PhD place available - Materials 4.0 CDT on digital modelling of material response to environmental and operational loading

FLAME-GPU accelerated agent-based modelling of material response to environmental and operational loading

This PhD links research in Mechanical Engineering with the latest computing developed in Computer Science in Sheffield. 

Mechanical Engineering supervisors: Prof. David Fletcher, and Prof. Roger Lewis

Computer Science supervisor: Prof. Paul Richmond

Partner company: British Steel

Project details

Summary: Research already underway Sheffield has developed an accelerated modelling technique for materials damage, applying it to the steel used in railway rails. This uses computing originally developed for graphics and gaming, but applies it to accelerate engineering modelling. This PhD aims to take this work forward and extend it.  

‘Agent-based’ modelling (ABM) simulates large numbers of autonomous yet interacting entities and in Sheffield has recently seen the first results in application to engineering materials for wear and plastic damage in steel. A significant opportunity is FLAME GPU general-purpose modelling framework developed in Sheffield to enable massively parallel processing of ABMs on NVIDIA graphics processing units (GPUs), without the need for specialist understanding of GPU programming or optimisation. This project will explore the technique and broaden its application to include additional materials for a range of environmental and operating conditions. The project aims to develop a data driven GPU accelerated agent-based materials modelling tool covering mechanical and thermal response, with a framework open to extension for impact, erosion, and corrosion

The research to-date has focused on mechanical loading for steel. It is planned that this PhD expand this for additional loading types and additional response mechanisms. A key factor is use of experimental data with the computing framework enabling a hybrid approach not possible in techniques such as finite element modelling. For example, plastic flow of steel is measured experimentally and the data embedded in the model to handle aspects where fundamental plasticity models struggle. Predictive application for alternative loading cycles can explore the material response, validated against additional physical tests.

Discretisation of the material in the ABM enables study of wear (loss of integrity in near surface agents), and crack initiation (discontinuity between agents), with peri-dynamic inspired interagent bonds defining the structural integrity of the material. In current research the GPU acceleration relative to CPU based modelling ranged 8 to 30 times depending on the hardware combinations considered.

Following success with plastic flow and wear modelling the motivation is to expand ABM as a high-speed technique for a range of materials. This includes additional loading types (environmental, manufacturing processes, operational), and additional response mechanisms (initially thermal response, and an ambition to explore others such as impact, erosion, and corrosion).

Enquiries

For project-related queries, please contact the lead supervisor, Prof David Fletcher ( d.i.fletcher@sheffield.ac.uk ). 

Application Web Page

This project is part of cohort 3 of the EPSRC CDT in Developing National Capability for Materials 4.0, with the Henry Royce Institute.

https://www.sheffield.ac.uk/postgradapplication/login.do 

After the personal details, you need to 'add research course', and select 'Doctoral Training Course', and then 'Developing National Capability for Materials 4.0'.  The Materials 4.0 CDT is committed to Equality, Diversity and Inclusion. We strongly encourage applications from underrepresented groups.