Method for determining the geometry of a defect and for determining a load limit

A method for determining a geometry of a real, examined defect in a magnetizable object in the form of a pipeline or tank for oil, gas or water that is subject to pressure loading during operation in order to accurately calculate a load limit of the magnetizable object, said method comprising the following steps executed in an electronic data processing (EDP) unit, said method comprising:

• receiving a first reference data record of the magnetizable object produced by a first magnetic flux leakage (MFL) measurement of a region of the magnetizable object including the real, examined defect;
• receiving a second reference data record of the magnetizable object produced by a second MFL measurement of the region, said second reference data record being linearly independent of the first reference data record;
• at least partially representing the region of the magnetizable object on a three-dimensional object grid;
• determining an initial defect geometry as a starting defect geometry on the three-dimensional object grid or an at least two-dimensional defect grid;
• for the first reference data record, determining a first MFL prediction data record as a first starting prediction data record on the basis of the starting defect geometry and parameters of the first MFL measurement;
• for the second reference data record, determining a second MFL prediction data record as a second starting prediction data record on the basis of the starting defect geometry and parameters of the second MFL measurement;
• iteratively adapting the starting defect geometry to the geometry of the real, examined defect by running multiple expert routines in competition with one another, each of said multiple expert routines executing at least one algorithm for adapting the starting defect geometry that differs from algorithms of the other of the multiple expert routines, each of said multiple expert routines producing an expert defect geometry that differs from the starting defect geometry;
• using the expert defect geometry to determine a first expert prediction data record by simulating an MFL measurement of the expert defect geometry or assigning an MFL data record corresponding to the expert defect geometry taking into account the parameters of the first MFL measurement;
• using the expert defect geometry to determine a second expert prediction data record by simulating an MFL measurement of the expert defect geometry or assigning an MFL data record corresponding to the expert defect geometry taking into account the parameters of the second MFL measurement;
• comparing the first expert prediction data record to the first reference data record and the first starting prediction data record;
• comparing the second expert prediction data record to the second reference data record and the second starting prediction data record;
• making the expert defect geometry produced by one of the multiple expert routines available as a new starting defect geometry for iterative adapting only when the first expert prediction data record is more similar to the first reference data record than the first starting prediction data record and the second expert prediction data record is more similar to the second reference data record than the second starting prediction data record or a fitness function taking into consideration both expert prediction data records is improved;
• repeating said steps of iteratively adapting and comparing until a stop criterion is met;
• and
• after the stop criterion is met, producing a data record from the new starting defect geometry useable in a calculation of the load limit.