CFD Analysis of Fire & Smoke in AR – Opera Theater Duisburg, Germany

The present article aims to encourage and support the discussion of combining Computational Fluid Dynamics (CFD) with IoT technologies in a more interactive and, perhaps, useful way. The ultimate goal is to complementary integrate mathematical modelling, real-time data acquisition and visualization in a more comprehensive approach by taking advantages of smart devices.

This work focuses on Fire & Smoke CFD Modelling in Augmented Reality, aiming to propose a useful cross-field correlation method to (hopefully) help and support everyone involved in fire prevention. Files for Download available at FetchCFD.com.

Introduction

In recent years, the numerical assessment of fluid dynamics, energy and, ultimately, safety performance of buildings has been constantly raising interest in the scientific community. The increased computational power commonly available an the impressively accurate numerical modelling, that has been recently developed and provided, has been helping the so called “numerical computing democratization”. In particular, the prediction of all the possible (critical) scenarios in Fire & Smoke modelling represents one of the main fields of research, where the integration of interdisciplinary methods may play a rather peculiar role. Both integration and cross correlation of IoTAIAugmented Reality and Digital Twin modelling, may represent the future of engineering in terms of prevention, performance prediction and design optimization.

Geometry pre-processing and CFD Analysis with FDS

The CAD models of Duisburg (Germany), where I currently live, have been downloaded from Google SketchUp Warehouse (see previous article) and then processed using FreeCAD.

The computational domain, the geometry of the Opera Theater and all the other boundary conditions have been imported into Blender, to be processed, meshed and exported for Fire Dynamics Simulator (FDS) using the plug-in BlenderFDS.

The mesh domain accounted for 640k hexa cells, with a spatial resolution of 0.5m in x, y and z direction. Even though the mesh was rather “coarse”, it was still sufficient to mimic and describe a realistic scenario. The simulation has run in parallel on my laptop on 4CPU till 30s of physical time was reached. During calculation, files in plot3D format have been exported every 0.25s. Four lamps for external lighting are assumed to catch fire at the same time. Even though (I admit) it was not correct, I used (a lot of) liquid Heptane (C7-H16) as fuel (easy and sooty option, with a soot yield value of 0.037).

Data Post-Processing and Android App deployment

Once the simulation terminated, plot3D files have been imported in Paraview for post-processing and then converted into a vectored data file format using a small piece of code I wrote myself some time ago.

The last step consists of importing the processed date into Unity3D, set up the scene and deploy the app for Android.

3D Iso-surfaces and 2D cross sections of smoke density have been selected as elements to be rendered. In order to visualize soot oxidation (occurring at the buffering interface between dense smoke and fresh air) while adding some realistic effect, virtual elements are contoured with velocity magnitude in grey scale, reducing the color scale intervals from 256 levels (default in Paraview) to (literally!) 50 shades of gray… which provided the smoothest color transition with the lowest intervals of color

AR Rendering

For the AR experience, a rather common Samsung S8+ was used. The CPU and the RAM provided a very smooth image transition. The recorded movies are very fluid featuring a crispy image resolution in HD.

As mentioned before, four lamps (or electrical devices, see the related picture) are assumed to catch fire at the same time (a kind of short circuit failure). In the animation 3D Iso-surfaces of smoke density and a 2D cross sections (passing through the middle of the lamps) are chosen as representative elements to depict smoke advection and spatial distribution. Furthermore, a no-wind condition was assumed (which was the real weather condition occurring during the recordings).

Indeed, in case of indoor fire and smoke, visibility plays a critical role, since a clear vision of the excape route may save lives. Rendering visibility in AR using parametric transparency is indeed a challenge to be tackled in the next future. For the time being, full transparency based on smoke density level cut-off values (thresold levels), represents an easy and viable solution.

Future Work

Future work will be oriented towards the integration of CFD, AR and the IoT realm. Next step will be to simulate an indoor fire condition to test, check and show the capability and usability of simple smart devices (which might still turn of practical use in critical situations and help out the evacuation procedure out of any building) to virtually overlay smoke visibility and ensure that escape routes will still be visible.

Leave a Comment