Linear Algebra and Its Applications · Problem 6.5.5

Computations with Matrices — Gilbert Strang

Mathematics · galerkin-method

Problem

Galerkin's method starts with the differential equation (say −u′′=f(x)-u'' = f(x)) instead of the energy PP. The trial solution is still u=y1V1+y2V2+⋯+ynVnu = y_1 V_1 + y_2 V_2 + \cdots + y_n V_n, and the yy's are chosen to make the difference between −u′′-u'' and ff orthogonal to every VjV_j: Galerkin∫(−y1V1′′−y2V2′′−⋯−ynVn′′)Vjdx=∫f(x)Vj(x)dx.\textbf{Galerkin} \qquad \int (-y_1 V_1'' - y_2 V_2'' - \cdots - y_n V_n'') V_j dx = \int f(x) V_j(x) dx. Integrate the left side by parts to reach Ay=fAy = f, proving that Galerkin gives the same AA and ff as Rayleigh-Ritz for symmetric problems.

Browse the problem library