A. Numerical Techniques

Lake Michigan—unsalted and shark-free.

A.1 Useful Integrals in Chemical Reactor Design

Also see www.integrals.com.

χ0dx1-x=ln11-x(A-1)

x2x1dx(1-x)2=11-x2-11-x1(A-2)

x0dx(1-x)2=χ1-x(A-3)

x0dx1+ɛx=1ɛln(1+ɛx)(A-4)

x0(1+ɛx)dx1-x=(1+ɛ)ln11-x-ɛx(A-5)

x0(1+ɛx)dx(1-x)2=(1+ɛ)x1-x-ɛln11-x(A-6)

x0(1+ɛx)2dx(1-x)2=2ɛ(1+ɛ)ln(1-x)+ɛ2x+(1+ɛ)2x1-x(A-7)

x0dx(1x)(ΘBx)=1ΘB1InΘBxΘB(1x)ΘB1(A-8)

W0(1-αW)1/2dW=23α[1-(1-αW)3/2](A-9)

x0dxax2+bx+c=-22ax+b+2bforb2=4ac(A-10)

x0dxax2+bx+c=1a(p-q)ln(qpxpxq)forb2>4ac(A-11)

where p and q are the roots of the equation.

ax2+bx+c=0,that is, p,q=b+b24ac2ax0a+bxc+gxdx=bxg+agbcg2 Inc+gxc(A-12)

A.2 Equal-Area Graphical Differentiation

There are many ways of differentiating numerical and graphical data (cf. Chapter 7). We shall confine our discussions to the technique of equal-area differentiation. In the procedure delineated here, we want to find the derivative of y with respect to x.

This method finds use in Chapter 5.

  1. Tabulate the (yi, xi) observations as shown in Table A-1.

  2. For each interval, calculate Δxn = xnxn-1 and Δynynyn-1.

    TABLE A-1 FINDING DIFFERENTIALS (dydx) FROM DISCRETE DATA

    xi

    yi

    Δx

    Δy

    ΔyΔx

    dxdy

    x1

    y1

    (dydx)1

    x2x1

    y2y1

    (ΔyΔx)2

    x2

    y2

    (dydx)2

    x3x2

    y3y2

    (ΔyΔx)3

    x3

    y3

    (dydx)3

    x4

    y4

    and so on.

  3. Calculate Δynxn as an estimate of the average slope in an interval xn-1 to xn.

  4. Plot these values as a histogram versus xi. The value between x2 and x3, for example, is (y3y2)/(x3x2). Refer to Figure A-1.

    Equal area differentiation is graphed.

    Figure A-1 Equal-area differentiation.

  5. Next, draw in the smooth curve that best approximates the area under the histogram. That is, attempt in each interval to balance areas such as those labeled A and B, but when this approximation is not possible, balance out over several intervals (as for the areas labeled C and D). From our definitions of Δx and Δy, we know that

    yn-y1=nΣi=2ΔyΔxiΔxi(A-13)

    The equal-area method attempts to estimate dy/dx so that

    yn-y1=xnx1dydxdx(A-14)

    that is, so that the area under Δyx is the same as that under dy/dx, everywhere possible.

  6. Read estimates of dy/dx from this curve at the data points x1, x2, … and complete the table.

An example illustrating the technique is given on the CRE Web site, Appendix A.

Differentiation is, at best, less accurate than integration. This method also clearly indicates bad data and allows for compensation of such data. Differentiation is only valid, however, when the data are presumed to differentiate smoothly, as in rate-data analysis and the interpretation of transient diffusion data.

A.3 Solutions to Differential Equations

A.3.A First-Order Ordinary Differential Equations

See the CRE Web site, Appendix A.3.

dydt+f(t)y=g(t)(A-15)

Using integrating factor = exp (fdt), the solution is

y=efdtg(t)efdtdt+K1e-fdt(A-16)

Example A–1 Integrating Factor for Series Reactions

dydx+k2y=k1ek1t

Comparing the earlier equation with Equation (A-15) we note

f(t) = k2

and the integrating factor =expk2dt=ek2t

Multiplying both sides by the integrating factor

d(yek2t)dt=ek2tk1e-k1t=k1e(k2-k1)t

Integrating

ek2ty=k1e(k2-k1)tdt=k1k2-k1e(k2-k1)t+K1y=k1k2-k1e-k1t+K1e-k2tt=0 y=0y=k1k2-k1[ek1tek2t]

A.3.B Coupled Differential Equations

Techniques to solve coupled first-order linear ODEs such as

dxdt=ax+bydydt=cx+dy

are given in Web Appendix A.3 on the CRE Web site (http://www.umich.edu/~elements/6e/appendix/AppA.3_Web.pdf).

A.3.C Second-Order Ordinary Differential Equations

Methods of solving differential equations of the type

d2yd2x-βy=0(A-17)

can be found in such texts as Applied Differential Equations by M. R. Spiegel (Upper Saddle River, NJ: Pearson, 1958, Chapter 4; a great book even though it’s old) or in Differential Equations by F. Ayres (New York: Schaum Outline Series, McGraw-Hill, 1952). Solutions of this type are required in Chapter 15. One method of solution is to determine the characteristic roots of

(d2dx2-β)y=(m2β)y

which are

m=±β

The solution to the differential equation is

y=A1eβx+B1e+βx(A-18)

where A1 and B1 are arbitrary constants of integration. It can be verified that Equation (A-18) can be arranged in the form

y=A sinh βx+B coshβx(A-19)

Equation (A-19) is the more useful form of the solution when it comes to evaluating the constants A and B because sinh(0) = 0 and cosh(0) = 1.0.

A.4 Numerical Evaluation of Integrals

In this section, we discuss techniques for numerically evaluating integrals for solving first-order differential equations.

  1. Trapezoidal rule (two-point) (Figure A-2). This method is one of the simplest and most approximate, as it uses the integrand evaluated at the limits of integration to evaluate the integral

    X0X1f(X)dX=h2[f(X0)+f(X1)](A-20)

    when h = X1 = X0.

    The trapezoidal rule is depicted.

    Figure A-2 Trapezoidal rule illustration.

  2. Simpson’s one-third rule (three-point) (Figure A-3). A more accurate evaluation of the integral can be found with the application of Simpson’s rule:

    X0X2f(X)dX=h3[f(X0)+4f(X1)+f(X2)](A-21)

    where

    h=X2X02 X1=X0+h

    Methods to solve

    0XFA0rAdX

    in Chapters 2, 4, 5, 12, and 0X(t)E(t)dt

    in Chapter 17

    A graph of f of x versus x is shown.

    Figure A-3 Simpson’s three-point rule illustration.

  3. Simpson’s three-eighths rule (four-point) (Figure A-4). An improved version of Simpson’s one-third rule can be made by applying Simpson’s three-eighths rule:

    X0X3f(X)dX=38h[f(X0)+3f(X1)+3f(X2)+f(X3)](A-22)

    where

    h=X3X03X1=X0+hX2=X0+2h

    A graph of f of x versus x is shown.

    Figure A-4 Simpson’s four-point rule illustration.

  4. Five-point quadrature formula.

    X0X4f(X)dX=h3(f0+4f1+2f2+4f3+f4)(A-23)

    where

    h=X4X04

  5. For N = 1 points, where (N/3) is an integer

    X0XNf(X)dX=38h[f0+3f1+3f2+2f3+3f4+3f5+2f6+...+3fN1+fN](A-24)

    where h=XNX0N

  6. For N = 1 points, where N is even

    X0XNf(X)dX=h3(f0+4f1+2f2+4f3+2f4++4fN-1+fN)(A-25)

    where h=XNX0N

These formulas are useful in illustrating how the reaction engineering integrals and coupled ODEs (ordinary differential equation(s)) can be solved, and also when there is an ODE solver power failure or some other malfunction.

A.5 Semi-Log Graphs

Review how to take slopes on semi-log graphs on the Web. Also see https://bolide.cs.uoguelph.ca/tutorials/GLP. Also see Web Appendix A.5, Using Semi-Log Plots for Data Analysis.

A.6 Software Packages

Instructions on how to use Polymath, MATLAB, Python, Wolfram, COMSOL, and Aspen can be found on the CRE Web site.

For the ordinary differential equation solver (ODE solver), contact:

Polymath Software

P.O. Box 523

Willimantic, CT 06226-0523

https://www.polymath-software.com/fogler

Wolfram

100 Trade Center Drive

Champaign, IL 61820-7237

Tel: +1-217-398-0700
       +1-800-WOLFRAM (965-3726)

Fax: +1-217-398-0747

www.wolfram.com

Python Software

206 East 9th Street

Floor 18

Austin, TX 78701

Tel: +1-512-222-5440

E-mail: [email protected]

www.python.org

MATLAB

The Math Works, Inc.

20 North Main Street, Suite 250

Sherborn, MA 01770

For ordinary and partial differentiating equation solvers, contact:

COMSOL, Inc.

1 New England Executive Park

Burlington, MA 01803

Tel: +1-781-273-3322

Fax: +1-781-273-6603

E-mail: [email protected]

www.comsol.com

For simulators, contact:

Aspen Technology, Inc.

10 Canal Park

Cambridge, MA 02141-2201

Email: [email protected]

www.aspentech.com

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset