METHOD FOR DETERMINING THE GEOMETRY OF AN OBJECT BASED ON DATA FROM NON-DESTRUCTIVE MEASURING METHODS

Abstract

A method for determining the geometry of a metallic object, with in particular one or more real, examined defects, with a reference data set of the object generated on the basis of at least one measurement by at least one non-destructive measuring method,

preferably comprising an at least partial representation of the object on or by an at least three-dimensional object grid by means of a computer unit,
wherein a classification of anomaly-free areas and anomaly-affected areas of the object is performed on the basis of at least parts of the at least one reference data set,
wherein an initial object grid is created, a prediction data set of the at least one non-destructive measurement method is calculated by a simulation routine using the initial object grid, at least parts of the prediction data set are compared with at least parts of the at least one reference data set, excluding the anomaly-afflicted regions, and the initial object grid is used as an object grid describing the geometry of the object as a function of at least one accuracy measure, or the initial object grid is iteratively adapted to the geometry of the object in the anomaly-free regions by means of the EDP unit.

Claims

1. A method for determining the geometry of a metallic, in particular magnetizable object, in particular a pipe or a tank, with in particular one or more real, examined defects, with a reference data set of the object generated on the basis of at least one measurement by at least one non-destructive measuring method, preferably comprising an at least partial representation of the object on or by an at least three-dimensional object grid by means of an EDP unit, wherein a classification of anomaly-free areas and anomaly-affected areas of the object is performed on the basis of at least parts of the at least one reference data set, wherein an initial object grid is created particularly based on previously known information about the object, a prediction data set of the at least one non-destructive measuring method is calculated by a simulation routine using the initial object grid, a comparison of at least parts of the prediction data set to at least parts of the at least one reference data set is performed, excluding the anomaly-afflicted regions, and, depending on at least one accuracy measure, the initial object grid is used as an object grid describing the geometry of the object, or an iterative adjustment of the initial object grid to the geometry of the object in the anomaly-free regions is performed by means of the EDP unit, wherein a new initial object grid is created during the iteration and for this a new prediction data set is calculated by the simulation routine, and a comparison of at least parts of the new prediction data set to at least parts of the at least one reference data set is performed, excluding the anomaly-afflicted regions, until a stop criterion is present and the initial object grid then present is used as an object grid describing the geometry of the object.

2. The method according to claim 1, characterized in that information of the reference data set and/or object grid is interpolated or extrapolated from the anomaly-free areas into the anomaly-affected areas.

3. The method according to claim 1 or 2, characterized in that the simulation routine is set up by presetting parameters representing material properties of the object, the geometry of a sensor used in the non-destructive measuring method, the distance of the sensor from the object surface, and/or operating conditions of the sensor.

4. The method according to any one of the preceding claims, characterized in that the classification is performed based on at least parts of at least two reference data sets obtained by different non-destructive measuring methods.

5. The method according to any one of the previous claims, characterized in that a first reference data set generated based on an MFL measuring method, particularly with axial or circumferential magnetization, and at least one other reference data set generated based on an eddy current, EMAT, or ultrasonic measuring method are used.

6. The method according to any one of the preceding claims, characterized in that an anomaly-free area is assigned to a predefined local element of the object during the classification and this element is used in the creation of the initial object grid or inserted into the initial object grid.

7. The method according to claim 6, characterized in that the local element, which is particularly formed in the form of a weld seam, is described by means of a parametric geometry model.

8. The method according to claim 7, characterized in that a change of at least one parameter of the parametric geometry model is performed during the iterative adjustment of the initial object grid.

9. The method according to any of the previous claims, characterized in that the iterative adjustment of the initial object grid is performed by grid modification.

10. The method according to any one of the preceding claims, characterized in that the initial object grid is created by calculating a new grid within a changed contour of the initial object grid in the iterative adjustment of the initial object grid.

11. The method according to any one of the preceding claims, characterized in that the object grid is used to determine defect geometries in the anomaly-afflicted regions.

12. The method according to claim 11, characterized in that the geometry of one or more real, examined defects of a metallic and in particular magnetizable object, in particular a pipe or a tank, is determined by means of at least two reference data sets of the object generated on the basis of different, non-destructive measuring methods, wherein the object is displayed at least partially on or through the at least two-dimensional, preferably three-dimensional, object grid in an EDP unit, comprising a determination of at least one starting defect geometry as the starting defect geometry, determination of respective prediction data sets as initial prediction data sets on the basis of the initial defect geometry by simulation or assignment of a measurement that matches the respective reference data set, iterative adjustment of the initial defect geometry to the geometry of the real defect(s) by means of the EDP unit and by means of at least one, preferably multiple, particularly competing expert routines (11) which further particularly run parallel to one another, wherein a respective expert defect geometry is generated in the respective expert routine(s) (11) by means of at least one own algorithm and based on the initial defect geometry, respective expert prediction data sets are determined based on the respective expert defect geometry by simulation or assignment of a measurement that matches the respective reference data set, and the expert defect geometry on which the respective expert prediction data sets are based is made available to at least one, in particular all of the expert routines (11) as a new initial defect geometry for further adjustment to the geometry of the real defect(s), if the expert prediction data sets of a respective expert routine (11) are more similar to the respective reference data sets than the initial prediction data sets and/or a fitness function that takes into account the at least two expert prediction data sets is improved, and then the expert prediction data sets belonging to the new initial defect geometry are used as new initial prediction data sets, wherein the iterative adjustment by means of the expert routines (11) takes place until a stop criterion is met.

13. The method according to claim 12, characterized in that a data set based on an MFL, eddy current, EMAT, or ultrasonic measuring method is used as the first reference data set, and a data set generated based on another measuring method from this group of measuring methods is used as at least one other reference data set.

14. The method according to any one of claim 12 or 13, characterized in that the expert routines (11) run in competition with one another in such a way that the resources of the EDP unit, particularly in the form of computing time, preferably CPU and/or GPU time, to a respective expert routine (11) are distributed as a function of a success rate, particularly taking into account the number of the initial defect geometries calculated by said expert routine (11) and made available to one or more other expert routines (11), and/or as a function of a reduction in the fitness function in which particularly the number of expert prediction data sets generated for the reduction is taken into account.

15. A method for determining a load limit of an object that is under pressure at least during operation and particularly designed as an oil, gas or water pipeline, wherein a data set describing one or more defects is used as an input data set in a calculation of the load limit, characterized in that the input data set is generated first according to a method according to any one of the preceding claims.

Description

[0049] Further advantages and details of the invention can be found in the following description of the figures. Wherein:

[0050] FIG. 1: shows a flowchart for a determination of an object grid representing the defect-free object

[0051] FIG. 2: shows an object grid representing the defect-free object

[0052] FIG. 3: shows an object grid representing the defect-free object in the area of a weld seam

[0053] FIG. 4: shows the object grid according to FIG. 3 in a 3D representation

[0054] FIG. 5: shows a parametric geometry model for the representation of the object in the area of a weld.

[0055] FIG. 6 shows a schematic representation of a further development of the method according to the invention,

[0056] FIG. 7 shows a more detailed explanation of a portion of FIG. 6.

[0057] FIG. 1 shows the flow chart of a possible implementation of the method according to the invention. A model for the non-destructive sensor 21 is created based on measurement data 20 from one or more calibration measurements with a non-destructive measuring method on a calibration object of known geometry, particularly with defects of known geometry. This is required for the simulation of measurement data or reference data sets. A simulation routine is set up in step 22 with an assessment of the relevant material properties of the examined object. This can be done by specifying known parameters that represent the material properties and properties of the sensor used. Alternatively or additionally, the parameters can be iteratively adjusted until the results of the simulation routine for the non-destructive measuring method used, based on the known geometry of the calibration object, match the measurement data of the calibration measurement with sufficient accuracy. The simulation routine can also be prepared and reused for multiple measurements using the non-destructive measuring method.

[0058] One or more reference data sets are created on based on one or more measurements with one or more non-destructive measuring methods. In step 2, FIG. 1 shows the creation of a reference data set based on the data sets 1 of multiple measurement runs. A classification into anomaly-free areas and anomaly-afflicted areas is carried out in step 23 based on the reference data set. By using two or more reference data sets that were obtained based on different non-destructive measuring methods, the classification can be improved again, since individual measuring methods are more sensitive to specific defects than to others.

[0059] An object grid representing the intact geometry of the object is created in step 24 based on the anomaly-free areas and using the simulation routine. For this purpose, information from previous measurement runs in the object with no or fewer defects can also be used. The object grid can be created in the anomaly-free areas and then completed by extrapolating and/or interpolating into the anomaly-afflicted areas. It is also conceivable to carry out an interpolation and/or extrapolation based on the reference data sets from the anomaly-free areas into the anomaly-afflicted areas.

[0060] The object grid is created in step 24 using an iterative process. A first initial object grid is guessed or, for example, specified based on an estimated object geometry. This is adjusted in an iterative process. For example, an initial object grid can be an object grid shown in FIG. 2. If the examined object exhibits a weld in this region, an iterative adjustment of the initial object grid will occur until it has a shape representing the weld, as shown in FIG. 3 and FIG. 4. This adjustment can be done by mesh morphing, i.e. adjusting the initial object grid by moving individual grid points.

[0061] A parametric description of the weld seam by means of a parametric geometry model can in particular also be used to accelerate the method. FIG. 5 shows such a parametric geometry model. In this model, the shape of the weld seam is described by a small number of parameters, in this case seven. The parameters describe the wall thickness of the object (Z.sub.5), the respective extension of the weld seam on both sides (z.sub.3, z.sub.6), the weld seam elevation (z.sub.1, z.sub.7), as well as the width and depth of notches on the weld seam (z.sub.2, z.sub.4). The object grid can thus be changed in the area of the weld seam by adjusting a small number of parameters. In this case, previously known information about a general shape of an object area, here a weld seam, is used. Additional boundary conditions can be specified for individual parameters. This rules out physically nonsensical or impossible results. In FIG. 5, for example, the parameters z.sub.2, z.sub.3, z.sub.5 and z.sub.6 cannot reasonably be negative, z.sub.4 cannot reasonably be greater than z.sub.5, etc. The parameter values can be determined by the following optimization problem:

[00002]$\left\{z_1...z_n\right\}=\arg \min \underset{i}{.Math.}.Math.Y_{\mathrm{cal}}^i\left(z_1...z_n\right)-Y_m^i.Math.$$\mathrm{under}\mathrm{boundary}\mathrm{conditions}\mathrm{for}\left\{z_1...z_n\right\}$

wherein Y.sub.m.sup.i—is the measured signal of the i-th measurement, Y.sub.cal.sup.i is the calculated signal for the i-th measurement. Values for the parameters can be determined using derivative-free optimization algorithms, for example by means of random search. The parameters can be changed in fixed steps, preferably defined as a function of the wall thickness, to accelerate the method. For example, a change can be made in steps that are 1% of the wall thickness.

[0062] The process sequence according to the invention is described at least in sections below according to FIG. 6, wherein a plurality of the parallel and competing expert routines 11 are described as having only one block 14.

[0063] For example, several runs of the same MFL pipeline pig can be combined as input data sets according to box 2. Both data sets 1 can be filtered beforehand for the purpose of better merging and adjusted to one another (method step 3), for example to reduce any artifacts or background noise. In addition, another data set 4 is processed based on another measuring method as an additional reference data set in the associated box 3 and filtered for the purpose of matching to identical grid structures, such that, according to method section 6, two matched reference data sets are available that were created on the basis of different non-destructive measuring methods.

[0064] Data sets that are precisely matched to one another can be treated jointly, wherein the method according to the invention implements the simultaneous treatment of the data sets by using a fitness function that takes into account the data sets to be considered together. In step 7, the reference data sets present in step 6 are accessed, for which purpose a starting defect geometry is first determined as the initial defect geometry in step 8. As described above, this takes place based on a neural network into which the reference data sets are read as input data sets.

[0065] The solution of the neural network is then made available as one or more initial defect geometries x.sub.1 . . . x.sub.n to the individual expert modules. In advance, the number of parameter values that describe the defect geometries can be kept as small as possible, with the aim of reducing computing time. This is achieved, for example, by a dynamic grid adjustment. Since the number of depth values corresponds to the number of node points in the defect grid 5, the number of nodes can at the same time also be the number of defect parameters. Starting with a comparatively coarse grid, this is gradually refined in relevant areas.

[0066] The refinement shown in the relevant grid area in FIG. 5 can be achieved for an exemplary specified node point distance of 14 mm, an associated grid cell size of 14 mm×14 mm, and defect limit values of 30%, 50%, and 80% of the wall thickness, for example, wherein those cells that exceed the above depth values are successively subdivided. The grid deformation then correlates with the assumed defect geometry, i.e. in areas of large gradients there is a larger number of grid points.

[0067] After a defect grid made available centrally to all expert routines has now been selected, a new expert defect geometry is then calculated in step 14 for specific defects in the respective expert routines, and it is checked under 14.1 whether this needs to be made available to the other expert routines. This is the case if, for example, a fitness function has been improved as described above and no stop criterion has yet ended the defect finding process. In this case, the iteration continues with the defect geometry or geometries made available to all expert routines. Otherwise, the method is ended in 14.2. with the determination of the defect geometries and, in particular, the specification of the accuracy of the solution. In addition, the burst pressure can be calculated based on the defect geometries found.

[0068] According to the method according to the invention, the sequence of the work flow of a group of expert routines 11 which are in competition with one another is simulated on the EDP unit. For this purpose, the program can have various modules which can set data in specific areas of the EDP unit independently of one another and particularly not synchronized with one another, so that they can be further processed there. This particularly takes place under the supervision of a monitoring routine 9 (FIG. 7). A plurality of expert routines 11 thus hold a number of computation slots 13 depending on the success defined above, i.e. for example the number of initial defect geometries written in a common memory area 12, to generate expert defect geometries and/or to be able to carry out associated MFL simulations or, in the case of an independent MFL simulation module, to have these simulations carried out. This corresponds to block 14 according to FIG. 6, wherein this block is an example of multiple expert routines 11 (FIG. 7). According to the present exemplary embodiment, the simulations of the measurement data that match the individual expert defect geometries are carried out based on the individual computation slots 13 in the simulation modules 16 for the purpose of creating the expert prediction data sets, also under the supervision of the monitoring routine 9. The more slots 13 are available for an expert routine, the greater the proportion of IT resources available to this expert routine. The number of program modules provided for carrying out simulations is preferably equal to the number of slots. The monitoring routine 9 monitors the number of iterations and the resulting changes in the initial defect geometry and further monitors whether an associated stop criterion has been reached. The result according to block 17, which corresponds to block 14.2 from FIG. 6, is then output.

[0069] The number of computation slots 13 available to an expert routine 11 and the simulation routines subsequently made available can vary in such a way that a first expert routine, for example, can utilize up to 50% of the total available for the computation slots and computing time available to simulation routines.

[0070] As shown, the initial defect geometries are stored in the memory area 12. This can be a memory area accessible to the expert routines 11. Log files of the expert routines 11 and monitoring routine 9 as well as instructions to the expert routines 11 can also be stored there, which are then independently implemented by them. For example, this can be an interrupt command that is set when the stop criterion is reached.

[0071] The expert routines 11 are preferably independent program modules which generate new expert defect geometries and place them in the simulation routines 16. Furthermore, the fitness function presented at the beginning can be generated in the expert routines 11 based on the expert prediction data set and compared to the initial prediction data set stored in the area 12. If the expert prediction data sets are overall more similar to the reference data sets than the data sets stored in area 12, these expert prediction data sets are then used as new initial prediction data sets.

[0072] For example, a new defect geometry is generated randomly in the expert routines 11. Machine learning algorithms or empirical rules can be used for this. Advantageously, however, the implementation of at least two basic expert routines working in a defect-specific manner based on the type of defect is provided to further improve the convergence of the solutions, as described below.

[0073] 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, the depth value of which results in a maximum reduction in the fitness function to determine a corrosion-based defect geometry. The probability function is used to identify N grid points (x.sub.n, y.sub.n). Instead of grid points x.sub.n, y.sub.n, the parametric representation of the group of defects (x.sub.1 . . . x.sub.n) already used above can be assumed as the subject of the probability distribution, with N grid points (x, y) or (x.sub.n, y.sub.n).

[0074] At each of the points under consideration, the depth function, which in the present case describes the depth D of the corrosion at the grid point, is changed by ΔD, wherein the sign of the change is distributed randomly. The number of selected points N can also be chosen randomly:

[00003]$D_{new}\left(x,y\right)=\{\begin{array}{cc}D\left(x_n,y_n\right)\mp \Delta D,& \mathrm{for}\mathrm{selected}\mathrm{points}\\ D\left(x,y\right),& \mathrm{otherwi}se\end{array}$

[0075] When selecting the probability function p (x, y), different expert strategies can be implemented, for example:

[00004]$p\left(x,y\right)=\frac{D\left(x,y\right)}{.Math.D\left(x,y\right).Math.}$

[0076] This algorithm implements a variation of the defect depth, in which the grid points with the greatest depth are preferred. Another strategy for a corrosion-based development of the expert defect geometry may be as follows:

[00005]$p\left(x,y\right)=\frac{H_{\mathrm{the}\mathrm{best}}\left(x,y\right)-H_m\left(x,y\right)}{.Math.H_{\mathrm{the}\mathrm{best}}\left(x,y\right)-H_m\left(x,y\right).Math.}$

[0077] Such an algorithm varies the defect geometry at positions at which the simulated MFL measurement signal H.sub.the best has the greatest difference to the measured signal H.sub.m for the best known solution.

[0078] On this basis, different expert routines or their algorithms can be set up by varying the number of grid points to be considered and the ΔD. As an example, the following six expert routines can be used for the development of corrosion-based defects:

[00006]$1.p\left(x,y\right)=\frac{D\left(x,y\right)}{.Math.D\left(x,y\right).Math.},N=1\mathrm{and}\Delta D=1\%\mathrm{wall}\mathrm{thickness}$$2.p\left(x,y\right)=\frac{D\left(x,y\right)}{.Math.D\left(x,y\right).Math.},N=2\mathrm{and}\Delta D=5\%\mathrm{wall}\mathrm{thickness}$$3.p\left(x,y\right)=\frac{D\left(x,y\right)}{.Math.D\left(x,y\right).Math.},N=3\mathrm{and}\Delta D=5\%\mathrm{wall}\mathrm{thickness}$$4.p\left(x,y\right)=\frac{H_{\mathrm{the}\mathrm{best}}\left(x,y\right)-H_m\left(x,y\right)}{.Math.H_{\mathrm{the}\mathrm{best}}\left(x,y\right)-H_m\left(x,y\right).Math.},$$N=1\mathrm{and}\Delta D=1\%\mathrm{wall}\mathrm{thickness}$$5.p\left(x,y\right)=\frac{H_{\mathrm{the}\mathrm{best}}\left(x,y\right)-H_m\left(x,y\right)}{.Math.H_{\mathrm{the}\mathrm{best}}\left(x,y\right)-H_m\left(x,y\right).Math.},$$N=2\mathrm{and}\Delta D=5\%\mathrm{wall}\mathrm{thickness}$$6.p\left(x,y\right)=\frac{H_{\mathrm{the}\mathrm{best}}\left(x,y\right)-H_m\left(x,y\right)}{.Math.H_{\mathrm{the}\mathrm{best}}\left(x,y\right)-H_m\left(x,y\right).Math.},$$N=3\mathrm{and}\Delta D=5\%\mathrm{wall}\mathrm{thickness}$

[0079] The following functional rules can be used for an expert routine that is suitable for the variation of a crack-based defect: [0080] the depth of the defect is randomly reduced or increased by a specific amount, preferably e.g. 1 or 2% of the wall thickness of the object, [0081] the position of all points of the crack is varied in a randomly selected direction, and/or [0082] a line describing the crack is lengthened or shortened by the position of the grid nodes on the object grid or defect grid.

[0083] An expert routine that describes a laminating defect can work according to the following functional rules: [0084] on the basis of the 2D parameter description of a laminating defect, the values associated with the grid nodes are varied step by step by 5% in one direction or the other with the aim of varying the position of the lamination; this can only be done for a subset of the known of the 2D description of the lamination, [0085] randomly selected points (grid nodes) with values not equal to zero, which are in the vicinity of points with values of zero, can be set to zero (reduction of the extent of the lamination), [0086] randomly selected grid points with values of zero, which are located in the vicinity of grid points with values not equal to zero, can be set to the corresponding neighborhood value, whereby the lamination is increased, and/or [0087] all values in the grid can be moved in a randomly selected direction, which is accompanied by a change in the position of the lamination along the pipeline surface.

[0088] As described, the monitoring routine 9 shown in FIG. 7 particularly has two functions: On the one hand, it checks if the stop criterion is reached, on the other hand, it allocates the resources of the EDP unit between the individual experts based on their successes. A measure of success is

[00007]$P=\frac{\Delta F}{N},$

wherein ΔF is the reduction of the fitness function F by the result of the respective expert routine, and in this case N is the number of simulations required for this. An assessment of the n expert routines can be assumed as

[00008]$R_n=\frac{P_n}{.Math.P_i}.$

. The number of computation slots Ns for an expert routine in one iteration then is N.sub.s=int(R.sub.n N.sub.all),
wherein N.sub.all is the number of all available slots.

[0089] The respective non-destructive measurements for the expert defect geometries are simulated in the simulation routines 16. An expert routine can iterate until it finds a solution whose expert prediction data sets are better than the initial prediction data sets stored in area 12. If this is the case, the expert routine 11 can attempt to achieve other better solutions on the basis of the already improved solution.

[0090] A resulting error E for the individual observations of the simulated and measured data sets can result from the errors of the respective data sets in the individual calculations:

E=Σ.sub.i∥Y.sub.cal.sup.i(D)−Y.sub.m.sup.i∥,

wherein Ym and Ycal represent the above-described respective measured and simulated measuring fields at the defect geometries (x1 . . . xn).

[0091] Based on the method according to the invention, the condition of a pipe and thus the pressure that can be specified for safe operation of the pipeline can be specified much more realistically, while operational reliability is still ensured. Such a result can be made available to the pipeline operators more quickly than, or at least in the same evaluation time as, in the prior art using the method according to the invention with the expert routines competing for resources of the EDP unit.