Asst. Prof. Songsak Onsuwan, Asst. Prof. Korapong Karnasuta, and Asst. Prof. Paiboon Kittikul
Abstract :
In the digital era, Computational Fluid Dynamics (CFD) has become indispensable for microclimate and architectural airflow analysis. However, the inherent complexity of such software has inadvertently created a barrier between engineering computations and the “physical memory” of designers. Hand-sketching, which serves as the physical trace of early cognitive design processes, is increasingly displaced in modern digital workflows. This paper presents ChromaFlow: Interactive Projection-Mapped Wind Tunnel for Architectural Airflow Analysis, a project that responds to the “Postmemory” conceptual framework by functioning as a bridge between inherited vernacular wisdom—invisible climate memory—and contemporary digital visualization technologies.
This project was developed to bridge the gap in the conceptual design phase by introducing a Tangible User Interface (TUI) that transforms a physical whiteboard and markers into a robust engineering analysis tool. The system is driven by the integration of three core technologies: 1) Computer Vision (OpenCV) for perspective correction and real-time contour extraction of physical sketches; 2) GPU-accelerated Fluid Dynamics Simulation (PixelFlow) for processing quantitative wind velocity rendered as a heatmap vector field; and 3) Projection Mapping (Keystone) to superimpose the digital airflow seamlessly onto the physical drawing.
The results indicate that ChromaFlow can instantly process and visualize aerodynamic behaviors around physical architectural forms in real-time with zero latency. This innovation not only accelerates rapid conceptual prototyping but also revives the “Memory of the Hand.” By creating an active learning environment, the tool empowers designers and architecture students to inherit vernacular architectural memories and tangibly interact with natural fluid dynamics through augmented reality, transforming isolated digital workflows into a collaborative, shared experience.
Objectives :
The research and development of the ChromaFlow project aim to dismantle the boundary between traditional physical architectural design processes and advanced computer science computations. Rooted in the “Postmemory” framework, the project holds the following primary objectives:
- To develop an Interactive Fluid Dynamics Wind Tunnel Simulation: A real-time system capable of processing microclimate data by utilizing computer vision technology to extract physical sketch traces. These traces are directly converted into computational physics boundaries, significantly accelerating the rapid conceptual prototyping phase.
- To create an innovative educational tool for architectural environments: Designed to foster Active Learning and Interactive Learning, this tool allows students to tangibly assess spatial impacts, observe multi-directional airflow behaviors around structures, and analyze quantitative velocity through a virtual probe within an augmented reality environment.
To apply Projection Mapping technology in preserving the Physical Memory of the design process: By transforming a traditional sketching board into an augmented space that reflects aerodynamic results back onto the original ink strokes. This revives the “Memory of the Hand” and ensures that the inherited philosophy of climate-responsive design coexists harmoniously with modern digital instruments.
Conceptual Framework :
ChromaFlow was developed by integrating two main theoretical frameworks: the Postmemory concept and David Kolb’s Experiential Learning Theory. This integration creates a design environment that connects the past, present, and future.
6.1 The Postmemory Dimension In the context of architectural design, hand-sketching represents an “Embodied Memory” that accumulates generations of wisdom regarding environmental understanding. ChromaFlow preserves this memory by avoiding total replacement with computer screens and mice. Instead, it employs a Tangible Interface as a medium, allowing the new generation of architects to connect with inherited traditional skills within a modern digital context.
6.2 The Experiential Learning Dimension To ensure the tool facilitates a comprehensive learning process regarding fluid dynamics, the system’s workflow is aligned with Kolb’s four-stage learning cycle:
- Concrete Experience (Doing): Users physically interact with the system by using markers to sketch architectural forms or placing physical obstacles on a whiteboard.
- Reflective Observation (Reviewing): Users immediately observe the outcomes through the projection-mapped vector field, watching how the wind collides and forms vortex shedding around their specific drawings.
- Abstract Conceptualization (Learning from experience): Users comprehend complex physics and fluid dynamics by reading quantitative wind speeds and heatmap data via a Virtual Probe, leading to a profound understanding of aerodynamic principles.
- Active Experimentation (Trying out what was learned): Users apply their conclusions by modifying the physical sketches (erasing or redrawing) or changing the wind direction and speed, instantly observing new results. This forms an effective and continuous Iterative Design Loop.
By merging these frameworks, ChromaFlow transcends being a mere simulation software to become an active learning space that systematically stimulates analytical thinking.
Process / Methodology :
For the software architecture of ChromaFlow, the Processing platform—an open-source Integrated Development Environment (IDE) designed for digital arts and visual computing—was selected as the core framework. The primary advantages of Processing are its capacity for rapid prototyping and its flexibility in seamlessly integrating hardware and software. It is exceptionally suited to act as the central hub connecting the input device (webcam) and the output device (projector). Processing’s robust ecosystem of libraries ensures stable, real-time fetching of live camera feeds, advanced physics processing, and projection mapping output at a continuous 60 Frames Per Second (FPS).
The software architecture is divided into three real-time integrated phases:
7.1 Phase 1: Computer Vision (Vision Processing) The system captures the physical architectural sketches via an overhead webcam, processing the feed using the OpenCV library to extract obstacle data:
- Perspective Transformation: A Homography Matrix is calculated from four user-defined corners to correct camera distortion (Perspective Warp), ensuring the sketch plane is perfectly orthogonal and accurately scaled to the computational grid.
- Contour Extraction: The image is converted to grayscale, followed by Thresholding and Morphological Closing (Dilation to seal ink gaps and Erosion to restore stroke proportions). This transforms the ink traces into a solid, transparent Obstacle Mask (Alpha 0) for physics calculations.
7.2 Phase 2: Fluid Dynamics Simulation The obstacle mask data is transmitted to the Graphics Processing Unit (GPU) via the PixelFlow library to simulate wind behaviors:
- Physics Computation: The system utilizes Navier-Stokes Equations to compute pressure and velocity as air masses collide with the architectural boundaries, accurately generating realistic deflection and vortex shedding.
- Vector Field Rendering: Quantitative velocity data is fetched back to the CPU to render a Vector Field. Wind directions are drawn as constant-length vectors and color-coded using a Heatmap scale based on velocity (from blue for stagnant air to red for high-velocity impacts).
7.3 Phase 3: Projection Mapping The physics simulation results are translated back into the physical world to create an Augmented Space:
- Corner Pinning: The Keystone library is utilized to distort the mesh surface of the airflow render, projecting the light perfectly onto the physical whiteboard’s boundaries (Seamless Alignment) to prevent any offset between the physical ink and digital wind.
Virtual Probe HUD: A digital sensor interface allows users to hover over any coordinate on the physical board to read real-time quantitative wind velocity (m/s).
Techniques and Materials :
Result / Conclusion :
9.1 System Performance Results By integrating computer vision and GPU-accelerated fluid dynamics, ChromaFlow successfully processes and projects the vector field onto physical sketches in real-time with zero latency. The Virtual Probe reads quantitative wind velocities with high precision, significantly reducing the time required for rapid iteration during the early conceptual design phase.
9.2 Collective Experience and Shared Postmemory A defining feature that distinguishes ChromaFlow from conventional CFD software is its “Social Affordance.” Standard engineering software is typically designed for individual use, where monitors and mice confine students to isolated interactions. Conversely, ChromaFlow transforms a sketching board into a collaborative space. This setup, which allows users to gather around, discuss, point at physical features, and collaboratively edit drawings, fosters a peer-to-peer learning environment. In the context of “Postmemory,” this collaborative workflow mirrors vernacular traditions—such as local carpenters and community members gathering to discuss house orientations and natural ventilation. Therefore, ChromaFlow not only preserves the individual memory of hand-sketching but also successfully revives the “Collective Memory” of shared architectural knowledge and group spatial judgment in the digital age.
9.3 Conclusion and Future Work ChromaFlow succeeds in synthesizing fluid dynamics knowledge with the physical memory of design through Augmented Reality technology. The tool concretely enables architecture students to observe and comprehend complex wind behaviors. For future development, the project aims to integrate real-world weather API data and develop a 3D building height analysis system based on ink color density, further enhancing the precision and capability of microclimate analysis.
References :
Kolb, D. A. (1984). Experiential Learning: Experience as the Source of Learning and Development. Englewood Cliffs, NJ: Prentice-Hall.
Batchelor, G. K. (2000). An Introduction to Fluid Dynamics. Cambridge University Press.
Stam, J. (1999). Stable Fluids. Proceedings of the 26th Annual Conference on Computer Graphics and Interactive Techniques (SIGGRAPH ’99), 121–128