Data-driven BPS frameworks

 

Traditionally, many stakeholders are involved in the BPS consulting process including architects, BPS modelers, and BIM modelers. During the design consulting process, architects manage building design options in CAD software based on the concept and programs of the buildings. With the increased use of BIM documents in the design process, the BIM modeler mainly focuses on conveying the appropriate information for future documentation. The BPS modeler takes the model information and creates a BPS workflow for the analysis results for the different performance metrics. Due to the expertise-dependent nature of physics-based BPS tools, the discrete workflow is inevitable in practice.

future.png
3.png
44.png
33.png

  • Replacing physics-based BPS models with data-driven ANN models can potentially reduce the number of steps between each expertise in the early design decision-making process. Recent ANN models can easily integrate into the CAD platform with less dependency. Therefore, they can be directly integrated into the modeling workflow for architects. Figure 8 represents the new workflow with ANNs models used as an alternative to existing physics-based models. To integrate ANNs models into the modeling workflow, it is necessary to develop data exchange protocol and methods for CAD and BIM software using their API. we introduce a different type of data processor to complete this workflow in this new framework.

    To realize this, it is necessary to look at the required data format for CAD and BIM with the existing BPS models. BIM-based geometry includes the detailed definition of buildings and related information not fully used by other BPS tools, such as EnergyPlus, and CFD. IFC-based building energy modeling tools define building geometry as a system of surfaces, also referred to as “space boundaries.” Examples of such surfaces include walls, slabs, roofs, columns, and beams (Bazjanac, 2010). Space boundaries are a critical part of the spatial building geometry definitions used by non-CAD tools. The current practice of modeling for BPS simulation usually involves the manual recreation of a building’s geometry to represent various physical properties of buildings (such as thermal, acoustic, structural, and airflow; see Figure 9), producing the ontology for BPS inputs of the building’s attributes for that simulation. Research by Autodesk (2015) has illustrated a mesh conversion process from Revit to CFD. Its use can verify the newly added CAD geometry and help convert a building’s geometry into meshes.

  • What are the data exchange requirements for ANNs based BPS tools for the proposed framework? The design of high-performance buildings using the CAD interface and interacting tools and surrogate models are complex and nonlinear. The sustainable nature of the process requires comprehensive guidelines to derive the necessary information efficiently throughout the design decision-making process. Figure 10 explains the ANN-driven BPS data processing flow for sustainable building design and compares it to the conventional method. Since data-driven building simulation software require different types of data representations as input, it is essential to find the adequate data conversion methods. Data conversion consists of two parts: pre-processing and post-processing. Pre-processing includes refined BPS requirements and geometry definitions for different data types (i.e., simulated and sensor data), and post-processing incorporates BIM requirements and performance metrics for entering the data into the BIM software. The BIM-toANN-based BPS interoperability framework covers data processing methods from the CAD software to ANN models (data pre-processing) and from ANN models to BIM documents (data post-processing).