1. EXAMPLES AND DEFINITIONS 1.1 Differential equations in civil engineering 1.2 Notation 1.3 Classification 1.3.

1 Ordinary vs partial differential equations 1.3.2 Order and degree 1.3.3 Linear and non-linear equations 1.4 Some canonical differential equations 1.5 Boundary conditions 1.6 Reduction of ODEs to first-order equations 1.6.1 Coupled equations 1.6.2 Higher-order ODEs 1.7 Solution as an integration problem

SPRING 2005

1.1 Differential Equations in Civil Engineering Many physical laws can be expressed as differential equations. In engineering, differential equations often describe the continuous variation of some physical quantity with respect to time or position. Field equations often arise from a conservation equation (e.g. mass, momentum, energy) and a constitutive law (e.g. viscous stresses in fluids, heat diffusion, or Darcys law in porous media). Many are partial differential equations of order 2. Examples mechanics (linear and rotational motion): d2 x d2 F =m 2 , T =I 2 dt dt oscillations (damped mass-spring system): mx + cx + kx = 0 structures (beam theory): d4 y EI 4 = w dx hydraulics (gradually-varied flow): dh S 0 S f = dx 1 Fr 2 heat diffusion in a solid: T ( cT ) = (k ) x x t

Computational Mechanics

1-1

David Apsley

potential flow, steady-state diffusion, electrostatic potential, : 2 2 =0 + x 2 y 2 fluid mechanics; mass (continuity) and momentum (Navier-Stokes) equations: p u v + = 0, ( u ) + ( u 2 ) + ( uv) = + 2 u x y t x y x

1.2 Notation
dy d 2 y d 3 y Spatial derivatives , , , are often denoted by y, y, y, dx dx 2 dx 3 dy d 2 y d 3 y Time derivatives , , , are often denoted by y , y , y , dt dt 2 dt 3 2 f f f Partial derivatives e.g. , and are often denoted by fx, fy, fxy etc. x y xy

The Laplacian operator 2 is the sum of second partial derivatives: 2 2 2 2 + 2 (2 dimensions) y x

2 2 2 + + x 2 y 2 z 2

(3 dimensions)

1.3 Classification 1.3.1 Ordinary vs Partial Differential Equations An ordinary differential equation (ODE) contains derivatives with respect to only 1 independent variable; e.g. dy = 3x + y independent variable x; dependent variable y dx
d2 x dx +2 2 dt dt
2

+ 6 x = e t

independent variable t; dependent variable x

A partial differential equation (PDE) contains derivatives with respect to 2 or more independent variables; e.g. F F +3 = 3xy independent variables x, y; dependent variable F x y
2 = x 2 t

independent variables t, x; dependent variable

Computational Mechanics

1-2

David Apsley

1.3.2 Order and Degree The order of a differential equation is the highest derivative present (after all derivatives of products have been expanded). Examples: df + 3 f = 3t 2 dt
d 2u du +3 +u =t 2 dt dt 2u 2 u 2 u u + 2 + 2+ + u = 3x 2 y 2 x xy y y d dy ( 4 x ) = 2e x dx dx
4

(order 1) (order 2)

(order 2) (order 2 - note)

The degree of a differential equation is the power to which the highest derivative is raised (after any square roots etc. have been multiplied out). Examples:
dy dx
2

d2 y + y 3 = 4x 2 dx

(degree 1) (degree 3 - note)

d2 y + y =3 x+ y dx 2
1.3.3 Linear and Non-Linear ODEs

The general linear ODE is of form dn y d n 1 y dy a n ( x) n + a n 1 ( x) n 1 + ... + a1 ( x) + a0 ( x) y = f ( x) dx dx dx where the ai and f may be functions of x, but not y. All linear equations are of degree 1, but the converse is not necessarily true. If the RHS of the above equation is identically zero the equation is called homogeneous; otherwise it is inhomogeneous. In general, non-linear equations are more difficult to solve and may have multiple solutions. For example, dy = y, y ( 0) = 0 dx 2 has two solutions: y = 0 and y = 1 4 x .

Computational Mechanics

1-3

David Apsley

1.4 Some Canonical Differential Equations

Ordinary differential equations: exponential decay: dy y = y0e t = y dt simple harmonic motion (SHM): d2 y = 2y y = A sin( t + B) 2 dt

or

C cos( t ) + D sin( t )

Partial differential equations: Laplace equation: Poisson equation: wave equation: diffusion equation: 2 2 =0 + x 2 y 2 (elliptic)

2 2 = f ( x, y ) (elliptic) + x 2 y 2
2 = x 2 = t

1 2 c 2 t 2 2 x 2

(hyperbolic) (parabolic)

(The terms elliptic, hyperbolic and parabolic will be defined later in the course).

1.5 Boundary Conditions

An nth-order linear ODE requires n additional pieces of information (boundary conditions) to fix the particular solution. Example: d2 y + 4 y = 4 sin 2t dt 2 Boundary conditions y(0) = 0, y(0) = 1 give a solution y = t cos 2t . Boundary conditions y(0) = 0, y(0) = 1 give a solution y = t cos 2t sin 2t .

All first-order ODEs with one dependent variable can be treated as initial-value problems. They have the form dy = f ( x, y ) , y ( x0 ) = y 0 dx Second-order ODEs may have information fixed at both ends of an interval these are called boundary-value problems. Boundary conditions for PDEs depend on the type of problem: elliptic, hyperbolic, parabolic see later. Boundary conditions may be given on a line (2d) or a surface (3d).

Computational Mechanics

1-4

David Apsley

1.6 Reduction of ODEs to First-Order Equations 1.6.1 Coupled Equations

Coupled first-order ODEs can be rearranged as a single differential equation for a vector dependent variable. Example: The coupled differential equations dy + 3 z = 3t dt dz + y = 2t 2 dt can be rewritten as: 3t 3z d y = 2t 2 y dt z or Y y dY where =F , Y 1 Y2 z dt

F=

3t 3Y2 2t 2 Y1

Important technicality: from a computer-programming point of view, the dependent variable Y simply becomes a vector (i.e. one-dimensional array) instead of a scalar.

1.6.2 Higher-Order ODEs An nth-order ODE can be reduced formally to a first-order ODE by treating the 1st to (n-1)th derivatives as dependent variables. Example: d3 y d2 y dy 6 + + 15 + 2 y = 0 3 2 dx dx dx dy d2 y Write Y1 = y , Y2 = , Y3 = 2 . Then the equation becomes dx dx dy y dx d dy d2 y = dx dx dx 2 2 2 d y d y dy 6 2 15 2 y 2 dx dx dx or
Y1 Y2 d Y2 = Y3 dx Y3 6Y3 15Y2 2Y1 which is of the form

Computational Mechanics

1-5

David Apsley

dY = F( x, Y ) dx

Thus, (i) a higher-order equation can be rearranged as a first-order equation with vector dependent variable; but (ii) this is not helpful if the boundary conditions are specified in different locations; e.g. in the above example with, say, y(0) = 0, y(0) = 3 but y(1) = 3. Thus, this rearrangement only helps for initial-value problems.

1.7 Solution as an Integration Problem

y y 1 y 0 x0 x1 x2 x3 x4 x y2 y3 y4

Formally, the initial-value problem (1) can be integrated between successive x values as yi +1 = y i +
x I +1

f ( x, y ( x)) dx
xi

(2)

Thus, the solution process may be regarded as a series of steps: as x progresses x0 x1 x2 x3 ... , then y is incremented y 0 y1 y 2 y 3 ... , with each increment in y being equal to the integral shown. But, since the functional form of y(x) is not known a priori between the limits of the integral, this integral cannot be evaluated analytically and some discrete approximation must be made for f(x,y) between these limits. The means of doing so gives rise to a wide variety of numerical methods for solving ODEs.

Computational Mechanics

1-6

David Apsley