Computational Structural Mechanics Lab (CSMLab)

Department of Engineering

a

Research Overview

In broad terms, our research is focused on computational modelling and analysis of light-weight structures originating from a wide range of applications, including structural engineering, aerospace and marine engineering. Specific areas of research emphasis include

  • advanced discretization methods for integrated design, optimisation and analysis of solids and structural components such as beams, membranes and shells,
  • computational fluid-structure interaction with application to micro-air-vehicle design and insect flight, and
  • mechanical modelling and computational analysis of large-scale structural membranes, such as parachutes and inflatable decelerators for outer-space applications.

cartoon overview

In terms of computational methods development two overriding themes throughout our research are

  • the integration of, traditionally disparate, computational analysis and geometry using immersed and b-spline based finite element discretization techniques,
  • the development of scalable computational methods for simulating challenging engineering problems on high-end parallel computers.

Selected Active Research Projects

Computational toolbox for fluid-membrane interaction

This research addresses the need for computational tools and techniques for the aeroelastic analysis of fluid-membrane systems. The focus is on development of a comprehensive computational toolbox for analysing the dynamics of fluids and embedded structures with the goal of obtaining high-fidelity predictions for aerodynamic design parameters, like drag and lift, and structural design parameters, like maximum stress and deflection. Although the equations for viscous, incompressible fluid flow and membrane dynamics are straightforward and well known, it is still challenging, if not impossible, to perform fluid and membrane coupled computations in the presence of large deformations. Besides fundamental differences in the mathematical structure of fluid and solid equations, progress is hampered by the vast disparity of the physical length and time scales involved. To address the first issue, the computational toolbox will include a mathematically rigorous and algorithmically robust formulation for representing the coupled dynamics of a fluid with an immersed membrane. For resolving the length and time scales (for Reynolds numbers of up to 3000) a two-pronged approach will be followed. First, the developed techniques will be scalable to large system-level, three-dimensional simulations, with up to billions of unknowns, which will be achieved through systematic utilization of high-performance computing platforms. Second, the influence of the unresolvable sub-grid scales on the large-scale motions of the fluid flow will be explicitly modelled with a multi-scale method.


High-performance computation of structural failure

In assessing the safety margins of engineering structures the ability to computationally predict structural and material failure is crucial. A prevalent form of structural failure is due to emergence of strongly localized or discontinuous deformations, for example in form of cracks or shear bands. The appearance of strongly localized deformations constitutes a major problem to the state-of-the-art simulation techniques. Within this effort the finite element discretization method will be used and the discontinuities will be, if required, adaptively inserted into the discretization. The proposed algorithms will not impose any constrictive limitations on the possible failure modes and their dynamic evolution. The parallel implementation will make effective use of open source software, such as PETSc (Portable, Extensible Toolkit for Scientific computation). Software development and validation will be performed on a small-sized local computing cluster. The actual scalability studies and production runs will be conducted using massively parallel supercomputers at the Cambridge High Performance Computing Facility.


Component and assembly level optimization of industrial electromechanical devices

We propose to facilitate a paradigm change from state-of-the-art Simulation-Based Design to novel Optimisation-Based Design of complex electromagnetically-driven industrial products. This new design philosophy will first enable automatic, physics-governed re-sizing / miniaturisation of complex device assemblies and at the same time enforce a re-thinking of industrial design in general. Our complete physics-based optimization system includes: component-level optimisation, automatic mesh smoothing/exporting, fast boundary element solvers, design-criteria evaluation, assembly-level optimisation and is importantly supported by high-performance computing.

This is a joint project with ABB R&D (Zurich), Technical University of Munich and Technical University of Graz. See also the CASOPT project web page.


Three-dimensional computational analysis of supersonic disk-gap-band parachutes

The development of new entry-descent-landing sequences for planned heavy payloads to be sent to Mars requires the use of new supersonic parachutes. Our understanding of the performance of these supersonic disk-gap-band (DGB) parachutes is limited. We propose to use high-performance three-dimensional coupled large-eddy and finite-deformation structural dynamics simulation to aide in the development of these new systems. The objective is to improve our understanding of the multiple stability domains and performance parameters, drag, lateral forces and structural loading, of the parachute-payload system in the entry, descent and landing sequence of Mars missions. To aide in mission design and planning, we propose to use the most accurate modeling and simulation techniques available today, including: large-eddy simulation, compressible subgrid scale modeling, shock capturing, finite element modeling of the parachute canopy and the suspension lines, adaptive mesh refinement, dynamic error control and parallelization. These techniques have been developed, in part, under the DoE funded ASC program at Caltech and have been applied to a number of verification and validation studies of flows/structure interaction problems.

This is a collaboration with Prof. Carlos Pantano, University Illinois at Urbana-Champaign. We use the Virtual Test Facility within this project.


Multiscale analysis of structural membranes

Abstract to be added soon.



Support by our sponsors is gratefully acknowledged
EPSRC Royal Society The Cambridge Trusts JPL/NASA
European Union (FP 7) ESA CUED Partnership for Advanced Computing in Europe