EP3692362A1

EP3692362A1 - Method for determining the geometry of a defect and for determining a load limit

Info

Publication number: EP3692362A1
Application number: EP18782929.6A
Authority: EP
Inventors: Andrey Danilov; Johannes PALMER
Original assignee: Rosen Swiss AG
Current assignee: Rosen Ip Ag
Priority date: 2017-10-06
Filing date: 2018-09-28
Publication date: 2020-08-12
Also published as: UA126485C2; US11624728B2; US20200340948A1; WO2019068588A1; AU2018344386A1; US20230258599A1; CN111448453A; EP3467489A1; CA3109990A1; RU2020114878A; RU2763344C2; RU2020114878A3; MX2020003580A

Abstract

The invention relates to a method for determining the geometry of a plurality of defect of a magnetizable object with a reference data set of the object, comprising determining an initial defect geometry as an output defect geometry, determining a first MFL prediction data set as an output prediction data set on the basis of the output defect geometry, and iteratively adapting the output defect geometry to the geometry of the real defect(s) by means of the EDP unit and by means of a plurality of expert routines (11) running in competition and preferably parallel to one another.

Description

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

The present invention relates to a method for determining the geometry of a defect and to a method for determining a load limit of an object that is at least loaded during operation.

One of the key tasks of pipeline inspections, particularly with so-called smart pigs, is to predict safe operating conditions resulting from the condition of the pipeline. In particular, pipeline operators are interested in the condition of any welds and the number and size of defects. Defects are, for example, areas with metal losses due to corrosion, cracks or other weakenings of a wall of an object provided in particular for storage or for the transport of liquid or gaseous media. These include, for example, pipes, pipelines or tanks.

Knowing the pipeline's maximum burst pressure at which the pipeline will be destroyed is relevant to the pipeline's adjustable operating pressures. Accordingly, the accurate prediction of this pressure is important. Currently, for the calculation of this limit, a defect is approximated only in terms of its length, width and depth and thus as a box considered. However, especially for metal losses due to corrosion on the outside or inside of a pipeline, this conservative approach is disadvantageous because the simplified geometric figures necessarily overestimate the current structure of the defect. This leads to the underestimation of the maximum burst pressure of the object and thus to the underestimation of the permitted operating pressures. However, an object that can be operated with higher pressure, such as a pipeline or a gas tank, can be operated much more economically.

It is state of the art for the measurement of corrosion of an object to have scans of magnetic flux data (MFL data) due to magnetic flux leakage measurements (MFL measurements) evaluated to determine the size of (corrosion) defects of specially trained persons. The signals displayed in the scans are parameterized in boxes and evaluated. On the one hand, the assumptions necessary for this evaluation, which is also known as sizing, are proprietary. On the other hand, the interpretation of the measurement results is strongly influenced by the experience of the evaluating persons. Ultimately, the quality of the forecasts can only be ensured through on-site pipeline inspections. This, in turn, is associated with high costs for the operators. The most widely used industry standard API 1 163 describes the adverse effects of this simplified approach. It is well documented that the quality of this approach depends heavily on the knowledge of the persons viewing it. Practical approaches are always an interpretation of the data obtained by an MFL measurement, influenced by subjective factors. In addition, from the scientific, ie theoretical view, approaches are known in which, via successive variation and iterative methods of an assumed flaw geometry, the most accurate simulation of the measured signals over forward models is to be achieved. Here, for example, neural networks are used. Theoretically, these approaches may provide solutions in terms of resulting hole geometry, but these solutions are not necessarily realistic. This is especially true for complex datasets, which lead to unexpected, exotic and wrong solutions under certain inverse problems. While in well-defined and delimited test scenarios such scientific models provide a solution to the described problem in the form of flaw geometries, this has not yet been successful for real measurement data that has a variety of perturbations.

It is therefore an object of the present invention to demonstrate a robust way of accurately reconstructing a real defect from its MFL measurement data set and to carry out the most accurate possible calculation of the load limit of a corrosion-prone object.

The object is achieved by a method according to claim 1 and with respect to the determination of a load limit by a method according to claim 15. Advantageous embodiments of the invention are described in the dependent claims to these claims and the following description. According to the invention, the determination of the geometry of one or more defects by means of a plurality of competing expert routines with in each case at least one own search strategy or at least one own algorithm, in particular, be carried out in parallel on a computerized unit.

Such a method according to the invention for determining the geometry of one or more real, examined defects of a magnetizable object, in particular a pipe or a tank, with a reference data set of the object generated on the basis of one or more MFL measurements preferably comprises an at least partial representation of the object by means of a computer unit, in particular on or by an at least three-dimensional object grid, and further comprises a determination, in particular a generation, of an initial defect geometry as initial defect geometry, in particular on the object grid or on an at least two-dimensional defect grid,

a determination, in particular a generation, of a first MFL prediction dataset as output prediction dataset on the basis of the output missile geometry, in particular by simulation of an MFL measurement or assignment of an MFL dataset,

and an iterative adaptation of the initial defect geometry to the geometry of the real defect or defects by means of the EDP unit, wherein this adaptation is carried out by means of several competing and preferably parallel to each other running expert routines, wherein in respective expert routines by means of at least one own algorithm or an own search strategy and Based on the original missile geometry, a respective expert defect geometry is generated.

The expert routine has its own algorithm when at least one of the algorithms available to the expert routine for adjusting the defect geometry at least partially differs from the algorithms of another expert routine. Preferably, stochastic processes can be used to differentiate the algorithms of different expert routines. Each expert routine has at least one algorithm for adapting the flaw geometry, preferably at least one expert routine has several algorithms available. Likewise, within an expert routine, the selection of an algorithm based on stochastic processes can be carried out or specified.

According to the invention, it is further provided that a respective expert prediction data set, in particular by simulation of an MFL measurement or assignment of an M FL data set, is determined on the basis of the respective expert defect geometry, the expert defect geometry on which the respective expert prediction data set is based at least one, preferably more and in particular all of the expert routines are provided as new source miss geometry for further adaptation to the geometry of the real defect or real defects if the expert predictive data set is more similar to the reference data set than the original prediction data set. Subsequently, ie for the next comparisons of the respective expert defect geometries with the new source defect geometry, this becomes the new one Outgoing missile geometry associated expert predictive data set used as a new baseline prediction record. A measure of similarity can be formed via a fitness function. The iterative adaptation using the expert routines takes place until a stop criterion is met. The expert predictive data set and the first MFL predictive data set both show MFL fields belonging to the respectively assumed flaw geometry. These can be calculated or simulated.

The determination of the expert predictive data record can take place within the workflow of the expert routine and / or via a program module controlled separately by a monitoring routine.

The determination of the expert forecast data set on the basis of the respective expert defect geometry continues to take place in particular when no sufficiently large databases with calculated or measured MFL data for respective defect geometries are available by simulation of a MFL measurement, which will be described below , Alternatively, the expert defect geometry can also be provided with an MFL data set from a sufficiently extensive database. Also, a combined approach in which a first search is made in a database for already existing MFL data and only then a simulation is performed in the case of negative search, is possible. Overall, this can lead to a rapid determination of the respective expert forecast data record. The inventive method is completely and in particular automated on a computer unit, which may optionally consist of several computers performed. The associated computation program can be a single program or it can be a program package comprising a plurality of program modules which, for example, run on resource-dependent distributed computer systems or subunits and can be stored there on respective EDP media. In particular, a computer has the typical means of a data processing unit, such as one or more processors, at least temporary memory (RAM), data communication means, display and / or input units. While the selection of the reference data record can preferably be user-controlled, the determination of the defect geometry is performed automatically during the iterations. Preferably, prior to the actual iteration, program parameters for selecting the algorithms available to the expert routines, determining an initial defect geometry, determining the first prediction dataset and / or an expert prediction dataset, each showing MFL fields, can be defined. For example, it can be determined whether the determination of the initial prediction data set is to be created via a simulation of an MFL measurement on the basis of a grid representing the object with the defect or to be loaded from a database via a regression. In particular, in the case of the simulation of an MFL field, the parameters necessary for comparison with the reference data record, such as, for example, the direction of the magnetization, the strength of the magnetization, the distance of the sensor from the object surface and / or the speed of the measuring device, can be defined. By competing with each other for resources of the computer unit expert routines, which seek their own search strategies each own solutions for determining the geometry of the real defect, in particular the existing in the theoretical approaches of the prior art problem is avoided that isolated solutions being found. Compared with the manual evaluation of the data sets, the solution found is not only much better but also more comprehensible and documented. Singular solutions, in which an algorithm of any kind gets stuck and nevertheless the defect geometry is not reproduced realistically, are avoided in this way.

In general, a representation of the object on or through a three-dimensional grid will be necessary for a simulation of the leakage flux data associated with a geometry. In this case, the representation takes place at least partially in the sense that at least the part of the object with the defect or the imperfections and preferably adjacent regions are represented by or on the object grid. Alternatively, the determination of the leakage flux data can also take place via a database query, for example by means of a regression function.

Although the method according to the invention can be carried out simultaneously for the determination of one or more defects within a reference data record, in the following, for the sake of simplicity, in most cases only a defect will be referred to. The defect geometry can be assigned to the MFL simulation in a skilful choice of the representation of the object, in particular by or on the at least three-dimensional object grid, the grid elements or grid points as a value. Depending on the geometry of the respective grids, interpolation or lattice adjustments may be necessary.

In particular, the flaw geometries that are the basis for determining the associated (MFL) and predictive prediction data sets, however, are defined by flawed depths representing, for example, the depth of corrosion of an object surface on the grid nodes of a two-dimensional defect grid. Due to the two-dimensional representation of the defects, the expert routines can work much faster than if the adaptation of the defect geometry is carried out on a three-dimensional grid.

For the simulation of the MFL measurement of a new defect geometry, the particular two-dimensional defect grid is preferably interpolated onto the grid points of the object grid, wherein the surface of the object to be displayed is adapted to the defect depths of the defect geometry. The simulation is then calculated on the particular three-dimensional object grid. Alternatively, the stray-flux simulation can also be performed on a two-dimensional grid or by means of a regression model that builds on a database of MFL data sets based on finite element method simulations and / or MFL measurements. On the basis of the initial defect geometry, which is obtained or specified, for example, via a look-up table, a database comparison or, in particular, a one-time execution of an expert routine, a first MFL forecast data record is determined as an initial prediction data record, in particular by simulation of an MFL measurement. The simulation of the MFL measurement is carried out, for example, by means of a finite element model as a forward calculation. In the simulation of the leakage flux measurement, the necessary parameters for this are determined according to the real measurement. This applies in particular to the magnetization direction, the magnetic field strength and / or the distance of the sensors above the surface of the object. Based on the initial defect geometry, an output prediction data set is then obtained as a simulated leakage flux measurement. This dataset could already be compared with the object's reference dataset, but this does not usually lead to meaningful solutions at the beginning of the iteration.

A separate routine for establishing an initial defect geometry as the output missile geometry is not essential, but reduces the computation time required in the subsequent program operations. Alternatively, the initial defect geometry may already be the result of a pass of an expert routine. The initial defect geometry can also be predetermined alternatively, for example by a completely flat, so to speak defect-free geometry. The initial defect geometry is taken as the initial defect geometry in the iterative approximation process of the competing expert routines. The expert routines themselves are, for example, independent of one another as separate program modules without direct interaction with one another and can be equipped with resources, in particular with computing time, depending on a monitoring routine or a main module.

Based on the expert defect geometry developed in a respective expert module, expert predictive data sets are again determined for each of these geometries in particular by simulation of an MFL measurement. Thus, each expert defect geometry, which is present in particular as a 2D data set of depth values of the defect, generates an expert forecast data set as a simulated MFL measurement. The simulation of the MFL fields on the basis of the respective expert defect geometries is carried out according to the above-described calculation of the output prediction data set. In particular, the calculations are performed in parallel based on the respective expert defect geometries. This can be accompanied by the construction of a database in which the leakage flux fields belonging to respective defects are stored with the aim of being able to save computing time later and for other similar data.

Before generating the expert prediction data set, it may be advantageous to adapt, in particular partially refine, the underlying grid, in particular the defect grid, possibly also the object grid, for the calculation of the expert defect geometry. For this purpose, in particular mesh morphing techniques are used, in which the object or defect grating is refined by lattice point shifting and / or division, in particular in regions of strong gradients, in order to allow a more accurate determination of the geometry or subsequently a more accurate simulation. In other areas with weaker gradients, the grating may become coarser to save computation time. Thus, the grid used is automatically adjusted for optimal evaluation of the defect geometry. At the same time a significant reduction of the number of unknowns is achieved, so that in turn calculation time is saved.

If the comparison between the reference data set and the expert predictive data set results in an expert routine being closer to the reference data set than the previous initial prediction data set, the associated expert miss geometry will be available as output missile geometry for the further as well as the corresponding expert routine posed. Starting from this solution, the further expert routines can then start from this geometry in a next iteration step, unless, for example, they have found a better solution during their own defect geometry determination, which is then available to further or all expert routines is provided.

In the case of the competitor routines that are in competition with one another, those with regard to the available resources of the EDP unit which are more successful in approaching as described below are preferably preferred to the real metrics are as other competing expert routines. The resources of the EDP unit are in particular the CPU or GPU time and / or the prioritization in the memory allocation.

Advantageously, the expert routines (on the EDP unit) compete in such a way that the distribution of the resources of the EDP unit, in particular in the form of computing time, to a respective expert routine depending on a success rate, in particular the number the output missile geometry calculated by the expert routine and provided for one or more other expert routines is taken into account, and / or depending on a reduction of a fitness function, in particular taking into account the number of expert predictive data sets generated for the reduction , The competition of the expert routines results in particular from the fact that the respective expert routines are then assigned more resources, in particular in the form of computing time, preferably CPU or GPU time, from the program part designed as a monitoring routine, if these are more successful than other expert routines. routines. An expert routine is successful if it has found a defect geometry provided with a more appropriate, for example, simulated stray field measurement to the reference data set, which is made available to the other expert routines.

This may, for example, show that individual, particularly successful expert routines account for more than 50% of the total available computing time obtained, which significantly reduces the overall duration of the process according to the invention. At the same time, it can be specified by the program that none or individual of the expert routines do not go below a certain percentage of computing time in order to avoid the problem of singular and exotic defect geometry or results from the individual routines. Thus, in the event that a hitherto successful expert routine finds only a local and not a global solution, a way out of the otherwise occurring in the prior art blockade situation can be found.

The adaptation by means of the expert routines takes place until a stop criterion is met. For example, this is a residual difference in the measured and simulated measurement data. It may also be an external stop criterion, for example, based on the available computing time or a particular specifiable number of iterations or a particular predetermined or predetermined or calculated from the available computing time computing time.

It has been found that the accuracy of the defect determination by the inventive method is qualitatively improved. A resulting calculation of the maximum load capacity shows that, for example, pipelines can be operated much longer. The accuracy of the defect detection is significantly increased. It is possible to obtain maximum operating pressures from the simulated defect geometry according to the method described above and described below Up to 50% higher, significantly reducing maintenance and maintenance costs for the operation of the pipeline and its operators. For the first time, an adequate determination of the ASME B31 G-2012 level 2 approach ("river bottom profile") for the "remaining strength algorithm" can now be realized for M FL data sets.

Preferably, a distribution of the resources of the computer unit, in particular in the form of CPU time to a respective expert routine depending on the number of provided by this expert routine for all expert routines output error geometries done. This may be, for example, a number of slots for calculating the expert forecast data sets in the form of simulated M FL data sets, the number of processor cores or the like working in parallel on the arithmetic task. Moreover, it can be implemented within the framework of the method according to the invention Computer program product can be provided that this adapts to the existing resources in the computer units in the form of processor cores, memory space, memory architecture, graphics cards, etc. By prioritizing particularly preferred expert routines and their algorithms, the recognition of the real defect geometry becomes significantly faster.

In order to further minimize the problem of singular, local solutions, it is additionally provided, in addition to determine the geometry of the defect or defects, to have a linearly independent from the first reference data set, to use another reference data set. The data sets are then linearly independent when generated by MFL measurements with magnetizations of the object angled to each other. The magnetizations are then angled to one another if the respective average induced magnetic field strengths in the investigated region are not parallel or congruent. In particular, the angle is between 40 ° and 140 °, preferably between 80 ° and 100 ° and particularly preferably at 90 °. On the basis of the output error location geometry, a further output prediction data set is determined, in particular by means of a further MFL simulation which takes account of the linear independence, ie especially the different magnetizations, and an expert error geometry is used as output error location geometry only if the corresponding expert data determined for both independent magnetizations Predictive data sets are more similar to the respective reference data sets than the output prediction data sets intended for the two magnetizations and / or a fitness function which takes into account both expert prediction data records. Due to the parallel processing of the two linearly independent data sets and the use of an output missile geometry whose simulated measurement data must be better overall with regard to a similarity or a fitness function, the danger of singularities is further reduced. At the same time, the quality of the output forecast data sets available for all expert routines improves. The number of iterations can thus be further reduced. In particular, the first reference data set is generated via an MFL measurement with axial magnetization and the second reference data record is produced via an MFL measurement with magnetization in the circumferential direction of the tube. In this case, the respective magnetizations of the pipe or of an object are at right angles to each other, so that a maximum information content can be obtained from the magnetic flux measurements, which is fully available through the simultaneous consideration of the corresponding reference data sets and the simulated expert forecast data sets during the calculation. Taking into account the above, the method steps described below proceed analogously when using two reference data sets resulting from linearly independent magnetizations.

The use of initial and / or expert predictive data sets based on a forward model to simulate MFL measurements, in particular using a finite element model, quickly performs the MFL measurement simulations. The simulation of the leakage flux measurements on the basis of the expert defect geometries can be realized by means of a separate program module, which in particular is controlled and / or monitored by a monitoring routine, called by the individual expert routines separately. It can also be several modules that are distributed over individual computer units of a respective expert routine provided.

Advantageously, the initial defect geometry is determined by means of a look-up table, by one of the expert routines and / or by a machine Learning algorithm generates, which as described above, the overall computing time improves, especially if in addition there is already a grid adjustment.

In particular, the refinement of the object and / or the defect grid can take place in the areas in which the depth of the simulated defect or defects exceeds a threshold value, wherein this threshold value can be predefinable, so that only gradients above a certain size result in a change of the grid to lead. With such a refinement, in particular the total number of gradients of a new expert defect geometry can be taken into account in order to achieve a balance between the adaptation of the respective grid, in particular of the object grid, and the subsequent arithmetic operations.

The refinement of the grid with the goal of a reduction in computing time can be carried out both on the basis of an initial reference data record and before the calculation of the expert forecast data record. Again, it is possible to provide a separate program module or individual sub-modules of the respective expert routines.

In particular, the refinement of the object and / or defect gratings particularly advantageously reduces the required CPU time by means of grid point shifting and / or division by significantly reducing the number of independent variables which must be used in the forward algorithm for simulating the MFL measurement. A lattice point shift can also be used to fit object or defect grids. Preferably, a fitness function is used as a measure for the similarity of the expert prediction and reference data records in order to effect a comparison of the simulated and measured data records on the basis of standard routines and correspondingly fast, ie with saving of computing time.

In particular, the output missile geometry is stored in the form of a two-dimensional data set or a pointer referring thereto in a memory area of the computerized unit accessible to all expert routines. In particular, this memory area is again under the control of a monitoring routine, so that prioritization of individual expert routines can also be undertaken in this regard.

Instead of using initial output geometry at the beginning of a new iteration, which is stored, for example, in a central memory area accessible to all expert routines, at least one expert routine at the beginning of a new iteration can waive assumption of source miss geometry Customize your own expert defect geometry. For this purpose, an expert routine can have a function rule in which, for example, depending on search strategies used in other expert routines, an opposing strategy is specifically selected. In such a case, the expert routines can influence indirectly. Such a procedure may be advantageous, in particular, if it is established that one has hitherto been the case always successful routine preferred an unrealistic solution. This can be recognized for example on the basis of impermissible values with regard to the depth of a defect. If an expert routine, which does not take over the output missile geometry, does not provide improved solutions, it is automatically down-prioritized, so that it is provided with increasingly less computation time.

As a stop criterion or also as a convergence criterion, it is possible to assume, preferably, a change in the initial defect geometry or, more generally, the geometry of the defect and / or the object lattice that is missing after a plurality of iterations and to be designated as substantial. The solution found so far is then the best. Preferably, the stop criterion is chosen such that the observed variations in the leakage flux simulation that would result in the refinement of the object or defect lattice are substantial, eg, with a factor less than 10, in particular 2, below the variations resulting from the individual measurement dispersion , in particular the measurement error, and which are individually specified, for example on the basis of so-called "Essential Variables" of the API 1 163 standard. This ensures that the accuracy of the final model is within the range specified by the measurement itself accuracy. Accordingly, a comparison of the variation of the expert forecast data set with the measurement spread of the real data set is preferably used as the stop criterion. The one stop of the program and in particular an output or storage of the hitherto calculated output error location geometry causing stop criterion can preferably be specified by pre-adjustable program parameters. In particular, several algorithms for adapting the expert defect geometry are available to an expert routine. These may be approaches in the field of machine learning, stochastic optimization, empirical and / or numerical model functions. In particular, experience values of evaluating persons can also be utilized in the expert routines. Preferably, in an expert routine, the different algorithms are either randomized or selected by a selection function. Thus, a sufficiently diverse approach is created, with which all solutions can be considered purposefully and under competitive conditions.

The object stated at the outset is also achieved by a method for determining a load limit of an object subjected to pressure at least in operation and configured in particular as an oil, gas or water pipeline, wherein in the method a data set described one or more flaws acts as an input data set in a forward modeling in particular trained calculation of the load limit is used, wherein the input data set is first generated according to a pre- or post-described method for determining the geometry of a defect. The advantageous representation of the flaw geometry, in particular as a non-parameterized true three-dimensional geometry or as a two-dimensional surface with respective depth values, has previously made redundant in the industry as necessary assumed simplifications, so that for this reason an increase in the accuracy of the defect detection as a whole in so far unavailable manner is ensured. Previously, the accuracy was limited to the indication of the point of the maximum depth of the defect, the entire profile is now determined with high accuracy. Typically, the accuracy of the maximum depth is reduced to the extent achievable as a function of the measurement accuracy, ie about ± 5% of the wall thickness in comparison to previously about ± 10% of the wall thickness in the sizing according to the prior art described above. However, the prediction of the load limit, depending on the geometry of the defect reaches just for critical cases in the accuracy increases from, for example, previously ± 50% to now ± 5%. The advantage according to the invention thus lies in particular in an adequate representation of the flaw geometry which has been achieved for the first time, which enables this increase in the first place.

Further advantages and details of the invention can be taken from the following description of the figures. Schematically shown:

1 shows a defect determination according to the prior art,

2 is a schematic representation of a method according to the invention,

3 is a closer explanation of a part of Fig. 2,

4 shows a comparison of a result of a method according to the invention with measured data, 5 a schematic representation of a lattice refinement as part of a method according to the invention,

Fig. 6 result of a method according to the invention.

Individual features of the embodiments described below, in combination with the features of the independent claims may also lead to further developments according to the invention.

In the prior art, the evaluation of MFL data of a tube according to FIG. 1 is carried out by means of the in particular also experience-based definition of boxes. The boxes shown in the figure have respective dimensions in length, width and depth. The x and y axes are shown in units of meters ([m]). A check of the actual flaw geometry by means of laser scanning, ie by means of a direct measurement, has shown that the maximum burst pressure of 4744.69 kPa determined on the basis of the flaw geometry assumed by the MFL data analysis is only 55.2% of that calculated on the basis of the actual geometry maxium burst pressure is. The state-of-the-art operating pressure for safe operation of the pipeline, which results from the experience-based evaluation to be 3621.29 kPa, is well below a possible safe operating pressure. In the method according to the invention, the surface of a tube is represented by a 2D mesh surface according to an embodiment. The flaw geometry can be described as a vector of depth values D that lie on a defect grid 5 (FIG. 5). This defect geometry is compared with the original defect geometry on the basis of a result of a fitness function F (D) taking into account the MFL fields belonging to the respective geometries. It is assumed that the lower the value of a fitness function, the closer the assumed error geometry to the real geometry is:

Here M is the number of data sets to be treated simultaneously (real MFL data sets), H _ca / the result of a simulation of the MFL measurement, H _m is the measured data from the MFL measurement (reference data set), and R (D) is a regula- tation term that can be stated as follows:

R (D) = \ VD \,

where α is a scaling term.

The method sequence according to the invention is described at least in sections below according to FIG. 2, wherein a plurality of the parallel and competing expert routines 1 1 is described only with a block 14. As input data sets, for example, several runs of the same MFL pipeline can be merged according to Box 2. Both sets of data 1 can be filtered in advance and matched to one another for better merging (method step 3), for example to reduce any artifacts or background noise. In addition, a further data set 4, which was generated on the basis of a linearly independent, further magnetization and was also filtered for equalization to identical grating structures, can be used, so that according to method section 6 two aligned reference data sets provided on the basis of linearly independent magnetizations Measuring runs were created exist.

Exactly matched data sets can be handled together, with the method according to the invention realizing the simultaneous treatment of the data sets by using a fitness function which takes into account the merged data sets.

In step 7, a first of the existing in step 6 reference data sets for further processing is selected. For this purpose, an initial defect geometry is initially assumed in step 8 as the output defect geometry, in particular generated in the present case, which is based, for example, on a normalized measurement signal S (x, y) / (max S). For example, the flaw geometry may be derived from a threshold function that takes into account the amplitude at grid points where the signal is greater than a certain threshold / (eg, 0.2):

maxi

The above approximation leads to a number of N defect depth values at the respective grid points:

D ± = i wt / N * G, where wt is the thickness of the wall of the tube, i is the index, which is also the value for

Determination of the defect depth value is used. For such

Flaw geometry calculates the fitness function and the profile with the lowest score is used as the input solution:

Dinit = arg min F Di)

This input solution is then made available as output missile geometry for the individual expert modules. In advance, with the aim of reducing the computing time, the number of parameter values (elements of vector D) describing the defect geometry can be kept as small as possible. This is achieved in particular via a dynamic grid adaptation. Since the number of depth values corresponds to the number of nodes in the defect grid 5, the number of nodes is also the number of defect parameters at the same time. Starting with a comparatively coarse grid, this is successively refined in relevant areas. By way of example, for a given node spacing of, for example, 14 mm, an associated grid cell size of 14 mm x 14 mm, and flaw limits of 30%, 50%, and 80% of the wall thickness, the refinement shown in FIG. 5 can be achieved in the relevant grid area Cells are successively divided, which exceed the above depth values. The lattice deformation then correlates with the assumed defect geometry, ie in regions of large gradients there is a larger number of lattice points.

On the computerized unit according to the inventive method, the flow of the workflow of a group of expert routines 1 1, which are in competition with each other, simulated. For this purpose, the program may have various modules that can set independently of each other and in particular not synchronized with each other data in certain areas of the computer unit, so that they are further processed there. This takes place in particular under the supervision of a monitoring routine 9 (FIG. 3 a). A plurality of expert routines 1 1 thus holds a number of arithmetic slots 13 in dependence on the above-defined success, ie, for example, the number of output missile geometries written in a common memory area 12, in order to generate expert defect geometries and / or associated MFL values. Simulations can be carried out or in the case of an independent MFL simulation module to perform. This corresponds to the block 14 according to FIG. 2, this being exemplary for several expert routines 11 (FIG. 3a). Starting from the individual rake slots 13 are According to the present embodiment, the MFL simulations of the individual expert defect geometries for the purpose of creating the expert forecast data sets are also performed under the supervision of the monitoring routine 9 in the simulation modules 16. The more slots 13 are available for an expert routine, the greater the proportion of computer resources for this expert routine. Preferably, the number of program modules for performing MFL simulations is equal to the number of slots. The monitoring routine 9 monitors the number of iterations and the resulting changes in the iterations

Output missile geometry and continues to monitor whether an associated stop criterion has been reached. Subsequently, the result is output in block 17.

The number of calculation slots 13 available for an expert routine 11 and the simulation routines subsequently provided can vary in such a way that a first expert routine, for example, up to 50% of the available for the calculation slots and simulation routines can use the entire computing time.

In memory area 12, the output missile geometries are stored as shown. This may be a memory area accessible to the expert routines 11. There log files of the expert routines 1 1 and monitoring routine 9 as well as instructions to the expert routines 1 1 can also be stored, which are then independently implemented by them. For example, this can be an interrupt command that is set when the stop criterion is reached. Preferably, the expert routines 11 are independent program modules which generate new expert defect geometries and place them into the simulation routines 16. Furthermore, in the expert routines 11, the fitness function described at the outset can be generated on the basis of the expert forecast data record and compared with the output prediction data record stored in the area 12. If the expert prediction data record is more similar to the reference data record or, in the case of linearly independent measurement data sets, then the data record stored in area 12 corresponding to the two reference data records, then this expert forecast data record is used as a new output prediction data record.

For example, in expert routines 11, a new defect geometry is generated randomly. For this machine learning algorithms or empirical rules can be used. Advantageously, however, the implementation of at least two basic expert routines is provided as described below in order to further improve the convergence of the solutions.

These search strategies, which are preferably always implemented in a method according to the invention, are based on an assumed probability distribution p (x, y) of grid points whose depth value results in a maximum reduction of the fitness function. The probability function is used to identify N grid points (x _n , y _n ). At each of the points considered, the depth function, which describes, for example, the depth of corrosion at the lattice site, changed to AD, with the sign of the change being randomized. The number of selected points N can also be chosen randomly:

With a selection of the probability function p (x, y) different expert strategies can be realized, for example:

This algorithm realizes a variation of the defect depths, in which the lattice points with the greatest depth are preferred. Another strategy may look like this:

Hthe best ( ^x > V) ^~ H _m (x, y)

p (x, y) =

\\ Hthe best (x, y) - _m {x, y) \\

Such an algorithm varies the defect geometry at positions where the simulated measurement signal for the best known solution has the greatest difference to the measured signal.

Based on this, different expert routines or their algorithms can be set up by varying the number of grid points to be considered and the AD. By way of example, the following six expert routines can be used: 1 pO, y) = -, N = 1 and AD = 1% wall thickness

2 pO, y) = 2 and AD = 5% wall thickness

3 pO, y) = -, N = 3 and Δΰ = 5% wall thickness

4 p (x, y) = N = = 1 and Δΰ = i% wall thickness

5 p (x, y) = N = = 2 and Δΰ = 5% wall thickness

6 p (x, y) = N = 3 and Δΰ = 5% wall thickness

As has been described, the monitoring routine 9 shown in FIG. 3a has, in particular, two functions: Firstly, the achievement of the stop criterion is checked, and secondly, the resources of the EDP unit are compared between the individual ones

Experts based on their successes. A measure of success

where AF is the reduction of the fitness function F by the result of the respective expert routine and N is the number of simulations necessary for this. An evaluation of the n expert routines can be considered

P _{n be} accepted. The number of arithmetic slots Ns for an expert routine in an iteration is then

where N _a // is the number of all available slots.

In the simulation routines 16, an MFL measurement for an expert defect geometry is simulated. An expert routine can iterate until they are find a solution whose expert predictive data set is better than the baseline prediction record stored in area 12. If this is the case, the expert routine 11 may process another linear independent data set or try to achieve further better solutions based on the already improved solution.

If several data sets from different iterations, which can not be matched in the same manner, were calculated by the expert routines, the resulting geometries can also be automatically superimposed in the context of carrying out the method according to the invention, wherein the maximum depth at the individual grid points is a conservative estimate is taken:

D (x, y) = max D _n (x, y)

N

for n = 1 ... N, where N is the number of records that must be processed one after the other. Starting from a resulting depth profile resulting from such a superimposition of flaw geometries, an MFL signal can again be simulated. The resulting error can result from the errors of the respective data records in the individual calculations:

E = \\ H _cal (D) - H _m \\

In order to dismantle the efficiency of the proposed method, a multiplicity of test scenarios have been carried out, whereby, according to FIG. 4, the data two MFL inspection runs carried out with linearly independent magnetization.

Fig. 4, numeral 21, represents an image of a real MFL measurement with magnetization running in the axial direction, while the image 22 results from a measurement made in the circumferential direction. The calculation result of the defect geometry, which was achieved according to the method according to the invention described above, is indicated at 23. As in Figure 24, the contours evenly divide the area between 0 and 60% metal loss depth. Figure 24 shows the actual scanned and thus directly measured outer surface of the pipe section associated with Figures 21 and 22. This results in a very high agreement of the laser scan measurement with the solution achieved by the method according to the invention. This is significantly better than the solution according to the known in the art evaluation. It can be assumed here that the deviations of the result according to the method according to the invention and that of the laser scan measurement are predominantly due to technical tolerances.

On the basis of the conventional consideration with determination of the flaw geometry established in the prior art and shown in FIG. 1 as a result, the abovementioned maximum burst pressure of 4744.69 kPa results. On the basis of the method according to the invention, the flaw geometry shown in FIG. 6 (contour lines at 2 mm depth) results for the MFL data set on which FIG. 1 is based, and based thereon a maximum burst pressure of 8543.46 kPa. This In the present case, up to 99.4% of the maximum burst pressure is used, which was determined on the basis of the actual defect geometry determined by laser scanning. Accordingly, a pipeline investigated by the method according to the invention can be operated with a safe operating pressure of 6520.53 kPa.

This results in considerable advantages for pipeline operators in comparison to the safe operating pressure of 3621, 29 kPa on the basis of the evaluation according to the prior art (FIG. 1). Due to the method according to the invention, the state of a pipe and thus the pressure that can be given for safe operation of the pipeline can be stated much more realistically, while operational safety is still guaranteed. By virtue of the method according to the invention with the expert routines competing for the resources of the EDP unit, such a result can be made available to the operators of pipelines faster or at least in the same evaluation time as in the prior art.

Claims

claims

1 . Method for determining the geometry of one or more real, examined defects of a magnetizable object, in particular a pipe or a tank, with a reference data set of the object generated on the basis of one or more MFL measurements,

preferably comprises an at least partial representation of the object by means of a computerized unit, in particular on or through an at least three-dimensional object grid,

and comprising a determination of an initial defect geometry as initial defect geometry, in particular on the object grid or an at least two-dimensional defect grid (15),

a determination of a first M FL prediction data set as output prediction data record based on the output missile geometry, in particular by simulation of an M FL measurement or assignment of an M FL data set,

and iterative adaptation of the initial defect geometry to the geometry of the real defect (s) by means of the EDP unit and by means of several expert routines (1 1) running in competition and preferably parallel to one another,

wherein in respective expert routines (11) a respective expert defect geometry is generated by means of at least one own algorithm and on the basis of the initial defect geometry,

a respective expert forecast data set is determined on the basis of the respective expert defect geometry, in particular by simulation of an MFL measurement or assignment of an MFL data record, and the expert defect geometry on which the respective expert forecast data record is based is then made available to at least one, preferably several, in particular all, of the expert routines (11) as new output missile geometry for further adaptation to the geometry of the real defect (s);

if the expert predictive data set is more similar to the reference data set than the initial prediction data set and then the expert prediction data set corresponding to the new output missile geometry is used as the new starting prediction data set,

wherein the iterative adaptation by means of the expert routines takes place until a stop criterion is met.

2. The method according to claim 1, characterized in that the expert routines (1 1) in such a way compete with each other, that a distribution of the resources of the computer unit to a respective expert routine in particular in the form of computing time, preferably CPU and or GPU time, depending on a success rate, in which in particular the number of output error geometries calculated by this expert routine and provided for one or more other expert routines (11) is taken into account, and / or depending on a reduction A fitness function, in particular taking into account the number of expert forecast data sets generated for the reduction, takes place.

3. The method according to any one of the preceding claims, characterized in that a fitness function is used as a measure of the similarity of the expert prediction and reference data sets.

4. The method according to any one of the preceding claims, characterized in that for determining the geometry of the defect (s) additionally from the first reference data set with respect to the magnetization linearly independent further reference data set is used and based on the output missile geometry another output prediction data set in particular by means of a linear Independence considering further MFL simulation is determined and an expert defect geometry is only used as output missile geometry if the associated for both independent magnetizations associated expert forecast data sets are more similar to the respective reference data sets than the determined for the two magnetizations Ausgangsvorssprungsensensätze and / or both experts Predictive data sets considering fitness function is improved.

5. The method according to claim 4, characterized in that the first reference data set was generated via an MFL measurement with axial magnetization and the second reference data set via an MFL measurement with magnetization in the circumferential direction of the tube.

6. The method according to any one of the preceding claims, characterized in that output and / or expert forecast data sets are generated on the basis of a forward model for simulating the MFL measurement and in particular by means of a finite element models.

7. The method according to any one of the preceding claims, characterized in that the initial defect geometry is generated by means of a look-up table, by one of the expert routines (1 1) and / or by a machine learning algorithm.

8. The method according to any one of the preceding claims, characterized in that the defect grid (5) in areas where the depth of the simulated defect in) exceeds a threshold, is refined.

9. The method according to any one of the preceding claims, characterized in that prior to the calculation of a respective expert predictive data set, the object and / or the defect grid (15) is refined.

10. Method according to one of the preceding differences, characterized in that the output error location geometry or a pointer pointing to it is stored in a memory area (12) of the EDP unit accessible to all expert routines (11).

1 1. Method according to one of the preceding claims, characterized in that as a stop criterion after a plurality of iterations missing substantive change of the Ausgangsweltenstellengeometrie and / or the geometry of the object and / or defect grating (15) and / or the Ausgangsvorhorsgestensatzes and / or at least an expert forecast record is accepted.

12. The method of claim 1 1, characterized in that as a stop criterion, a comparison of the variation of the expert predictive data set with the measurement spread of the real data set is used.

13. The method according to any one of the preceding claims, characterized in that an expert routine (1 1) are associated with several algorithms for adaptation of the expert defect geometry including machine learning, stochastic optimization, empirical and / or numerical model functions.

14. The method according to claim 13, characterized in that in an expert routine (1 1) an algorithm is randomly selected or selected and / or changed by a selection function.

15. A method for determining a load limit of an at least in operation pressure-loaded and in particular designed as an oil, gas or water pipeline object, wherein a one or more defect (s) descriptive data Tensatzatz is used as input data set in a calculation of the load limit, characterized in that the input data set is first determined according to a method according to one of the preceding claims.