Geomechanics and Geoengineering

RESEARCH ARTICLE

Coupled FEMDEM/Fluids for coastal engineers with special reference to armour

stability and breakage

John-Paul Lathama*, Julian Mindela, Jiansheng Xianga, Romain Guisesa, Xavier Garciaa, Chris
Paina, Gerard Gormana, Matthew Piggotta and Antonio Munjizab

a
Department of Earth Science and Engineering, Imperial College, London, UK
b
Department of Engineering, Queen Mary London University, London, UK

Abstract

Sea-level rise and increased storminess present huge challenges to coastal engineers worldwide.
The seaward slope of many breakwaters and shoreline defence structures consists of thousands of
interlocking units of concrete or rock making up a massive granular defence against wave attack.
The units are placed freely to form an armour layer which is intended to both dissipate wave energy
and remain structurally stable. Design guidance on the mass and shape of these units is based on
empirical equations derived from Froude scale physical model tests. The two main failure modes
for concrete armour layers are displacement (hydraulic instability) and breakage (structural
instability) which are strongly coupled. Breakage mechanisms cannot all be faithfully reproduced
under scaled physical models. Fundamental understanding of the forces governing such wave-
structure interaction remains poor and unit breakages continue to baffle the designers of concrete
armour units. This paper illustrates a range of DEM and FEMDEM methods being developed to
model the granular solid skeleton of freely packed brittle units. Such discrete element methods are
increasingly being used by engineers for solids modelling. They are especially powerful when
coupled with a CFD model which can resolve ocean wave dynamics. The aim is to describe a
framework for coupled modelling technologies applicable to coastal engineering problems.
Preliminary simulation test cases, still at proof of concept stage, but based on a wealth of validation
studies are presented. Thus, we report a snap-shot of progress towards a future where designers
combine multi-physics numerical technology with knowledge from scaled physical models for a
better understanding of wave energy turbulence, block movement, and internal stresses within
armour units.

Keywords: concrete armour units, breakwaters, discrete element, finite element, CFD, coupling
_____________________
*Corresponding author. Email: [email protected]

1
1 Introduction

The changing climate and associated increases in forcing factors, namely higher sea levels

and more severe storms, present huge challenges worldwide. Coastal engineering is where

policy makers will increasingly look for practical solutions to adapt. Perhaps surprisingly

for such a fluids-dominated subject, it is not only the hydraulic loading considerations that

are important. An understanding of the mechanical response of the geomaterials used in

construction of coastal structures is also critical. In fact, to model the problem numerically,

the system requires coupled fluids-solids interactions where the solids, which are often

discrete units, are also deformable and breakable elements. A numerical approach being

adopted to look at this problem, in essence, defines the scope of this paper.

In practice, the most widely employed design for breakwaters and coastal structures is to

armour the seaward slopes with rock blocks or concrete units so massive that they remain

static or have only limited movement during predicted storm wave conditions. In exposed

oceans, rock pieces above 20 tonnes are often needed but are difficult to source locally as

nearby quarries may be unsuitable. For stability, concrete units with a mass as high has 125

tonnes may be required for design storms with waves exceeding 10 m. Many different

special shape units have been invented 1 such as Tetrapod, Dolosse, Accropode™,

Coreloc®, Xbloc® etc. (Fig. 1). When cast in concrete, the units have been shown to

interlock well but may be prone to breakage with movements and impacts during

construction, storms or by fatigue mechanisms. This has encouraged use of cube-like

shapes if very massive units are needed, but the scientific basis of such a preference is far

from proven. Thousands of such units may be cast and placed in one breakwater (Fig. 2).

They provide a granular outer armour layer of pseudo-random units that interlock well

while providing the voids in which wave energy is dissipated through turbulence. Guidance

1
Information on the various units can be found via websites of the inventors and agents for the units
2
on the use of concrete armour units is largely reliant on empirical equations, perhaps the

most famous of which is the Hudson formula (Eq. 1) based on calibration of laboratory

wave flume results for armour stability. The maximum ratio of significant wave height Hs

to nominal diameter Dn (or cube root of the unit’s volume) to prevent a certain

unacceptable amount of block movement is a function of the armour layer seaward slope α

and a stability factor KD that covers the multitude of remaining effects most notable being

unit shape. Buoyant density (Δ = (ρc/ρw)-1, where ρc and ρw are concrete and sea-water

density) also contributes to degree of stability, usually expressed by the symbol Ns, termed

the damage number or hydraulic stability number.

Ns = Hs/ΔDn = (KD cotα)1/3 (1)

KD is found to increase greatly with the degree of interlock as a consequence of shape,

going from a value of 2 typical for natural rock to >10 for many concrete units. Using

mainly experimental approaches and Froude scale physical models, (i.e. to avoid bias in the

hydraulic response at the surface of the breakwater, the ratio of the inertia forces and

gravity forces are made similar in the prototype and model by setting the velocity ratio

equal to the square root of the length ratio), researchers and designers have continued to

refine empirical stability equations employing further parameters and extensive testing

such as for different bathymetries and water depths in front of the structure. For a

comprehensive guide to rock armour, concrete units in use and recent stability equations,

see CIRIA/CUR/CETMEF (2007).

3
 Xbloc®

Figure 1. Examples of armour units and a 25 tonne piece of rock used as armourstone

Figure 2. Use of concrete units. (a) Rectangular blocks of 100 tonnes for Barcelona breakwater (stacked in
casting yard) (b) 125 tonne rectangular units on seaward slope of Bilbao breakwater facing the Atlantic,
(during construction). (c) and (d) Accropode™ units, Le Havre, France.

One aspect of the problem for designers, is the large wave climate uncertainty associated

with the design storm (e.g. 100-year prediction), but putting this topical problem to one

side, another source of uncertainty with Froude scale model results is the difficulty of
4
building a laboratory model that correctly scales the physical and geometric properties to

they key forces affecting stability. The stress levels inside scaled units induced by waves in

models are too small to cause any unit breakage because conventional model-unit making

materials are disproportionately strong, but these stresses can be measured by transducers.

Burcharth et al. (2000) measured the stresses inside six model units under a suite of wave

structure conditions. To predict breakage conditions for a typical concrete tensile failure

criterion they recognised the importance of applying two different scaling laws one for

static (and pulsating) and one for impact stresses. They concluded that keeping Ns as given

in Eq. (1), to less than 2 in the design will ensure that the contribution of the impact stress

to the maximum principal tensile stress is generally less than 10%. This conclusion, based

on a study of dolosse is interesting, given that Ns values of about 2.7 and 2.8 are often

recommended by the manufacturers of concrete units such as Accropode™, Coreloc®,

Xbloc® Accropode™ II, as a suitable measure to adopt for their hydraulic stability. These

recommended Ns values refer to units having a more robust (chunkier) geometry than

dolosse. However, taken together with Burcharth et al.’s conclusion for dolosse, they

underline the need for confidence in the unit’s structural stability under dynamic loads

when employing designs with Ns values higher than 2. For armour layer stability of layers

made from dolosse units, as an alternative to the essentially empirical hydraulic stability

equation, Eq. (1), Burcharth et al. (2000), presented a prediction equation for relative

breakage as a function of wave height, concrete tensile strength, armour unit mass, and four

fitting parameters. Hald and Burcharth (2000) continued to pursue the idea of taking wave

loading out of the Hudson formula’s black box, this time by evaluating the wave loads

directly for rock breakwater layers. Records of wave-induced forces on model units of rock

tested in a flume were found to be fitted well by the Gumbel distribution (a special case of

the Weibull distribution often used by coastal engineers for extreme value prediction) and

5
this was used to describe the disturbing forces in their new approach. Pull-out forces

normal to the slope in static dry conditions were measured with a pulley wire and hook

device to give the restraining forces. By comparing disturbing and restraining forces they

derived another formula which was found to have a comparable reliability to Hudson’s in

terms of best-fit criteria applied to the experimental scatter plots. This attempt to address

force measurement is a step in the right direction but has been greatly overshadowed by the

pragmatism offered by analysis of flume test results and the introduction of more

meaningful dimensionless parameters.

Unfortunately, physical scale modelling of wave-structure interaction, understandably

undertaken throughout the world’s major hydraulics laboratories since it continues to

provide the best available design validation, rarely obeys the similitude requirements for all

the significant forces that may be influencing the prototype behaviour of interest. This is

not only true for the model material’s strength to resist breakages, but is also considered to

be the case for hydraulic forces and flows within the permeable core of a breakwater

affected by viscous forces (poorly represented by small scale models) as the core outflows

must link with the incident wave flows. Owing to this kind of limitation of the force

scaling, there is a trend amongst the research community to seek near-prototype-scale

experiments with model to prototype length ratios of ~1:5 or less. There are only a handful

of such massive and costly wave flume facilities available in the world. The large scale

tests are used to validate major breakwater project designs and for international research

consortia to test the validity of empirical guidance formulae based on extensive testing

programmes conducted at smaller length scales for which ratios are typically ~1:40. The

consequential costs of going to the larger flume tests, because of uncertainties posed by

small scale physical tests, are widely recognised by coastal engineers. Other differences

between prototype and scale model that are of concern to the client and designer of a
6
breakwater scheme include factors such as machine operator construction technique,

frictional properties and the effect of local crushing of corner, edge and face contacts in

concrete units, fatigue, dynamic stress transmission and the variability of brittle properties

of real geomaterials.

As numerical modelling does not suffer from these problems (notably it suffers from

others: discretization induced effects, errors and computational cost), it is a safe bet to say

that numerical simulation with multi-physics and multi-scale capabilities will play an

increasing role to complement flume tests and field monitoring in our search to understand

how stability is achieved through turbulence generation (energy dissipation), unit interlock,

mass and strength.

In this paper we illustrate the modelling of problems that hopefully, can be recognised as

components to be mastered on the way to addressing the 3D coupled breakwater armour

stability problem which is the ultimate goal. In the methods proposed, 3D solid behaviour

of the armour units is handled by discretizing the separate armour units into finite elements

and using combined finite and discrete element (FEMDEM) modelling technology to track

unit movement and deformation under breakwater construction and wave loading. The

paper takes a thematic problem, and by showing visualized simulation results where key

diagnostic parameters could clearly be the target of further simulation study, it aims to

illustrate the power of combining a generic computational fluid dynamics (CFD) platform

(see Pain et al. 2001, Piggott et al. 2008) with discrete solids modelling using DEM or

FEMDEM (see Munjiza 2004, Xiang et al. 2008a). The paper is in four parts.

Concrete unit and rock particle geometry capture is discussed in the first part. In the second

part, a DEM model has been implemented using a multi-sphere approach (e.g. see Favier et

al. 1999) to enable granular heaps with different 3D shapes to be modelled. The third part

describes the FEMDEM approach. The internal stresses and breakability of concrete units,
7
assuming unit depth in the third dimension (Munjiza et al. 1999), is presented as a foretaste

of the potential that a 3D FEMDEM model with breakability would provide in the context

of coastal engineering. Both elastic deformation and possible fracture of units are taken

into account for different loading conditions. Results using a new 3D FEMDEM

formulation for stress development within complex-shaped multi-body dynamics are shown

for two hypothetical breakwater units, the Vcross and the VRcross concrete units, during a

simple collision. Part four briefly introduces wave simulation and wave interaction with

granular systems. Fluid behaviour and interaction between air and water is modelled using

the CFD code, Fluidity 2 and results of one-way and two-way coupling of 3D DEM with

CFD are shown. As the paper aims at this wider scope, it is considered appropriate to

include proof of concept simulation results in addition to results of simulations that can be

compared directly with experimental and analytical results for validation purposes.

2 Geometry capture

To model interactions between complex shapes, DEM with multisphere methods, or

alternatively FEMDEM can be used. In either case, the particle geometry has to be meshed

and this can be a bottle-neck for realistic modelling. Fig. 3 shows meshes generated at high

resolution from laser scanning methods. Data capture, mesh formation and mesh

coarsening are summarised in Latham et al. (2008b). For the DEM or FEMDEM model, a

surface mesh is selected from a library of natural particles or CAD-generated fundamental

or engineered shapes.

2
(e.g. see http://amcg.ese.ic.ac.uk/FLUIDITY)
8
Figure 3 Laser scanning methods to capture and mesh the geometry of model armour units were discussed in
Latham et al. (2008b). (a) Rendered surface for an elongate rock and (b) high resolution optimized mesh. (c)
Rendered surface for an EcopodeTM block, a new unit from a specialist company for concrete units, Concrete
Layer Innovations and (d) its high resolution mesh

3 DEM

3.1 Test Case 1: Packing of irregular shaped grains

Using a new algorithm to cluster overlapping non-uniform spheres efficiently within

particle meshes from the shape library, (Garcia et al. 2008), it has been possible to examine

several mono-shaped packs with different particle shapes. An in-house multi-sphere DEM

code was used to produce packs from gravitational settling. Results presented in Fig. 4 are

for four particle shapes selected from the rounded class of a shape library. It is encouraging

to note that these highly rounded high sphericity ‘pebbles’ have a realistic range of pack

porosities found to be 33.8, 33.0, 34.3 and 33.8 looking from left to right in the figure.

However, their absolute volumes are 86.5, 80.3, 65.1 and 71.7 (x 103) mm3 respectively

and hence each pack of 1000 pebbles produces column heights that reflect slight variations

mainly in particle sizes but also in shapes. The methods are in principle also applicable to

angular particles and concrete armour unit shapes. But, the computational cost in terms of

the numbers of spheres required to give a good approximation to the real shape becomes

prohibitive. Whereas the DEM method with multi-spheres looks promising for modelling

9
armour layers built with relatively rounded rock, concrete armour units and rock armour

are invariably angular and facetted.

Figure 4. Test Case 1: Packs of 1000 grains deposited using four different particles selected directly from a
shape library (i.e. the slightly different particle sizes used were not normalised to a reference volume, see
text).

3.2 Test Case 2: Collapsed heaps

As an exercise in proof of concept using DEM a comparison of the likely behaviour and

differences between cube-shaped units and cross-shaped units in terms of stability was

investigated using mono-sized multi-sphere representations. In terms of armour units, a

parallel can be drawn with the concrete cube unit and the Xbloc® unit. Fig. 5 shows simply

formed slopes generated from collapsing columns of previously packed units. Emergent

average properties such as angle of repose and porosity were measured. Numerical methods

using DEM were adapted to also measure the average pull-out forces normal to the slope

for surface units (Latham et al. 2008a) to compare them with force distributions determined

experimentally for the Xbloc® (Muttray et al. 2006). Results of such investigations will

have more pertinence to real structures when the armour unit slopes are constructed by

individual placement from toe upwards as though placed by crane, rather than by modelling

a collapsed heap. The numerical methods do yield a realistic range of results with a

distribution of forces that also depend upon rate of pull-out, giving confidence that such

10
numerical tests can be used in future. That said, multi-sphere DEM has known limitations

when dealing with contact behaviour and friction between faceted units so that FEMDEM

methods are ultimately more likely to give realistic pull-out force determinations. The

increased computational cost of using a FEMDEM approach rather than DEM has yet to be

studied in this context.

(a) (b)

Figure 5. Test Case 2: Visual comparison of collapsed heaps (a) of cubes (b) crosses modelled with single-
sized multisphere DEM.

4 FEMDEM

4.1 Test Case 3: 2D Collapse of crosses

The modelling reported in this section is generated using code based on the methods

reported by Munjiza (2004) with further algorithms and implementations to include the

effect of friction (Xiang et al. 2008a). It is interesting to observe the FEMDEM simulation

of a “dam-break” problem using a granular material made up of cross units. Using

FEMDEM with each block discretised coarsely into finite elements, the dynamical stress

development as the wall of units topples, (see Fig. 6) can be observed. For this illustrative

case the material has been given elastic properties appropriate for concrete but with no

criteria implemented for plastic or brittle deformation, i.e. they cannot break.

11
 (a) (b)

(c) (d)

Figure 6 Test Case 3: Dam-break of cross units with initial pack formed in rectangular container shown; (a)
with the wall removal, (b) toppling, (c) run-out and (d) final rest state. In colour, this figure displays
differential stress level. The highly stressed elements are seen in red and the relatively unstressed elements
(some deep within the mound) are shown in dark blue.

4.2 Test Case 4: 2D Breakage

To model cracking, the 2D single smeared crack model of Munjiza et al. (1999)

implemented in FEM/DEM (Munjiza 2004) for fracture and fragmentation, is applied to

concrete unit behaviour. The 2D FEMDEM implementation uses a fracture mechanics

Mode I (i.e. opening mode) tensile breakage model with cracks propagating along pre-

existing element boundaries. A simplistic failure criterion is also implemented for shear

failure in compression, when the shear stress exceeds a certain material-dependent strength

and cracks propagate along pre-existing element boundaries. The pre-peak regime is

modelled with FEM and the softening behaviour upon single crack opening displacement is

modelled with three curve fitting parameters based on real data from the full stress-

displacement curve for direct tension tests on concrete (see Munjiza et al. 1999). The stress

resisting opening of the crack is made to vary smoothly (i.e. is smeared) from zero in

already cracked material up to the tensile strength of the material ahead of the crack tip and

energy is expended in generating crack surface in accordance with the fracture energy

release rate.

12
As an important test case, we illustrate the individual collision response for a drop test with

a block of 2.4 tonnes striking a rigid substrate. The method considers unit depth in the third

dimension. The experimental data with which to validate this result is not readily available

from the literature. The difficulty for validation is largely that field site drop test conditions

do not allow for controlled parallelism of anvil and cube base nor an effectively rigid anvil.

Also, the plane stress and strain assumption would not be applicable to full-scale tests on

cubes. (Further discussion of drop testing is given in Section 4.3). Results in Fig. 7

illustrate a test with the drop height set at 0.5 m (impact velocity of 3.1 ms-1) with ample

potential energy to result in well defined brittle fragmentation phenomena in both

compressive and tensile modes. The mesh is refined to sizes meeting the criterion

suggested by Munjiza and John (2002) and realistic concrete properties for shear and

tensile failure and fracture energy of concrete are applied (see Table 2). The stresses within

such concrete units when they are packed into granular armour layers can in principle be

investigated using these 2D methods. However, computational cost quickly escalates if

many units are to be modelled with meshes sensitive enough to predict fracture

realistically. The simulation in Fig.7 shows the initial stress wave front travelling upwards

at about 4 km/s prior to stress wave reflection. Cracks initiate at about 0.01 seconds after

impact and propagate to create separate fragments that are projected outwards as the block

bounces.

Table 1. Numerical model parameters for 2D FEMDEM shown in Fig. 7

Parameter Units Value
Young’s Modulus GPa 26.6
Poisons Ratio - 0.205
Tensile strength MPa 5.0
Shear strength parameter MPa 17
Fracture energy release rate N/m 150
Density Mg/m3 2.4

13
Figure 7. Test Case 4: Illustration of plane stress, plane strain FEMDEM drop test of 1.0 x 1.0 m, mass scaled
to represent depth of 1 m in the third dimension. Sequence (a) to (g) showing milliseconds after initial impact
of concrete block on rigid substrate. Note stress wave propagation and reflections at free faces, tensile and
shear crack initiation and propagation, branching, block bouncing, fragmentation and projection of pieces,
with some final fragments containing partially traversing cracks.

4.3 Test Case 5: Collision of Vcross and VRcross unit in 3D

FEMDEM simulations of the packing of cubes using linear tetrahedral elements to compute

the contact interactions of each cube were compared with packing experiments (Latham

and Munjiza 2004). The similarities of the emergent bulk density as well as qualitative

features of patterns of stacking were noted in their study. The 3D FEMDEM method used

in this paper is more robust, (see Xiang et al. 2008c). It is based on higher order tetrahedral

elements to overcome the locking problems associated with linear tetrahedral finite

elements. The results discussed below and illustrated in Figs. 9 and 10 have no plastic,

brittle or viscous dissipation and are for purely elastic deformations (i.e. infinite strength).

The problem addressed however is of direct relevance to the design and application of

concrete armour units. The stresses generated when a unit is dropped onto a massive anvil

using the contact detection and contact interaction features of FEMDEM provide an

informative test case for the numerical methods. In the near future, the force interaction and

14
stress transmission within an armour layer system consisting of many concrete armour

units will be studied with these methods.

(a) (b)

Figure 8. Geometry and mesh used to model impact by drop of concrete unit on massive anvil using
FEMDEM. The square cross section arms are 1.26 m x 1.26 m and the target anvil is 8 m x 8 m x 4 m. Left:
Vcross (14 m3); Right: VRcross (18 m3)

Figures 8, 9 and 10 illustrate two hypothetical units which only exist in a virtual sense for

the purpose of providing a non-commercial test case. They are termed the Vcross and

VRcross units, the latter having the corners substantially reinforced and a greater mass but

the same square cross section for the base of the cross arms. The model parameters used are

given in Table 2. Figure 8 shows the shape and mesh of the two units. If statically

positioned on the anvil, the units would induce a compressive stress through the same

square base areas (1.26 m x 1.26 m) of 0.202 MPa and 0.260 MPa respectively. The

maximum dynamic stress generated can be considerably higher than the static stress,

depending on the impact velocity. Figs. 9 and 10 show the simulated stresses on a cut plane

resulting from each unit impacting at 0.5 ms-1. The dynamic impact of a symmetric flat-

against-flat, with the very rapid arresting of the momentum, lack of rotation, and the fact

that dissipative crushing phenomena is excluded from the modelling, results in large

transient stress predictions. Some of the concrete units deployed in the industry have

geometries with features in common with the hypothetical geometries considered here and

15
indeed, the flat-on-flat and other drop–test are sometimes performed on real units to

demonstrate robustness, (Muttray et al. 2005), (Turk and Melby 1997). However, for

practical reasons, it is difficult to compare the simulated stresses with prototype drop test

results because of the inherent difficulties in instrumentation, signal noise and

reproducibility of test set-up. Relatively expensive dynamic loading tests on prototype units

are still considered useful. The fracture and damage patterns observed during increasingly

severe impact tests are therefore the subject of detailed interpretation along side stress level

predictions, since direct stress measurement is usually unsuccessful. The problem with

using many of the available numerical analysis codes has often been that boundary

conditions in the model can only partially replicate prototype test conditions. Conversely,

from the numerical modeller’s viewpoint, it has often been considered impossible to use

standard drop geometry as a means to validate code predictions, there being other collision

geometries such as pendulum and horizontally arranged train-track tests, considered better

suited for comparing numerical and experimental results (e.g. see Burcharth, 1993). With

the new FEMDEM model illustrated below, it would be possible to record with high speed

photography the precise conditions of a full-scale test and to set up the boundary conditions

of impact and the structure and properties of the anvil relatively accurately.

With these provisos, the FEMDEM stress analyses of Figs. 9 and 10 are briefly discussed

below. They indicate what is widely known already about tensile stress generation and

designs to reduce it which, from the modelling viewpoint, is reassuring.

The Vcross unit, 4.56 ms after impact, develops a maximum tensile stress of 3.89 MPa

from the corners above the arms while differential stresses never exceed 9.35 MPa. The

results indicate that although the compressive strengths of typical armour unit concrete

(~35-60 MPa) could withstand the dynamic differential stresses, the tensile stresses

approach the typical concrete tensile strengths of 3.5-4.5 MPa (Franco et al. 2000)
16
suggesting the initiation of major tensile cracks and complete failure and loss of the arms

could occur from such an impact, rendering the unit useless. By contrast, the VRcross unit

with about 30% more energy on impact than the Vcross develops potentially damaging

stresses only in the bottom impacted arm. After 4.8 ms the differential stress reaches a

maximum of 12 MPa while maximum tensile stresses are 3.49 MPa suggesting that tensile

spalling cracks would develop in the walls and base of the bottom arm with relatively little

likelihood of significant compressive shear cracking at all and the likelihood of no major

damage being transferred to the unit as a whole.

(a) (b) (c)

(d) (e) (f)

Figure 9. Vcross unit of 14 m3 volume and height of 3.78 m. Simulation results are for a vertical impact of the
unit striking normal to a massive anvil at 0.5 ms-1 using FEMDEM. (a) just after the impact; (b) instant when
highest stresses are developed; (c) instant just before unit bounces up and leaves the anvil. The values in (a),
(b) and (c) show the differential stress (σ1-σ3) with a scale from 0 to 10 MPa and (d) (e) (f) show least
principal stress (σ3) with tensile stress positive on a scale from -3.5 MPa to +3.5 MPa. Maximum differential
stress (MPa), maximum tensile stress (MPa), time after impact (milliseconds) are: (a) 0.96, 0.00, 2.0; (b) 8.77,
3.89, 4.6; (c) 1.49, 1.02, 6.7.

It is worth pointing out that the dynamic stresses within prototype units during extreme

storm conditions are tempered somewhat by the mobility of the micro-crush zones of unit

contacts within an armour layer and the neighbouring units’ non-rigidity during dynamic

impact events, as well as damping by the surrounding water. Such considerations underline
17
the many means by which collision energy will be more benignly absorbed than by the

collision of a flat face against relatively immovable flat base as modelled in Figs. 9 and 10.

Further improvements to these purely elastic stress development predictions will be

possible in the future within the FEMDEM framework by implementing energy dissipating

constitutive behaviour such as an elastic-plastic relation and ultimately a fracture model, as

seen in the 2D case. In conclusion, the results of this test case provide confidence that the

FEMDEM method has the capability to tackle the dynamics of complex-shaped geometries

and massive granular systems typical of concrete armour and rock armour layers.

(a) (b) (c)

(d) (e) (f)

Figure 10. VRcross unit of 18 m3 volume and height of 3.78 m. See Fig. 9 caption for key to stress contours.
Maximum differential stress (MPa), maximum tensile stress (MPa), time after impact (milliseconds) are: (a)
0.88, 1.47, 2.4; (b) 12.0, 3.49, 4.8; (c) 2.0, 0.65, 6.8.

Table 2. Numerical model parameters for 3D FEMDEM simulation of

Vcross unit collision (see Figs. 9, 10)
Parameter Units Value
Height m 3.78
Square cross section area of arms m2 1.59
Mass Mg 32.8, 42.1
Impact velocity ms-1 0.5
Young’s Modulus GPa 26.6
Poisons Ratio - 0.205
Density Mg/m3 2.34

18
5 Coupling fluids and solids

The coupling work is realized using a dual mesh approach. One mesh is used across the

whole solution domain on which the fluids equations are solved and the second mesh

contains a finite element representation of the solid structures. Adaptivity resolves down

onto the complex geometry of the solids at the level of detail necessary defined by some

error metric, hence addressing one of the main challenges - the accuracy of the flow field

near the solid surfaces and the capture of boundary layer effects. The forces and volume

fraction from the FEMDEM or DEM structure model are mapped onto the fluids mesh

using FEM mapping and updated hydraulic forces are returned to the explicit transient

dynamic FEMDEM or DEM modelling of the solids. The adaptive meshing method is fully

robust and the two-fluid approach has been chosen to represent the particles/structures on

the fluids mesh, one fluid being the water/air the other being the solid. This approach

allows the solids to overlap as is necessary to calculate the particle interaction forces using

a penalty function approach adapted in the FEMDEM contact interaction formulation. To

compute drag forces directly, volume integration methods are used. This means that in the

modelling application to simulate storm damage of breakwaters, for example, the

adaptive/moving mesh will follow the armour units should they move within the domain,

putting the necessary fine scale resolution around the armour units and conforming

accurately to the boundaries of these unusual shaped concrete blocks.

Compared to alternative approaches for multi-fluid-structure interactions with free surfaces

and large solid movements, the approach followed here has some distinct advantages in

that it can employ an adaptive approach in combination with solid FEMDEM models. As

shown in Section 4 below, these are beginning to produce meaningful results in 3D. The

approach is underlined by being FEM-based and associated with a variational principle and

higher order accuracy. Such FEMDEM/CFD approaches are complementary to others that

19
may be numerically less computationally costly such as Lattice-Boltzmann methods which

have also been coupled with DEM particle methods, (e.g. see Han et al. 2008). Also,

relatively simple to couple with DEM solids are the particle-based SPH fluids approaches

(Cleary et al. 2006). However, the building of a numerical wave flume with realistic

random waves requires detailed capture of the two-fluid phases at the very energetic water

surface. While Reynolds Average Navier-Stokes solvers in combination with VOF methods

(Lara et al. 2006) have proved effective, the results achieved with Fluidity using adaptive

unstructured moving meshes, shown below, also look to be a promising basis for wave

modelling. In summary, our approaches have several features in common with those

proposed by Takizawa et al (2007) with the significant addition that advanced FEMDEM

solids methods can be exploited.

5.1 Test Case 6: One-way coupling of wave-structure interaction

Figure 11. Test Case 6: Wave breaking on a slope armoured with two hand-meshed concrete units. In this
simulation the units are fixed.
20
Details of the wave and free-surface modelling using the CFD code Fluidity, (e.g. Pain et

al. 2001) are given in Mindel et al. (2008). The interface tracking approach differs from

VOF and exploits several methods including a compressive advection scheme, node

movement, and general mesh optimization. Preliminary results of one-way coupling of

waves breaking over fixed unit geometries are shown in Fig. 11.

5.2 Test Case 7: One-way coupling of flow in granular system with single phase CFD

The solid skeleton of packs of armour units or rock blocks making up the primary or

secondary armour layer or internal core of a breakwater structure can be modelled by

sequentially depositing particles/units using DEM or FEMDEM. In Fig. 12, the solid

structure, similar in some ways to the core of a breakwater, is built using multisphere DEM

(see Fig. 4). The grains are created assuming Hertz-Mindlin interaction forces with elastic

constants to match quartz. The pack has porosity of 34%, and is created with periodic

boundary conditions in the xy and yz planes to avoid unrealistic porosities and associated

high flow velocities that would otherwise occur near artificially created container walls.

z
y
x

Figure 12. Test Case 7: Flow field through 1680 non-spherical cluster particles of typically ~60 mm diameter
where each cluster is created with 38 spheres with optimized sizes. The inflow velocity is 0.1 mm s-1 and the
snapshot is after only 0.5 s. The method is implicit in time and the maximum Courant number is ~ 4. The
number of nodes is 440000. The computational domain boundaries form a cube with sides that cut through
particles, inflow is through the front left xz plane.
21
The solid mesh is again superimposed on the fluid domain and the micro-scale fluid flow is

captured by a Navier-Stokes solver with an unstructured adaptive grid run in parallel. The

same capability, but extended to a fluid mixture carrying a concentration field of water and

air and a carefully tracked water-air interface, was seen in Fig.11. It is therefore suggested

that to capture the flow characteristics of the waves penetrating into the porous mound of a

breakwater and the outflow back through the armour (when armour instability is often

greatest if in resonance with the trough of the incoming wave), is also within the scope of

this modelling approach.

5.3 Test Case 8: One-way coupling of flow responding to fragmenting particle

Figure 13. Test Case 8: 2D water flow in response to motion of a fragile coarsely meshed ‘concrete’ triangle
of base width 1 m, breaking up on impact with base of water container. (Bottom left enlarged).

22
In this example, the simulation is 2D, where the flow captured by the Fluidity CFD code is

responding to multi-particle movement driven by a falling fragile solid represented using

2D FEMDEM (Fig. 13). The triangle breaks on impact with the base of a water tank. The

FEMDEM code is interfaced to the CFD code although in the case shown, the solid motion

and behaviour is independent of the fluid flow.

5.4 Test Case 9: Two-way coupling of DEM and CFD, one disc

Two test cases have been used to illustrate the two-way coupling capability of the method

adopted for superimposing solids behaviour within the fluids domain, the classical Stokes

problem was considered for the case of one cylinder falling through water. The results are

shown in Table 3 and Fig. 14.

Table 3. Simulation results compared to

experiments.
Case Terminal velocity
ms-1
Simulated -6.78
Clift et al. (1978) -6.53

(a) (b)

Figure 14. Test Case 9: Velocity of cylinder accelerating to terminal before hitting base of tank. (a) snapshot
of the disk falling through the fluid (b) Plot showing velocity agrees with experimental data indicating an
accurate two-way coupling and drag force formulation using powerful adaptive mesh methods.

23
5.5 Test Case 10: Two-way coupling of DEM and CFD, colliding spheres

To go to the next step the adaptive CFD coupled with DEM was applied to a problem of

two steel balls colliding in a column of silicone oil and falling (see Xiang et al, 2008 for

details). The resulting behaviour is encouragingly realistic while the necessary further

comparison with experiments and validation is ongoing (Fig. 15).

(b)
(a)

(c) (d) (e) (f) (g)

Figure 15 Test Case 10: Coupled DEM/CFD simulation of two steel spheres impacting in silicone oil; (a)
computational domain and initial position and relative velocity of the spheres, (b) mesh near spheres at
t=0.002s after impact, spheres positions and fluids velocity contours at (c) t=0.02s, (d) t=0.054s, (e) t=0.105s,
(f) t=0.155s, (g) t=0.183s
24
6 Discussion

In coastal engineering, just as in many engineering disciplines today, there are high hopes

for advances in numerical modelling of the complex mechanical processes occurring over

different length and time scales. In the CIRIA/CUR/CETMEF Rock Manual, (2007, Fig

2.3) fourteen different mechanical failure modes are identified for a typical rock or

concrete unit armoured breakwater. Only one of these is erosion by armour unit removal or

breakage as highlighted in this paper. Some have a more classical geotechnical origin such

as rotational and translational failure in the core. Ship and ice collision are also listed.

Arguably, all failure modes are mechanical problems amenable to fluid-solid coupled

continuum-discontinuum approaches. The modelling technology needed (shown in Fig. 16)

requires a seamless means to couple the continuum behaviour of the sea state as it is

transformed over topography, to tackle the discontinuous armour layer zone where highly

energetic dissipation occurs. Furthermore, as well as capturing flow in the armour layers,

fluid pressures and inter-particle forces within the breakwater core, best approximated by

continuum approaches, are also needed. In the next ten years or so, it is considered highly

probable that coupled CFD and FEMDEM will be capable of modelling high energy waves

impacting on engineered coastal defences. Furthermore, modelling natural solid-fluid

processes such as beach pebbles moving in the swash zone and other finer sediment

transport behaviour using FEMDEM/Fluids or DEM/Fluids models has great potential.

Coastal engineers dealing with soft defence strategies will be more confident when using

continuum models that have been developed and calibrated using these kinds of

DEM/Fluids models and sub-grid scale multi-phase models that capture the physics of

solid-fluid mixtures.

25
Where stresses within concrete units are important such as in breakwater armour units

suffering wave shock transients, and where angular and faceted shapes are involved,

FEMDEM will be required and it is recognized that this brings a considerable extra

computational cost. This paper, a work in progress, does not address the real constraints

imposed by computational cost, but focuses on the feasibility of modelling the physics.

Wave-structure coupled interaction

Continuum: Air/water
interface, wave over Discrete Discrete or Continuum:
bathymetry concrete units continuum rock porous media
armour
underlayer
Figure 16 Conceptual framework for continuum and discontinuum modelling of coupled FEMDEM/Fluids
for breakwater stability. Figure adapted from de Rouck et al. (2005).

Computer resources, parallelization, resolution and errors are also paramount in guiding the

development of the modelling strategies being developed here. This paper has attempted to

show our vision for multi-material computational modelling in the context of challenges

faced by coastal engineers. The tools are still some way off engineering solutions that will

supersede empirical formulae such as Eq (1). However, they may soon throw new light on

the complex interplay between rubble mound geotechnics, flow in porous media, armour

unit block shape and fragility, block interlock, the variability of contact forces and the

wave loading and turbulent flow associated with different wave climate and structural

settings.

26
7 Conclusions

Practically any particle shape can be captured by laser scanner or alternative device and

converted to surface and volumetric computational meshes for further mechanical

modelling with DEM or FEMDEM. The modelling capability of FEMDEM is more

versatile and can give accurate geometry representation for further multi-body applications.

Rounded forms of particles can be efficiently handled by DEM multisphere approaches

using relatively few spheres per particle. Their computational efficiency over FEMDEM

suggests there are particulate applications where the rigid behaviour assumption of DEM

simulation is not restrictive, for example in creating a realistic granular solid skeleton or in

granular flow simulation. For faceted and angular concrete units and rock blocks used in

armour layers, FEMDEM provides excellent shape representation and deformability for

static and dynamic problems. It also provides a powerful tool for examining stress chains

within granular packs of armour units, e.g. showing where units in the toe of a structure are

carrying excessively high stresses while other units are carrying very little.

The introduction of breakage by tensile and shear failure and the forward modelling of

fragmentation in FEMDEM is required for the ultimate expression of concrete unit

breakage in a numerical model. This has been illustrated in a simple 2D case. The

modelling of localized break-up in 3D is extremely challenging and is currently a subject

being widely researched in the geomechanics community. A great deal can be learned

about likely failure modes in concrete by examining the surface and internal principal stress

components for a purely elastic analysis such as was illustrated here for a drop collision of

a hypothetical unit or new unit design under development.

The generic CFD code, Fluidity has been illustrated and found to be a powerful platform

for coastal engineers wishing to capture wave breaking and structure interaction when

coupled with FEM/DEM or DEM.

27
ACKNOWLEDGMENTS
The authors are grateful for funding from EPSRC under grant GR/S42699/01 GR/S42705/01 and industrial
support from Sogreah, CLI and Baird Associates. We wish to thank ICT-HPC at Imperial College for
computer resources and support.

8 REFERENCES

Boutt, D. F., B. K. Cook, B. J. O. L. McPherson, and J. R. Williams (2007), Direct simulation of

fluid-solid mechanics in porous media using the discrete element and lattice-Boltzmann
methods, J. Geophys. Res., 112, B1029-2004.
Burcharth, H. F. 1993 Structural integrity and hydraulic stability of dolos armour layers. Hydraulics
and Coastal Engineering Laboratory, Department of Civil Engineering Aalborg University,
Series Paper No 9. ISSN 0909-4296.
Burcharth, H. F. d’Angremond, K. van der Meer J. W. and Liu, Z . 2000. Empirical formula for
breakage of dolosse and tetrapods. Coastal Engineering (40), Issue 3, 183-206.
Cleary, P. W., Sinnott, M. and Morrison,. R. 2006. Prediction of slurry transport in SAG mills using
SPH fluid flow in a dynamic DEM based porous media, Minerals Engineering Volume 19,
Issue 15, 1517-1527.
CIRIA, CUR, CETMEF. 2007. The Rock Manual. The use of rock in hydraulic engineering (second
edition). C683, CIRIA, London.
Clift, R., Grace, J.R., Weber, M.E., 1978, Bubbles, drops, and particles, Academic Press, London.
De Rouck, J., Geeraets, J., Troch, P., Kortenhaus, A., Pullen, T. and Franco, L. 2005 New results on
scale effects for wave overtopping at coastal structures. In: International Conference on
Coastlines, Structures and Breakwaters 2005. Institution of Civil Engineers. London. 29-44.
Favier, J. F. Abbaspour-Fard M H., Kremmer M. and Raji A.O. 1999. Shape representation of axi-
symmetrical, non-spherical particles in discrete element simulation using multi-element
particles. Engineering Computations, 16(4), 467–480.
Franco, L, Noli, A., De Girolamo, P. and Ercolani, M 2000. Concrete strength and durability of
prototype tetrapods and dolosse: results of field and laboratory tests Coastal Engineering
Volume 40, Issue 3, June 2000, Pages 207-219.
Garcia, X., Xiang, J., Latham, J-P. and Harrison, J.P. 2008. DEM simulation of real particles
represented by clustered spheres. Geotechnique submitted.
Hald, T and Burcharth, H. F. 2000. An alternative stability equation for rock armoured rubble
mound breakwaters. Proceedings of the 27th International Conference on Coastal
Engineering (ICCE 2000), Sydney, Australia. 1921-1934.
Han, K., Feng, Y.T. and Owen, D.R.J. 2008, Numerical Simulations of Irregular Particle Transport
in Turbulent Flows Using Coupled LBM-DEM. Computer Modelling in Engineering &
Science. 18(2):87-100.
28
Lara, J.L., Garcia, N. and Losada, I.J.. 2006. RANS modelling applied to random wave interaction
with submerged permeable structures Coastal Engineering 53, 395–417.
Latham, J.-P and Munjiza., A. 2004. The modelling of particle systems with real shapes,
Philosophical Transactions of the Royal Society A, 362, 1953–1972.
Latham, J.-P., Mindel, J., Guises, R., Garcia, X., Xiang, J., Pain. C and Munjiza A. 2008a Coupled
FEM-DEM and CFD for coastal structures: application to armour stability and breakage.
Proceedings of the Fifth International Conference on Coastal Structures, Venice, July
2007, ASCE.
Latham, J.-P., Munjiza A., Garcia, X., Xiang, J., Guises, R. 2008b. Three dimensional particle
shape acquisition and use of shape library for DEM and FEMDEM simulation (submitted
to Mineral Resources, DEM Special Issue)
Mindel, J.E., Collins G.S. Latham, J.-P. Pain C.C and Munjiza, A. 2008. Towards a numerical wave
simulator using the two-fluid interface tracking approach combined with a novel ALE
scheme. Proceedings of the Fifth International Conference on Coastal Structures, Venice,
July 2007, ASCE.
Munjiza, A., 2004. The combined discrete finite element method. Wiley.
Munjiza, A., Andrews, K.R.F., White, J.K. 1999. Combined single and smeared crack model in
combined finite-discrete element analysis, International Journal for Numerical Methods in
Engineering, 44: 41-57.
Munjiza, M and John, N.W.M. 2002. Mesh size sensitivity of the combined FEM/DEM fracture
and fragmentation algorithms. Engineering Fracture Mechanics 69 281-295.
Muttray, M., Reedijk, J., Vos-Rovers, I., Bakker, P. 2006. In: International Conference on
Coastlines, Structures and Breakwaters 2005, Institution of Civil Engineers. London. 556-
567.
Pain, C.C., Umpleby, A.P., de Oliveria, C.R.E., Goddard, A.J.H. 2001 Tetrahedral mesh
optimisation and adaptivity for steady-state and transient finite element calculations.
Computer Methods in Applied Mechanics and Engineering, 190, 3771-96.
Piggott, M.D., Gorman, G.J., Pain, C.C., Allison, P.A., Candy, A.S., Martin, B.T., and Wells, M.R
2008. A new computational framework for multi-scale ocean modelling based on adapting
unstructured meshes, International Journal for Numerical Methods in Fluids, Vol: 56, 1003
– 1015.
Takizawa, K., Yabe, T., Tsugawa, Y., Tezduyar, T.E. and Mizoe, H. 2007. Computation of free-
surface flows and fluid–object interactions with the CIP method based on adaptive
meshless soroban grids. Comput Mech 40:167–183.
Turk, G. F., and Melby, J. A. 1997. Dynamic structural response of CORE-LOC®. The REMR
Bulletin. 14.

29
Xiang, J., Munjiza, A., Latham, J.-P., and Guises, R. 2008a. On the validation of DEM and
FEM/DEM models in 2D and 3D. (submitted to Engineering Computations, DEM Special
Issue)
Xiang, J., Mindel, J.E., Latham, J.-P., Pain C.C, Piggott, M.D. Guises, R., Garcia, X. and Munjiza,
A. 2008b. A coupled fluids-particulates model for waves interacting with granular media
using FEM and DEM. Proceedings of the Fifth International Conference on Coastal
Structures, Venice, July 2007, ASCE.
Xiang, J., Munjiza, A. and Latham, J.-P., 2008. Finite strain, finite rotation quadratic tetrahedron
element for the combined finite-discrete element method. Submitted to IJNME.
http://amcg.ese.ic.ac.uk/ FLUIDITY (accessed 30.03.08)

30