV) format is used in this research to represent the collected data. The Data Processing step consists of two sub-divisions, viz. Data Aggregation and Data Standardisation. Resource Description Framework (RDF) as a SemanticWeb technology is employed to aggregate the data. The aggregated data is then standardised into Industry Foundation Classes (IFC), which is subsequently used to generate BIM models by importing the created IFC file into any IFC-compliant application. The steps above are carried out by the implementation of two key algorithms, including CSV-TO-RDF and RDF-TO-IFC. In terms of framework validation, the HES BIM project specifications are initially utilised to implement the developed framework and generate BIM models, including wall objects. A more complex application consisting of multiple building objects, such as external & internal walls, floor slabs, window and door openings, is also implemented in order to validate the applicability and scalability of the developed framework.

The key finding is the feasibility of incorporating geometrical and non-geometrical data for the use in BIM models. The use of RDF as a standard unified data format (aggregated data) facilitates data management, in particular, large-scale data, i.e., it simplifies the data storage, sharing and reuse. The use of RDF also facilitates data merging and linking, i.e. the geometrical and non-geometrical data presented in the form of RDF can be linked to other corresponding data sources as Linked Data (LD) if required. The proposed framework could provide a potential solution to the challenges and limitations involved in generating semantically enriched parametric models as well as the information exchange and interoperability within the AEC industry. It may also make a valuable contribution to the Asset/Facilities Management (AM/FM) domain and could be beneficial for a variety of FM practices for existing assets, such as the building information/knowledge management for design, construction, and O&M stages of an asset’s life-cycle.

Date of Award	2021
Original language	English
Awarding Institution	Glasgow Caledonian University
Supervisor	Bimal Kumar (Supervisor), Michael Tong (Supervisor) & Warren Chan (Supervisor)