The Benefits Cloud Simulation Can Bring to CFD

Smoke Control in Subway Tunnels: Revolutionizing Safety through HPC-Based...

September 21, 2023 Comments Off on Urban Microclimate Modelling Using CFD Views: 159 CFD, HELYX, Success Story

Urban Microclimate Modelling Using CFD

At ENGYS, we take immense pride in delivering cutting-edge Computational Fluid Dynamics (CFD) solutions that drive success for our valued clients. One such instance involves our long-term partner SoftSim, a Bulgarian wind engineering consultancy, with yearlong experience in urban physics modelling and industrial process automation. Together, we embarked on the venture of developing a comprehensive, Cloud-based CFD solution, tailored for Urban Microclimate Modelling applications. This article delves into how the recent implementations in ENGYS’ flagship product, HELYX, helped SoftSim optimize its’ day-to-day processes, boost the company’s productivity and maintain its’ competitive edge.

Battling Limitations: SoftSim’s Quest For Enhanced CFD Capabilities

As the Architecture, Engineering, and Construction (AEC) industry is rapidly becoming more digital, it increasingly relies on Urban Microclimate Modelling and Computational Fluid Dynamics (CFD) for informed decision-making. Together with the technological advances in the field, the volume and sophistication of the required CFD analyses is also increasing. The assessment of different site configurations, multiple wind conditions and directions (typically up to 36) on large computational grids (up to 150 – 200M cells) is a requirement in modern wind microclimate investigations. This necessitates not only the employment of state-of-the-art high-performance computers but also a decent level of analyst expertise. As a result, many SMEs in the field, including SoftSim, struggle to cope with these demands and are seeking for more efficient solutions that will allow for sustainable growth and competitiveness.

Strength By Numbers: FF4EuroHPC and Cloud-SOPHIA Consortium

Aimed at addressing these challenges, a partnership was established between ENGYS, SoftSim and the High Performance Computing Centre of Stuttgart (HLRS). This venture was named Cloud-SOPHIA (which stands for Cloud-based Simulation of Pedestrian Urban Microclimate for Health Impact Assessment and Architectural planning). The 15-month operations were funded by the FF4EuroHPC project, a European initiative with the scope of facilitating innovation and business development through connecting European SMEs with HPC-related technologies.

Image shows members of the consortium of the CLOUD-Sophia project
Members of the Consortium

The Solution: UPM Tailored Process Chain and Cloud-Based HPC Integration

Together with the other members of the consortium, we devised a highly scalable Urban Physics Modelling (UPM) solution based on three key ingredients: a) enhanced algorithms and methodologies tailored for UPM applications, b) automated workflows integrated in HELYX and c) a user-friendly client/server interface that would allow utilizing HELYX capabilities on HLRS’ HAWK cluster.

Through offering seamless on-demand access to High-Performance Computing (HPC) resources, the developed solution not only allows SoftSim to perform more and more sophisticated CFD analyses on finer computational grids but also substantially reduces turnaround times through the high level of automation introduced. These newly developed workflows encompass all key atmospheric studies, from pedestrian comfort analysis and pollutant concentration forecasting for commercial space viability assessment, to wind driven rain deposition, and structural analysis for preliminary calculations during early design stages.

Honing In On The Details: Validating & Fine-tuning The Solution

To quantify the accuracy of the developed methodologies and overall performance of the solution, an extensive validation study was conducted by SoftSim. The CFD results were compared against experimental data from: a) Architectural Institute of Japan (AIJ) benchmark cases, b) Nova Fluid Mechanics (FM) wind tunnel measurements and c) other publicly available data. Two of the most interesting cases that were modelled during our validation study were the pedestrian comfort assessment of Palmerston Road recent developments and the pollutant dispersion prediction over the Atsugi campus of Tokyo’s Polytechnic University (TPU).

Pedestrian Comfort Assessment – Palmerston Road Developments

Accurately predicting how the introduction of new developments will affect the urban microclimate  within residential areas is key to ensuring wind and thermal comfort for pedestrians. Such an assessment was of particular interest for the case of Wealdstone Heights developments at Palmerston Road, Harrow. The new configuration comprised 2 high-rise, 2 mid-rise and 3 low-rise buildings, and were expected to impact the developing wind speeds and thus the local microclimate.

Actual Palmerston Road developments (on the left), CAD representation of the existing site and proposed developments (on the right). Image source.

An experimental investigation was performed by the wind engineering experts of NOVA FM by the means of wind tunnel testing. For the purposes of this study, a scaled geometrical representation (1:400) of the actual site was created. A total of 66 Irwin probes were used for monitoring wind speed at various locations within the site, both at street level and on terraces.

Scaled model (1:400) inside the wind tunnel (on the left), Irwin probes configuration on key locations (on the right).

With the wind tunnel testing data at hand, courtesy of Nova FM, we proceeded with validating our methodology for pedestrian wind comfort analysis on the proposed developments. The full CFD study required the assessment of 16 different wind directions (every 22.5 degrees). In terms of performance, the newly developed Multi-Instance Framework (MIF), which allows running multiple configurations with a single execution, along with the employment of a cylindrical fluid domain, which erases the need for rotating the geometry and re-meshing, allowed SoftSim to efficiently setup and complete the study. Additionally, distinct were the benefits of employing the more efficient HELYX pressure-velocity block-coupled solver which offers substantial reduction of turnaround time, when compared to the traditional segregated approach.

Performance comparison between Coupled and Segregated for different wind directions

In terms of accuracy, the model’s results demonstrated good correlation with the experimental data, accurately predicting the Lawson categories for 57 and 58 out of 66 measuring locations for winter and summer seasons respectively. From the points that were in no agreement with wind tunnel measurements, more than 60% were within 10% of the experimental prediction in both cases. The following image presents how the CFD-predicted Lawson categories (presented on a plane normal to the ground) compare to the wind tunnel measurements (presented by colored points).

Comparison between Lawson categories predicted by CFD and wind-tunnel experiment.

Pollutant Dispersion Forecast – Tokyo Polytechnic University, Atsugi Campus

Another principle that is increasingly becoming more important within the UPM application spectrum is pollutant dispersion modelling. For validating the related methodology that was developed during the early stages of Cloud-SOPHIA experiment, we selected the latest benchmark case that was published by AIJ late 2022. This case was based on the thorough experimental study that was carried out by Mr. Tachibana and others, aimed at creating a solid validation case for pollutant dispersion prediction. The study included not only field measurements but also wind-tunnel testing data. Tokyo’s Polytechnic University Atsugi campus was selected as the target region.

Actual Atsugi Campus site (on the left), Wind tunnel model configuration (in the middle – image source), CAD representation (on the right).

The employed CFD model was targeted at replicating the wind-tunnel experiment, rather than the full-scale case. Velocity components and dimensionless concentration of Ethylene (C2H4) were monitored at 15 different locations scattered throughout the campus.

The numerical results, once again, demonstrate not only sufficient approximation of the overall trend both for velocities and concentration but also exceptionally fast turnaround times. Especially, when compared to the more accurate, yet computationally expensive Large Eddy Simulation (LES) approach, the newly developed methodology based on steady-state solution of Reynolds Averaged Navier-Stokes (RANS) equations can offer a viable solution for preliminary calculations during early stage assessments.

Velocity contour plot at 2m above ground level (one the left), Ethylene concentration contour plot at 2m above ground level (on the right)
Graph showcasing comparison of dimensionless concentration prediction between wind tunnel experimental data, field measurements, RANS and LES predictions.

The good correlation between numerical results and experimental data that was demonstrated throughout the validation cycle, not only confirmed the reliability and fast turnaround times of the developed process chain but also paved the way for further refining the original approach.

Putting It To The Test: Sofia City Center

Aimed at assessing the scalability of the developed solution, the new process chain was adopted on a large-scale case, which requirements were representative of the demanding contracts that wind engineering consultancies are often faced with.  For the purposes of our study, a 2 km x 2 km area around the city center of Sofia, Bulgaria was considered as target region. SoftSim’s experts created a detailed geometric representation of Sofia (including elements like, detailed canopies, trees, fences etc.) and more than 160M cells were required for properly discretizing the fluid domain.

Sofia city center (upper left), 3D CAD model of Sofia (upper right) and computational grid generated in HELYX® (lower left and right).
Velocity contour plot at 2m above the ground level.
Iso-surface of NO2 concentration.

A total of 36 wind directions (every 10 degrees) were assessed. Fast convergence rates offered by the HELYX Coupled solver together with the high level of automation introduced by the developed methodologies made it possible to complete the full investigation (pedestrian comfort analysis, pollutant dispersion forecast, wind driven rain deposition) within approximately 48 hours on 1024 cores. Of course, this achievement would not have been possible if it wasn’t for HLRS’ state-of-the-art computational resources that are now easily accessible via HELYX.

Video: Velocity contour plot 2m above the ground for different wind directions.

Empowering Growth: The Business Impact and Benefits

With approximately 60 person-months and over 500,000 core-hours allocated for the successful completion of this endeavor, the results of our collaboration were nothing short of remarkable. A thoroughly validated, fined-tuned by experts for reliability and performance CFD tool based on HPC has been brought to the table. The developed solution does not only streamline the typical processes that are required for various UPM applications, but also reduces the complexity for conducting multi-disciplinary atmospheric studies, ensures faster turnaround time and alleviates the cost of financing on-premises computational resources. Furthermore, the open-source character of HELYX allows experts to access the code, add value to the tool through their expertise and fine-tune it based on their specific needs. All these, add up to a comprehensive, highly-scalable and versatile solution that enables engineers and urbanists make better-informed decisions, faster.


SoftSim’s website:

FF4EuroHPC’s websit:

AIJ website:

Nova Fluid Mechanics website:

Comments are closed.