£6,000 FindAPhD Scholarship | APPLICATIONS CLOSING SOON! £6,000 FindAPhD Scholarship | APPLICATIONS CLOSING SOON!

Deep Learning (DL) based methods for Pedestrian-Level Wind (PLW) assessments

   Cranfield Defence and Security (CDS), Shrivenham Campus

   Saturday, August 20, 2022  Funded PhD Project (UK Students Only)

About the Project

With the complexity of modern urban areas, pedestrian wind environment analysis becomes a critical factor in urban and building planning design, helping to ensure the overall wellbeing, safety, and comfort in pedestrian zones.

The fluid flow simulations enable architects and engineers to predict and optimise the performance of buildings in the early stages of the design process. Traditional Computational Fluid Dynamics (CFD) methods produce high-accuracy results, but they are computationally expensive and do not work well in the design process of new prototypes in different domains.

To obtain reliable results, it often takes several hours or days depending on the prototype’s complexity. We aim to explore Deep Learning (DL) with the objective of creating an interactive tool for testing new designs, even when they are getting computationally hard for physical solvers.

In particular, we plan to go in a similar direction and define Deep Learning based (DL-based) architectures that can generate wind flows for arbitrarily shaped buildings in scenarios with different levels of complexity (city maps) with the motivation of building a surrogate model that can be used in an interactive tool for smart building assessments.

Novelty and commercialisation

As the CFD approach lacks flexibility and leads to high pricing, online tools are starting to appear in the market. Existing solutions require, however, significantly higher work in term of external data gathering, pre-processing, user expertise, and manual input implementation. While this could be an attractive way to solve complex problems, even with cloud computing and unlimited power, this solution does not lead to fast enough results that can allow the user to work interactively on their models during the design process.

Pedestrian-level wind (PLW) assessments for new buildings are particularly complex and costly. The technical accuracy, resolution, and confidence will be in line with the consultant's ability. Realistic alternatives for a customer are either to use an engineering consulting company or establish an in-house solution by purchasing or subscribing to expensive modelling and simulation tools, and extensively training staff to perform this task. Nabla Flow is working on a full-service process to streamline the simulation workflow by automating the whole chain from model placement to report delivery. This new development represents a game-changing innovation for architects, designers, city planners and the whole engineering consultancy industry.

Objectives of the PhD project 

The primary objective of the PhD project is to develop an DL-based technique as a fast alternative of CFD for PLW assessments calculations in an urban environment. This will be accomplished by creating a fully tested set of DL-based methods replacing the current CFD-based method for PLW calculations.

The following secondary objectives (SOs) will support the realisation of the primary objective.

• SO1: Design and development of a complete ML-pipeline for testing and validating the DL-based methods for PLW calculations. This workflow includes a complete software framework for:

o SO1.1: Data preparation (i.e. data collection, pre-processing, cleaning, transformation) 

o SO1.2: Definition and implementation of DL-based model as surrogate model 

o SO1.3: Training, Test and Validation 

o SO1.4: Model deployment including versioning, provisioning and access control 

• SO2: Integration and Testing: all the tested models will be integrated and deployed in a final product

• SO3: Define a framework for cross-validation of the results of the proposed prototype through the support of domain experts (i.e. AI architects; aerodynamicists; civil engineers; urban architects and planners)

Cranfield Defence and Security (CDS) provide unique educational opportunities to the Defence and security sectors of both public and private sector organisations.

Based at the UK Defence Academy at Shrivenham in Oxfordshire, CDS is the academic provider to the UK Ministry of Defence for postgraduate education at the Defence Academy, training in engineering, science, acquisition, management and leadership.

The PhD project is expected to develop an DL-based technique as a fast alternative of CFD (without any compromise on accuracy) for PLW assessments calculations in an urban environment. The PhD project findings are expected to have academic, economic, and societal impact.

In addition to tuition fees, stipend and consumables, travel & subsistence (including the registration fees if applicable) for external training, international conferences, and industrial partner are covered. 

The student will have access to subject experts in aerodynamics and machine learning to successfully complete the PhD research project in a timely manner. This studentship will enable the PhD student to develop technical, programming, and transferrable skills, which will improve his/her employability.

This is a fully-funded PhD studentship, which is funded by Nabla Flow (Norway) and Cranfield University (UK).

SupervisorDr Karthik Depuru Mohan

Dr Luca Oggiano and Dr Knut Erik Teigen Giljarhus will act as industrial advisers at Nabla Flow, Norway.

Funding Notes

This is a fully-funded PhD studentship, which is funded by Nabla Flow (Norway) and Cranfield University (UK). It covers tuition fees (£4,500 per year); additional fee element for research projects (£7,500 per year); stipend (£18,000 per year); travel and subsistence (£3,000 per year); and consumables (£500 per year) over the whole period (3 years) of PhD studies.


Entry requirements
Upper Second Class MEng in computer science or aerospace/mechanical engineering
• Knowledge of computational methods and their implementation using relevant programming languages
• Upper Second Class MSc in advanced computational methods
• Basic knowledge of fluid mechanics or aerodynamics
• Basic knowledge of machine learning techniques

Email Now

Search Suggestions
Search suggestions

Based on your current searches we recommend the following search filters.

PhD saved successfully
View saved PhDs