Composite Analysis, Modelling and Prediction
Over the last thirty years, the Composites Centre has pioneered analysis and modelling of composites, and now are world-leading in this field. Models developed within the Composite Centre underpin many of the commercial predictive finite elements codes, such as ABAQUS. The research into analysis and prediction in the Composite Centre is diverse including composite failure prediction, damage growth modelling, impact behaviour and material design.
Development of failure models and criteria for composites
We have developed physically-based failure models for various composite materials, including UD plies and woven plies. Our models were ranked top in an international benchmark exercise (the 2nd World-Wide Failure Exercise). Many of our models have been Integrated in the commercial releases of the finite element codes Abaqus and LS-Dyna.
We have also developed original analytical models to predict the mechanical response of unidirectional composites under longitudinal tension. These models are able to predict damage accumulation with progressive loading, stochastic size effects on the strength of composites, and the influence of fatigue on the tensile failure process and load-carrying ability.
Soraia jointly organised and ran (together with Y. Swolfs (from KU Leuven) and H. Mortan (from the University of Southampton)) the first international benchmark exercise for tensile failure of composites. In this exercise, Soraiaâ€™s analytical model was 300 to 20,000 times faster than the other (FE-based) models (Mines ParisTech and KU Leuven, respectively), and its blind predictions had the best correlation with the wide range of independently-obtained experimental data (obtained at the University of Southampton).
Randomly-oriented tow-based discontinuous composites
Randomly-oriented tow-based discontinuous composites (also known as carbon-fibre SMCs, randomly-oriented strands, and randomly-oriented chopped-prepreg) and a new class of composites which can be processed in less than 5 min, and are as stiff and even tougher than conventional laminates. Several industries are deeply committed to these materials, and Lamborghini, Airbus, Hexcel, ELG Carbon Fibre and Simulia are official partners in this research. Soraia’s group has experimentally demonstrated (for the first time in the literature) the equivalence between randomly-oriented discontinuous composites and layer-by-layer laminates, and used this equivalence to develop models for the response of these materials. These models have been used to design improved material microstructures, and have been implemented in an FE framework to design automotive components, in a collaboration with Lamborghini. This work has been recognised by the Tsai Award for Best Student Presentation (Yizhuo Li, 2017).
Design of Polymeric Armour
A continued need exists for high performing, lightweight and cost effective protection for personnel and vehicles to improve survivability and reduce injury, when subject to a wide range of potential threats. Such protection must cover threats which generate extreme loadings, e.g. blast and ballistic impacts. Typical protective armours can be either ceramic or polymer based.
Most polymeric armours use high performance fibres, which do not exhibit the brittle nature of high stiffness carbon and glass fibres, but have a greater strain to failure. To understand the behaviour of such complex polymeric systems a material modelling approach has been developed capable of predicting the ballistic performance of polymeric armours and includes the scissoring of the laminates observed during ballistic impacts. The approach has been successful used in the design several polymeric systems.
Stochastic modelling of heterogeneous materials - size dependence and uncertainty quantification in strength of composites
Dr V. L. Tagarielli, Andreas Schiffer
How do we predict the mechanical response of large composite structures using as input the measured response of small coupons? Can we quantify the uncertainty in the mechanical response of composite structures? What is the relation between tensile strength and fracture toughness in brittle and quasi-brittle materials? Is it possible to measure (rather than guessing) damage evolution laws for fibre composites? How do we deal with micro-mechanical modelling of materials displaying a strain-softening response?
To answer these questions we conducted experimental and modelling work focusing on stochastic aspects of the mechanical response of materials, and in particular of fibre composites loaded through the thickness. We introduced a new revolutionary modelling philosophy by which mechanical properties are not assumed, as usual, to be uniform, but rather to be described by correlated and uncorrelated random fields. We treat materials as a N-phase composites, to capture the effects of microstructural heterogeneity upon the macroscopic material response. By doing so we construct predictive numerical models containing a set of physically meaningful length-scales; such models are stochastic (they predict a different outcome every time they are run) and size dependent. We also develop highly instrumented new mechanical testing protocols to measure directly the damage evolution laws of composites and the spatial variability of this law, at both low and high strain rates. Such models predict realistic outcomes, do not necessitate 'calibration' of non-physical parameters (hence they are truly predictive), and capture the scatter and size dependence of the mechanical response of composites.
Modelling the electro-mechanical response of CNT-polymer composites
Recent technological advances have allowed economical manufacturing of relatively cheap polymer composites reinforced by carbon nanotubes,and these materials have attracted great interest from industry. A small volume fraction of reinforcing CNTs (of order 1%) confers electrical conductivity to the polymeric composite, making it suitable for a number of applications, such as for example flexible electronics and electromagnetic shielding. Furthermore, these materials are self-sensing: the electrical conductivity is strongly dependent on the applied strain field, such that there is a potential to use these materials also for the manufacturing of inexpensive sensors and in structural health monitoring applications. We have developed theoretical and numerical physically-based models able to predict the strain sensing response of these materials. The models, based on the Finite Element method, are quite general: they capture the details of electron tunneling transport between adjacent nanotubes, are calibrated only by measurable parameters, and predict the dependence of conductivity upon an arbitrary multiaxial strain state. We have then developed machine learning techniques to perform such predictions with savings in computation time of at least 5 orders of magnitude compared to the FE simulations.
Blast response of composite structures: measurements and modelling
Dr V. L. Tagarielli, Andreas Schiffer
We have developed new experimental techniques to observe, at laboratory scale and without the use of explosives, the response of composite structures to blast in air and water.
In the case of air blast, experiments involved structural impact by soft projectiles, with mechanical properties tailored to reproduce the dynamic loading pulse exerted by impinging shock-waves in air. To reproduce underwater explosions we developed a transparent water shock tube to allow simultaneous dynamic observations of the structure and the fluid, including the evolution of cavitation processes, to deduce the prominent role of fluid-structure interaction in such events.
We used these techniques to validate theoretical and numerical predictive models of these events. Models allowed optimization of monolithic and sandwich composite structures subject to the threat of explosion, providing precious tools to design engineers.
I am considered a world-leader in this area, and a full list of relevant publications is available in Google Scholar.
Development of image-based numerical models for predicting the microstructure-property relationship in particulate composites
Numerical methods that can provide predictions of the mechanical response of particulate polymeric matrix composites as a function of volume fraction and particle mean diameter are needed as a design tool for materials composites engineers. A generic methodology has been derived which has been applied to special cases such as a ceramic particle reinforced polymer matrix. Representative Volume elements are determined through images obtained from Scanning Electron Microscopy (SEM). The model takes into account the possibility of failure through interface debonding as well as cracks through the matrix. The model predictions for the modulus and fracture strength of the composites are validated through independent experiments on the composite. The numerical results are also used to qualitatively explain the trends measured regarding the fracture toughness of the composites. Compared to other literature on particulate composites, our work is the first to report accurate stress-strain distributions as well as fracture predictions whilst all the necessary model parameters defining the failure criteria including that of the interface are all derived through independent experiments. This paves the way for a relatively simple methodology for determining structure-property relationships in composites design.
Multi-scale models for damage and deformation process prediction in highly filled polymer matrices
Modelling the deformation and failure processes occurring in highly filled composites (approximately 95% volume fraction) is extremely complex as the very wide particle size distribution necessitates the development of multiscale models in the form of hierarchical simulations. An example of such materials are polymer bonded explosives (PBX) and other energetic materials; predictive models for their mechanical response are of great importance for processing methods and lifetime storage purposes. Crystal debonding is undesirable since this can lead to contamination and a reduction in mechanical properties. The particle/matrix interface is characterised with a bi-linear cohesive law whereas the soft matrix is modelled through visco-hyperelastic or visco-plastic models which take into account the extreme time dependence of such soft polymers. Once calibrated, the material laws are implemented in finite element models to allow the macroscopic response of the composite to be simulated. Real SEM images as well as simulated microstructures with the same size distribution and volume fraction can be accommodated. The models are validated using experimental data. Their accuracy enables the quantification of parameters such as strain rate, cohesive parameters and contrast between filler and matrix moduli to be readily examined, replacing costly and time-consuming trial – and error experimental methods.
Determination of Mixed-Mode Cohesive Zone failure parameters using Digital Volume Correlation and the Inverse Finite Element Method
Experimental determination of the mix-mode cohesive zone model parameters is very challenging. A collaboration with Prof John Lambros at the University of Illinois at Urbana-Champaign is currently underway where the possibility of using digital volume correlation (DVC) data and the inverse finite element method is examined. DVC data are derived from compression experiments of a PDMS cylinder with a spherical inclusion. The test is also modelled using the finite element method. A bilinear traction separation law with a linear mixed-mode relationship is used to describe the interfacial behaviour. Full factorial experiments are being performed for the four cohesive parameters in order to determine the target cohesive fracture energies and damage initiation stresses through an inverse optimisation technique.
Mechanical and electrical modelling of multifunctional composites
We are developing models to predict the electrical and mechanical performance of structural power composites. The mechanical models are investigating the compaction of the laminates when we assemble the devices and subsequently the in-plane shear loading response. The electrical models are characterising the charge/discharge behaviour of the devices. These models utilise the constituent properties, and hence facilitate parametric studies into the critical factors (such as electrode spacing, electrolyte properties, etc) for structural supercapacitor design and optimisation. In the longer term this activity will couple the mechanical and electrical models to predict interactions, and ultimately provide a framework to support qualification of structural power composites.
High Strain Composites
The use of thin composite flexures as compliant hinges for deployable space structures is extremely attractive due to their capability to self-deploy through the release of stored strain energy and to self-lock upon deployment through controlled buckling. Ongoing research in this field includes: numerical modelling of folding and deployment; characterizing the failure of extremely thin composite structures; high-precision manufacture of high strain composite flexures. The effects of long-term stowage are also currently poorly understood, in particular relaxation due to viscoelastic effects, and this is also the subject of investigation.
The directional material properties of composites can be harnessed to shape the large-scale deformation of aerostructures under aerodynamic forces. This becomes a passive method for control of aerodynamic loads, known as aeroelastic tailoring. A typical situation is the bend/twist coupling introduced on the largest wind turbine blades, which are tailored to pitch down when they bend upwards under high wings, therefore reducing the overall strength requirements. At Imperial, we are building high-fidelity computational models for optimization of composite layout in coupled fluid-structure interaction environments, and seek combine passive/active solutions in which aeroelastic tailoring is complemented by feedback control systems in the high-frequency load spectrum.
Details of the Aeroelastics research group can be here.
Floating Node Method
Silvestre T Pinho, B-Y Chen (TU Delft), Nelson V de Carvalho (NASA Langley)
The floating node method allows for complex networks of interacting cracks to be simulated efficiently. We have implemented this method in a commercial FE code and demonstrated the effective simulation of 3D problems involving hundreds of cracks in complex networks involving delamination, splitting, migration of delaminations and fibre failure.
Multiscale analysis of composites
We work on various multi-scale modelling approaches for the analysis of failure in composites. Recently, we developed a Mesh Superposition Technique which can link different types of elements without creating artificial stress concentrations and without reflecting stress waves in dynamic problems. For further details, see here.
Molecular Dynamics Finite Element Method
We have developed a molecular dynamics finite element method which allows for efficient simulation of complex atomistic structures atom by atom. This method represents the force fields from molecular dynamics exactly, but runs (as superposed non-linear user-elements) inside a finite element architecture. This makes it ideal for multiscale simulations with multiscale transitions to continuum descriptions. Further details can be found here.
Modelling the Threat of Runway Debris
Runway stones thrown up by aircraft undercarriage wheels can cause considerable damage to the aircraft composite structure. Models for runway debris lofting have been developed at Imperial College which capture the wheel encounter with the stone, the subsequent physics of the lofting process, the subsequent trajectory of the lofted stone and ultimately predicts the severity of the impact on the aircraft structure. The output from this model are ‘impact threat maps’ (see opposite) which predict the relative likelihood and severity of the impacts from lofted stones on the underside of the aircraft, thus informing design decisions on composite fuselage design. This modelling capability is unique, and has successfully supported the development of a nose wheel debris deflector for an aircraft now in service. These models have direct relevance to vehicles from other transportation sectors which are exposed to lofted debris during service. Further details of the methodology and model development are given here.
Topology optimisation for multifunctional material design
Composite materials are highly regarded for their excellent mechanical performance. Recently, there has been a move towards taking them beyond their more traditional load-bearing role and imparting functionalities that translate in capabilities such as: self-healing, self-sensing, energy harvesting and energy storage. This project sets out to address a key question “can we design the optimal microstructure to fulfil both structural and functional requirements?” We are currently developing a design-optimisation framework capitalising on topology optimisation methods to enable the realisation of multifunctional materials. Preliminary investigations into the design of composite materials that offer both high stiffness and high ionic transport to maximise structural power functions (i.e. load bearing energy storage systems) have been promising.