From: Ha Le <ha.7@gm...>  20100108 22:40:26

Hi guys, I'm using libmesh to solve a basic structural mechanics problem: Just using 1 element (QUAD4) for now.  1 side (Node 0 1) is constrained, 3 other frees,  1 force (0,20) (xy coordinate) acting down face 3 (Node 30). 0force down y3   constrained  constrained   constrained      12 Long story short, after assembling, I've got a matrix look like this: *3.34825e+09** **3.32591e+09** **1.66e+09** ** 1.66832e+09* 9.77308e+06 6.6e+06 1.52308e+06 825000 3*.32591e+09** **3.36083e+09** **1.66914e+09** **1.64206e+09* 6.6e+06 1.67856e+07 1.65e+06 1.38981e+07 *1.66e+09** ** 1.66914e+09** **3.34412e+09** **3.33663e+09* 1.52308e+06 1.65e+06 5.64808e+06 4.125e+06 *1.66832e+09** **1.64206e+09** **3.33663e+09** **3.35938e+09* 825000 1.38981e+07 4.125e+06 1.53418e+07 9.77308e+06 6.6e+06 1.52308e+06 825000 2.00538e+07 8.25e+06 1.18038e+07 2.475e+06 6.6e+06 1.67856e+07 1.65e+06 1.38981e+07 8.25e+06 3.82038e+07 3.3e+06 3.53163e+07 1.52308e+06 1.65e+06 5.64808e+06 4.125e+06 1.18038e+07 3.3e+06 1.59288e+07 2.475e+06 825000 1.38981e+07 4.125e+06 1.53418e+07 2.475e+06 3.53163e+07 2.475e+06 3.67601e+07 AND the following RHS # Value 0 0 1 20 2 0 3 0 4 0 5 0 6 0 7 20 Solve for the following u1x u1y u2x u2y u3x u3y u4x u4y (xy) displacements at 4 nodes The matrix is correct because I used the example from the book First Course to Finite Elements and coefficients are correct. Only the *LARGE* numbers (i.e. 10^9) are not sure  because I use penalty method  not like in the book. My question is:  Does the matrix looks right in term of penalty method? Because node 0 and 1 is constrained, so the large numbers are on these, therefore node 0 and 1 should be ZERO? Libmesh gave me very weird answers though!  I use "u" and "v" ( 2 variables) in a LinearImplicitSystem with FIRST order appx, to make the DOF became "8" (because eventhough it's 2D, libmesh still provide DOF of 4 !!!! (for a QUAD4 element)). I think this might be a problem! Does system>solve() just simply inverse the matrix and multiply to the RHS?? Does the "2 variables" affect the "solve()" in any way? Many thanks guys. I really appreciate this if you could help. Regards, Ha Le  Intern at University of Nebraska  Lincoln COHESIVE project Lincoln, NE Phone: 402  613  9268  Undergraduate B.E. Robotics and Mechatronics B.S. Computer Science (Software Engineering) Swinburne University of Technology, Melbourne, Australia PH 0433 595 307 