Evaluation of inverse algorithms in the analysis of magnetic flux leakage data

We evaluate the use of linear and nonlinear inverse algorithms (maximum entropy method, low resolution electromagnetic tomography, L/sub 1/ and L/sub 2/ norm methods) in the analysis of magnetic flux leakage (MFL) measurements commonly used for the detection of flaws and irregularities in gas and oil pipelines. We employed MFL data from a pipe with well-defined artificial surface breaking flaws at the internal and external wall. Except for the low-resolution electromagnetic tomography, all algorithms show, on average, similar accuracy in the flaw extent estimation. Maximum entropy and the L/sub 1/ norm have a tendency to yield better results for smaller flaws, while the L/sub 2/ norm performs slightly better for larger flaws. The errors of the flaw location estimation are comparable for the maximum entropy and the L/sub 2/ norm algorithm. The L/sub 1/ norm performs worse for those flaws situated on the internal pipe wall. Linear methods (L/sub 2/ norm) are easier to implement and require less computation time than nonlinear methods (maximum entropy method, L/sub 1/ norm). In conclusion, inverse algorithms potentially provide a powerful means for the detection and characterization of flaws in MFL data.