The accuracy of the EddyCentre advanced models is controlled by the number of elements into which the defect volume is divided and the precision to which the matrix coefficients and field are computed. The EddyCentre Inspection tool will determine appropriate defect volume divisions and precision values automatically when you choose Low, Medium or High precision properties, but you are free to set them yourself by choosing the Advanced property setting. To find out more see the Inspection tool's channel control documentation.
The major influence on the accuracy, problem size and solution time is the number of defect elements in the defect volume. The number of unknowns (degrees of freedom) is
nu = 3 x nw x nl x nd,
where nw, nl and nd are the number of cells across the width, length and depth of the defect volume, respectively. For cylindrical geometries: depth is in the radial direction, length in the axial direction and width is around the circumference. The factor of 3 occurs because of the 3 components of the field in each cell. The system of equations (matrix) which results is nu x nu. As you can see you can run out of memory if the number of elements is not controlled.
The dimensions of each cell need to be of the order of local skin-depth of the region, which of course is controlled by the conductivity and frequency of the scan (remember, the defect region must have a relative permeability of one). The Inspection tool determines the number of elements (nxi x nyj x nz), based on the following algorithm, where k is the skin-depth, dm is the smallest cell dimension in each of the 3 coordinate directions and di represents either of the remaining cell dimensions:
Therefore, all cell dimensions are of the order of a skin-depth, with a maximum of 2 skin-depths and a minimum of 1/3 of a skin-depth.
You are free to determine the number of cells in each or all directions yourself by choosing the Advanced precision level; however, care must be taken not to be too coarse, which could cause the EddyCentre models to provide inaccurate results, or too fine, which could cause a high demand on system resources..
The aspect ratio, which is the ratio of the largest cell dimension to the smallest, has a significant influence on time needed to solve for the matrix coefficients. The higher the aspect ratio the longer the solution time. However there is no accuracy trade-off here, it just takes longer to solve for the matrix coefficients. In some instances, a shorter overall solution time for a scan can be achieved by decreasing the aspect ratio and increasing the number of unknowns, thus the matrix solution time is shortened, but the time used for each scan point is increased.
If you do not have enough memory to hold the entire matrix and the fields your
system will start to thrash, continuously swapping parts of the matrix
in and out of memory, which will dramatically increase the solution time. You
may have to decrease the frequency, precision level, or defect size to obtain
a solution. If you are consistently thrashing or cannot run a particular Inspection
File, you may have to increase your virtual memory setting or buy more memory
for your machine.
|
|