74 . 75 76 Partial Differential Equations (PDEs) 385 ‘The unsteady isothermal TRAM equation with an irreversible first order reaction taking place (Eq. 7.3) can be non dimensionalized to de_1de_de Or Peaz? az —(Daje £ GL" and 5 in terms of actual variables}. Use the N+ 1° grid-line FD where c,z and ¢ are now nondimensionalized variables [ technique to obtain a set of ODE-IVPs. Attempt to solve these using a GEAR package with Pe = 10', Da = 1, N = 10 and 100. Very sharp changes in the c profile will be observed [14], and you may have to use very large values of N to get good results. [Moving finite-element techniques [10-12, 14] are better suited to handle such problems]. Complete Eq. (b) in Example 7.3 using the given boundary conditions for T. Use N, = N, = 3, and simplify to obtain only four equations corresponding to the internal grid points. Line Jacobi Method (3, 4}: Assuming values for T}1;'~' and T}}"', etc., develop the algorithm for solving the set of N, — 1 equations in Eq. 7.9 using the Jacobi technique. Thus, develop the following flow-chart: Know all Tf; Assume all Ti} " (Do j=2,3, Assume Th !*= 2,3, ...ytill convergence oN, [Sassy FT AHO) Panyeriy]386 V7 78 7.9" 7.10 Numerical Methods for Engineers After this full set of computations, we update the values of all the Ti} using the results. obtained, and repeat this set of. ‘computations. The line Gauss-Seidel technique is also easily written, and con- verges (usually) even faster. In this technique, updated values of T#?', and Ti, are to be used, Consider Example 7.3 for the 2-D steady-state case, i.e., for &r or ax? dy? x20: T=T 3 x=L, : T=7, y=0 2 T=T, ; yal, : Ter, Obtain the set of algebraic equations to be solved and write the algorithm for solving these using (a) the line Gauss-Seidel technique (b) the ADI method Use the Crank-Nicholson scheme to obtain algebraic equations for the FD equations in Example 7.1 (for a single species). Then write the algorithm involving a line Gauss-Seidel technique to solve these cavations. You may consider only the equations for the internal grid points. Alternating Direction Explicit Scheme [3]: Consider the following Scheme to convert Eq. (a), Example 7.3,to.aset of algebraic equations: + (Ar + + Ti Th AO rh arts tert OO ress ort ert) yy Show that (with all 7!; known, and with boundary values also spe- cified) if we solve the above equations in the Sequence : increasing i; decreasing j (ice, p—Do ie ere. We get an explicit set of computations, Confirm the explicit nature of the computations using a grid with W, = N, = 6. Consider the PDR.TAL 7.12 713 Partial Differential Equations (PDEs) 387 eas 7=f@y) Use the FD technique with unevenly-spaced grid points and obtain an algebraic equation for T,,; [see Prob. 6.3 (a)]. This technique can also be used for treating Dirichlet boundary conditions (T specified) with irregular boundaries. Obtain the FD form of the 1-D wave equation (hyperbolic PDE): oT OT sr =a Sxsl aP ax T(x,0)= sin(cx) ar Sr ees0)=0 T,1)=0 TU, =0 Use the ot formulae for both ¢ and x. Obtain numerical results 15 using Ax ==, At = + and obtain results tilly = 72. Compare your results with the analytical solution: T(x, t) = sin(ex) cos(mr) Jain et. al. [4] discuss the use of implicit and ADI schemes for solving this equation, as well as the 2~D wave equation. Use the explicit Euler formula on Eq, (a), Example 7.3, to obtain a set of algebraic equations. - For the first-order PDE TT no; osrse ar ’ TO x)=f&) + T¢,0)=g@). deduce (using the implicit Buler technique in the x direction, and a first order formula for %)- Titec tere388° Numerical Methods for Engineers Show that this can be solved such that no trial and error is required. Similarly, deduce the leap-frog scheme : pitta tot ‘ =r -Thy H vel This is stable for 0 Ond.0)=0 Wr =0,8,9,0)=1 (Note that r is the dimensionless reciprocal of the radial Position from the center of the sphere, so that r = 1 is at the sphere and_r = 0 is far away): Use the finite difference scheme and develop a set of algebraic ? equations to be solved. Use an implicit scheme for the a term, : and an-explicit scheme for the other terms (with respect to 1). Also, study the boundary conditions, as well as any special problems that may arise at r = 0 and r = |. Note also that we do not have ‘boundary’ conditions for @ and 6 in this problem, but these are not required. 7.15 The convective diffusion equation for an isothermal, first-order reaction occurring in a laminar flow reactor is described by (Ref. 36; p. 180) 2 AX dc ao at r=0 and at r=17.16 TAT 7.18 Partial Differential Equations (PDEs) 389 Obtain FD results for (ki) = 1 and a0 The 1—D flow of a fluid between two flat plates spaced a distance, h, apart, with one plate at rest, and the other starting to move impulsively at t = 0, with velocity i, is described by (Ref. 37, p.83) au _(v \eu or \h? Jax? u(x,0)=0 u(Q0,t)=1 u(1,)=0 where u is the dimensionless velocity, and x is the dimensionless position. Set up the’ FD equations and solve’ for yess {also see Prob. 7.1). The unsteady 1—D heat transfer in an infinite slab (see Prob. 6.12) of thickness L, with thermal conductivity being a linear function of temperature, is described by the following dimensionless equation: Taran eer(Z) T(x,0)=0 TO,)=1 T(t) =0 where the non-dimensionalization is slightly different than in Prob. 6.12. Apply the FD technique and obtain a set of ODE-IVPs to be solved. The unsteady reaction-diffusion in a spherical catalyst particle is described by [3, ac o,%¢ -42 |x [4% }- @R(c,T) or Lal or Ma = Baxl* [4z eee, T) a0 ee0390 Numerical Methods for Engineers rel ¥aBige(.)- a0) =BilT(1,t)-g,0)) £=0 :¢(,0)=c,(x) T(@,0)=Tyx) Set up the ODEs to be integrated using the OC method. Ref. 38 gives the results obtained: using a Gear integration package, for R= c exp [y-2] + Y= 20, B= 0.6, 6° = 0.25, M, = 176, M, = 199, Bi = Bin, = 22, To = 1.05, C= 1.0, T (1, 1) = 8, () = 1.0, ¢ (1, ) = gy (0) = 1.0: 7.19 The desorption of a gas from a liquid stream in a wetted-wall column can be described by [7] the following PDE: ar_a ane 54 Zo at n=l f=0 at n=0 f=1l at C=0 Note that the problem, as defined above, is NOT symmetric about 1) = 0 (do NOT redefine variables to make them symmetric). Carry out an orthogonal collocation solution in the 1 direction, using N = 2 (i.e, two internal OC points), and simplify to obtain a set of two ODE-IVPs. Do not solve these. 7.20. An axially dispersed isothermal chemical reactor can be described by the following dimensionless equations: of laf 1 [26% a2 Pe,dz?Pe,Lrar\" ar J] * OOF ae mz=0 ISB eae7.21 Partial Differential Equations (PDEs) 391 =0 at r=1 (no flux) Note symmetry in the r-direction, and lack of symmetry in, the z-direction. Use 1 (internal) collocation point in the r-direction and 2 internal collocation points in the z-direction, and set up all the alge- braic equations necessary to solve the problem . Define F),7 = f(r,2))- Use actual numbers for the matrix elements. Collect all your equations into a neat matrix form : (AA)gxs Fpx1 = Bsxt With F = (Fy, Fis Fis, F yay Fats Faas Fass F nl, $0 that one could use LU Decomp easily. Do not solve your final set of algebraic (matrix) equations. The dimensionless equation describing steady-state heat transfer to a fluid under fully developed laminar flow inside a cylindrical pipe (with walls heated electrically) is given by (Ref. 21, pp. 295, 296; also see Example 4.10): 2°68 100 06 ig 90,190, 28 Vows at ra 3, 2=0, 6=0 306 r=l1, 3p} 30 r=0, 570 Note the symmetry in the r-direction. (a) Apply orthogonal collocation with one internal collocation point and solve to get @(z) at the collocation points. Solve the ODE obtained analytically. (b) Compare your numerical values from your OC result with the following solution given in Ref. 21 (p. 296): 4 ana @=-42-r'+ 7 + mu at appropriate values of r and for z = 0.5. (c) For the physical situation described here, the dimensionless bulk temperature is given by392 Numerical Methods for Engineers A fea-ryrar 6, == 24 f ea —Pyrar fe =P \rdr reo Obtain a general equation to get 6, (z) using 8, (z) and 6, (z) of part a. 7.22 Consider er _ er ax?” ay? . with BCs: x=0 3. T=f() x=l ; T=fQ) . yal 5 T=g(x) or y=0 ; aa Apply OC exploiting symmetry in the y direction about y = 0. Use. internal collocation point in the y direction, and 1 internal collocation « Point in the x direction. Set up the 2-D OC algebraic equations. Substitute the values of the matrix elements as appropriate, Solve to get the numerical value of the Tat the collocation points if fy) =0.1y Bole) =0.1+0.1x AQ)=0.2y 7.23 Consider the following PDE [3] (see Prob. 7.17): a_a ae ae, fac? 3-H reo] +00 240%) cQ,1)=1 c@,0)=0 ac he =O atx=1 Using ‘wo internal grid points, obtain the ODEs using the OC tech- nique. Do not solve the ODEs.on the computer, (Hint : Redefine the x-axis so that symmetry at x,y = 0 is attained)7.24 7.25 7.26 Partial Differential Equations (PDEs) 393 Obtain the ODE- IVPs with 2 internal OC points (in the z-direction) in the case of the unsteady, isothermal TRAM (Prob. 7.4). The profile is so sharp that integration of these ODEs wiil give poor results; however, these equations will prove handy when we apply the OCFE technique. Formulate Prob. 7.3 (packed bed reactor with radial dispersion) using the OC technique. Deduce the 1 internal collocation point equations by hand. Solve for the parameter values given in Prob. 7.3. Finlayson [3, 34] gives numerical solutions of the ODEs so obtained. He finds that the N =.2 results are almost indistinguishable from higher-N results. However, when parameter values are such that sharp profiles are present, we need to use N = 6. Interestingly, the stiffness of the ‘ODEs is found to increase as N increases. Gupta et. al. [39] present the equations characterizing the third-stage wiped-film reactor used for polyester production. In this there is a ‘bulk’ phase which moves downstream fromthe feed-end to the product end (C direction). In between, rotating blades take up the reaction mass, coat it ina thin layer (reaction and diffusion take place), collect the material and mix it back into the bulk phase. The equations can be summarized as (see Ref. 39 for details of model): Bulk-Phase Film394 Numerical Methods Sor Engineers T20 5 y= FCO) ug &=0: “BE =O forall x S=1 : c,=0 Interconnection between bulk and film Mem fleas ~6)~ lye) ab do Write the OC equations for the wiped-film and obtain a set of - ODE-IVPs to be solved. Attempt to make a computer code (or read up Ref. 39) which solves the bulk-phase equations for a small. distance, AG, then solves the film phase equations (using the computed bulk Phase values) from t = 0 to t= 1, computes N, and uses this in the next stage of the bulk-phase computations, Use Hao = 50 K=05 a,=2000 a,= 1000 a; =10* a,=2 : . ‘on op 27! -27 You can write the OC expression of 52], in two ways: x; ar) me! a |, 2 Bet cs 2 (ar) "2 (ap) ve wea a (3) Wr 3 af} = 2A 2 Aat Confirm for N = 2 (non symmetrical case) if yar Aj A, Ain By= Use this to simplify the PDE (see Prob. 7.23) a) dc 5-2 Po%] c@N=1; c()=0; c@,0)=0. Use N= 1. 7.28 Deduce Eq, (a) of Example 7.6. Hint: Assume7.29 7.30 7.31 7.32 7.34 7.35 Partial Differential Equations (PDEs) 395. Nya 2Ny 42 Te y= Ld ddpxt ty! we te with d, and d; as constants, Write matrix equations for the subsets, (Wij Tajo Trysad” and (Ti. Ti2 + Ti.w,sal” separately in terms of d and d” respectively. Deduce Eq. 7.15, and then write the expression for T for the problem discussed in Example 7.4, with N, = 1,N, = 2. Reduce the PDE of Example 7.3 into a set of ODE-IVPs, using the 2-dimensional OC technique. In this case, the coefficients, d, and dj" in Prob. 7.28, become functions of time. Write the OCFE (2FEs, 2 internal OC points in each) equation for the PDE in Prob. 7.19. Solve using a DDASSL package if available; else use the Crank-Nicholson scheme. Obtain the OCFE equations (2FEs, 2 internal OC points in each) for the system of equations given in Prob. 7.18 (unsteady reaction- diffusion in a porous catalyst sphere (3, 38]). One explicit technique [3] of solving the DAEs obtained on applying the OCFE method to PDES is described here for Eq, (a) of Example 7.7. We first take only the ODEs, using values at z = z, on the right and use any explicit scheme to obtain values of T at OC points 2, 3, Sand6atz =z, +Az. We then write the algebraic equationsand expr Ti"', TY"! and T7*" in terms of all other variables at n+ 1 : (13°, 731,78", Te *"). These algebraic equations can be solved using Thomas’ algorithm (called smoothing). Illustrate this for Eq. (a) of Example 7.7 for T only (you can solve these equations only after performing a similar decoupling of the equations for c)). S Apply the OCFE techniques (WV, FEs, 2 internal OC points in each FE) to the PDE in Prob. 7.4 (unsteady, isothermal TRAM), in the z-direction Consider the isothermal TRAM with an irreversible second order reaction taking place, operating under unsteady conditions. The PDE is given by (for Pe = 6, Da = 2) ae _1ae a . 2 ‘Ot. 622 dz396 Numerical Me (ethods for Engineers z=0 1=0 : c=0 (see Prob. 7.4 and Example 6.9). Use the GFE technique and obtain aset of ODE-IVPs fora. general N. Use the ‘triangular’ basis-functions [Hint : the results of Example 6.9 can easily be extended]. 7.36 Derive Eq. 7.18(c). 7.37 Set up and solve the GFE equations for the heat conduction problem (3) oT aT ae tsel T=Oall around c. Fig. P.7.37 Use the FEs shown in Fig. P. 7.37. Note that there is only one unknown variable, 7, 738 Consider the steady, 2-D heat conduction problem with internal ‘generation’; or ant 5378— 7.40 Partial Differential Equations (PDEs) 397 atrayeo at exo (0,0)2 atvayes Fig. P.7.38 with the boundary conditions shown in Fig. P.7.38. Obtain the GFE for j = 1 using the FEs shown. Obtain your results in the form : Ea, = a6 with actual values for the coefficients ay, a, ds arid dg. Consider the steady, 2-D heat conduction in the I-beam shown in Fig. P7.39: . eT eT _ ax? dy? The face 4-8-7-11-12-16 is insulated while T= 0 on the other three faces (1-2-3-4, 1-5-6-10-9-13, and 13-14-15-16). Q =— 2 onelements (7) and (8) and Q = 0 elsewhere. Set up the GFE equations in terms of B, D, and E. Do not evaluate these coefficients. Assume the T = 0 condition to be valid for nodes 4 and 16. The first order irreversible reaction cum-diffusion in the finite, cylindrical, porous catalyst pellet under non-isothermal conditions can be described by the following dimensionless PDE [40]: ae lac Fe 1de, nde _ yp [ mig Ht rar tt a2 96 PLY Ty Bde) | BOLO398 — Numerical Methods ‘for Engineers 12(3,2) araxeo 60,0) ‘ 2 * 0) (2,0) 3,0) where a is the diameter/length ratio, z is the dimensionless axial {ocation, and the other terms have the same meaning as in Example 6.3. Use the GFE technique and obtain the set of algebraic equations, 1 Davis [7] presents results for this problem using code, DISPL, for B = 0.1, 7=30,6=2,a=1, TAL Expressions for a and x have been obtained for the oc - "y formulation. Obtain an expression for the mixed derivative, = ha at location x,, y, in terms of the appropriate OC matrices. If N, = 1 and N,= 2 (a = 1) write, with matrix elements substituted, a series 3 expression for T ordy Insu. 13, 14 15. 16. 17. 18. 19. 20. 21. 22. 24, 25. 26. 27. ct 23. Partial Differential Equations (PDEs) 399 | References F.B. Hildebrand, Advanced Calculus for Applications, Prentice Hall, Englewood Cliffs, NJ, 1962. D. Pal and S.K. Gupta, Polymer, 30, 1918 (1989). B.A. Finlayson, Nonlinear Analysis in Chemical Engineering, McGraw Hill, New York, 1980. .. MAK. Jain, S.R.K. Iyenger and R.K. Jain, Numerical Methods for Scientific and Engineering Computation, Wiley Eastern, New Delhi, 1985. [a very crisply-written book. Discusses several techniques] D.W. Peaceman and H.H. Rachford, JSIAM, 3, 28(1955). D.W. Peaceman, Fundamentals of Numerical Reservoir Simulation, Elsevier, New York, 197. ME. Davis, Numerical Methods and Modeling for Chemical Engineers, Wiley, New York, 1984. R.E. Bank and D.J. Rose, SIAM J. Numer, Anal., 17, 806 (1980). B. Wendroff, Theoretical Numerical Analysis, Academic, New York, (1966). R. Alexander, P. Manselli and K. Miller, Moving Finite Elements for the Stefan Problem in Two Dimensions, Rend, Acad. Naz. Lincei (Rome), Serie VIII, Vol. LXVIL, fase. 1-2, pp. 57-61 (1979). K. Miller and RN. Miller, SIAM J. Numer. Anal., 18, 1019, 1033 (1981). B.A. Finlayson, Numerical Methods for Problems with Moving Fronts, Ravenna Park Pub. Inc., 6315, 22nd Ave. NE, Seattle, WA 98115-6919 (1991). [Refs. 10-12 discuss moving finite elements for the solution of PDEs}. AX. Jana and S.K. Gupta, J. Polym. Eng., 9, 23 (1990). C. Sereno, A. Rodriguez and J. Villadsen, Comp. Chem. Eng., 15, 25 (1991). [a good discussion of the moving finite element method] . JN. Reddy, An Introduction to the Finite Element Method, McGraw Hill, New York, 1984, .C. Zienkiewicz and R.L. Taylor, The Finite Element Method, 4th ed., McGraw Hill, London, 1989. . S.S. Rao, The Finite Element Method in Engineering, Pergamon, Oxford, UK, 1982. (Refs. 15-17 are some classic texts on FE methods). M. Tumer, R. Chough, H. Martin and L. Topp, J. Aero. Sci., 23, 805 (1956). A. Hrenikoff, J. Appl. Mech., Trans. ASME, 8, 169 (1941). H. Mavridis, A.N. Hrymak and J. Vlachopoulos, Polym. Eng. Sci., 26, 449 (1986). R.B. Bird, W.E. Stewart and E.N. Lightfoot, Transport Phenomena, Wiley, New York, 1960. \V. Gupta and S.K. Gupta, Fluid Mechanics and its Applications, Wiley Eastern, New Delhi, 1984, E. Mitsoulis, Ph. D. Thesis, Dept. of Chemical Engineering, McMaster University, Hamilton, ON, Canada, 1984. 1H. Mavridis, A.N. Hrymak and J. Viachopoulos, J. Rheol., 30, 555 (1986). R. Schmidt, Polym, Eng. Sci:, 14, 797 (1974). DJ, Coyle, TJ. Tulig and M. Tirrell, Ind. Eng. Chem. Fundam, 24, 343 (1985). JR. Rice and RF. Boisvert, Elliptic Problem Solving with ELLPACK, Springer Verlag, New York, 1983.400 28. 29, 30. 31. 32. 33. 34, 35. 36. 37. 38. 39, 40. Numerical Methods for Engineers G. Sewell, in Adv. Computer Methods for PDEs ~ Ill, (eds., R. Vishnevetsky and RS. Stepleman), IMACS (AICA), Rutgers University, New Brunswick, NJ, 1979. (Details of code 2DEPEP] G2. Leaf, M. Minkoff,G.D. Byme,D. Sorensen, T. Bleakney and J. Saltzman, Rept. ANL-77-12, Argonne National Lab., Argonne, IL, 1977, (Details of code DISPL] Ri. Melosh, Structural Engineering Analysis by Finite Elements, Prentice Hall, Englewood Cliffs, NJ, 1990. C. Taylor and T.G. Hughes, Finite Element Programming of the Navier-Stokes Equations, Pineridge, Swansea, UK (1981), Pow. Nicholson and N.W. Nelson, Rubber Chem. Tech., 63, 368 (1990), Jd. Carberry, Chemical and Catalytic Reaction Engineering, McGraw, Hill, New York, 1976. B.A. Finlayson, Chem. Eng, Sci., 26, 1081 (1971). D.D. Ravetker, Ph.D. Dissertation, Univ. of Bombay, Mumbai, 1989 EB. Nauman, Chemical Reactor Design, Wiley, New York, 1987, H. Schlichting, Boundary Layer Theory, th ed, McGraw Hill, New York, 1968, NB. Ferguson and B.A. Finlayson, Chem. Eng. J. 1, 327 (1970), SE, Gupta. A.K. Ghosh S.K. Gupta and A. Kuma, J. Appl. Polym. Sc, 29, 3217 J. Villadsen and M.L. Michelsen, Solution of Differential Equation Models by Polynomial Approximation, Prentice Hall, Englewood Cliffs, NJ, 1978,