Nondestructive Testing Problems with Mesh Layer Feature

Infolytica's products have been used to generate reliable nondestructive evaluation (NDE/NDT) results for years. Our new Mesh Layers feature makes these tasks easier to set up and faster to solve.

Often, an NDT problem involves the excitation of the material under test using one or more coils, and the detection of a change in impedance of these coils in the presence of a defect in the workpiece (the exciter and detector need not necessarily be the same).

There is a need to model the magnetic fields and eddy currents accurately in the regions under inspection. The fields decay rather quickly to zero in the surrounding regions. Further, the fields tend to change very quickly (exponential decay) with increasing depth into the test specimen (skin depth effects), and not nearly as rapidly in the two perpendicular directions.

The Mesh Layers feature allows the construction of a mesh that is optimized for these types of problems. This feature allows for anisotropic elements, which produce the best modeling of the fields, prevents "bleeding" of mesh refinement into areas where it is not required, and involves much less setup work than competitive techniques.

Results

Click image to see full size version

Starting from the WFNDEC Eddy Current Benchmark #2, the region directly below the coil was remeshed with the Mesh Layers feature. All other mesh refinements were removed. The resulting mesh had only 42,274 tetrahedra, compared with the 243,068 tetrahedra generated by the previous mesh refinement techniques (nearly 6X smaller). The image to the right shows the initial attempt at mesh refinement using the standard controls that Infolytica provides. Note the refinement in the coil and the "bleeding" of the refinement into the tube region.

Click image to see full size version

As shown to the right, the layered meshing concentrates many elements along the depth of the tube (in the direction where the fields change fastest). It is also clear that this refinement has not "spread" into the surrounding air and tube regions as it did in the previous image showing the initial mesh used. The mesh is only refined in the region of interest (the section of pipe located directly under the coil). Third-order elements must be used to obtain accurate eddy current modeling.

Click image to see full size version

The reduced mesh complexity produced a 7X speed-up in solution time. After solving the problem, it was important to verify that the same accuracy of solution was obtained. The graph at right shows that this solution is comparable in accuracy to the original, which took 7X longer and 3X more peak memory.

Click image to see full size version

The Mesh Layers feature allows the creation of a much smaller mesh, which is ideally suited to problems of this type. The problem is easier to set up (only one parameter is changed to refine the mesh as shown). The result is a faster solution time, less memory requirements, a smaller file size, all without any loss of solution accuracy.

Problems solved: 22
CPU: AMD Athlon 3500+ 2.20 GHz.
Peak RAM used: 22.3 MB (vs. 72 MB)
Average number of unknowns: 345,000 (vs. 4,620,000)
Average number of tetrahedra: 42,000 (vs. 694,000)
Solving time: 37 minutes 57 seconds (vs. 4 hours, 9 minutes)