HW Problem 1: Summary of Martin J. Gander and Gerhard Wanner’s paper “From Euler, Ritz, and Galerkin to Modern Computing*” section 3 and 4: In engineering there are many problems modeled in partial differential equations, however many of these differential equations cannot be solved in closed form solution. Approximations are generally needed to solve these problems and are referred to as the weak or variational form of the solution. This requires introducing arbitrary test or approximate functions into the differential equation and integrating by parts. Good choice in arbitrary functions will result in more accurate solutions. In general the more terms introduced in the sum the more accurate the results will be. The variational form can be then taken by introducing boundary conditions such as the dirichlet condition. The deformation of an elastic plate was a very difficult problem to solve and was first considered by Sophie Germaine and through the years corrections and improvements were made. However it wasn’t until 1850 that a breakthrough was achieved by Kirchhoff. Kirchhoff formed the double partial differential equation of the form below. To obtain a unique solution to the above problem suitable boundary conditions are necessary. These boundary conditions are listed below: where ῼ denotes the domain range. Ritz proposed solving the elastic plate problem by converting the above differential equation into variational form. This was accomplished by using the “wie man ohne weiteres einsieht” procedure. Once done, the following formulation is obtained: The next step is to choose a sequence of functions to approximate the solution to the above formulation as a linear combination. Coefficients a1,a2..,am will need to be determined. The quality of the solution will depend on the choice of ψi(x, y). ψi(x, y) should generally take into account the boundary conditions of the problem being analyzed. After the ψi(x, y) is obtained it can be reinserted into the J formulation. This will result in a finite -dimensional expression in a1,a2..am. Next J will be differentiated to obtain when J is minimum. This will result in a linear system which may be tedious to solve by hand but easy with today’s computers. The coefficients a1,a2..am are obtained as a result of this linear system. It should be noted that as m approaches infinity the error of a will be reduced. HW Problem 2: Galerkin Finite Element Method is a technique to solve approximate solutions to these problems. The steps to performing the Galerkin Finite Element Method are as follows: 1. Variational Weak Formulation 2. Finite Mesh and Basis Functions 3. Galerkin Method 4. Solve system of algebraic equations 5. quality of approximate solution Engineering problems are generally represented in differential form. To use the Galerkin Finite Element Method the differential equation will need to be written in its variational weak form. To obtain a test function v(x) will be needed. v(x) shall lie inside realm of [0,1] such that the second derivative of function u(x) with respect to x will equal a known function f. f is generally from given data. To solve the differential equation for the unique problem a set of boundary conditions will be necessary. These boundary conditions are generally in the form of Neuman or Dirichlet B.C.s. Neumann B.C.s are those that have the differential of u(x) =0. While Dirichlet will have some given value of x in u(x) equal to zero. After introducing the test function v(x) it will be necessary multiply both RHS and LHS of the differential equation by v(x). Generally the double differential will be written as a different letter or different notation such that it is now a single differential. This simplification will allow for the process of integration by parts (LHS only). After the simplification the BCs will be introduced and the equation will be simplified. It should be noted that BCs of Dirichlet type will require v(x) to be zero. The result is the weak formulation. The next step is the mesh step where the domain will be subdivided into continuous elements. Where elements meet will be called nodes. The next sub-step is introducing the basis functions, generally in the form of Lagrange Polynomials. These are interpolating polynomials to obtain basis or weight functions. Each element will contain local finite element basis functions that will be used for the global basis functions of the nodes. These functions are in the form of roof functions. The third step is to introduce the Galerkin Method. While u(x) is desired, an approximate solution uh(x) will be obtained. The test function v(x), will be represented as the summation of the weighted nodal functions and uh(x) will be the summation of the product of ui and vi(x). These equations form a collection of terms in a system of equations that can be solved with linear algebra. The system of equations is of the form KU=F, where K is the stiffness, U the desired solution, and F the forcing function. Once the desired solution is obtained the validity of the solution can be obtained by comparing the approximate solution with known data. This will give the error. HW Problem 3. 𝑑2 𝑢 = −𝑢′′ = 𝑓(𝑥) 𝑑𝑥 2 𝑢′ (0) = ℎ(𝑥) 𝑢(1) = 𝑔(𝑥) 𝑣(1) = 0 (𝑑𝑖𝑟𝑒𝑐ℎ𝑙𝑒𝑡) 1 1 ∫ 𝑢′ 𝑣 ′ 𝑑𝑥 − {𝑣(1)𝑢′ (1) − 𝑣(0)𝑢′ (0)} = ∫ 𝑓(𝑥)𝑣(𝑥)𝑑𝑥 0 0 1 1 ∫0 𝑢′ 𝑣 ′ 𝑑𝑥 − {𝑣(1)𝑢′ (1) − 𝑣(0)𝑢′ (0)} = ∫0 𝑓(𝑥)𝑣(𝑥)𝑑𝑥 : direchlet condition 1 1 ∫ 𝑢′ 𝑣 ′ 𝑑𝑥 − {𝑣(0)ℎ(𝑥)} = ∫ 𝑓(𝑥)𝑣(𝑥)𝑑𝑥 0 0 HW Problem 4. 𝛻 2 𝑢 = 𝑓(𝑥, 𝑦) 𝑢=0 𝛻2𝑢 = 𝜕2𝑢 𝜕2𝑢 + 𝜕𝑥 2 𝜕𝑦 2 𝑏 𝑏 𝑏 𝑏 𝑏 𝑏 𝑏 𝑏 𝑏 𝑏 ∫𝑎 ∫𝑎 ∇𝑢∇𝑣𝑑𝑥𝑑𝑦- {𝑣(1)𝑢′ (1) − 𝑣(0)𝑢′ (0)}=∫𝑎 ∫𝑎 𝑓(𝑥, 𝑦)𝑣(𝑥, 𝑦)𝑑𝑥𝑑𝑦 ∫𝑎 ∫𝑎 ∇𝑢∇𝑣𝑑𝑥𝑑𝑦- {𝑣(1)𝑢′ (1) − 𝑣(0)𝑢′ (0)}=∫𝑎 ∫𝑎 𝑓(𝑥, 𝑦)𝑣(𝑥, 𝑦)𝑑𝑥𝑑𝑦 : direchlet condition 𝑏 𝑏 ∫𝑎 ∫𝑎 ∇𝑢∇𝑣𝑑𝑥𝑑𝑦= ∫𝑎 ∫𝑎 𝑓(𝑥, 𝑦)𝑣(𝑥, 𝑦)𝑑𝑥𝑑𝑦 b) Energy functional 𝑏 𝑏 1 𝐽 = ∫ ∫ ( ( ∇𝑢)2 − 𝑓 ∗ 𝑢)dxdy − −>> minimum 𝑎 𝑎 2