Anda di halaman 1dari 26

Sulistyono,MetodeBedaHinggaSkemaEksplisit| 41

APLIKASIMETODEBEDAHINGGASKEMAEKSPLISITPADAPERSAMAAN
KONDUKSIPANAS
BambangAgusSulistyono
ProgramStudiPendidikanMatematikaFKIPUNPKediri
bb7agus1@gmail.com
Abstrak
Paper ini mengkaji bentuk numerik dari persamaan konduksi panas yang model
matematikanya berbentuk persamaan diferensial parsial tipe parabolik dengan
menggunakan metode beda hingga skema eksplisit. Dikaji pula stabilitas skema
tersebut dengan cara mengambil beberapa nilai . Dari kajian ini diperoleh bahwa
skema eksplisit akan memberikan hasil yang baik (stabil) bila cukup kecil atau

kondisihitunganakanstabilbilanilai0 .

Katakunci:pdpparabolik,metodebedahingga,skemaeksplisit,stabilitas
PENDAHULUAN
Persamaan diferensial adalah suatu persamaan yang mengandung fungsi dan
turunannya yang tidak diketahui. Jika terdapat satu variabel bebas dan turunannya
merupakan turunan biasa maka disebut dengan persamaan diferensial biasa, dan jika
terdapat dua atau lebih variabel bebas dan turunannya adalah turunan parsial maka
persamaannyadisebutdenganpersamaandiferensialparsial.Persamaandiferensialparsial
menurut nilai koefisiennya dibedakan atas tiga persamaan, yaitu persamaan parabolik,
persamaaneliptik,danpersamaanhiperbolik.
Kebanyakan permasalahan dalam ilmu pengetahuan dan teknologi dapat
dipresentasikan dalam bentuk persamaan diferensial parsial. Persamaan tersebut
memegangperananpentingdidalampenggambarankeadaanfisis,dimanabesaranbesaran
yangterlibatdidalamnyaberubahterhadapruangdanwaktu.
Persamaan parabolik biasanya merupakan persamaan yang tergantung pada waktu
(tidak permanen). Penyelesaian persamaan tersebut memerlukan kondisi awal dan batas.
Persamaan eliptik biasanya berhubungan dengan masalah keseimbangan atau kondisi
permanen (tidak tergantung waktu), dan penyelesaiannya memerlukan kondisi batas di
sekeliling daerah tinjauan. Persamaan hiperbola biasanya berhubungan dengan getaran,
atau permasalahan di mana terjadi ketidakkontinyuan dalam waktu, seperti gelombang
kejut yang terjadi ketidakkontinyuan dalam kecepatan, tekanan dan rapat massa.
Penyelesaiandaripersamaanhiperbolikmiripdenganpenyelesaianpersamaanparabola.
Dalam paper ini akan dikaji bentuk numerik dari persamaan konduksi panas yang
model matematikanya berbentuk persamaan diferensial parsial tipe parabolik dengan
menggunakan metode beda hingga skema eksplisit. Pada skema eksplisit, variabel
(temperatur)padasuatutitikdihitungsecaralangsungdaribeberapavariabeldibeberapa
titikdisekitarnyapadawaktusebelumnya,yangsudahdiketahuinilainya.Denganmetode
ini,penurunanpersamaandiferensialkedalambentukbedahinggaadalahmudah,namun
42 | JurnalMathEducatorNusantaraVolume01Nomor01,Mei2015

kendalautamanyaadalahkemungkinanterjadinyaketidakstabilanhitungan.Olehkarenaitu
dikajipulastabilitasskemaeksplisitdengancaramengambilbeberapanilaidari .
PERSAMAANKONDUKSIPANAS
Perhatikan tafsiran grafis secara intuisi aliran panas yang melalui medium padat
berikutini:


Gambar 1. Tafsiran secara intuisi aliran
panas
Panasmengalirdaribendabertemperaturlebihtinggikebendabertemperaturlebih
rendah. Laju perpindahan panas yang melewati benda padat sebanding dengan gradien
temperaturataubedatemperaturpersatuanpanjang.
L

A B

perambatan panas



Gambar 2. Perambatan panas pada batang
logam


Jikasebuahelemenbatanglogamyangpanjangdantipisdipanaskanpadasalahsatu
ujungnya,sedangujungyanglainadalahtetap,sepertiditunjukkandalamGambar2,maka
dapatditurunkansuatupersamaanuntukmemprediksipanas/temperaturyangmerambat
pada batang dari ujung A ke ujung B dalam suatu interval waktu tertentu ( ). Dengan
menggunakan prinsip keseimbangan inputoutput panas yang di serap, maka dapat
dituliskanpersamaan:

(1)

membagipersamaan(1)denganvolumedarielemen( )dan ,

denganmengambillimitnya,diperolehbentuk:
Sulistyono,MetodeBedaHinggaSkemaEksplisit| 43

dandarihukumkonduksipanasFourier,makadihasilkanpersamaankonduksipanassebagai
berikut:

dengan Tadalah temperatur, Kadalah koefisien konduktifitas, tadalah waktu dan x adalah
jarak(ruang).
Persamaan(4)berlakuuntukdaerah0 dan0 ,dengan adalahwaktu
hitungantotal,sedangkankondisiawaldanbatasadalah
,0 ;0
0, ;0 (5)
, ; 0
Dalam persamaan (5), , 0 adalah kondisi awal sedangkan dan adalah
kondisibatas.
METODEBEDAHINGGASKEMAEKSPLISIT
Penyelesaian persamaan tipe parabolik dengan menggunakan metode beda hingga
dapat dibedakan menjadi dua metode (skema) dasar, yaitu skema eksplisit dan skema
implisit.Padaskemaeksplisit,variabelpadawaktun 1,dihitungberdasarkanvariabelpada
waktunyangsudahdiketahui(gambar1).


n 1

n
penyelesaian diketahui
sampai waktu n
n 1


i 1 i i 1
Gambar 3. Skema eksplisit

DenganmenggunakanskemasepertiyangditunjukkanpadaGambar3,fungsivariabel
(temperatur)T(x,t)danturunannyadalamruangdanwaktudidekatiolehbentukberikut:

,
,


, 2


44 | JurnalMathEducatorNusantaraVolume01Nomor01,Mei2015

Denganmenggunakanskemadiatas,Persamaan(4)dapatditulisdalambentukberikut:


2 6

Penyelesaian Persamaan (4) dan (5) terhadap batang logam yang dipanaskan (AB)
dilakukan dengan membagi batang logam tersebut menjadi sejumlah pias. Selanjutnya
dibuat jaringan titik hitungan dalam bidang xt. Jarak antara titik hitungan (panjang pias)
adalah ,denganMadalahjumlahpiassedangintervalwaktuhitunganadalah .
Dengan Persamaan (6) dan kondisi batas di kedua ujung batang, memungkinkan untuk
menghitung (i 1, 2, ..., M1) berdasarkan nilai (i 1, 2, ..., M ) yang telah
diketahui.
Padaawalhitungan,nilaiawaldaritemperatur diketahuisebagaikondisiawal.Dari
nilaiawaltersebutdankondisibatas,dapatdihitungnilaiTdisepanjangbatanglogam(i 1,
2, ..., M ) pada waktu berikutnya. Nilai yang telah dihitung tersebut digunakan untuk
menghitungTi(i 1,2,...,M)untukwaktuberikutnyalagi.Prosedurhitunganinidiulangi
lagisampaiakhirnyadidapatnilaiTi(i 1,2,...,M)untuksemuanilaiwaktu.
HASILDANPEMBAHASAN
Penerapanmetodebedahinggaskemaeksplisitpadapersamaankonduksipanassatu
dimensi dapat dilakukan dengan memberikan sebuah contoh kasussebagai berikut,
diberikan sebuah batang logam yang pada kedua ujungnya dipertahankan temperaturnya
konstanyaitu0 .Akandicaripenyebarantemperaturdisepanjangbatanglogamdanuntuk
setiaplangkahwaktu.Secaramatematispermasalahantersebutdapatdigamabrkansebagai
penyelesaiannumerikdaripersamaan:


yangmemenuhikondisiawal
2 , 0 0.5
,0
2 1 , 0 1 ,
Dannkondisibatas
0, 1, 0 .
Dalam kasus ini dianggap bahwa K 1, sehingga bentuk persamaan beda hingga
skemaeksplisit(Persamaan6)menjadi:

2

atau
1 2 7
Sulistyono,MetodeBedaHinggaSkemaEksplisit| 45

dengan
Perhitungan dilakukan menggunakan Matlab terhadap Persamaan 7 dengan
mengambil beberapa keadaan, yaitu 0.1 dan 0.001, dan hasil plot untuk
0, 10 , 50 , 100 adalahsebagaiberikut:

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.5 1 0 0.5 1

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.5 1 0 0.5 1

Gambar 4 Solusi numerik dari persamaan konduksi

BerdasarkanhasildariplotdalamGambar4dengan 0.1menunjukkanbahwaperubahan
temperatur dari waktu ke waktu terjadi secara berangsurangsur. Hal ini sesuai dengan
kondisifisikdanmencerminkankondisiyangterjadidialam.
Selanjutnya dilakukan perhitungan menggunakan Matlab dengan mengambil
beberapa keadaan yang lain yaitu 0.1 dan 0.001, dan hasil plot untuk
, 2 , 3 , 4 adalahsebagaiberikut:
Berdasarkan hasil dari plot dalam Gambar 5 dengan 1 menunjukkan bahwa
perubahan temperatur dari waktu ke waktu tidak terjadi secara berangsurangsur.
Dibeberapatitikhitunganterjaditemperaturmelebihitemperaturawaldanjugaterjadinilai
negatif. Secara fisik perubahan semacam itu tidak benar dan hasil hitungan tidak
mencerminkan kondisi yang terjadi di alam. Hal ini disebabkan karena adanya
ketidakstabilanhitunganpadakondisitersebut.

46 | JurnalMathEducatorNusantaraVolume01Nomor01,Mei2015

1 10

5
0.5
0

0 -5
0 0.5 1 0 0.5 1

4
x 10
200 1

0.5
0
0
-200
-0.5

-400 -1
0 0.5 1 0 0.5 1

Gambar5Ketidakstabilansolusipersamaankonduksi
panasdenganskemaeksplisit.

Hasil diatas sesuai dengan teori yang mengatakan agar perhitungan dengan skema
eksplisit ini konvergen dan stabil, maka besarnya delta t harus diambil sedemikian rupa
sehinggamemenuhi


2
apabilatidakmemenuhimakahitunganmenjaditidakstabil.
SIMPULAN
Dari hasil penelitian ini diperoleh kenyataan bahwa distribusi temperatur pada
sebatanglogampanjangdantipismengikutifungsiparabolikdengantemperaturmaksimum
berada dibagian tengah dan temperatur berubah terhadap waktu. Dari kajian ini pula
diperolehbahwaskemaeksplisitakanmemberikanhasilyangbaik(stabil)bila cukupkecil

ataukondisihitunganakanstabilbilanilai0 .

DAFTARPUSTAKA
Triatmodjo,B,(2002)MetodeNumerikdilengkapidenganprogramkomputer,PenerbitBeta
Offset.
Paolo, B. (2006) Numerical Methods in Finance and Economics A MATLABBased
Introduction,JohnWiley&Sons,Inc.
Ibrahim,KI.danHisyam,A.(2003)MetodeNumerikuntukSainsdanTeknikdenganMatlab,
UADPress.
Study air pollutants
Case Study: Finite

Informationsteknologi
Difference Methods
Air Quality Modelling
(and the advection
diffusion equation)

We want to study how our environment is affected


from the pollutants from an industrial plant?

Institutionen fr informationsteknologi | www.it.uu.se

Study air pollutants Mathematical model

ut=Du- (vu)

Informationsteknologi

Informationsteknologi

Diffusion Transport

u=u(x,y,z,t) concentration of pollutant


v=[vx,vy,vz] wind vector (weather forecast)
D-Diffusion constant

Note: The model can only be solved


analytically in some special cases, for a
Or we want to track the radioactivity from general wind vector the only possibility is
a nuclear plant accident (Tjernobyl 1986). to use numerical methods.

Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

Numerical Method Numerical Method


FDM approach: Reduce and simplify the problem:
One dimension, u=u(x,t)
Informationsteknologi

Informationsteknologi

Discretize the physical domain Constant wind, v=v=const


Represent the solution in the discrete nodes Zero diffusion, D=0
Replace the derivatives with finite differences
Compute the solution in the nodes ut=-vux
How do approximate the derivatives and how Has the analytical solution u(x,t)=u0(x-vt)
do we compute the solution, need to consider: Where u0(x)=u(x,0) is initial data, i.e., the
Accuracy solution follows the characteristic lines x-vt=c
Stability
Efficiency
But also other problem specific properties
(e.g. transport or diffusion properties)
Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

1
Numerical Method Numerical Method
Upwind method, take the derivates from
left (from the direction of the wind):
Informationsteknologi

Informationsteknologi
(ujn+1-ujn)/dt=-vj(ujn-unj-1)/dx
t

x
The solution follows the characteristic lines,
x-vt=c, and moves unmodified to the right as
t increases (the wind blows to the right, v>0).
Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

Matlab implementation Matlab implementation


dx=0.1; dt=0.05;
Informationsteknologi

Informationsteknologi

x=-15:dx:15;
u=u0(x); % Function u0.m

% Upwind method
c=v*dt/dx;
for t=dt:dt:T
u(2:end)=u(2:end)-
c*(u(2:end)-u(1:end-1));
end

plot(x,u);
Upwind method
Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

Analysis Improved method


Leap-Frog, take centered derivates both in
Accuracy: Taylor expansion gives order (1,1),
the space and time directions:
Informationsteknologi

Informationsteknologi

i.e, first order in space and time.

Stability: Von Neumann analysis (Fourier (ujn+1-ujn-1)/2dt=-vj(unj+1-unj-1)/2dx


mehod) gives the stability condition vdt/dx1 t

Efficiency: Not an issue, 1D problem,


generous stability condition.

Problem properties: Upwind method gives un-


physical damping (|Q|<1 for all frequencies). x

Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

2
Matlab implementation Matlab implementation
dx=0.1; dt=0.05;
x=-15:dx:15;
Informationsteknologi

Informationsteknologi
u=u0(x); % Function u0.m
unew=u0(x-v*dt);

% Leap-Frog method
c=v*dt/dx;
for t=2*dt:dt:T
uold=u;u=unew;
unew(2:end-1)=uold(2:end-1)-
c*(u(3:end)-u(1:end-2));
end

plot(x,u); Leap-Frog method


Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

Matlab implementation Analysis


Accuracy: Taylor expansion gives order (2,2),
i.e, second order in space and time.
Informationsteknologi

Informationsteknologi

Stability: Von Neumann analysis (Fourier


mehod) gives the stability condition vdt/dx1

Efficiency: Not an issue, 1D problem, generous


stability condition.

Problem properties: Leap-Frog preserves


solution amplitude (|Q|=1 for all frequencies
without diffusion term) but gives some un-
Leap-Frog with dx=dx/4, dt=dt/4 physical oscillations.

Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

Crank-Nicolson Crank-Nicolson
Crank-Nicolson is implicit
But, we can not use this method for our
Informationsteknologi

Informationsteknologi

equation with diffusion term D0. ujn+1+vjdt/(4dx)(un+1j+1-un+1j-1)=


von Neumann analysis gives |Q|>1. ujn-vjdt/(4dx)(unj+1-unj-1)

t
Use an implicit non-damping scheme,
e.g., Crank-Nicolson:

(ujn+1-ujn)/dt=-vj(D0un+1+D0un)/2
With D0u=(uj+1-uj-1)/2dx
x
Gives similar solution as Leap-Frog Need to solve a tri-diagonal linear system of
equations in each time step, Aun+1=Bun.
Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

3
Matlab implementation Analysis
% Crank-Nicolsons method Accuracy: Taylor expansion gives order (2,2)
e=ones(n,1); c=v/(4*dx)*e;
Informationsteknologi

Informationsteknologi
Stability: Unconditionally stable
A=spdiags([-c e c],[-1 0 1],n,n);
B=spdiags([c e -c],[-1 0 1],n,n); Efficiency: Can take arbitrary dt but need to
[L,U,P]=lu(A); Bp=P*B; solve a tri-diagonal system in each step.
Problem properties: Crank-Nicolson preserves
for t=2*dt:dt:T solution amplitude (|Q|=1 for all frequencies)
z=L\(Bp*u); but gives some un-physical oscillations. Similar
u=U\z; solution as for Leap-Frog but with D0 we get
end correct damping.

Summary: Crank-Nicolson is accurate, stable,


efficient and preserves application properties.
Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

Mathematical model Mathematical model


Extend the model into two dimensions, D=0,
Extend the model with diffusion D=1 and and circular wind, v=[-y/R,x/R]. At t=0 a
varying wind: v=5x2/(1+x2) for x>0
Informationsteknologi

Informationsteknologi

concentrated cloud of pollutant is released.

Numerical solution
Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

Numerical solution Numerical solution


Informationsteknologi

Informationsteknologi

Upwind method, one sided differences Crank-Nicolson, centered differences


Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

4
Numerical solution Numerical solution
Increase wind, : v=10x[-y/R,x/R] Decrease space and time steps
Informationsteknologi

Informationsteknologi
Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

Numerical solution Summary


Finally, add diffusion D=0.1 Need to have control on accuracy (Taylor) and
stability (Fourier) => Convergence
Informationsteknologi

Informationsteknologi

For effiency:
Rule 1: Choose higher order methods if
accuracy is insufficient, more efficient than
increasing the number grid points.
Rule 2: Choose implicit methods for parabolic
problems, unconditionally stable (with correct
boundary conditions).

(For high performance, write in a more


efficient language and parallelize the code.)
Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

Summary Mini project 1


How do we distinguish between physical Can work in groups with up to 3 students
and numerical phenomena in the solution, Full report required, see instructions
Informationsteknologi

Informationsteknologi

e.g., non-physical damping and oscillations? Use INBOX #9, floor 2, house 2 (no email!)
How do we know that we have a correct Deadline October 2
solution?
Repeat the experiments with smaller and
smaller discretization steps. If the solution
converges the results are correct.
(Stability + Consistency => Convergence)
Convergence rate proportional to the order of
accuracy and Richardson extrapolation gives
an estimate of the global error.

Institutionen fr informationsteknologi | www.it.uu.se Institutionen fr informationsteknologi | www.it.uu.se

5
SCIENTIFIC PUBLICATIONS OF THE STATE UNIVERSITY OF NOVI PAZAR
S ER . A: A PPL . M ATH . I NFORM . AND M ECH . vol. 2, 1 (2010), 29-43.

Application of Finite Difference Method to Study of the


Phenomenon in the Theory of Thin Plates

C. B. Dolicanin, V. B. Nikolic, D. C. Dolicanin

Abstract: The term thin plate implied elastic body with cylindrical or prismatic shape of small
thickens in relation to other two dimensions. The basic dependencies between geometrical and
physical properties come to mostly to setting up of relations between stress and strain condi-
tions, which has been described by differential equations, simple and partial. Methods used for
solving of established equations, with respect of outline and initial conditions, may be classified
in analytical and numerical. In case of complex and big construction systems subjected to the
arbitrary loads, including a complex boundary conditions, solving of differential equations by
analytical methods is almost impossible. Then the solution is application of numerical methods.
One of the basic numerical methods is Finite Difference Method (FDM) based on replacing of
differential equations with corresponding difference equations. Using of this method, the prob-
lem come to solving of system of paired algebraic equations, making the problem more easier
for solving. In the end, more comments and farther directions of investigations are given.
Keywords: Numerical analysis, thin plate, finite difference method, strain

1 Introduction

Many problems in the field of deformable bodies are described by parametars of the con-
dition of the systems which depend on coordinates, time, temperature, ect. Such systems
are given in mathematical models by partial differential equations, for example, relation
between stress condition and strain condition and external load in mechanics of continuum
and others for solving these equations there are mainly, 1, 2, 3, 4, 5, 6, 7, 10 , two ap-
proaches: analytical and numerical. Analytical methods are of an earlier date and imply
determination of mathematical functions which define solution in a closed form. The basic
characteristic of numerical methods is that the fundamental equations which describe the
problem, including the boundary conditions are soved in approximate numerical way. In
this paper, Finite Difference Method (FDM) and the finite element method (FEM) are used
for analysis of stress strain condition in the thin plates. Finite difference method (Diferencne
method-FDM) is a numerical method based on mathematical discretization of differential
Manuscript received December 15 2009; accepted May 24, 2010.
C. B. Dolicanin, V. B. Nikolic, D. C. Dolicanin are with the State University of Novi Pazar, Serbia.

29
30 C. B. Dolicanin, V. B. Nikolic, D. C. Dolicanin

equations, which are the finite difference equations are translated into algebraic. The fi-
nite element method (FEM) implies discretization in physical model and the final result is
obtained by solving a system of algebraic equations. FEM uses various forms of variation
methods applied to discrete model. In this paper uses the finite difference method for the
determination of strain state analysis of thin plates.

2 Review of finite difference method

Finite difference method is numerical methods based on mathematical discretization of the


equations of boundary problems. By using this method, continuous process is studied in a
finite number of sufficiently small time intervals. So it is possible, in these small intervals,
the function approximated by approximate expressions. In each elementary interval is the
integration, with the results of integration in the previous interval are taken as initial for the
next time interval.
The function f (x) is continuous in the interval AB, and some are also continuing her
running f (i) (x) to that order, i = 1, 2, 3, ... that we need. Interval AB, where A (a, 0) and
B(b, 0), divide into equal divisions x = h, Fig. 1. Then the first derivative of f at x ,
[1, 3, 7, 10] , is given by the expression

f = f |x /2h; F|x = ( f (x + h) f (x h))/2 (1)

where f is the first central difference.


We can say that the first derivative is the slope of a graphical point of function f (x)
calculated based on the value of the function in neighboring points.
Approximate value of the first statement of f (x) at intervals from x to x + h, ie the
part of x h to x is obtained by dividing the difference with the corresponding first step
interpolation h. The second derivative is

d2 f 2 f f (x + h) 2 f (x) + f (x h)
f = 2
= 2
= (2)
dx h h2
Analogue, can be determined, third, fourth and differences n-th row and the corresponding
running.

Fig. 1. The method of finite differences


Application of Finite Difference Method to Study of the Phenomenon ... 31

If we replace the corresponding statements in the differential equations that we study,


will get to a certain point (x) space (a, b) a linear equation. The total number of equations
equals the number of pre-selected points nodesof a given space. To arrive at a solution,
these equations should be added to the initial and boundary condition.
Border procedures, when the distance between the points fall to zero, gives the same
threshold in the respective partial derivatives of functions given point.
In the case of two-dimensional problem that describes the partial differential equations
in (x, y), .ie. in the case of f (x, y) that depends on two variables require the same in most
cases approximated by applying the finite difference method, to solve the problem using a
computer, [1, 2, 3, 7, 9, 10]. To calculate the partial statement, suppose that the known values
of the function fu network nodal points of the budget, etc. Fig. 2.

Fig. 2. Networks for solving two-dimensional problem

In the point i(x, y) observed rectangular grid with power ranges x = h (on the x axis)
and y = l (on the y axis), the first partial derivatives, analogous to the previous procedure,
are
f x f f1 f3 1
= = = ( f (x + h, y) f (x h, y))
x 2h 2h 2h
f y f 1
= = ( f (x, y + l) f (x, y l)) (3)
y 2l 2l
where are x f and y f the first partial derivatives for x and y directions. Analogously, the
partial derivatives of second, third and fourth order we have
( )
2 f 2x f f1 2 fi + f3 1
= = = = 2 ( f (x + h, y) 2 f (x, y) + f (x h, y)) ;
x 2 x x h 2 h 2 h
( )
2 f y f
2
f2 2 fi + f4 1
= = 2 = = 2 ( f (x, y + l) 2 f (x, y) + f (x, y l)) ; (4)
y 2 y y l l 2 l
( )
3 f 2 f 3x f 1
= = = 3 ( f (x + 2h, y) 2 f (x + h, y) + 2 f (x h, y) f (x 2h, y)) ;
x 3 x x 2 2h 3 2h
( 2 )
f
3 f y f
3
1
= = 3 = 3 ( f (x, y + 2l) 2 f (x, y + l) + 2 f (x, y l) f (x, y 2 l)) ;
y 3 y y 2 2l 2l
32 C. B. Dolicanin, V. B. Nikolic, D. C. Dolicanin

( )
4 f 3 f 4x f f9 4 f1 + 6 fi 4 f3 + f11
= = = =
x4 x x 3 h 4 h4
1
= ( f (x + 2h, y) 4 f (x + h, y) + 6 f (x, y) 4 f (x h, y) + f (x 2h, y))
h4
( )
4 f 3 f 4y f f10 4 f2 + 6 fi 4 f4 + f12
= = = =
y4 y y 4 l 4 l4
1
= ( f (x, y + 2l) 4 f (x, y + l) + 6 f (x, y) 4 f (x, y l) + f (x, y 2l))
l4
In a similar way can be derived in terms
( ) ( )
2 f f 1 f f
= = (5)
x y y x 2l x)2 x)4

Substituting these expressions into the corresponding partial differential equations of the
problem at the same transform the finite difference equation or difference equation . Also,
the total number of equations equals the number of pre-selected points or nodes of a given
space on the x and y axes.

3 Examples of solution for strain in thin plates

Under a thin plate in the theory of elasticity of the elastic body includes a cylindrical or
prismatic shape thickness h in relation to the other two dimensions, a and b, [3, 6, 7, 10] .
Plate thickness should be less than 1/10 of other dimensions. Medium flat panel takes up
the 0xy plane, and axis 0z is directed downward in the direction of the load which the unit
of surface is p, N/mm2 , Fig. 3 and Fig. 4.

Fig. 3. Bending of thin plates.

Due to the small size h plate can be considered as points of attack power loads in the
very high level. Due to loads p perform deformation plate. It is assumed that the deflections
are very small, so that the upper limit of the size of the strain should not exceed the size of
h/5. Some high points of the plane deflections w = w0 = w(x, y) and functions are just the
coordinates x and y. The mean plane by deformation exceeds the elastic surface.
In the plate theory is introduced hypotheses linear element as follows: normal to the
middle plane remains the administrative and the elastic surface, which means that the linear
elements remain unchanged by deformation length. The plate theory assumes that the z =
Application of Finite Difference Method to Study of the Phenomenon ... 33

0, and the mean plane of normal stress and strain in the flat, zero, neutral surface (plane). It
is further assumed that no body forces act.
M'(x)=M'yx
O x O x
dx dy
z tx y M'(y)=M'xy p Z'x
ty z y tx z
sx M'xy
dz dz
y tyx

sy z M'yx Z'y
1 1
dz
z

Fig. 4. Force reduction.

To perform differential equations of equilibrium, ie. differential equation of the elastic


surface plate, it is enough to observe the element dxdydz separated from the plate. In this
isolated element acting bending moments Mx and Mx , or My and My , moments torsion Ttxy
or, T
and Ttxy
tyx and Ttyx , transversal forces Qx and Qx , Qy and Qy and the given external
load, etc.
By placing the link between stress and moments, stress and strain, we get the equations
of equilibrium in the form of

4 4 4 p
= + 2 + = (6)
x 4 x y
2 2 y 4 D
Eh 3
where is D = 12(1 3)
N/mm2 plate bending stiffness or cylindrical plate bending stiffness
(flexural rigidity of the plate). It is analogous savojnoj stiffness beams. Finite difference
form, of difference equation (6), [3, 6, 7, 10], according to the scheme on Fig. 2, is:
p 4
= 20i 8 (1 + 2 + 3 + 4 ) + 2 (5 + 6 + 7 + 8 ) + (9 + 10 + 11 + 12 ) = h (7)
D
When the known dimensions and known when the load p(x, y) rectangular plate, it should
solve the partial elastic surface equation (6), and the system of difference equations (7). The
solution is required to satisfy boundary (boundary) conditions on the side panels parallel
basis 0x and 0y axes. There are four types of suspension lateral basis (side) panels: the flex
side (base), simply supported (supported) pages, free pages and elastically supported and
elastically flex side [3, 6, 7, 10] .
In the paper given, apply finite difference method, three examples of solutions for the
deflection plate under a given load .
Example 1: Rectangular plate flex sides AB and CD, simply supported by the ad and
free side BC is loaded as in Figure 5. It is known: F = 1kN ; a = 4m; b = 6m; = 0.1m;
= 0, 3; E = 2.1105 MPa.
By Finite difference method we will be certain values of strain at nodal points of the
adopted network.
34 C. B. Dolicanin, V. B. Nikolic, D. C. Dolicanin

For the flex side of the deflection equals zero, the free, supported by foreign side of the
deflections and bending moments are equal to zero and on the free side, there is no bending
or twisting, and transversal forces do not act.

Fig. 5. A rectangular plate is loaded in the concentrated force of gravity .

Deflection equation is given by (6) and Diferencne form expression (7). If you introduce
tags
l = b/4 = 6/4 = 1.5m; h = a/4 = 5/5 = 1m;

l2 9 F 1000 2000 N
= = = 2.25; pi = = = = 666.667 2
h2 4 h l 1 1.5 3 m
E 3 2.1 1011 0.13
D = = = 19.23077 106 Nm
12 (1 2 ) 12 (1 0.32 )
obtained by members of equations (7) in the form of

6 2 + 8 + 6 = 54.375; 4 ( + 1) = 29.25; 4 ( + 1) = 13; 2 = 4.5; 2 = 5.0625

Concentrated force F, pressure is defined as p = F/hl and a force that increases the demand
node F = phl.
Taking into account the symmetry of the plate relative to the y-axis and the boundary
conditions of reliance, there is a system of algebraic equations (according to the scheme in
Fig. 2):
node 1:

54.3751 29.25 (3 + 5 ) 13 (2 + 4 ) + 4.5 (6 + 7 + 8 + 9 ) +


+ 5.0625 (0 + 11 ) + 0 + 0 = 175.499
Application of Finite Difference Method to Study of the Phenomenon ... 35

node 3:
54.3753 29.25 (0 + 1 ) 13 (7 + 8 ) + 4.5 (0 + 4 + 2 + 0) +
+ 5.0625 (5 3 ) + 0 + 0 = 0

node 5:
54.3755 29.25 (1 + 11 ) 13 (6 + 9 ) + 4.5 (4 + 12 + 10 + 2 ) +
+ 5.0625 (3 + 15 ) + 0 + 0 = 0

node 11:
54.37511 29.25 (5 + 15 ) 13 (10 + 12 ) + 4.5 (9 + 16 + 14 + 6 ) +
+ 5.0625 (1 + 19 ) + 0 + 0 = 0

node 7:
54.3757 29.25 (0 + 2 ) 13 (0 + 3 ) + 4.5 (0 + 1 + 0 + 0) +
+ 5.0625 (6 7 ) + 7 + 8 = 0

node 2:
54.3752 29.25 (7 + 6 ) 13 (1 + 0) + 4.5 (3 + 5 + 0 + 0) +
+ 5.0625 (0 + 10 ) + 4 + 2 = 0

node 6:
54.3756 29.25 (2 + 10 ) 13 (5 + 0) + 4.5 (1 + 11 + 0 + 0) +
+ 5.0625 (7 + 14 ) + 9 + 6 = 0

node 10:
54.37510 29.25 (6 + 14 ) 13 (11 + 0) + 4.5 (5 + 15 + 13 + 0) +
+ 5.0625 (2 + 18 ) + 10 + 12 = 0

Helpfully sizes:
( )

= 0.3
2.25 = 0.1333 4 1 + 2 = 7.022

( ) ( )
2 1 + = 2.267 22
2
1
4 = 2.18074

( ) 2
4 1 + 2 + 23 22 = 8.16
2
2 = 1.5111

2 2
= 0.10074
2
36 C. B. Dolicanin, V. B. Nikolic, D. C. Dolicanin

1 line: 13 = 2.267 0 0 0.1333 (10 + 10 )


14 = 2.267 10 6 0.1333 (11 + 0) ;
15 = 2.267 11 5 0.1333 (10 + 12 )

2 line: 18 = 8.16 10 7.0226 2.18074 (11 + 0) + 1.511 (5 + 0) +


+ 2 + 0.10074 (10 + 12 )
19 = 8.16 11 7.0225 2.18074 (10 + 12 ) + 1.511 (9 + 6 ) +
+ 1 + 0.10074 (0 + 0)

After ordering we get a system of equations whose solving in Mathcad using the Find
function Given the values of deflection in the requested budget network nodes:
1 line
given
-6
54.375w 1 - 26w 2 - 29.25w 3 - 29.25w 5 + 9w 6 + 9w 7 + 0w 10 + 5.0625w 11 175.499 10

- 29.25 w 1 + 9w 2 + 49.3125w 3 + 5.0625w 5 + 0w 6 - 26w 7 + 0w 10 + 0w 11 0

- 29.25 w 1 + 9w 2 + 5.0625w 3 - 49.3125w 5 - 26w 6 - 0w 7 + 7.6503w 10 - 17.773w 11 0

10.125w 1 + 0w 2 + 0w 3 - 35.5489w 5 + 15.299w 6 + 0w 7 - 19.8799w 10 - 28.1755w 11 0

4.5w 1 - 29.25w 2 - 13w 3 + 0w 5 + 5.0625w 6 + 51.31252w 7 + 0w 10 + 0w 11 0

- 13w 1 + 56.375w 2 + 4.5w 3 + 4.5w 5 - 29.25w 6 - 29.25w 7 + 5.0625w 10 + 0w 11 0

4.5w 1 - 29.25w 2 + 0w 3 - 13w 5 + 51.3125w 6 + 5.0625w 7 - 17.7733w 10 + 3.825w 11 0

0w 1 + 10.125w 2 + 0w 3 + 7.6499w 5 - 35.5489w 6 + 0w 7 + 29.9958w 10 - 11.07297 w 11 0

-6
3.3782068488235194989 10
-6
1.0332885227400467117 10
-6
2.5752521219211802998 10
-6
-2.4220254509186598866 10
( )
find w 1 , w 2 , w 3 , w 5 , w 6 , w 7 , w 10 , w 11
-2.5056705797307480518 10-7

9.6991010744365241995 10-7
-6
1.1884074553557873549 10
-6
3.2952685171458499366 10
Application of Finite Difference Method to Study of the Phenomenon ... 37

The values of deflection can be considered relevant, because this method, previously,
tested on many examples.
Example 2. Rectangular plate flex sides AB and CD, simply supported sides AD and
BC is loaded as in Figure 6 continuous load q. It is known: E = 2.1105 MPa, = 0.3;
= 0.1m, q = 2kN/m2 , a = 4m, b = 6m.
With Finite difference method will be determine the values of deflection at nodal points
of the adopted network.

Fig. 6. A rectangular plate partially loaded continuous load.

Note: Analogous to the previous example, the case when the surface of the panel op-
erates continuously load q, a node that is located inside the area is burdened by power qlh,
and for node 0, Figure 7 a) p = q. In other words, node 0 receives a continuous burden on
themselves by acting on the surface of the rectangle width height h and l, where it is located
in the center of the rectangle.In contrast nodes 1, 2, 3 and 4 received half so defined load,
Figure 7 b), and nodes 5, 6, 7 and 8, only one quarter, Figure 7 c). So the pressure values

a) b) c)
Fig. 7. Scheme for determining force in the nodal point

in the corresponding nodes will be:

q q
p0 = q, p1 = p2 = p3 = p4 = , p5 = p6 = p7 = p8 =
2 4
38 C. B. Dolicanin, V. B. Nikolic, D. C. Dolicanin

4 4 4 p
+ 2 + =
x 4 x y
2 2 y 4 D

Applying MKR is obtained by the equilibrium equations of the form:

( )
6 2 + 8 + 6 i 4 ( + 1) (1 + 3 ) 4 ( + 1) (2 + 4 ) +
pi l 4
+ 2 (5 + 6 + 7 + 8 ) + 2 (9 + 11 ) + 10 + 12 =
D

l = b/4 = 6/4 = 1.5m; h = a/4 = 4/4 = 1m

l2 9 q kN N
= 2
= = 2.25 p1 = p2 = = 1 2 p3 = p5 = q = 2 2
h 4 2 m m
E 3 2.1 1011 0.13
D = = = 19.23077 106 Nm
12 (1 2 ) 12 (1 0.32 )

6 2 + 8 + 6 = 54.375; 4 ( + 1) = 29.25; 4 ( + 1) = 13; 2 = 4.5 2 = 5.0625

node 1: 54.3751 29.25 (2 + 2 ) 13 (3 + 4 ) + 4.5 (5 + 5 + 6 + 6 ) +


+ 5.0625 (0 + 0) + 0 + 0 = 0.26325 103

node 2: 54.3752 29.25 (0 + 1 ) 13 (5 + 6 ) + 4.5 (0 + 0 + 3 + 4 ) +


+ 5.0625 (2 2 ) + 0 + 0 = 0.26325 103

node 3: 54.3753 29.25 (5 + 5 ) 13 (0 + 1 ) + 4.5 (0 + 0 + 2 + 2 ) +


+ 5.0625 (0 + 0) + 3 + 4 = 0.5625 103

node 4: 54.3754 29.25 (6 + 6 ) 13 (0 + 1 ) + 4.5 (0 + 0 + 2 + 2 ) +


+ 5.0625 (0 + 0) + 4 + 3 = 0

node 5: 54.3755 29.25 (0 + 3 ) 13 (0 + 2 ) + 4.5 (0 + 0 + 0 + 1 ) +


+ 5.0625 (5 5 ) + 5 + 6 = 0.5625 103

node 6: 54.3756 29.25 (0 + 4 ) 13 (0 + 2 ) + 4.5 (0 + 0 + 0 + 1 ) +


+ 5.0625 (6 + 6 ) + 6 + 5 = 0
Application of Finite Difference Method to Study of the Phenomenon ... 39

given

54.375w1 - 58.5w2 - 13w 3 - 13w4 + 9w 5 + 9w6 0.26325 10- 3

-29.25w1 + 54.375w 2 + 4.5w 3 + 4.5w4 - 13w5 - 13w 6 0.26325 10- 3

-13w 1 + 9w2 + 55.375w 3 + w4 - 58.5w5 + 0w 6 0.5625 10- 3

-13w 1 + 9w2 + w 3 + 55.375w4 + 0w 5 - 58.5w 6 0

4.5w1 - 13w2 - 29.25w 3 + 0w4 + 55.375w 5 + w6 0.5265 10- 3

4.5w1 - 13w2 + 0w 3 - 29.25w4 + w 5 + 55.375w6 0

-5
8.4135943688372158218 *10
-5
6.0658239523563842244 *10
-5
8.3348495453246540610 *10
find *
* 1,*w* 2,*w* 3, w
find (w 1 2 3
** 4, w
4
** 5, w
5 6
*
** )6 *
-5
3.4062732779655031070 *10
-5
6.0498163354981516936 *10
-5
2.4303063434015049735 *10

Example 3. Rectangular plate simply supported on all its four sides is loaded as in Figure 8
triangular continuous load q. Apply Finite differences Method to find the values of deflec-
tion at nodal points of the adopted network. It is known: q= 4kN/m2 ; a = 6m; b = 8m; =
0.1m; = 0,3; E= 2105 MPa. The governing equations

4 4 4 p
+ 2 + =
x 4 x y
2 2 y 4 D

( 2 )
6 + 8 + 6 i 4 ( + 1) (1 + 3 ) 4 ( + 1) (2 + 4 ) +
pi l 4
+ 2 (5 + 6 + 7 + 8 ) + 2 (9 + 11 ) + 10 + 12 =
D
40 C. B. Dolicanin, V. B. Nikolic, D. C. Dolicanin

Fig. 8. A rectangular plate with a continuous load.

Under the given conditions, we get

l = b/4 = 8/4 = 2m; h = a/4 = 6/4 = 3/2m;


l 2 16
= 2
=
h 9
E 3 2 1011 0.13
D = = = 18.315 106 Nm
12 (1 2 ) 12 (1 0.32 )

Pressure values in the corresponding nodes:


q : q = 4h : h q = q / 4 q4 = q6 = q / 4
q : q = 4h : 2h q = q / 2 q1 = q3 = q / 2
q : q = 4h : 3h q = 3q / 4 q2 = q5 = 3q / 4

3174
6a 2 + 8a + 6 =
81
1600
4a (a + 1) =
81
100
4 (a + 1) =
9
32
2a =
9
2 256
a =
81
Application of Finite Difference Method to Study of the Phenomenon ... 41

Starting from the equation in the form (7):


( 2 )
6 + 8 + 6 i 4 ( + 1) (1 + 3 ) 4 ( + 1) (2 + 4 ) +
pi l 4
+ 2 (5 + 6 + 7 + 8 ) + 2 (9 + 11 ) + 10 + 12 =
D
Can be written the equation for individual nodes:
node 1: 31741 1600 (2 + 4 ) 900 (3 + 3 ) + 288 (5 + 5 + 6 + 6 ) +
4
+ 256 (0 + 0) + 0 + 0 = 81 pD1 l

node 2: 31742 1600 (0 + 4 ) 900 (5 + 5 ) + 288 (0 + 3 + 3 + 0) +


4
+ 256 (2 + 4 ) + 0 + 0 = 81 pD2 l

node 3: 31743 1600 (5 + 6 ) 900 (0 + 1 ) + 288 (5 + 0 + 6 + 1 ) +


4
+ 256 (0 + 0) 3 + 3 = 81 pD3 l

node 4: 31744 1600 (1 + 0) 900 (6 + 6 ) + 288 (3 + 0 + 0 + 3 ) +


4
+ 256 (2 4 ) + 0 + 0 = 81 pD4 l

node 5: 31745 1600 (0 + 3 ) 900 (2 + 0 ) + 288 (0 + 1 + 0 + 0) +


4
+ 256 (5 + 6 ) + 5 5 = 81 pD5 l

node 6: 31746 1600 (3 + 0) 900 (4 + 0 ) + 288 (1 + 0 + 0 + 0) +


4
+ 256 (5 6 ) + 6 6 = 81 pD6 l
After ordering we get a system of equations whose solving in Mathcad using the Find
function Given the values of the stress functions in the requested budget network nodes:
given

3174w1 - 1600w2 - 1800w3 - 1600w4 + 576w5 + 576w6 0.14152

-1600w1 + 2918w2 + 576w3 + 256w4 - 1800w5 0.21228

-900w1 + 288w2 + 3174w3 + 288w4 - 1600w5 - 1600w6 0.14152

-1600w1 + 256w 2 + 576w3 + 2918w4 - 1800w6 0.07076

288w1 - 900w2 - 1600w3 + 2918w5 + 256w6 0.21228

288w1 - 1600w3 - 900w4 + 256w5 + 2918w6 0.07076


42 C. B. Dolicanin, V. B. Nikolic, D. C. Dolicanin

9.304734180966319352210-4

7.2847116725436574292 10-4

-4
6.878809912749644516310
(
find w1 , w2 , w3 , w4 , w5, w6 )
-4
6.129508501957053538110

5.4320903651006013122 10-4

-4
4.509895453933079335510

4 Conclusion

Summarizing the results presented in this paper, the conclusion that the finite difference
metod (MCD) can effectively solve the problems of bending of thin plates and bulges and
at various loads. The developed methodology makes it much easier to find solutions for the
plate deflection, moments, stresses, strains, etc., which provides a significant advantage over
conventional analytical methods. Further study may include temperature strains, dynamic
load and the like By using this method can be dealt with more complex problems, [5, 8, 11,
12], what will be the subject of the future,

References

[1] PARVIZ M OIN, Engineering Numerical Analysis, Cambridge, University press, 2001.
[2] C D OLI CANIN , S OLOVLJEV P ETROVICH, Differential geometry, ETF in Pristina, Pristina,
1993.
[3] V. N IKOLI C , C. D OLI CANIN , M. R ADOJKOVIC, Aplication of Numerical Methods in Solvong
a Phenomenon of the Theory of Thin Plates, Scientific Technical Review, Vol. 60, 1 (2010),
61 - 65.
[4] C. D OLI CANIN, Mathematics II, University textbook, Pristina, 1995.
[5] K. M AKSIMOVI C , V.NI KOLI C , S.M AKSIMOVI C, Efficient computation method in fatique
life estimation of damaged structural components, Facta Universitatis, Series Mechanics, Au-
tomatic control and Robotics, Vol. 4, 16 (2004), 101-114.
[6] D.R AKOVI C, The Theory of Elasticity, Beograd, 1985.
[7] S. T IMO SENKO , J. N. D UDIER, The Theory of Elasticity, 1962.
[8] K. M AKSIMOVI C , V. N IKOLI C , S. M AKSIMOVI C, Modeling of the surface cracks and fatique
life estimation, ECF 16 2006, Minisymposium integrity of dinamical systems, Alexandroupo-
lis, JULY 3-7, 2006, 113-116.
Application of Finite Difference Method to Study of the Phenomenon ... 43

[9] K. (S TEVANOVIC ) H EDRIH , V. N IKOLIC -S TANOJEVI C, A model of gear trans-


mission: fractional order system dynamics, (Research Article), Mathematical
Problems in Engineering, (http://www.hindawi.com/journals/mpe/aip.972873.html),
http://www.hindawi.com/journals/mpe/aip.6.html
[10] V. N IKOLI C, Mechanical analysis of gears, Monograph, Cimpes, Kragujevac, 1999.
[11] S. M AKSIMOVI C , S. P OSAVLJAK , K. M AKSIMOVI C , V. N IKOLI C - S TANO -
JEVI C , V.D JURKOVI C, Total fatigue life estimation of notched struc-
tural components using low cycle fatigue properties, Journal STRAIN,
http://onlinelibrary.wiley.com.proxy.kobson.nb.rs:2048/doi/10.1111/j.1475-
1305.2010.00775.x/full
[12] D. S TAMENKOVI C , S. M AKSIMOVI C , K. M AKSIMOVI C , V. N IKOLI C -S TANOJEVI C , S.
M AKSIMOVI C , S. S TUPAR , I. VASOVI C, Fatigue Life Estimation of Notched Structural
Components, Strojniki vestnik - Journal of Mechanical Engineering, 56, 12 (2010), 846-852.

Anda mungkin juga menyukai