College of Engineering and Computer Science Mechanical Engineering Department Mechanical Engineering 501B Seminar in Engineering Analysis Spring 2009 Class: 14443 Instructor: Larry Caretto First Midterm Exam Solutions 1. Solve the diffusion equation u/t = 2u/x2 for u(x,t) with the boundary conditions that u(0,t) = 0 and u/x = 0 at x = L. a) Find the general solution and the expression for the expansion coefficient for any initial condition, u(x,0) = f(x). Applying separation of variables to the diffusion equation gives the following result. u( x, t ) e t C1 sin( x) C2 cos(x) 2 [1] Taking the first derivative of this equation with respect to x gives. 2 u( x, t ) e t C1 cos( x ) C2 sin( x ) x [2] The boundary condition that u = 0 at x = 0 gives u( x, t ) e t C1 sin( 0) C2 0 2 [3] This requires that C2 = 0. The boundary condition that the ∂u/∂x = 0 at x = L gives the following result. u ( x, t ) x 0 e t C1 cos(L) 0 2 x L [4] We can satisfy this condition by having cos(L) = 0 which requires L to be an odd integer times /2. We can express this condition as follows. L 2n 1 or 2 2n 1 2L integer n [5] We can now write the general solution for our problem as the sum of all solutions that satisfy the differential equation and the boundary conditions. Each solution can have a different constant which we call Cn in the equation below. u ( x, t ) Cn e nt sin( n x) 2 n 0 n 2n 1 2L [6] As usual we find the coefficients, Cn, by using an eigenfunction expansion for the initial condition. Jacaranda (Engineering) Room 3333 Email: lcaretto@csun.edu Mail Code 8348 Phone: 818.677.6448 Fax: 818.677.7062 First midterm solutions L. S. Caretto, Spring 2009, ME 501B 2n 1x n 0 L u ( x,0) f ( x) Cn sin Page 2 [7] If we multiply both sides of this equation by sin[(2m+1)x/2L, and integrate the result from 0 to L, we obtain. L f x sin 0 2m 1x dx L sin 2m 1x 2L Cn sin n 0 2L 0 2m 1x dx 2L [8] Because of the orthogonality of the sine eigenfunctions, there is only one term in the summation on the right side of this equation. L 0 2m 1x dx C f x sin 2L x 2L 22m 1x 2m 1x dx C sin m sin m 2L 2L 2 42m 1 0 [9] 0 L 2m 1L 0C LC m L sin m L 2 2 22m 1 L L 2 Solving for Cm gives 2 Cm L L f xsin 0 2m 1x dx [10] 2L b) Write the solution for the specific initial condition that u(x,0) = f(x) = U, a constant. In this case, equation [10] for Cn becomes (after replacing m by n) 2n 1x dx 2U 2 L cos 2n 1x 2 Cn U sin L 2L L 2n 1 2L x 0 x L L 0 2n 1L cos0 4U 4U cos 2n 1 2n 1 2L [11] Substituting this result into equation [6] gives. u ( x, t ) 4U n 0 e 2 n12 2 t 4 L2 sin 2n 1x 2n 1 2L [12] c) What are the first three terms in your equation for part b)? 9 2t 252t 2t 1 x 1 3 x 1 5x 2 2 2 4 L 4 L 4 L u ( x, t ) 4U e sin e sin e sin 2 L 3 2 L 5 2L [13] First midterm solutions L. S. Caretto, Spring 2009, ME 501B Page 3 d) Find the answer to part a) change if the original boundary conditions were replaced by the boundary conditions that u(0,t) = A and u/x = B at x = L, where A and B are constants? (Continue to use the initial condition that u(x,0) = f(x).) In this case our solution for u(x,t) would be written as the sum of two functions. u( x, t ) u( x, t ) wx [14] Here, v(x,t) is the solution to the original problem, with zero boundary conditions. The solution for v(x,t) is the same as the solution for u(x,t) found in equation [6]. The differential equation and boundary conditions for w are found from the requirements that u has to satisfy the diffusion equation and boundary conditions, while v satisfies the diffusion equation with zero boundary conditions. These requirements can only be satisfied if w(x)is given by the following differential equation and boundary conditions. d 2w dx 2 0 w0 A dw dx B [15] xL This gives the following result for w. w Bx A [16] (Note that dw/dx =B so that for long times the gradient is a constant for all x values.) The solution for u = v + w would then be given by a combination of equations [6] and [16]. u ( x, t ) Cn e nt sin( n x) Bx A 2 n 0 n 2n 1 [17] 2L Here, the equation for Cn, given by equation [10] would have to be modified to include the Bx + A term. 2 Cm L L f x Bx Asin 0 2m 1x dx [18] 2L e) What is the answer to part d if f(x) = u(x,0) = U, a constant? (This question was not included on the exam!) In this case equation [18] for Cm = becomes 2 . Cm U Bx Asin 2m 1x dx 2U A sin 2m 1x dx 2B x sin 2m 1x dx [19] L 2L L 2L L 2L L L L 0 0 0 The first integral above, with the (U – A) factor, is similar to the one found in equation [11]. The result for this integral is given below. First midterm solutions L. S. Caretto, Spring 2009, ME 501B Page 4 2m 1x dx 4U A 2U A sin 2m 1 L 2L 0 L [20] The second integral in equation [19] is found from integral tables. 1 xL 2 L 2m 1x 2B 2m 1x 2B 2 Lx 2m 1x 2 L sin x sin dx cos L 2L L 2m 1 2L 2L 2m 1 0 x 0 2 2 2m 1L 2 L sin 2m 1L 0 sin 0 2B 2L cos . [21] 2m 1 L 2m 1 2L 2L 2 m 2m 1 8BL 1 2B 2L sin 0 L 2m 1 2 2m 12 2 Combining equations [19], [20], and [21] gives 2U A 2m 1x 2B 2m 1x 4U A 8BL 1m [22] Cm sin dx x sin dx 2m 1 2m 12 2 L 2L L 2L L . L 0 0 WE can replace m by n in this equation and then substitute the result for C n into equation [17] to get the problem solution. 4U A 8BL 1 e u ( x, t ) 2n 12 2 n 0 2n 1 2 n 12 2 n 4 L2 t sin 2n 1x 2n 1 2L Bx A [23] We can divide this equation by U to obtain the following result. e u ( x, t ) A 4 BL 8 1m 1 2 2 U U 2m 1 U 2m 1 n 0 2n12 2 t 4 L2 sin 2n 1x 2n 1 2L BL x A U L U [23] As before the solution for u(x,t)/U depends on x/L and /L2, but this solution also depends on the ratios BL/U and A/U. 1 x 1 x sin axdx a cos ax a 2 sin ax C First midterm solutions L. S. Caretto, Spring 2009, ME 501B Page 5 2. Solve the following problem which is a model of heat transfer in an electronic component. 2T x 2 T x 0 x 0 T x x L 2T y 2 0 0 x L, 0 y H h Tx L T 0 k T y GS y 0 T y yH h T y H T 0 k In this problem, T, k, h, and GS are constants. We can solve this problem by defining a new variable, = (T – T). With this substitution we have the following problem. 2 x 0 x 0 2 0 0 x L, 0 y H x 2 y 2 h x L 0 GS x x L k y y 0 y yH h yH 0 k [2] This problem has homogenous boundary conditions at all boundaries except the boundary at y = 0. The general separation of variables solution for the Laplace equation with an eigenfunction expansion in the x direction to fit the boundary condition at y = 0 is ( x, y) Asin( x) B cos(x)C sinh( y) D cosh( y) [3] Taking the first derivative of this solution gives. A cos(x) B sin( x)C sinh( y ) D cosh( y ) x [4] We can satisfy the boundary condition this derivative is zero at x = 0 by setting A = 0. The mixed boundary condition at x = L requires the following relation (after setting A = 0). h x L B sin( L)C sinh( y ) D cosh( y ) x x L k h B cos(L)C sinh( y ) D cosh( y ) 0 k [5] We can satisfy this boundary condition by the following equation that can be solved for the eigenvalues, = L. The eigenvalues will depend on the problem parameter ratio hL/k. h sin( L) cos(L) 0 k k L cot L hL [6] We now have to match the boundary conditions at y = 0 and y = H. Taking the y derivative of the solution in equation [3] (after setting A = 0 as noted above) gives [1] First midterm solutions L. S. Caretto, Spring 2009, ME 501B Page 6 B cos(x)C cosh( y ) D sinh( y ) y [7] The boundary condition at y = H requires that the following equation be satisfied y yH h y H B cos(x)C cosh( H ) D sinh( H ) k h B cos(x)C sinh( H ) D cosh( H ) 0 k [8] This equation will be satisfied if C cosh( H ) D sinh( H ) h C sinh( H ) D cosh( H ) 0 k [9] Dividing this equation by cosh(H), using the definition that tanh(H) = sinh(H)/cosh(H) and rearranging gives h h C tanh( H ) D tanh( H ) 0 k k [10] Thus the constants C and D are related by the following equation. D C h k tanh( H ) h tanh( H ) k [11] Substituting this result and into equation [3] (with A = 0) gives h n tanh( n H ) k ( x, y ) BC cos(x) sinh( y ) cosh( y ) h n tanh( n H ) k [12] This is our desired eigenfunction solution. We can replace the product of constants, BC, by a single constant, Bn for each eigenvalue and write the general solution as the sum of all the eigenfunctions h n tanh( n H ) k ( x, y ) Bn cos( n x) sinh( n y ) cosh( n y ) h n 1 n tanh( n H ) k [13] We have to match the condition that /y = GS.at y = 0. Taking the y derivative of equation [13] and setting y = 0 gives the following results. First midterm solutions L. S. Caretto, Spring 2009, ME 501B Page 7 h n tanh( n H ) k Bn cos(n x)n cosh( n y ) sinh( n y ) h y n 1 n tanh( n H ) k y [14] h n tanh( n H ) k GS Bn cos(n x)n cosh( n 0) sinh( n 0) Bn n cos(n x) h n 1 n1 n tanh( n H ) k y 0 Multiplying this equation by cos(mx), where m is a different eigenvalue, and integrating from 0 to L gives the following equation that can be solved for the expansion coefficients, Bn. L L 0 0 n 1 GS cos(m x)dx Bn cos(n x)n cos(m x)dx [16] As usual, the orthogonality of the eigenfunctions leaves only one term in the sum, the one for which n = m. L L GS cos(m x)dx Bmm cos 2 (m x)dx 0 [17] 0 The integrals in this equation are found as follows. xL 1 1 0 cosm xdx m sin m x m sin m L x 0 L [18] xL x 1 L 1 0 cos (m x)dx 2 4m sin( 2m x) 2 4m sin( 2m L) x 0 L 2 [19] Combining equations [17], [18], and [19], and multiplying top and bottom by L to obtain the eigenvalue, m = mL, gives the following result for Bm. GS Bm 1 m sin( m L) L 1 m sin( 2m L) 2 4m 4GS sin( m L) 4G L sin( m L) 2 2 S 2 L m sin( 2m L) 2m L m L sin( 2m L) 2 m Substituting this expression for Bm into the general solution in equation [13] gives [20] [15] First midterm solutions L. S. Caretto, Spring 2009, ME 501B Page 8 h n tanh( n H ) 4GS L sin( m L) cos(n x) k ( x, y ) 2 2 cosh( n y ) [21] sinh( n y ) h n1 2m L m L sin( 2m L) n tanh( n H ) k b) Show that the solution to this problem can be expressed in terms of the ratio (T – T)/GSL. Show that this ratio is a function of the following variables: x/L, y/H, H/L, and hL/k. We can substitute the definition of = T – T in equation [21] and divide that temperature difference by GSL to get the desired ratio. In addition we can multiply the fraction with the tanh terms top and bottom by L and introduce the roots of the eigenvalue equation [6], n = nL throughout equation [21] to obtain the following result. x hL H sin( n ) cos n n tanh n T T y H y H L k L 4 sinh cosh n n 2 H hL GS L H L H L 2 sin 2 n n n n 1 n tanh n L k We see that the variables that are explicitly shown in this equation are the ones listed in the problem statement. In addition, the eigenvalues, n, depend only on the ratio hL/k. (See equation [6].) Thus, equation [22] shows the desired functional behavior: (T – T)/(GSL) = f(x/L, y/H, H/L, hK/k). [22]