0 Suka0 Tidak suka

6 tayangan9 halamanComminution

Dec 23, 2016

© © All Rights Reserved

PDF, TXT atau baca online dari Scribd

Comminution

© All Rights Reserved

6 tayangan

Comminution

© All Rights Reserved

- The Law of Explosive Growth: Lesson 20 from The 21 Irrefutable Laws of Leadership
- Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race
- Hidden Figures Young Readers' Edition
- The E-Myth Revisited: Why Most Small Businesses Don't Work and
- Micro: A Novel
- The Wright Brothers
- The Other Einstein: A Novel
- State of Fear
- State of Fear
- The Power of Discipline: 7 Ways it Can Change Your Life
- The Kiss Quotient: A Novel
- The 10X Rule: The Only Difference Between Success and Failure
- Being Wrong: Adventures in the Margin of Error
- Algorithms to Live By: The Computer Science of Human Decisions
- The 6th Extinction
- The Black Swan
- The Art of Thinking Clearly
- The Last Battle
- Prince Caspian
- A Mind for Numbers: How to Excel at Math and Science Even If You Flunked Algebra

Anda di halaman 1dari 9

International Journal of

Rock Mechanics & Mining Sciences

journal homepage: www.elsevier.com/locate/ijrmms

with the Discrete Element Method

Danilo A. Estay n, Luciano E. Chiang

lica de Chile, 4860 Vicun

a Mackenna Avenue, Santiago, Chile

Department of Mechanical Engineering, Ponticia Universidad Cato

a r t i c l e i n f o

abstract

Article history:

Received 21 November 2011

Received in revised form

6 August 2012

Accepted 28 December 2012

Available online 1 February 2013

Comminution processes are not fully understood today and more research is necessary towards the

improvement of existing comminution equipment. Presently, the best suited modeling technique is the

Discrete Element Method (DEM) but in its conventional form it is not suitable for simulating industrial

comminution processes where size reduction is an important issue, if not the main issue, such as in

crushers. Attempts for simulating such mining equipment with DEM have been carried out previously

applying empirical models for rock rupture. The foundations of the Discrete Crack Model (DCM) are

presented here as a new method for efciently modeling rock fracture within a DEM work frame. In

DCM, random cracks seeds are generated inside the rock specimens and the stress eld in the vicinity of

the cracks is calculated by the Convex Polygons Stress Approximation (CPSA), the basis of DCM.

A rupture criterion is then applied to determine if or when the rock splits. Resulting CPSA stress elds

are compared here to those obtained by a more precise but much slower Finite Element Method (FEM)

solution. Next, to validate DCM methodology, three illustrative loading cases are analyzed comparatively. It is shown that the failure mechanism predicted by the DCM methodology agrees to similar

documented cases in the literature. The proposed approach is applicable both in 2D and 3D.

& 2013 Elsevier Ltd. All rights reserved.

Keywords:

Discrete Element Method

Comminution

Rock fracture

1. Introduction

Simulation of comminution process is still not well understood

and although much effort has been made in the past, many

aspects of the process still remain poorly understood. Authors

such as Powell and Morrison [1] identify essentially four

approaches to study comminution processes. The rst approach

is Power based models, for example the Bond index [2,3]

provides a good correlation for controlled laboratory work in ball

mills. However, an abuse of Bonds excellent work is often made

by trying to cover most comminution machinery with little

consideration of the limitations of the original work. Improvements have been made to cover todays machines, but as in all

empirical relationships its strength relies on the experimental

database, and accurate extrapolation to different machines, ores,

or production rates is not guaranteed.

The second approach is Population balance-based model,

which is good at predicting size reduction, production rates, and

power drawn. The problem with these kinds of models is that

they do not describe the breakage process, and only keep track of

the output. It is also difcult to obtain some ore-specic parameters needed for simulation.

vertical and horizontal impact crushers where impacts can be

assumed to be independent. Simulation results are in good

agreement with real data, but cannot predict output for more

complex machines such as the Barmac impact crusher.

Finally, the fourth kind of approach are the Computationally

intensive techniques such as the Discrete Element Method

(DEM), which is good for modeling kinematics of granular

material. In comminution machinery, kinematics is not enough

to describe the process, fracture modeling must be included. New

models for this purpose are available but some difculties exist

when applied, such as excessive computation time requirements

in physic based models, or extensive experimental data in

phenomenological models.

Powell and Morrison [1] highlight that the future of comminution modeling must be focused in the process and not just the

output. A good mathematical model should be able to describe

every stage of the comminution process, this is why the bestsuited model is DEM, but an appropriate rupture model needs to

be developed for accurate and fast simulations.

n

E-mail addresses: destay1@uc.cl, danilo.estay@usm.cl,

axia277@gmail.com (D.A. Estay).

1365-1609/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.ijrmms.2012.12.041

fracture process in DEM simulations.

126

D.A. Estay, L.E. Chiang / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 125133

Traditionally, the approach to include fracture is to consider

that one body is formed by many elements bonded together.

When forces are applied to the body, the bonds must resist a force

to hold the particles in place. If the stress limit is exceeded then

the bond breaks.

The Bonded Particle Model (BPM) developed by Potyondy and

Cundall [4] ts this kind of model. It uses disks in 2D and spheres

in 3D. The bonds can resist axial and shear forces and moments.

It has been used to simulate rock fracture with good results, but

one of the disadvantages of this method is that when fracture

occurs the volume is not dened because of empty spaces

between particles. In comminution machinery, it is really important to estimate accurately the production rate so that no

ambiguities can be allowed. This problem is overcome by the

Discrete Grain Breakage method (DGB) prop. Herbst and Potapov [5]. It uses triangles in 2D and tetrahedrons in 3D so that no

spaces between particles are left.

Although these methods, BPM and DGB, give very good results

as big problem arises when used for simulating comminution

machinery. They require large number of particles to describe one

rock, and considering that a crusher or mill handles many tons

per hour, the time needed for the simulation is huge. Whittles

et al. [6] report simulations of as many as 5 days for a conventional uniaxial compressive test where 12,047 spherical elements

are needed for rock samples simulation, so it is evident that

application of these methods in crushers is difcult with todays

computational power.

Djordjevic et al. [7] and Lee et al. [8] used glued particles

simulations to describe the impact crusher comminution process.

Impact crushers accelerate rocks against the walls, so impacts

occur independently. This fact allows an important simplication

that it is not necessary to simulate all the rocks processed. DEM

technique provides data for single particle impacts then it is

extrapolated according to the crusher production rate, hence

results can be obtained in a reasonable time.

2.2. Population Balance Model based

The Fast Breakage Model (FBM) developed by Potapov et al. [9]

is a hybrid method that uses both DEM and Population Balance

Model (PBM). DEM with polyhedral convex polygons is used to

determine kinematics and PBM for breakage. PBM divides the

problem in three stages, each one associated with a specic

function, selection function, describes which particles will break,

appearance or breakage function, describes the degree in which

the selected particle undergoes breakage and discharge function

which describes the particles that leave the process.

FBM uses DEM to determine the forces and energy applied to

particles, when a certain level is exceeded the appearance

function is used to describe the rock breakage. Lichter et al.

[10] use FBM to simulate a cone crusher. They reported good

results using this method, and that a typical simulation takes

3 weeks.

Powell and Morrison [1] highlight that the appearance function in PBM is always an issue. It is ore specic, and prior

intensive laboratory test work is needed which cannot be done

in an industrial setup.

In this section, we explain the foundations of our proposed

approach. Let us consider a rock that contains randomly oriented

cracks. These cracks are generated by loads that have been

place at the original site and those induced by the process used to

extract it, normally by rock blasting. When the rocks are comminuted, compression forces are applied to them. These forces

create internal stresses which activate existing cracks that grow

and therefore rock subsequently undergoes rupture and splits.

The above ideas are the basis of the Discrete Crack Model

(DCM). We consider that rocks contain random crack seeds that

due to the comminution process get activated and eventually

cause the rock rupture. In DCM, crack seeds are placed in

randomly and which one is activated is detected. Enough crack

seeds must be used to assure failure detection.

Modern methods such as the Rock Failure Process Analysis

(RFPA) use the Finite Elements Method (FEM) to determine the

stress eld on the rock. The MohrCoulomb strength criterion

with tensile cutoff is also used to determine if an element fails.

When this happens, the stress distribution changes, and at some

points the stress can increase so that more failures may occur.

When no more failures occur, time is increased and the simulation continues. It is not the intention here to introduce RFPA

method, hence only a small description is given. For a detailed

discussion see [11,12].

A combined FEMDEM method proposed by Munjiza [13], has

been successfully used for simulations. However FEM based

methods cannot be used in large DEM simulations because it is

too CPU intensive. So although with FEM it is possible to

determine how individual rocks will break under certain loads,

it is not practical for large simulations.

A convenient shortcut would be to estimate the rocks stress

eld without using FEM (i.e. analytically), but this is a really

difcult task and some simplications must be made to achieve

acceptable results. Hence in our method, we consider the rock

shape exclusively as convex polygons or polyhedrons; concave

shapes have stress concentrations that are too difcult to solve.

Note that this shape consideration is also required in FBM.

Although rocks are anisotropic in nature here they are

considered as isotropic for simplication. The last consideration used to simplify the model is that cracks exist in an innite

medium and do not change material properties or interact with

each other.

These considerations have important implications. As stated

previously, when RFPA detects a failed element the stress eld is

computed again. Such consideration is not used by our model.

Cracks do not interact so in a dynamic simulation the stress eld

is not affected by activated cracks. This implies that the fracture

path is derived using the stress eld before failure. Zacharopoulos

and Kalaitzidis [14] claim that this is not correct and the authors

agree on this, but nevertheless Ostoja-Starsewski and Wang [15]

present evidence that experiments on seven samples with

nominally identical shape presents different crack patterns.

Hence crack paths cannot be exactly replicated even by carefully

controlled experiments, and numerical simulations with state of

the art algorithms can only determine the major crack path. It is

obvious that DCM cannot be proposed to exactly determine the

complex crack path generated during a rock breakage process, but

it can give enough information for determining failure type, crack

initiation, and propagation direction so approximate rupture

information is obtained to simulate comminution.

Even if it was possible to determine the exact crack path in

rock rupture, the shape of the remaining pieces would be a

complex polygon with large number of vertex, increasing computation time for performing contact detection. The best choice

according to todays computational power is therefore to obtain a

roughly approximated crack path, so rocks can be represented

with a small number of vertex and contact detection can be

performed fast and efciently.

D.A. Estay, L.E. Chiang / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 125133

127

DCM. In the following analysis: compressive stress is positive and

s1 4 s2 4 s3 , where si are the three principal stresses. It was

shown by Hoek [16] that only s1 and s3 are responsible for crack

propagation.

Three situations may occur when a crack propagates.

If s1 r 0and s3 Z 0, a tensile stress on the tip exists and if the

materials tensile stress is overcome, the crack propagates in a

direction perpendicular to the minimum stress, the stress at the

tip will continue in tension and the crack will propagate until the

axial splitting occurs.

When s1 Z 0 and s3 Z 0, the crack propagates perpendicular to

the minimum stress. When the crack is aligned with the mayor

principal stress direction, the stress at the tip will be compressive

and the crack will no longer propagate. The crack growth is in the

order of the crack length so the rock will not break.

In the same case of s1 Z 0 and s3 Z 0 a second possibility

exists. Experimental evidence obtained by Hoek and Bieniawski

[17] show that shear failure occurs when

Flamant [18] proposed a solution for the stress of a lineal

elastic wedge by modifying the three dimensional solution

proposed by Boussinesq [19]. The stress eld is simple compression in the radial direction:

sxy ZC 0 msy

cohesion, and m is the internal friction angle.

Let us recall that in a standard DEM simulation, four basic

stages are performed. The rst stage is Contact Detection

ParticleParticle contact and ParticleWall contact is determined

at current time step, contact detection is the most time consuming

stage. The second stage is Force Calculation particles in touch

have reaction forces in the contact point, the force magnitude is

determined according to overlap. The third stage is Time Integration where the Newtons Law is used to calculate acceleration,

and time integration is carried in order to determine velocity

and position in the next time step. The nal stage is Position

Update the model is updated with new positions and the cycle

starts again.

When DCM is incorporated for rock rupture consideration,

three new additional stages are included. The rst step in DCM

cycle is the Stress Calculation, CPSA method is used to

determine the stress eld according to the contact forces.

The next stage is the Breakage Determination, here the

stress eld is evaluated according to a breakage criterion.

Two conditions are evaluated: tensile failure and shear failure.

The third and nal stage is Rock Fragmentation where

subdivision is performed according to which mechanism is

determined.

The newly added stages from DCM allows to simulate rock

breakage when performing a DEM simulation. The advantages of

DCM are that opposite to phenomenological methods like FBM,

breakage is based on physical laws by approximating the stress

eld. Also the rocks are considered as a single element so

simulations with DCM are faster than using Glued Particle

method (previously explained in Section 2.1).

The advantage of using CPSA over FEM is that there is no need

for mesh calculation or solving a large set of system equations, so

calculation time is much faster. CPSA is similar to an analytical

solution in that once the boundary stress is computed the stress

inside the rock can be computed at one point or in as many as

wanted without the need to solve a grid.

Nevertheless CPSA has some disadvantages with respect to

FEM. It may generate a larger error than FEM, and it is not

possible to directly nd the maximum stresses since the stress

eld is constructed by computing the stress at several selected

points.

srr

2C 1 cos y 2C 3 sin y

,

r

r

sry 0, syy 0,

conditions according to

Z b

F1

C 1 cos y C 3 sin ycos ydy 0

a

F2

where F1 and F2 are the applied forces, and a and b are the wedge

angles shown in Fig. 1.

This analytical solution is used to determine the stress eld in

a disk as in the case of the conventional Brazilian test for example.

As depicted in Fig. 2, it can be shown that a solution can be

obtained by applying the Flamant solution for each force. This will

create a compressive stress state at the perimeter of the disk,

adding a constant radial traction force at circumference of the

disk, the perimeter becomes stress free and it looks as below:

X F

si,j x,y

si,ji x,ysTi,j x,y

4

Fig. 1. Angles and forces for Flamant linear elastic wedge solution.

Flamant solution.

128

D.A. Estay, L.E. Chiang / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 125133

where si,j x,y is the stress in the disk, sFi,ji x,y is the stress

predicted by Flamant solution, and sTi,j x,y is the stress generated

by the traction force at the perimeter

Timoshenko and Goodier [20] analized more complicated load

conguration. The solution has a similar form, which is to apply

the Flamant solution for each force and apply a radial tension

solution.

3.2.2. Convex Polygons Stress Approximation (CPSA)

When we change from a disk to a more complex shape like

polygons the same principle can be applied, for each force we use

the Flamant solution. This will generate a compressive stress at

the boundary, the only difference is that the applied force at the

perimeter to eliminate the compressive stress is not constant and

because of this the stress distribution sTi,j x,y generated inside

the rock is not easily found:

sxy nx syy ny syz nz ty

sxz nx syz ny szz nz tz

solution at the perimeter must satisfy the natural boundary

conditions formulated in Eq. (5). This is done with Eq. (6).

If sTi,j x,y can be solved, then the stress inside the rock is directly

determined by Eq. (4). Considering that the shape of the rock can

change and that the forces can be at any point in the perimeter

then it is really difcult to develop an analytic expression for

sTi,j x,y. The problem could be solved numerically by FEM or an

equivalent method, but then it becomes more convenient to solve

directly the original problem without resorting to additional

computations. By using FEM at this stage the only advantage

would be that a coarse mesh can be used.

X

!

n 0

6

sFi sG d!

In our work, a much simpler but effective approach is used.

When a point force acts on a surface a large stress gradient exist,

specially near the force application point. When distributed forces

are applied, such gradients are not as high and a soft function

solution exists, this is the reason why a coarse mesh may be used

to solve it by FEM. It is known that the force that generates

sTi,j x,y is distributed with known values at the perimeter, then

the function is soft and a simple interpolation is used.

In the interpolation process, rst nT points are selected,

distributed at the boundary, then the Flamant solution is applied

at these points to calculate the stress. Once the stress is calculated, the natural boundary conditions are used to determine

sG xi ,yi according to Eq. (6). Finally the stress at any point inside

the rock can be approximated by Eqs. (4) and (7):

sTi,j x,y

nT

X

wi x,ysG x,y

i1

where sTi,j x,y is the stress caused by the tension solution on the

perimeter, wi x,y is the weight function, and sG xi ,yi is the stress

in the ith point of the perimeter.

It is not necessary to use all the nT points in the perimeter, a

subset with the n closest points can be used with good results.

The weighting function used depends on the distance from the

crack seed point to the boundary point and it is represented by

Qn q

xj x2 yj y2

j1

q

wi x,y

8

xj x2 yj y2

where x,y is the crack seed coordinates, and xj ,yj is the point at

the perimeter.

and computation time is proportional to the number of crack

seeds. Conversely, if FEM is used, a large percentage of computation time is used for solving the system equations and all node

stresses are computed every time.

3.2.3. Convex Polyhedron Stress Approximation

As mentioned previously in Section 3.2.1, Flamant solution

predicts a stress eld with compression in the radial direction and

null components in polar direction as shown in Eq. (2). As the

stress is exclusively radial, the solution can be applied to wedges.

The 3D solution found by Boussinesq [19] for a concentrated

normal load and by Cerruti [21] for a tangential load is not as

general, it is applicable only for half spaces. As shown in Eq. (9),

the BoussinesqCerruti solution in spherical coordinates has nonzero azimuthal and polar components. Analytical solution for

concentrated load on wedges or corners has not been found.

sinf

12n

srr

A 2n

F z 12n22ncosf

2

1 cosf

pr

12n cotfcosf

12n cos2 f

sff

A

1cosf F z

2

1 cosf

2pr

2pr2 1 cosf

12n cotf1cosf

12n cosfsin2 f

A

2 cosf F z

2

1 cosf

2pr

2pr2 1 cosf

12n cotf1cosf

12n cosfsinf

srf

A

Fz

1 cosf

2pr2

2pr2 1 cosf

12n cotf1cosf

sry

B

1 cosf

2pr2

12n 1cosf

syf

B

2pr2 1 cosf

syy

the concentrated point force, and n is the Poisson ratio.

Without an analytical solution for concentrated forces acting

on corners, it is not possible to extend CPSA to 3D. To overcome

this problems two approaches can be employed. It can be noted

that when Poisson ratio n 0.5, azimuthal and polar components

are zero. Stress becomes radial and BoussinesqCerruti solution

can be applied to wedges and corners.

The second approach is to consider Poisson ratio not equal to

0.5. In this case the stress at the faces conforming the wedge or

corner does not satisfy the natural boundary conditions. This can

be xed by using the same interpolation principle used in 2D

CPSA described in Section 3.2.2.

Both approaches can be used with acceptable results. The

algorithm is exactly the same as in 2D CPSA, the only difference is

that in 2D Flamant solution is used and in 3D BoussinesqCerruti

solution is used.

3.3. Stress calculation performance

Convex Polygon Stress Approximation (CPSA) is a new method

we propose to approximate the stress eld inside a convex

polygon for comminution process simulation. It is based on the

Flamant solution for linear elastic wedges and an important

assumption is made, which is that stress caused by a distributed

traction force on the perimeter can be approximated by simple

interpolation. It is important to validate the CPSA results to check

its validity. Hence in this section CPSA stress results are compared

with stress results obtained by Salome-Meca [22,23], a FEMbased program.

In the FEM models of Fig. 3, displacements are prescribed to

simulate the forces shown. Then stresses at the nodes and the

reaction force from imposed displacements points are calculated.

The computed forces are then used as input to the CPSA model,

D.A. Estay, L.E. Chiang / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 125133

the relative error is calculated as in Eq. (10). Three loading cases

are examined: (a) point to point; (b) point to plane and (c) plane

to plane loading.

Error

9sFEM sDCM 9

sFEM

10

Here a point load is applied at the top and bottom corners of a

diamond shaped rock as shown in Fig. 3a, the stress is computed

by CPSA method and results are shown in Fig. 4a.

The solution shows that the stress is higher at the loaded corners

and smaller at the right and left side. For direct comparison, Fig. 4b

shows that the relative error is small at the center and over 100% at

the sides, this large error causes no problem for DCM simulation

because as shown later all activated crack seeds fall in the low error

zone, not in the high error zone.

129

Here a triangle shaped rock is loaded at the top corner and at

the base as shown in Fig. 3b. When this model is solved by FEM,

the distributed force at the bottom is simulated as multiple point

forces applied at the nodes. The same principle is applied in CPSA,

so both models have exactly the same input forces. The stresses

are calculated by the CPSA method and results obtained are

shown in Fig. 5a.

As expected, stress is higher at the top corner, and lower at the

bottom side where the applied force is distributed. As in Case 1 a

low stress area is located at both sides. In this area the relative

error is large as shown in Fig. 5b. This large error causes no

problem for DCM simulation because as shown later all activated

crack seeds will occur in the low error zone.

3.3.3. Case 3: plane to plane load

A rectangular shaped rock is loaded at the top and bottom

faces as shown in Fig. 3c. In FEM models distributed loads are

discretized and applied at nodes, the same practice is applied

here. So the input for both models is the same. Stresses are

computed by CPSA method and the results are shown in Fig. 6a.

The stress eld in this case is smooth mainly because of the

absence of point forces, or stress concentration points. No areas

with small stress relative to the maximum exist, so it can be

expected that the error with respect to FEM results is small as

shown in Fig. 6b.

Fig. 3. Loading cases examined: (a) point to point loading; (b) point to plane

loading; and (c) plane to plane loading.

each activated crack generates changes in the neighborhood of

the stress eld. The CPSA method only allows us to approximate

the stress eld when no crack exists, so in this section we

Fig. 4. Stress eld for a diamond shaped rock: (a) Von Misses stress, solved by the CPSA method in point to point loading case; and (b) relative error between FEM and

CPSA solutions for point to point loading case.

Fig. 5. Stress eld for a triangle shaped rock: (a) Von Misses stress, solved by the CPSA method in point to plane loading case; and (b) relative error between FEM and CPSA

solutions for point to plane loading case.

130

D.A. Estay, L.E. Chiang / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 125133

Fig. 6. Stress eld for a square shaped rock: (a) Von Misses stress, solved by the CPSA method in plane to plane loading case; and (b) relative error between FEM and CPSA

solutions for plane to plane loading case.

Fig. 7. Activated cracks seeds in fracture process for the three analized loading cases: (a) point to point loading; (b) point to plane loading; and (c) plane to plane loading,

green cracks seeds are activated by shear and red by tensile failure, red arrow represents the propagation direction of tensile cracks. (For interpretation of the references to

color in this gure legend, the reader is referred to the web version of this article.)

the crack, enough information for characterizing the rupture

process is obtained. This is based mainly on the fact that activated

cracks affect the stress eld only on the vicinity and hence the

simplication of no interaction between cracks can be used. This

implies that we only need to know the stress eld before fracture

occurs to apply DCM.

The rock specimen will split when one of two predened

conditions is reached, that is to say when s1 r sTensile and s3 Z0,

a tensile stress on the tip exists and the crack propagates in a

direction perpendicular to the minimum stress, breaking the rock

into two pieces. When both principal stresses are compressive

shear failure occurs when Eq. (1) is satised.

CPSA allows to evaluate the stress at any point in a way similar

to an analytic solution, but still the maximum compressive or

tensile stress cannot be directly determined. Hence enough crack

seeds must be considered in order to detect correctly which crack

is activated. In the following examples for illustrative purposes

1000 random cracks seeds are generated inside each rock. The

stress eld is calculated by the CPSA method. At each crack, as

load increases, the rupture criterion is applied to evaluate

potential failure along the crack seeds.

Three loading cases are examined. Results are shown in Fig. 7

where crack seeds that satisfy predened conditions are painted

in red if tensile failure is reached and in green for the shear

condition. When rock reaches breakage condition, the direction of

crack propagation is shown with a red arrow for tensile failure

and green for shear failure.

4.1. Case 1: point to point load

Fig. 7a shows how cracks activate when breakage condition is

achieved. As load is increased the stress eld is calculated using

CPSA method. When the load is small stress concentrations at the

loading points create high stress points. Cracks located near the

force application point have shear failure. When load increases

more shear failure points occurs near the loading points, and

nally tensile cracks appear at the center of the rock. The red line

indicates the direction in which the crack should propagate, so in

this case clearly the rock will be divided in two main parts. This

behavior is similar to the Brazilian test where a disk is loaded

with opposite point forces. Khanal et al. [24] studied breakage in

the Brazilian test and reported that failure is by axial splitting.

It is also possible to say that the area near the loading point is

damaged due to contact and the multiple cracks activated by

shear are proof of this.

It is important to compare Fig. 7a with Fig. 4b. We can clearly

see that all activated crack fall in a zone with a low stress error

between FEM and CPSA, so the approximation for the stress eld

using CPSA is really helpful given the simulation time reduction

achieved.

4.2. Case 2: point to plane loading

Fig. 7b shows activated cracks in plane-to-plane loading at the

nal stage of loading. When load is small, shear failure occurs at

the point force application point. The cracks grow quickly as the

load increases. When breakage condition is achieved tensile

cracks appear, and given the crack propagation direction it can

be concluded that the cracks will propagate in an inclined

direction towards the loading point. Similar to point to point

loading (Case 1), it can be concluded that contact damage will

occur at the top.

As in point to point loading, we can identify the zone of

activated cracks. From Fig. 5b it is important to note that all

activated cracks are found in a zone with low error between CPSA

and FEM.

4.3. Case 3: plane to plane load

Fig. 7c shows activated cracks in plane-to-plane loading. Since

no point forces exist, no stress concentration points or shear

D.A. Estay, L.E. Chiang / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 125133

131

Fig. 8. Fracture pattern for the three different loading cases examined: (a) point to point loading, the rock is subdivided in two major parts by tensile failure, shear

activated cracks at the contact points are proof of contact damage; (b) point to plane loading, shear fractures are generated at the contact point and tensile cracks are

generated towards the loading point; and (c) plane to plane loading, a shear crack propagates at the center of the rock, only a few cracks seeds present shear failure at the

contact surface so no damage is assumed.

Fig. 9. DCM simulation of a disk Brazilian Test: (a) loading principle of Brazilian test in a cylindrical shaped rock; (bd) Von Mises stress before and after breakage;

and (e) rock fragments at the nal time step.

Table 1

DCM simulation parameters for Brazilian test

Parameter

Value

Loading rate [mm/s]

Compressive strength [Mpa]

Tensile strength [Mpa]

Friction angle [deg]

Cohesion [Mpa]

Youngs modulus [Gpa]

Specimen radius [mm]

2.5 10 3

0.01

129.5

13

50

4.23

36

50

early stage of loading there are no activated cracks. When load

keeps increasing shear failure points begin to appear but not

enough to assume damage in the contact area.

At the nal stage of loading tensile failure appears. In the same

gure there is also evidence of shear cracks in the center of the

specimen. This is consistent with the behavior in the Uniaxial

Compressive Test (UCT) where two fracture modes can be

recognized, as reported by Li et al. [25]. Their laboratory tests

show that an specimen in UCT develops shear and tensile cracks.

They also found that no matter what the nal failure method,

axial splitting is always present. When comparing Fig. 7c

with Fig. 6b all activated crack are in a zone of low relative

error between FEM and CPSA, as found also in the previous

loading cases.

4.4. Fracture pattern

Using activated crack seeds allows to identify the fracture

mechanism and a general description of the rock fracture. Fig. 8

similar documented experiments. Fig. 8a represents Case 1: point

to point loading. Tensile crack seeds located at the center indicate

a tensile failure, the crack split the rock in two main parts. The

area near the force application points shows high stress and shear

failures. This is evidence for contact damage. Fig. 8b represents

Case 2: point to plane loading. It can be seen that two groups of

tensile activated cracks seeds appear, the cracks directions are

inclined towards the point force. As in point to point loading, a

high stress area is generated near the loading point and again this

is evidence for contact damage. Fig. 8c represents Case 3: plane to

plane loading. Shear activated crack seeds at the center show that

the specimen undergoes shear failure. An inclined crack splits the

rock, where the angle of the crack depends on material properties.

As only distributed forces act there are no high stress zone and no

contact damage occurs, results that are in agreement with Wang

et al. [26] experiments reported in the literature.

The Brazilian test is a standard laboratory experiment to

indirectly determine the tensile strength of brittle materials.

Specications has been established by the American Society for

Testing and Materials (ASTM D 3967-86) and by the International

Society for Rock Mechanics (ISRM).

The Brazilian test consists of loading a cylindrical rock sample

of radius R and thickness L as shown in Fig. 9a. Previous attempts

for a computer simulation with DEM has been carried with

remarkable results, but the computation time requirements are

substantial.

A simulation of Brazilian test using DCM is presented in Fig. 9.

The simulation parameters used can be found in Table 1. It is

132

D.A. Estay, L.E. Chiang / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 125133

However, for the purpose of illustration, shown in Fig. 10 is the

convergence behavior for the Brazilian test simulation. It can be

seen in this example that convergence is reached when using in

the order or more than 600 crack seeds.

Fig. 11 shows the loaddisplacement curve. According to ASTM

the splitting tensile strength is calculated with Eq. (11). For this

simulation the calculated tensile strength using the peak force in

the loaddisplacement curve is 13.62 MPa which is 4.8% greater

than the input tensile strength. This difference is reasonable given

that DCM is an approximated method to obtain the stress eld.

After breakage the elements cannot resist the compressive force

and the load falls to nearly zero.

st

Fig. 10. Convergence behavior of tensile strength when the number of crack seeds

changes in a Brazilian test simulation.

2P

pLD

11

applied load, L is the thickness of the specimen, and D is the

diameter of the specimen.

Thus, DCM is able to determine the main crack path and the

secondary damage at the contact points. It can also predict when

the breakage occurs. Computation time for this simulation is

5 min and 32 s. This compared to computation times of previously

published reports is the outstanding improvement generated

by DCM.

Munjiza et al. [28] used the combined nitediscrete elements

method (FEM/DEM) to study the behavior of rocks in the Brazilian

test. Munjiza points out that the simulation loading rate he used

(0.1 m/s) is much higher than that in actual laboratory experiments (0.01 mm/s), because in combination with the small time

step used (2.5 10 5 ms) would require several weeks of computation. For comparison purposes a DCM simulation carried with

the same time step takes 308 min. This amazing time reduction

characteristic is possible because rocks are treated as a single

element.

6. Conclusions

Fig. 11. Loaddisplacement curve for a Disk Brazilian test simulation. Note that

loading platen was not in contact with the disk at start.

stability depends of the rigidity and mass of the elements. As new

and smaller elements are generated during a DCM simulation the

program automatically calculates the required parameters for

stability after each breakage. In this simulation a xed time step

is used in order to be able to compare elapsed computation time.

The time step chosen is small enough so no instabilities occur

after breakage.

Fig. 9b shows that no crack growth exists. When breakage

condition is achieved the specimen is instantaneously subdivided

as shown in Fig. 9d. Fig. 9e shows the remaining pieces of the

rock. The major crack is vertical due to axial splitting. The smaller

fragments are generated due to damage from the contact points,

this behavior is described by Jaeger et al. [27] and states that

when forces are distributed the likelihood of shear failure at

contact points is decreased.

We must point out that crack seed density is an important

operational parameter in DCM. A compromise exists between

accuracy and speed. If not enough crack seeds are placed, there is

a higher probability that DCM will fail to detect the true failure

point. On the other side, too many cracks will increase computation time. The effect of crack density is part of ongoing research

efciently model rock fracture within DEM simulations, and it

has been shown that enough information to predict rock rupture

is obtained when used, making it an appropriate tool for modeling

comminution processes.

The most signicant contribution of the DCM method is that it

is based in an efcient analytical solution (CPSA) found on

physical rules by approximating the stress eld inside an intact

convex shaped rock. It has been shown, by comparison with FEM,

that CPSA gives a really good approximation at the points of

interest, this meaning that all relevant high stress points are

computed with low error. CPSA may give larger errors in zones of

low stress, but this fact is not important since cracks do not get

activated in these zones, so rock rupture simulation delity is not

affected by this characteristic.

Previous work by Tang et al. [29] introducing the Rock Failure

Process Analysis (RFPA) method, which is used here to assess

DCM validity, and similar results were obtained. For the analyzed

cases it is possible to conclude that when opposed point forces

(Case 1) are applied to the rock the rupture mechanism is axial

splitting. As force increases, cracks near the loading points

activate by shear. At the nal stages of loading, tensile cracks

activate at the center of the rock. This mechanism is consistent

with the axial splitting mechanism, and cracks predicted near the

loading point indicate crushing due to stress concentration.

In the plane to point (Case 2) loading case, at the early

loading stages, shear activated cracks appear near the point force

D.A. Estay, L.E. Chiang / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 125133

At the nal stages of loading, tensile cracks appear at the center of

the rock specimens and the cracks propagation direction indicate

that they have an inclination towards the loading point.

In the plane to plane loading case (Case 3) only a few cracks are

activated near the loading point, and it is assumed that they are not

enough to establish crushing occurring in this zone. When the load

increases, shear and tensile cracks appears at the center of the rock. If

tensile failure is assumed, cracks should be vertical, and if shear

failure is assumed then an inclined crack must split the rock. As

stated before, the shear condition is necessary but not enough to

initiate shear movement, so both fracture modes can occur.

When shear cracks appear near the loading zone, crushing can

be assumed, and when they are in the middle of the specimen

they are an indication of shear failure. Tensile cracks must always

be associated with splitting.

For our purpose, which is to model comminution processes,

predicting the exact crack path is not important. We must bear in

mind that real cracks follow such a complicated path that even if we

could exactly determine its resultant geometry, it would be extremely

complicated to do so, and conventional contact detection algorithms

would be too CPU intensive to be of any practical value. Exact crack

path computation is not cost-effective in large scale simulations with

todays CPU power for the purpose of comminution simulations.

As postulated by Zacharopoulos and Kalaitzidis [14], real crack

paths cannot be determined alone from the stress elds known prior

to failure, and experiments also demonstrate that even with nominally equal specimens, cracks will have different paths. Hence DCM is

not intended for determining the exact complex crack path generated

during the rock breakage process, but instead it gives enough

information for determining failure type, crack initiation, and propagation direction, so approximate rupture characteristics can be

obtained to simulate comminution within DEM.

It is possible to extend DCM to 3D and good results have been

achieved. Our current work is focused on integrating DCM in to a 3D

DEM software, and results are in the preliminary stages of publishing.

Supplementary data associated with this article can be found in

the online version at http://dx.doi.org/10.1016/j.ijrmms.2012.12.041.

References

[1] Powell MS, Morrison RD. The future of comminution modelling. Int J Miner

Process 2007;84:22839.

[2] Bond FC. The third theory of comminution. Trans Soc Min Eng AIME

1952;193:48494.

133

[3] Bond FC. Crushing and grinding calculations parts I and II. Br Chem Eng

Process Technol 1961;6:37885.

[4] Potyondy DO, Cundall PA. A bonded-particle model for rock. Int J Rock Mech

Min Sci 2004;41:132964.

[5] Herbst JA, Potapov AV. Making a Discrete Grain Breakage model practical for

comminution equipment performance simulation. Powder Technol

2004;143144:14450.

[6] Whittles DN, Kingman S, Lownde I, Jackson K. Laboratory and numerical

investigation into the characteristics of rock fragmentation. Miner Eng

2006;19:141829.

[7] Djordjevic N, Shi FN, Morrison RD. Applying discrete element modelling to

vertical and horizontal shaft impact crushers. Miner Eng 2003;16:98391.

[8] Lee H, Kwon JH, Kim KH, Cho HC. Application of DEM model to breakage

and liberation behaviour of recycled aggregates from impact-breakage of

concrete waste. Miner Eng 2008;21:7615.

[9] Potapov AV, Herbst JH, Song M, Pate WTA. DEMPBM fast breakage model for

simulation of comminution processes. In: Proceedings of discrete element

methods, Brisbane, Australia, 2007.

[10] Lichter J, Lim K, Potapov A, Kaja D. New developments in cone crusher

performance optimization. Miner Eng 2009;22:6137.

[11] CRISR, RFPA Users Guide, Northeastern University, PR China, 1997.

[12] Tang C. Numerical simulation of progressive rock failure and associated

seismicity. Int J Rock Mech Min Sci 1997;34:24961.

[13] Munjiza AA. The combined Finite-Discrete Element Method. New York:

Wiley; 2004.

[14] Zacharopoulos DA, Kalaitzidis PA. Crack path stability in center cracked

rectangular plate. In: Proceedings of the fth international conference for

mesomechanics. Tokyo; 2628 August 2003, p.2118.

[15] Ostoja-Starzewski M, Wang G. Particle modeling of random crack patterns in

epoxy plates. Probab Eng. Mech 2006;21:26775.

[16] Hoek E. Brittle fracture of rocks. In: Rock, editor. Mechanics in engineering

practice. New York: Wiley; 1968 p. 99124.

[17] Hoek E, Bieniawski ZT. Brittle fracture propagation in rock under compression. Int J Fract 1965;1:13755.

[18] Flamant A. Sur la repartition des pressions dans un solide rectangulaire

charge transversalement. Compt Rend Acad Sci 1892;114:1465.

[19] Boussinesq J. Application des potentiels a letude de lequilibre et du

mouvement des solides elastiques. Partis: Gauthier-Villars; 1885.

[20] Timoshenko S, Goodier J. Theory of elasticity. New York: McGraw-Hill; 1951.

[21] Soutas-Little RW. Elasticity. New York: Dover; 1999.

[22] Salome-Meca [computer program], 2010.

[23] Code Aster [computer program], 2011.

TDEM. simulation of diametrical compression

test on particle compounds. Granular Matter 2005;7:8390.

[25] Li C, Prikry R, Nordlund E. The stressstrain behaviour of rock material

related to fracture under compression. Eng Geol 1998;49:239302.

[26] Wang Y, Forssberg E, Klymowsky R. Fine comminution of limestone by roller

press-stirred ball milling. Aufbereit-Tech 1998;39:26778.

[27] Jaeger JC, Cook NGW, Zimmerman RW. Fundamentals of rock mechanics. 4th

ed. Oxford: Wiley-Blackwell; 2007.

[28] Munjiza A, Mahabadi O, Grasselli G. Numerical modelling of a Brazilian Disc

test of layered rocks using the combined nite-discrete element method.

In: Proceedings of the 3rd CANUS rock mechanics symposium, Toronto, 915

May 2009, p. 878.

[29] Tang CA, Liu HY, Zhu WC, Yang TH, Li WH, Song L, et al. Numerical approach

to particle breakage under different loading conditions. Powder Technol

2004;143144:13043.

- XFEM for reinforced concreteDiunggah olehRabee Shammas
- Ghazvinian Ehsan 201009 MAScDiunggah olehFrancisco Javier Villaseca Ahumada
- 255 Geomech ReprintDiunggah olehgutian259
- Influence of casting condition on the anisotropy of the fracture properties of Steel Fibre Reinforced Self-Compacting Concrete (SFRSCC)Diunggah olehRodrigo Lameiras
- Chong 2014Diunggah olehmarcusin
- Carpinteri 2003 n.308 JamDiunggah olehL.a. Mayor
- 430_Fracture Assessment of a Ship StructureDiunggah olehtt3340
- cijepanje drvaDiunggah olehMatija Kelčec Pester
- MRI_4_75_2001Diunggah olehr_simoes
- A Review Paper on Study of Progressive Damage of Composite Structure under Tri Axial Loading by Using Micromechanical based Failure TheoriesDiunggah olehEditor IJTSRD
- ZZZZDiunggah olehKlisman A. Flores Durand
- PreambleDiunggah olehManish Raj
- CoursDiunggah olehAbdel Daa
- Strut-And-Tie Modeling Provisions- What When and HowDiunggah olehHildebrando Santos
- Introduction Fracture Mechanics 1Diunggah olehaap1
- 22-3-3kosaDiunggah olehmohammed_fathelbab
- 14-Stress Analysis Seminar-En.pdfDiunggah olehlaminars
- Composite Materials Technology 31-07-2010 25Diunggah olehaannbb12
- 03EE_ISRMCongress-RockStrength.pdfDiunggah olehiqbal
- Dynamic Behavior of a Mediterranean Natural Stone Under Tensile LoadingDiunggah olehShahab Yazdegerdi
- MohanDiunggah olehmohan_rapaka
- s6-4-DriverDiunggah olehShanice McAlister-Bishop
- yang2018.pdfDiunggah olehAngga Fajar Setiawan
- concrete modelingDiunggah olehRicardo Escobar
- annual_report_2013-14_final.docxDiunggah olehManoj
- 2008 Three Dimensional Combined CervenkaDiunggah olehLucas Marques
- paliwal2008.pdfDiunggah olehJorge Quispe
- Prediction of the load carrying capacity of bolted.pdfDiunggah olehHusseinali Hussein
- Mg alloyDiunggah olehFauzi Mahdy
- 2014_Pittsburgh_Geomechnics_Workshop_final.pdfDiunggah olehshabraye2

- sdffDiunggah olehPreso UL
- 254711249-Balances-de-Materia-y-Energia.pdfDiunggah olehYesi Ceballos
- Molienda.docxDiunggah olehYesi Ceballos
- Copper and Cyanide Recovery in Cyanidation EffluentsDiunggah olehYesi Ceballos
- Cyclones FilterDiunggah olehMarco Palezzato
- Merrill Crowe Flowsheet Zinc Precipitation on Gold RecoveryDiunggah olehUser20141
- Carbide Design HandbookDiunggah olehDidier Marneffe
- Water Testing Procedures NALCODiunggah olehYesi Ceballos
- Final JKDW and SMC Report for Veladero ProjectDiunggah olehYesi Ceballos
- CSC Operator - Laboratory - VeladeroDiunggah olehYesi Ceballos
- Column Leach Testing Guideline v2Diunggah olehYesi Ceballos
- CSC Assistant - Metallurgical Accounting - Lumwana v1Diunggah olehYesi Ceballos
- A Method of Predicting Breakage Properties of a Particulate Material When Subjected to ImpactDiunggah olehYesi Ceballos
- Cip Plant SaimmDiunggah olehgloffers
- GEMI Water Balance CalculatorDiunggah olehYesi Ceballos
- WaterBalance_onlinetextDiunggah olehYesi Ceballos
- Filter Precoating Better Airflow Longer Life Uas Technical ArticleDiunggah olehYesi Ceballos
- Design Requirements of Precoat Filters for Water FiltrationDiunggah olehYesi Ceballos
- Zeta potential of air bubbles conditioned with typical froth flotation reagents.pdfDiunggah olehYesi Ceballos
- Process mineralogy.pdfDiunggah olehimran58175
- hand bookDiunggah olehJhordyEfrainBejarMamani
- Alan Lu Challenges of Large Heap Leach Pad Design in Northern ChinaDiunggah olehYesi Ceballos
- MassBalance SpreadsheetDiunggah olehFrankcito Gallegos Cardenas
- SILV AssetHandbook FINAL Web Nov2016Diunggah olehWilliam Thomson
- Key Technologies of Drilling Process With Raise Boring MethodDiunggah olehYesi Ceballos
- Froth Zone Characterization of an Industrial Flotation Column in the Rougher CircuitDiunggah olehYesi Ceballos
- Towards a virtual comminution machine.pdfDiunggah olehYesi Ceballos
- Zeta Potential of Air Bubbles Conditioned With Typical Froth Flotation ReagentsDiunggah olehYesi Ceballos
- Effect of Comminution on Particle Shape and Surface Roughness and Their Relation to Flotation ProcessDiunggah olehYesi Ceballos

- Creo Simulate SeminarDiunggah olehAnonymous P8Bt46mk5I
- Drama Lesson PlansDiunggah olehEdelin Moscoso
- FDS+EVAC_5_GuideDiunggah olehAntonio Fuentes
- Olga Dynamic Multiphase Flow SimulatorDiunggah olehArif Mokhtar
- FX Blue Trading Simulator v3 for MT4 - User GuideDiunggah olehHariyanto Liem
- Good Random Number GeneratorsDiunggah olehblablablou
- AIAA_ebooksDiunggah olehsarah86304325
- 1676Diunggah olehafe2000must9083
- The_Instructional_Media_for_Different_Le.pdfDiunggah olehisabela
- 01-GettingStarted.pdfDiunggah olehGiuseppe
- Ejemplo SysML MecatronicaDiunggah olehorbedelodo
- Cst Studio Suite - Getting StartedDiunggah olehAaliah Zedek
- ANSYS Workbench LS-DYNA Users InstructionsDiunggah olehNenad Nestorovic
- 3-D Simulations of Ignition Trasients in the RSRMDiunggah olehAugusto De Biassi
- APQP Checklist Report.version 1.1.Dtd. 13.11.08Diunggah olehrobbie862
- Mixer Example Using ADSDiunggah olehNgô Văn Đức
- MGMT3101 International Business Strategy S22014Diunggah olehBonnieBao
- CFX-Intro 14.5 Appendix B Profile BCsDiunggah olehShaheen S. Ratnani
- Eda ExamplesDiunggah olehTamilinbaa
- The Solution CFD Analysis of Air Flow inside Oil TanksDiunggah olehMehul Patel
- An Enhanced Simulation Model for Building Envelopes With Phase Change MaterialsDiunggah olehRam Ramisetti
- 2009 - Automated Box–Jenkins Forecasting ModellingDiunggah olehRobson M Costa
- Reflection Justin KillianDiunggah olehAnar Huseynov
- MBA Retail ManagementDiunggah olehAsbagri
- agsys-scidirectDiunggah olehGil Resplandor
- Gas Well Production Optimization Using Dynamic Nodal AnalysisDiunggah olehTopiksarip05
- Comandos y Funciones de PromodelDiunggah olehCherezade Karine Saud Soto
- 56cf284f08ae85c823448358Diunggah olehStefan Tudorache
- VLSI Lab Mannual(1)Diunggah olehSadananda Krrish
- The Multifunctionality of Dreaming and the Oblivious AvatarDiunggah olehMajaMarsenic

## Lebih dari sekadar dokumen.

Temukan segala yang ditawarkan Scribd, termasuk buku dan buku audio dari penerbit-penerbit terkemuka.

Batalkan kapan saja.