Developing a Framework for Data-Driven Generation of Building Information Modeling from Sketches: Enhancing Efficiency in Space Configuration and Building Performance Analysis

Abstract

The demand for compact housing is on the rise, driven by the need for floor plans that accommodate stakeholders’ preferences. However, clients frequently struggle to convey their spatial needs to professionals, such as architects, due to a lack of means to present evidence, such as spatial configurations or cost projections. This study seeks to develop a methodology that translates sketched, data-driven spatial requirements into 3D building components within BIM (Building Information Modeling) to enhance spatial comprehension and offer building performance analysis, assisting in budget considerations during the initial design stages. The research methodology encompasses the formulation of a process model, its implementation, and subsequent validation. The process model outlines the data flow within the system and delineates necessary functionalities. Implementation includes the creation of systems and user interfaces for the integration of various components. Validation confirms the system’s capability to automatically transform sketched spatial requirements into BIM model elements, such as walls, floors, and roofs, and to autonomously compute material and energy expenses based on the BIM model. This system enables clients to effectively generate 3D building components from sketches, aiding stakeholders in spatial understanding and building performance evaluation through the generated BIM models.

Keywords

• building information modeling
• deep learning
• sketched data retrieval
• building performance analysis
• cost estimation