COMPUTER-IMPLEMENTED METHOD FOR OPTIMIZING THE DETERMINATION OF MEASUREMENT DATA OF AN OBJECT

Description

[0001] The invention relates to a computer-implemented method for optimizing the determination of measurement data of an object.

[0002] During the production of objects, such as components or workpieces, the objects are manufactured according to a nominal geometry. However, manufacturing tolerances and inaccuracies during production can cause the objects to deviate from the nominal geometry and contain internal defects. Therefore, the objects are inspected either randomly or on a general basis. Inspection of the produced objects can be carried out by measuring the objects, for example, to determine whether the dimensions of the measured object lie within the manufacturing tolerances.

[0003] The measurement can be performed non-contact using image processing techniques. Furthermore, the interior of the object can be examined, for example, using radiography. In such measurements, very generic or non-specific acquisition parameters are usually used for recording the measurement data and evaluation algorithms for evaluating the measurement data, even though the measurement tasks, e.g., the geometry and material of the measurement object and the measurement tasks to be performed, including the relevant tolerances, vary greatly. Instead, the generic or non-specific acquisition parameters are used for a wide range of different measurement tasks. The requirement is that these acquisition parameters achieve the best possible results for all measurement tasks to be performed.

[0004] The following is disclosed in document D1 (DE102017100594A1): The invention relates to an arrangement and a method for the automatic or semi-automatic determination of setting parameters for a computed tomography on the basis of at least one model equation for the computed tomographic determination of at least one dimensional measured variable, wherein at least some of the device parameters of the computed tomography device used for the measurement are taken into account in the model equation(s). In this case, the model equation(s) take into account the specified parameters to be observed, which include at least the measurement task parameters describing the one or more measurement tasks. The setting parameters are determined by simulation and / or mathematical calculation on the basis of the one or more model equations and / or computed tomographic test measurements in the available parameter space of the setting parameters.for which at least one target parameter is optimized and / or the target parameter falls below or exceeds a limit value.,

[0005] The object of the invention is to provide a computer-implemented method that achieves better results for the measurement tasks to be performed.

[0006] Main features of the invention are defined in claims 1 and 8. Embodiments are the subject of claims 2 to 7.

[0007] In one aspect, the invention relates to a computer-implemented method for optimizing the determination of measurement data of an object according to claim 1.

[0008] With the invention, the recording parameters are optimized for the specific task before a measurement of the object is carried out. This also optimizes the evaluation methods for carrying out the measuring task. For each measuring task, a separate recording parameter is initially determined, which is then specifically optimized for this measuring task. For example, a trajectory encompassed by the recording parameters, which a measuring device uses to measure the object, can be designed in such a way that it is optimally designed for this measuring task. This can mean, for example, that the geometries of the object relevant for the measuring task are only recorded with an accuracy that is just sufficient for meaningful yet fast evaluation. Instead of determining measurement data with maximum resolution, for example, if necessary.locally, a lower resolution can be used if the underlying evaluation method with the lower resolution delivers approximately the same good results as with the high resolution. In this case, since a higher resolution is generally more time-consuming, time can be saved for the acquisition of the measurement data. Furthermore, the setting options included in the recording parameters, with which the properties of the device for measuring the object can be changed, can be optimized for the measurement task. For example, the device for measuring the object can be set so that the geometries relevant to the measurement task are mapped with just sufficient accuracy for the measurement task. This further optimizes the time required to acquire measurement data of the object, without significantly compromising the meaningfulness of the results of the evaluation method.The acquisition parameters are therefore initially optimized in order to adapt the acquired measurement data to the measurement task as optimally as possible. It should be noted that optimizing the acquisition parameters does not necessarily mean that the acquired measurement data will have the best quality. Rather, the optimization of the acquisition parameters aims to select the acquisition parameters so that the measurement data acquired based on the acquisition parameters have just a minimum quality to obtain meaningful results using the evaluation method underlying the optimization of the acquisition parameters. This can mean, for example, finding acquisition parameters with which the measurement task can be carried out with the required accuracy in the shortest time and / or with the fewest number of measurement geometries.However, this does not rule out the possibility of finding acquisition parameters that allow the measurement task to be carried out with the highest accuracy within a defined time and / or with a defined number of measurement geometries. When radiative measurements are carried out, for example with X-rays, the optimized acquisition parameters can alternatively or additionally result in a minimization of the radiation dose used. However, combinations of these goals are also possible. After optimizing the acquisition parameters, a measurement of the object is carried out using a device for measuring the object. The measurement data obtained in this way are then used by an evaluation process to evaluate the geometric properties of the object. The computer-implemented process produces improved results for the measurement tasks to be performed.The invention provides further advantages when a large number of objects with the same nominal geometry, which are inspected for the same characteristics—the so-called measurement task—are analyzed very frequently. This is the case, for example, with inline testing during production, where the invention results in significant time savings and improved results for the measurement tasks to be performed. The same measurement task with the same tolerance specifications can thus be performed in a shorter time. This saves valuable machine time and increases machine capacity.

[0009] Optimization can, for example, further mean that a knowledge-based system or artificial intelligence suggests certain acquisition parameters based on the measurement tasks. Simulations that use the acquisition parameters can also be performed to evaluate the optimization. For example, an initial evaluation procedure can enhance the measurement data and compare the evaluation results with reference values, which can be a ground truth derived from the simulated geometry. From this evaluation, a typical measurement error for determining the measurement data can be derived, which can then be used to evaluate the acquisition parameters.The evaluation method used for this purpose may already be working with optimized acquisition parameters or may have been optimized using the optimized acquisition parameters to achieve the best possible or stable results from low-quality measurement data, for example, when a small number of measurement geometries were used. Other methods that do not use simulations and perform the optimization based on a preliminary evaluation method may also be used.

[0010] The measurement task can include or require one or more analyses. Examples of analyses can be: a defect analysis in 3D or 2D; an inclusion analysis in 3D or 2D; a surface or interface determination, which is used, for example, for an analysis in the field of dimensional metrology, where an interface can also be a surface that defines a transition from material to air and an interface between two materials of the measurement object is analyzed, e.g. in the case of multi-material objects; an analysis of geometric properties such as size, shape, position, waviness or roughness; an analysis of fibers or fiber structures, both geometric properties of fibers that can be individually segmented and analyses of fibers whose diameters are below the resolution limit of the measuring system, but which can still be analyzed as a composite, e.g. with regard to their orientation; an analysis of powder properties, e.g.Diameter, volume, surface area of ​​grains; identification of any type of defect, e.g., unmelted powder in additive manufacturing or cracks, and / or analysis of any material properties, e.g., density.

[0011] Furthermore, several acquisition parameters can be optimized for several analyses simultaneously.

[0012] The measurement task can also be defined for multi-material objects, which is particularly advantageous for the determination of interfaces, since this is difficult according to the state of the art.

[0013] The measurement task can be defined for specific areas of the object, so that the analyses only need to be performed there. In this case, it may also be useful for optimization purposes to limit the analysis to these areas or to prioritize them.

[0014] In axial computed tomography, projections are acquired at equidistant angular increments, with the same settings being selected for all projections. In this example, the following, potentially global, acquisition parameters can be optimized: number of projections, orientation of the component with respect to the acquisition geometry or beam path, geometric magnification, and settings such as tube voltage, tube current, and exposure time. This has the advantage that only a small number of acquisition parameters need to be optimized.

[0015] In the example of robotic computed tomography, where the measurement geometries can be freely selected as radiographic geometries, these acquisition parameters can be optimized individually for each projection. In addition, the radiographic geometries of the individual projections and the number of projections are optimized. This allows for more optimization options, but also increases the required effort due to the larger number of acquisition parameters.

[0016] The measurement may, for example, include an optical measurement, such as fringe projection or photogrammetry, or a radiographic measurement, such as computed tomography, radiography or ultrasound measurement, or a tactile sensor, e.g. using a probe.

[0017] In the case of optical measurements, the measurement geometry can, for example, describe the viewing direction of an optical sensor in relation to the measured object. In the case of a transmission measurement, the measurement geometry can be a transmission geometry, which describes the spatial relationship between the radiation source, the object, and the radiation detector. In this case, the measurement geometry describes the direction in which the object is irradiated, but also the position of the irradiated area and the magnification. This can be described using nine geometric degrees of freedom: three degrees of freedom for translation each for the radiation source and the radiation detector, and three degrees of freedom for rotation for the radiation detector. A transmission geometry can be defined with respect to the measured object and / or the device for measuring the object. In the case of a tactile measurement, the measurement geometry can, for example,describe the probing direction or the orientation of the object to be measured in the measuring volume.

[0018] In the case of a radiographic measurement, measuring an object can mean capturing one or more radiographic images or projections of the object. In the case of an optical measurement, measuring an object can mean capturing one or more images of the object with a measuring camera or an optical sensor. In the case of a tactile sensor, measuring an object can mean capturing one or more measuring points on the object.

[0019] In the case of computed tomography, setting options can be, for example, the voltage and current of an X-ray tube or the exposure time, which can also vary for each individual radiographic image. In the case of photogrammetry, a setting option can be a camera's exposure time. In the case of fringe projection, a setting option can also be, for example, a pattern projected onto the object to be measured. In the case of a tactile sensor, a setting option can be, for example, a contact force.

[0020] To perform a measurement with optimized measurement geometries or acquisition parameters, the object can be placed in a suitable holder, for example, which ensures a defined pose of the object. Alternatively or additionally, the object's pose can be determined from the initially acquired measurement data, and the device for measuring the object can be moved accordingly to the desired transmission geometries. This ensures that the object's measurement data is recorded with the desired measurement geometries.

[0021] According to one example, the step of optimizing the at least one acquisition parameter for the at least one measurement task may further comprise the following substep: providing a set of predefined measurement geometries; selecting a subset of the set of predefined measurement geometries based on the measurement task.

[0022] With the set of predefined measurement geometries, especially for radiography measurements, measurement geometries can be selected that are likely to have suitable acquisition parameters for the object under consideration and with regard to the evaluation method underlying the optimization of the acquisition parameters. This can accelerate the optimization of the acquisition parameters, as the acquisition parameters initially used for optimization are already favorable. Furthermore, these measurement geometries can be used for which the measuring device has already been calibrated or measured, thus enabling increased accuracy of the measurement results using these measurement geometries.

[0023] For example, measurement geometries that are not relevant for the evaluation procedure can be omitted. Alternatively or additionally, the smallest number or subset of existing measurement geometries can be selected for which the measurement task can still be fulfilled.

[0024] According to another example, after the step of determining the measurement data for the object, the method may further comprise the following step: determining a digital three-dimensional object representation from the measurement data; wherein the step of performing the at least one measurement task comprises the following substep: analyzing the digital object representation based on the measurement task.

[0025] In the case of a radiographic measurement in the form of computed tomography, this can be a reconstruction of the volume data from projection data or radiographic images. In connection with the sub-steps of providing a set of predefined measurement geometries and selecting a subset of the set of predefined measurement geometries based on the measurement task, the optimized measurement geometries can be equidistant or non-equidistant, depending on the measurement task. An equidistant measurement geometry can be advantageous for axial computed tomography, for example, because the more widespread filtered backprojection can then still be used for a reconstruction of the digital three-dimensional object representation, so-called volume data. Non-equidistant measurement geometries can be used, for example, if an iterative algorithm reconstructs the volume data. The setting options for the individual measurement geometries are typically left constant ordoes not vary.

[0026] In an optical measurement, the three-dimensional object representation can be a representation of the surface or an interface of the object, which is calculated based on the images taken by the measuring camera.

[0027] Furthermore, the step of optimizing the at least one recording parameter can be carried out by means of at least one simulated radiographic measurement of the object.

[0028] In this example, a radiograph of a geometry is simulated using defined acquisition parameters. The results are usually virtual radiographs that can be evaluated using the methods used in real radiographic measurements, such as reconstruction and / or analysis. Such a radiograph simulation can be based, for example, on ray tracing, Monte Carlo methods, or image-based forward projection.

[0029] The measurement task can, for example, include at least one defect analysis to identify and analyze possible defects in the object, whereby the measurement is carried out as a radiographic measurement.

[0030] Furthermore, for each of the analyses, minimum requirements for the analyses or analysis results can be defined that must still be achieved with the optimized acquisition parameters. For example, a minimum accuracy can be defined with which a geometric parameter must be determined. Another example could be a minimum certainty with which a critical property in the object must be determined.

[0031] Furthermore, the step of optimizing the at least one recording parameter for the at least one measurement task can, for example, further comprise at least one of the following sub-steps, wherein the measurement is a radiography measurement: changing the at least one recording parameter until each defect in the object having a predefined minimum size is detected with a probability that lies within a predefined probability interval for defects; changing the at least one recording parameter until geometric parameters of the defects in the object are determined with a probability that lies within a predefined probability interval for geometric parameters, with a predefined minimum accuracy for defects.

[0032] Since the probability that all defects will be detected cannot yet reach 100%, in this example, probability intervals can be specified within which the probability that each defect in the object has been detected must lie. The probability interval can be defined such that, for example, a defect of the defined size must be detected with a 95% probability, i.e., in 19 out of 20 measurements. This is then considered reliable.

[0033] The minimum size can be defined locally and may vary. To assess whether defects of the corresponding minimum size can be reliably identified with given, possibly optimized, acquisition parameters, real and / or simulated test measurements can be carried out. From these, the probability that defects at a defined location and of a defined size will be identified can be determined using Monte Carlo methods. For simulated measurements, the input geometry of the simulation or the simulated defects can be used as a reference or ground truth. This can also be determined based on properties of the volume data, e.g. noise level and resolution. The higher the noise level and the poorer the resolution, the less likely it is to correctly identify small defects. In another example, the evaluation procedure can.for example, an artificial neural network, will be trained to make a corresponding prediction of the reliability of the analysis results.

[0034] Furthermore, the measurement task can, for example, include at least one determination of an interface of the object.

[0035] Furthermore, minimum requirements can be defined for determining the object's interface, which must still be met with the optimized acquisition parameters. For example, a minimum accuracy for determining the interface or its position can be defined. Another example could be a minimum degree of certainty with which the interface must be determined or identified.

[0036] In all of the above examples, an analysis of properties of fibers or fiber structures in the object can be carried out alternatively or additionally.

[0037] According to another example, the step of optimizing the at least one recording parameter for the at least one measurement task may further comprise at least one of the following sub-steps: changing the at least one recording parameter until interfaces in the object are determined with a probability that lies within a predefined probability interval for interfaces, with a predefined minimum accuracy for the interfaces.

[0038] The minimum accuracy can also be defined for interfaces. Since, as explained above, a probability cannot yet reach 100%, in this example, probability intervals can be specified within which the probability that each interface in the object has been determined with a predefined minimum accuracy must lie. The probability interval can be defined such that, for example, in a considered interface area, the true interface lies with 95% probability, i.e., in 19 out of 20 measurements, at most a value corresponding to the minimum accuracy away from the determined interface. This is then considered reliable. The minimum accuracy can be defined locally, possibly varying.

[0039] To assess how accurately the local accuracy of the interfaces can be determined with given, possibly optimized, acquisition parameters, real and / or simulated test measurements can be performed. These can be used to determine the probability of identifying local accuracy at a defined location and with a defined size using Monte Carlo methods. For simulated measurements, the input geometry of the simulation or the simulated local accuracy can be used as a reference or benchmark.

[0040] Ground truth can be used. This can also be determined based on properties of the volume data, such as noise level and resolution. The higher the noise level, the lower the local accuracy of the interface. Poor resolution compromises the accuracy in interface regions of small structures. In another example, the evaluation method, for example, an artificial neural network, can be trained to make a corresponding prediction of the reliability of the analysis results.

[0041] Geometric parameters can be, for example, the defect volume or the diameter of a defect, which may be equivalent to a sphere of the same volume.

[0042] The minimum accuracy can also be defined for defects of a specific size, e.g., as a volume of defects in the range 200 µm 3< to 300 µm 3< that should be determinable to within 10%. Since, as explained above, a probability cannot yet reach 100%, corresponding probability intervals can be defined, e.g., that in 95% or 19 out of 20 cases, the measurement deviation of the defect volume is a maximum of 10%. The minimum accuracy can be defined locally (even varying).

[0043] To assess how accurately the geometric parameters can be determined with given, possibly optimized, acquisition parameters, real and / or simulated test measurements can be carried out. From these, the local accuracy of the determination of the geometric parameters can be determined using Monte Carlo methods. For simulated measurements, the input geometry of the simulation or the simulated defects can be used as a reference or ground truth. Furthermore, the local accuracy can be determined based on properties of the volume data, e.g., noise level and resolution. The higher the noise level, the lower the local accuracy of the determination of the geometric parameters typically is. Smaller defects can only be detected imprecisely, especially with poor resolution.In another example, the evaluation method, for example an artificial neural network, can be trained to make a corresponding prediction of the reliability of the analysis results.

[0044] In all of the above examples, other artificial intelligence or machine learning methods can be used instead of or in addition to artificial neural networks, e.g., deep learning, support vector machines, Bayesian classifier, nearest neighbor classification, random forest, support vector machine, etc.

[0045] Furthermore, in a further example, after the step of determining the measurement data for the object, the method may further comprise the following step: optimizing the evaluation method using measurement data determined by means of a measurement using the at least one optimized recording parameter as training data.

[0046] This optimization of the evaluation process should not be confused with the optimization of the acquisition parameters. In this way, the evaluation process is trained to evaluate measurement data that was acquired with the optimized acquisition parameters and which can therefore exhibit their own or very specific characteristics, as effectively as possible. The process learns, for example, how defects are represented in measurements using these acquisition parameters. This makes it easier for the process to distinguish whether an anomaly in the data is an image error, which can be caused by the poor quality of the measurement data, or a real geometry or property of the component, e.g., a defect. This makes it possible to perform the analysis successfully despite low or locally highly variable data qualities.

[0047] The optimization of the evaluation procedure is typically performed using adaptive methods that require training data, including ground truth. The starting point for this optimization can be an evaluation procedure already explained above. Alternatively, a different method can be used, for example, the artificial intelligence or machine learning methods mentioned above. The result of the optimization is an optimized evaluation procedure.

[0048] This training data can be generated, for example, through real measurements or simulations. For real measurements, the ground truth can be determined using a reference measurement, such as optically acquired micrographs or a very high-quality computed tomography measurement. For simulated measurements, the input geometry of the simulation can be used.

[0049] To ensure that the evaluation procedure is not too focused on a specific measurement task, measurements of the object with different or non-optimized recording parameters can be taken into account in the training data.

[0050] The evaluation method to be optimized may again be that mentioned in claim 1.

[0051] Preferably, the evaluation that has already been used as a basis or as a measure for optimizing the recording parameters is optimized.

[0052] Optimizing the evaluation procedure means that the evaluation procedure is changed so that the measurement tasks are fulfilled as well as possible on the training data.

[0053] Fulfilling the requirements as well as possible can mean, for example, that as many defects as possible are detected without causing too many type I and type II errors, or that the interface can be determined with the highest possible accuracy.

[0054] To avoid overfitting of the training data, regularization techniques such as data augmentation or dropout can be used.

[0055] To enable the most effective training possible, defect geometries or shapes, defect sizes, and defect distributions can be used in the training data for defect detection, which may later also occur in the objects to be analyzed. This knowledge can be derived from a database, e.g., depending on the manufacturing process such as die casting, from simulations of the manufacturing process, or from existing measurements.

[0056] Furthermore, the properties of the defects and other geometric properties of the objects in the training data can be varied as much as possible to avoid overfitting the evaluation.

[0057] Furthermore, the optimization of recording parameters and evaluation procedures can also be carried out simultaneously.

[0058] The method may, for example, further comprise the following step: optimizing the evaluation method using simulated measurement data, which were determined by means of a simulation using the at least one optimized recording parameter, as training data.

[0059] In this example, the same simulation methods or previously performed simulations can be used that were also used in the step of optimizing the at least one recording parameter or the at least one optimized recording parameter for the at least one measurement task to optimize the determination of measurement data.

[0060] This involves simulating a radiography of a geometry using defined acquisition parameters. The result is usually virtual radiography images, which can be evaluated using the same methods used for real measurements, such as reconstruction and analysis. Such a radiography simulation can be based, for example, on ray tracing or image-based forward projection.

[0061] According to a further example, the method may further comprise the following step: determining a probability value using the training data which indicates whether the optimized evaluation method identifies a defect of a defined size.

[0062] The probability value can be determined for the given acquisition parameters. A probability of detection diagram (POD diagram) can also be used for this purpose. This shows the probability of detecting a defect depending on the size of the defect, e.g., in number of voxels or defect volume. This probability is usually 0 for extremely small defects and 100% for large defects. The course of the POD diagram in between can be used to assess whether the measuring system or the selected acquisition parameters or the measurement capabilities including evaluation are suitable for the measurement task. The probability or the POD diagram can also be determined locally for different measurement areas. A POD diagram can be defined for various measurement methods that can be used to detect defects in the object.Such a POD diagram can be used as a parameter to be optimized when optimizing the recording parameters and / or the evaluation procedure.

[0063] It can further be provided in one example that the method further comprises the following step: optimizing the at least one recording parameter or the at least one optimized recording parameter based on the optimized evaluation method, wherein at least one further optimized recording parameter results.

[0064] Before the step of optimizing the at least one acquisition parameter based on the optimized evaluation method, a predefined termination condition can be checked. The step of optimizing the at least one acquisition parameter based on the optimized evaluation method is only carried out if the predefined termination condition is not met. If the step of optimizing the at least one acquisition parameter based on the optimized evaluation method is carried out, the evaluation method optimized using the optimized acquisition parameters is used to determine further optimized acquisition parameters. The further optimized acquisition parameters are generally better than the previously optimized acquisition parameters. Due to the optimized evaluation method, the basis for the optimization has changed. For example,The defects to be identified can now be better detected by the optimized evaluation procedure, which enables further optimization of the acquisition parameters, e.g., the use of even fewer projections.

[0065] According to another example, after the step of optimizing the at least one recording parameter based on the optimized evaluation method, the method may further comprise the following step: optimizing the evaluation method or the optimized evaluation method using measurement data determined by means of the at least one further optimized recording parameter to determine a further optimized evaluation method.

[0066] The optimized evaluation method can now be further optimized using the further optimized acquisition parameters. This results in a better evaluation method than the already optimized evaluation method. This optimization can start from the original evaluation method, the optimized evaluation method, or a completely different evaluation method. This means that existing evaluation methods can be further improved using the further optimized acquisition parameters. Alternatively, a completely new evaluation method can be improved using the further optimized acquisition parameters to avoid the further optimization remaining too close to the already optimized evaluation methods, which could represent a local optimum on a curve of optimized evaluation methods.

[0067] Furthermore, after the step of optimizing the evaluation method or the optimized evaluation method using the at least one optimized recording parameter to determine a further optimized evaluation method, the method can, for example, further comprise the following step: checking whether a predefined termination condition is met; if the predefined termination condition is not met: repeating the steps of optimizing the at least one recording parameter based on the optimized evaluation method and optimizing the evaluation method or the optimized evaluation method using measurement data that were determined by means of the at least one optimized recording parameter to determine a further optimized evaluation method until a predefined termination condition is met.

[0068] The steps for optimizing the evaluation procedure using the optimized acquisition parameters can be performed iteratively. Further optimized acquisition parameters and further optimized evaluation procedures are then determined alternately. The process is repeated until a termination criterion is reached, e.g., a maximum computing time for the duration of the optimization or convergence of the acquisition parameters or evaluation algorithms to be optimized.

[0069] In a further aspect, the invention relates to a computer program product with computer-executable instructions which, when executed on a computer, cause the computer to carry out the method according to the preceding description.

[0070] Advantages and effects, as well as further developments of the computer program product, arise from the advantages and effects, as well as further developments of the method described above. Reference is therefore made to the preceding description in this regard. A computer program product can be understood, for example, as a data storage medium on which a computer program element is stored that contains instructions executable by a computer. Alternatively or additionally, a computer program product can also be understood, for example, as a permanent or volatile data storage device, such as flash memory or RAM, that contains the computer program element. However, this does not exclude other types of data storage devices that contain the computer program element.

[0071] Further features, details, and advantages of the invention will become apparent from the wording of the claims and from the following description of exemplary embodiments with reference to the drawings. It shows: Fig. 1 a flowchart of an example of the computer-implemented method.

[0072] In the following, the computer-implemented method for optimizing the determination of measurement data of an object is designated by the reference symbol 100, as in Figure 1 shown.

[0073] Measurement data whose acquisition is to be optimized is determined by measuring the object, for example, using a device for measuring the object. The device for measuring the object uses at least one recording parameter, which can include a measurement geometry that describes a spatial relationship between the device for measuring the object and the object, and / or a setting option of the device for measuring the object. The measurement data obtained in this way are evaluated for the geometric properties of the object.

[0074] The method 100 serves to optimize the determination of the object's measurement data. For this purpose, in a first step 102, at least one measurement task is determined for the object. The measurement task describes which analyses are to be performed on the measurement data and which areas of the object are to be analyzed. The areas of the object exhibit geometric properties of the object at the position of the areas.

[0075] In a further step 104, the at least one acquisition parameter is optimized. The optimization takes place for the at least one measurement task determined in step 102. This means that the acquisition parameters determined in step 104 are optimized for the measurement task determined in step 102. If measurement data are determined using the at least one acquisition parameter from step 104, these measurement data optimally match the measurement task, allowing a highly efficient statement to be made about the object with regard to the analyses to be performed.

[0076] Step 104 may include optional substeps 110 and 112. In substep 110, a set of predefined measurement geometries is provided. The measurement geometries serve to describe the relative position of the device for measuring the object and the object. If different measurement geometries are used, the object is measured at different relative positions to the device for measuring the object.

[0077] In sub-step 112, a subset of the set of predefined measurement geometries is selected based on the measurement task. The subset of the set of predefined measurement geometries provides optimal measurement data with respect to the measurement task if these measurement geometries are used in determining the measurement data by the device for measuring the object.

[0078] If the measurement is a transmission measurement, for example performed using axial computed tomography, the axial computed tomography can be performed or simulated first. Based on the measurement data available as projections, the optimization can then be carried out to determine which projections actually need to be used for the measurement task. Since each projection is linked to a measurement geometry, which in this case is a transmission geometry, this can be used to determine the measurement geometries to be used. This selection of measurement geometries can be used for further measurements. During optimization, no further measurement data needs to be determined or simulated, since only a selection is made. In this way, the search space for the measurement geometries to be selected is extremely limited, which simplifies and speeds up the optimization process.

[0079] Step 104 can alternatively or additionally further comprise the optional sub-step 118 if the measurement task comprises at least one defect analysis for determining and analyzing possible defects in the object. The measurement is then a radiographic measurement that also records the internal volume of the object. In this case, the at least one recording parameter to be optimized is changed until every defect in the object that has a predefined minimum size has been detected. The detection must occur with a probability that lies within a predefined probability interval for the defects. The probability interval can, for example, be defined such that the defects should be detected with a probability of 70%. This means that the defects must be detected in 7 out of 10 analyses of similar measurement data.

[0080] Step 104 further comprises sub-step 120 if the measurement task includes at least one defect analysis to identify and analyze potential defects in the object. The measurement is then a radiography measurement that also captures the object's internal volume. In this sub-step, at least one acquisition parameter is changed until geometric parameters of the defects in the object have been identified with a predefined minimum accuracy for defects. The predefined minimum accuracy must be determined with a probability that lies within a predefined probability interval for the geometric parameters. For example, it may be required for the geometric parameters that the geometric parameters were identified with a 90% probability and the minimum accuracy.In this case, 9 out of 10 analyses of similar measurement data must identify the geometric parameters with this minimum accuracy.

[0081] Furthermore, step 104 additionally comprises sub-step 122, in which the at least one recording parameter is changed until interfaces in the object have been determined with a predefined minimum accuracy for interfaces. The determination of the interfaces with a predefined minimum accuracy must be carried out with a probability that lies within a predefined probability interval for interfaces. The probability interval for interfaces can, for example, be defined such that the interfaces are detected with a probability of 85% with the minimum accuracy. This means that, for example, in 17 out of 20 analyses of similar measurement data, the position of the interface was detected with the minimum accuracy. The method 100 further comprises step 106, in which measurement data for the object is determined.The measurement data is determined using at least one optimized acquisition parameter. This may mean that, compared to the non-optimized acquisition parameters, only specific measurement geometries are used to determine the measurement data.

[0082] Furthermore, the method 100 can include an optional step 114 in which a digital three-dimensional object representation is determined from the measurement data obtained in step 106. This can be done, for example, if the measurement data was obtained using a radiographic measurement. Volume data can then be determined from the measurement data available as radiographic images, which are usually projections of the object, using tomographic calculations.

[0083] In a further step 108, the at least one measurement task for the object can be performed based on the measurement data determined from step 106. If the method 100 includes the optional step 114, step 108 includes the optional substep 116. In substep 116, the digital object representation is analyzed based on the measurement task.

[0084] With at least one optimized recording parameter, the number of measurements can be reduced to the point where the measurement task can be performed with sufficient accuracy. This saves time for the additional measurements that would otherwise be required, which would only marginally or not significantly increase the accuracy of the measurement task being performed, and allows the measurement task, including the acquisition of the measured data, to be performed more efficiently.

[0085] After step 106, and in this example before step 108, the method 100 may further comprise the optional step 124. In this optional step, the evaluation method defined in the measurement task is optimized using measurement data determined by using the at least one optimized acquisition parameter. The at least one optimized acquisition parameter is used in this optional step to determine measurement data. Due to the optimization of the at least one acquisition parameter, this measurement data is of better quality than measurement data determined without optimizing the at least one acquisition parameter.

[0086] This higher quality measurement data is then used to optimize the evaluation process.

[0087] In a further optional step 126, the method 100 can optimize the evaluation method using simulated measurement data. The measurement data is simulated using the at least one optimized acquisition parameter. In this case, too, more optimal measurement data is determined than without optimization of the acquisition parameter. This measurement data also exhibits a higher quality than the measurement data determined without optimization of the acquisition parameter. It can therefore be used to optimize the evaluation method.

[0088] In both step 124 and step 126, the measurement data determined with the at least one optimized recording parameter are used as training data for optimizing the evaluation method.

[0089] In a further optional step 128 of method 100, a probability value is determined using the training data. This probability value indicates whether the optimized evaluation method identifies a defect of a defined size. This means that it is checked whether the optimized evaluation method meets a minimum requirement for detecting defects in the object. Preferably, the measurement used to determine the measurement data is a radiographic measurement.

[0090] The optimized evaluation method can be used in a further optional step 130 of the method 100 to optimize the at least one recording parameter. This means that the evaluation method optimized with the original at least one recording parameter is then used in turn to optimize the at least one recording parameter. At least one arbitrary recording parameter can be optimized. This can be the recording parameter originally used. Alternatively, this can be the already optimized recording parameter or a recording parameter not previously used in this method 100. Since the determined measurement data can be further improved due to the optimized evaluation method, the recording parameters can also be further optimized in order to reduce the number of measurement geometries required to perform the measurement task, if necessary.

[0091] Simulated or real measurement data can be used to optimize the acquisition parameters, for example, which is already available for optimizing the evaluation algorithm, and vice versa. This can further reduce computing time.

[0092] In a further optional step 132, the evaluation method or the optimized evaluation method can now be optimized again. Measurement data determined using the further optimized acquisition parameter are used. This leads to a further improvement of the at least one acquisition parameter for the device for measuring the object with respect to the measurement task to be performed.

[0093] In order to be able to carry out a successful evaluation with the evaluation method when only a few radiographic images are available, prior knowledge of the nominal geometry of the object can be used in the reconstruction, i.e. in determining the three-dimensional digital object representation. In this example, an iterative reconstruction method can converge more quickly and better to a correct result. Furthermore, a reference can be generated from a simulation of a measurement of the nominal geometry or from an averaging of previously performed measurements of objects with the same nominal geometry. In order to carry out or facilitate an analysis, e.g. a defect analysis, the two- and / or three-dimensional measurement data of the object to be examined can be compared with the reference.

[0094] Furthermore, the evaluation method can take into account which location on the measured object is currently being evaluated. This allows knowledge about the object's local properties with respect to the measurement geometry to be incorporated. This knowledge can describe an area where higher noise is expected, so a conservative approach should be used when identifying defects to avoid incorrect identifications.

[0095] In a further optional step 134, a check is performed to determine whether a predefined termination condition regarding the optimization of the evaluation method is met. A termination condition may, for example, require that the optimized evaluation method and the optimized acquisition parameters can be used to acquire measurement data in a sufficiently short time to achieve time savings without compromising the quality of the evaluation. However, a termination condition may also be defined differently.

[0096] If the predefined termination condition is met, the process can continue with step 108. If the predefined termination condition is not met, in a further optional step 136 of the method 100, steps 130 and 132 can be repeated until the predefined termination condition is met. In this way, the evaluation methods and acquisition parameters can be optimized iteratively.

[0097] The order of steps described in this example can be changed as desired, as long as this is reasonably practicable.

[0098] The computer-implemented method 100 can be executed on a computer using a computer program product. The computer program product includes computer-executable instructions. When executed on a computer, these instructions cause the computer to perform the method.

[0099] The invention is not limited to one of the above-described embodiments, but can be modified in a variety of ways. All features and advantages apparent from the claims, the description, and the drawings, including structural details, spatial arrangements, and method steps, may be essential to the invention both individually and in a wide variety of combinations.

Claims

1. Computer-implemented method for optimizing a determining of measurement data of an object, wherein the measurement data are evaluated for geometric properties of the object to be analysed, wherein the measurement data are determined by way of a measurement of the object using at least one recording parameter, wherein the at least one recording parameter comprises at least one measurement geometry and / or at least one setting option for the measurement, wherein the method (100) includes the following steps: - determining (102) at least one measuring task for the object, wherein the measuring task is carried out by way of an evaluation method and defines the geometric properties of the object to be analysed; - optimizing (104) the at least one recording parameter for the at least one measuring task to optimize the determination of measurement data; - determining (106) the measurement data for the object by way of a measurement using the at least one optimized recording parameter; and - carrying out (108) the at least one measuring task for the object on the basis of the determined measurement data, wherein, according to a first approach, the measuring task comprises at least one defect analysis to determine and analyse possible defects in the object, wherein the measurement is a radiographic measurement, and wherein the step of optimizing (104) the at least one recording parameter for the at least one measuring task furthermore includes at least one of the following sub-steps: - changing (118) the at least one recording parameter until every defect in the object which has a predefined minimum size is detected at a probability which is within a predefined probability interval for defects; - changing (120) the at least one recording parameter until geometric parameters of the defects in the object are determined at a probability which is within a predefined probability interval for geometric parameters, with a predefined minimum accuracy for defects, or, as an alternative to the first approach in a second approach, the measuring task includes at least one determination of a boundary surface of the object, and wherein the step of optimizing (104) the at least one recording parameter for the at least one measuring task furthermore includes the following sub-step: - changing (122) the at least one recording parameter until boundary surfaces in the object are determined at a probability which is within a predefined probability interval for boundary surfaces, with a predefined minimum accuracy for the boundary surfaces.

2. Method according to Claim 1, characterized in that the step of optimizing (104) the at least one recording parameter for the at least one measuring task furthermore includes the following sub-step: - providing (110) a set of predefined measurement geometries; - selecting (112) a subset of the set of predefined measurement geometries on the basis of the measuring task.

3. Method according to Claim 1 or 2, characterized in that the method (100), after the step of determining (106) the measurement data for the object, furthermore includes the following step: - determining (114) a digital three-dimensional object representation from the measurement data; wherein the step of carrying out (108) the at least one measuring task includes the following sub-step: - analysing (116) the digital object representation on the basis of the measuring task.

4. Method according to any of Claims 1 to 3, characterized in that the step of optimizing (104) the at least one recording parameter is carried out by way of at least one simulated radiographic measurement of the object.

5. Method according to any of Claims 1 to 4, characterized in that the method (100) furthermore includes the following step: - optimizing (130) the at least one recording parameter or the at least one optimized recording parameter on the basis of the optimized evaluation method, wherein a further optimized recording parameter results.

6. Method according to Claim 5, characterized in that the method (100), after the step of optimizing (130) the at least one recording parameter on the basis of the optimized evaluation method, furthermore includes the following step: - optimizing (132) the evaluation method or the optimized evaluation method using measurement data, which were determined by way of the at least one further optimized recording parameter from step (130), to determine a further optimized evaluation method.

7. Method according to Claim 6, characterized in that the method (100), after the step of optimizing (132) the evaluation method or the optimized evaluation method using the at least one optimized recording parameter to determine a further optimized evaluation method, furthermore includes the following step: - checking (134) whether a predefined abort condition is met; if the predefined abort condition is not met: - repeating (136) the steps of optimizing (130) the at least one recording parameter on the basis of the optimized evaluation method and optimizing (132) the evaluation method or the optimized evaluation method using measurement data, which were determined by way of the at least one optimized recording parameter, to determine a further optimized evaluation method until a predefined abort condition is met.

8. Computer program product having instructions executable on a computer which, when executed on a computer, prompt the computer to carry out the method according to any of the preceding claims.