6.2. WEM efficiency
The advantage of using the present methodology is that it requires less computational resources to achieve these positive results, compared to alternative numerical schemes and/or flow impedance assumptions, which makes it interesting for many practical applications. To demonstrate this, the accuracy of the WEM is compared to a finite elementmethod(FEM) solution(computed using the commercial solver COMSOL Multiphysics) with linear shape functions. Using the same domain as the GIT, this time with hard walls, and varying the frequency, the error relative to the exact reference solution was determined and is shown in Fig. 13. Included in this plot is the WEM error for different numbers of wave vectors, i.e. varying N in the discetisation. As can be seen, the relative error is considerable lower for the WEM for a given number of points per wavelength. The WEM requires roughly half number of points per wavelength for the same level of accuracy. This translates into a significant saving in computational resources as the size of the overall problem to be solved – the size of the overall stiff matrix ndof to be inverted – will be 1=2 the size, where D is the number of spatial dimensions.