Computer-implemented method for condition monitoring of an object inspection device and corresponding computer program product

Description

[0001] The invention relates to a computer-implemented method for monitoring the condition of a device for examining objects.

[0002] Manufactured objects can be measured using device testing, e.g., computer tomography or optical sensors such as photogrammetry or fringe projection systems, in order to record or assess their functionality. Adaptive measurements can be performed, which make it possible to make a statement about the object's functionality even before the object has been fully measured. As soon as a sufficiently reliable statement can be made, the adaptive measurement can be aborted to save time and thus costs. Conventional measurement, on the other hand, performs a complete measurement of the object. The quality of the statement about the measured objects depends on the functional status of the device for testing objects.

[0003] There are known methods for monitoring the functionality of a computed tomography system when recurring measurement tasks are to be performed. Using these previously known methods, monitoring the functionality of computed tomography systems is only possible with considerable effort.

[0004] The following is disclosed in document WO 2019 / 230040 A1: This X-ray thickness measuring device (12) is connected to a prediction data server (14) and equipped with an X-ray control power supply, an X-ray generator, and a detection unit. The prediction data server (14) is equipped with an detection unit (62), a prediction unit (64), and a storage unit (48). The detection unit (62) detects measurement information (56) from the X-ray thickness measuring device (12) that includes at least one of the following elements: a drive voltage value; a drive current value; a tube voltage value; a tube current value; and a detection signal. The prediction unit (64) generates statistics for measurement information (56) detected by the detection unit (62) as prediction data (68) for diagnosing abnormalities in the X-ray thickness measuring device (12).The storage unit (48) stores prediction data (68) generated by the prediction unit (64).

[0005] The following is disclosed in document US 2019 / 0200948 A1: The present disclosure relates to systems and methods for monitoring a medical device. The medical device may include a tube configured to generate rays and a detector configured to receive rays emitted by the tube. The tube may include an anode target and a filament. The detector may include a plurality of detection units. The method may include obtaining imaging data acquired by the detector via detection of the rays emitted by the tube. The method may also include determining a first feature parameter associated with the rays based on the imaging data. The method may further include monitoring the medical device based on the first feature parameter associated with the rays.The object of the invention is therefore to provide a computer-implemented method with which the functionality of devices for examining objects during adaptive and conventional measurements can be monitored with comparatively little effort. Main features of the invention are defined in claims 1 and 14. Further embodiments are the subject of claims 2 to 13.

[0006] In a first aspect, the invention relates to a computer-implemented method for monitoring the condition of a device for examining objects, wherein the examination of an object comprises determining measurement data by means of a measurement of the object and operating data of the device are determined during the examination of the object, wherein the method comprises the following steps: determining measurement data of the object by means of the device; determining operating data of the device during the step of determining measurement data of the object; determining at least one quality parameter from the measurement data; analyzing the operating data and the at least one quality parameter; and determining a condition characteristic value based on the analysis of the operating data and the at least one quality parameter for condition monitoring of the device, wherein the condition characteristic value indicates a condition of the device.

[0007] The invention thus provides a computer-implemented method in which the condition of the device is monitored by means of a condition characteristic value during or after a measurement of the object is carried out, based on an analysis of operating data of the device and a quality parameter of the measured data. This means that while the measurement of the object is still in progress, operating data of the device is already determined and analyzed together with the measurement data collected so far. The result of this analysis is a condition characteristic value that provides information about the condition of the device. The measurement of the object can be an adaptive measurement that can be stopped early as soon as sufficient data on the condition of the object is available, and / or optimized recording parameters for the further execution of the measurement can be determined based on the available measured data.Furthermore, the computer-implemented method according to the invention provides automatic monitoring of the device for examining objects. The computer-implemented method according to the invention automatically detects an error condition when the device for examining objects no longer functions correctly to measure objects with sufficiently high accuracy during an adaptive or conventional measurement process. This reduces the user's required involvement in monitoring the device for examining objects.

[0008] For monitoring purposes, all information that is present and available during operation of the device for examining objects, for example a computed tomography system, can be evaluated, even if it does not appear helpful at first glance. By simultaneously recording the operating data of the device for examining objects and the measurement data about the object using the device for examining objects, information from the monitoring can be used in adaptive or conventional measurements. Furthermore, the simultaneous recording of the operating data and the measurement data of the adaptive or conventional measurements ensures optimal monitoring and, if necessary, error states and their causes can be identified. By determining the condition parameter, the functional status of the device is monitored during the adaptive or conventional measurement process.

[0009] The status value can, for example, indicate whether the device's condition is OK or faulty, i.e., whether the device is in a normal state or a faulty state. Intermediate states are also possible, e.g., when individual detector pixels are defective, so-called bad pixels, which, however, do not yet have a significant impact on the measurement results. In this case, the status value would indicate that there is a certain, but not yet problematic, deviation from a normal state, which prevents the device from entering a faulty state. In this case, countermeasures can be initiated early to prevent a faulty state from occurring.

[0010] The measurement data can be final measurement results in the case of a dimensional measurement, e.g., the diameter of a bore. Furthermore, measurement data can also be raw data that is processed in the measurement chain—for example, in the case of a radiographic measurement, radiographic images or projection data, reconstructed volume data and / or surface data; in the case of a measurement using photogrammetry or fringe projection, images from the measuring camera or calculated surface data, for example.

[0011] According to one example, the step of determining a condition characteristic value may comprise the following sub-step: comparing the operating data and the at least one quality parameter with predefined comparison values ​​for the operating data and the at least one quality parameter to determine a condition characteristic value.

[0012] Predefined comparison values ​​are used for the operating data and the at least one quality parameter so that the computer-implemented method can decide whether a state of the device is already a faulty state or is still too normal. The comparison is not limited to predefined fixed comparison values ​​for individual variables. It can also be implemented in such a way that several variables are considered together. For example, if a first variable A or a second variable B individually deviates by 5% to 10% from a respective normal value, this is not necessarily a faulty state. However, if in this example both variables deviate by 5% to 10% from the nominal value at the same time, this can be identified as a faulty state. The first variable A and the second variable B can be operating data and / or quality parameters.

[0013] Alternatively or in addition to the use of predefined comparison values, a comparison can be performed based on pre-determined sample data for the operating data and the at least one quality parameter, wherein the sample data is derived from training data. The step of determining a condition characteristic can, for example, further comprise the following substep: comparing the operating data and the at least one quality parameter with pre-determined sample data derived from training data for the operating data and the at least one quality parameter to determine a condition characteristic.

[0014] The pre-determined sample data thus serves as a kind of reference for the operating data and the at least one quality parameter determined during the measurement. This training data or sample data can preferably be used to train a machine learning method, e.g., a neural network. The sub-step is then performed by this machine learning method. Measurements of a reference object, for example, can be used to generate training data.

[0015] The training data can therefore be collected from various sources. For example, error states can be identified locally on the monitored computed tomography system, e.g., due to unusually high measurement deviations during the measurement of a master part. This error state and the characteristics of the monitored variables at that time can be incorporated into the training data. Furthermore, training data from computed tomography systems of the same or at least similar design can be collected in the same way. This expands the data basis, although the significance may be partially limited due to possible differences in the properties of the computed tomography systems. Furthermore, with the same advantages, data from computed tomography systems with significantly different configurations can also be incorporated into the training data.

[0016] In order to create a certain degree of comparability for the monitored data from different sources, they can be normalized to typical value ranges.

[0017] A certain period before or after the identified error condition, the time course and the values ​​of the monitored variables can also be included in the training data.

[0018] Furthermore, not only error states can be included in the training data, but also functional states in order to identify them and to differentiate them from the error states.

[0019] Various pattern recognition or machine learning methods can be used to determine from the monitored variables and the available training data whether a fault condition is currently present or imminent, and if so, its cause. These can include artificial intelligence methods, e.g., artificial neural networks, deep learning, support vector machines, Bayesian classifiers, nearest neighbor classification, etc. The choice of at least one method used is based on the training data to identify patterns of fault conditions in the comparatively large volume of sometimes confusing monitored data.

[0020] The training data can further be generated, for example, from measurement data which are determined by means of the device by means of at least one measurement within a predefined time interval before and / or after the determination of measurement data of a reference object, wherein the reference object has a known geometry and / or a known property.

[0021] The measurements from which the training data is derived do not have to be measurements of a reference object, nor do they have to have the same or a similar geometry to the object to be examined or tested. According to one example, the reference object can have the same or a similar geometry to the object to be examined or tested. Only the geometry or the results of the measurements of this reference object need to be known. By comparing the measurement data and the results of the measurement of the reference object, the condition of the device can be deduced. In this case, larger deviations are an indication of a fault. In addition to the geometry, other properties of the measured object can also be relevant, e.g. defects inside the reference object, properties of any fiber composite materials used such as fiber diameter, fiber density and fiber orientation, or material properties such as the possiblelocal density.

[0022] A reference object can also be referred to as a master part. Master part measurements are performed regularly on a variety of computed tomography systems to test the functional status of the system. Typical repetition rates for such measurements range from several times a day to once a month.

[0023] A master part is an object with known geometric properties. These can include external and internal geometric dimensions as well as material properties such as density, fiber density, thickness, fiber orientation, porosity, or other properties. To verify the functional status of the computed tomography system, a measurement of this part is performed, and the measurement results, which may be task-specific, are compared with the reference values ​​or calibration values, which ideally come from a measurement with lower measurement uncertainty, e.g., with a more precise sensor. The extent of the deviations encountered allows conclusions to be drawn as to whether the computed tomography system is functional for the defined task.If the computed tomography system can successfully perform this measurement task, it can be assumed that it is in "good" condition and can also perform other similar, and sometimes less similar, measurement tasks. To be able to make a statement about the condition by comparing the measurement results with the reference values, limit intervals for the measurement error are usually defined.

[0024] These measurements can be used for the training data according to this example.

[0025] The information to be included in the training data or correlated to determine whether a fault condition existed can be derived from the measurement deviations of the master part measurement. For example, the assessment of a "good" or "bad" condition can be made based on whether defined limit intervals of the measurement deviation are not met. Furthermore, instead of such a "binary" classification, it is also possible to assess the condition of the measurement system with more fine resolution, for example, "on average, 47% of the measured values ​​are within the permitted limit intervals" or "30% of the tested measured values ​​are outside the limit intervals."

[0026] The respective information regarding whether a fault condition exists is assigned to the corresponding measurement data of the monitored variables. The measurement data during the measurement of the master part can be used for this purpose, as this is where the analyzed measurement deviations originated.

[0027] If it can be assumed that the state of a computed tomography system will not change significantly over a shorter period of time, measurement data from measurements in a certain temporal vicinity of the master part measurement can also be included in the training data, with the assignment to the state derived from the master part measurement. This has the advantage that data from types of measurements that are not represented by the master part can also be included in the training data, e.g., if the master part is made of aluminum while the "temporally surrounding" measurements are made of plastic. This generates training data from plastic measurements. How large this "temporal environment" can be reasonably selected depends on the stability of the computed tomography system.

[0028] In particular, if data collected before the master part measurement, in which a fault condition may be detected later, is included in the training data, a prediction of fault conditions to be expected in the future can be made, for example, based on patterns in this data.

[0029] Alternatively or additionally, the cause of the deviations can also be assigned to the individual master part measurements, especially those in which error conditions were detected, so that the system can be trained to predict these deviations as well. This can also be done manually.

[0030] The step of determining a condition characteristic may further comprise, for example, the following substep: determining whether the condition characteristic indicates a fault condition of the device.

[0031] In this substep, the condition characteristic is read to determine whether it indicates a fault condition in the device. This serves to identify the fault condition, if necessary. Identifying the fault condition can, for example, indicate which of the operating data or at least one quality parameter forms the basis for the fault condition. In this way, the fault in the device for examining objects can be automatically located using the computer-implemented method.

[0032] An error state is defined as a condition in which the measurement accuracy decreases so significantly that the measurement tasks to be performed can no longer be carried out sensibly. This can therefore be defined on a measurement-task-specific basis, e.g., based on the tolerances to be checked according to the measurement plan, but also specifically for a computed tomography system, e.g., based on a global specification or minimum accuracy. Depending on the definition, a moderate deterioration in the condition of the computed tomography system, which only results in a slight deterioration in measurement accuracy, may not be considered an error state in a computed tomography system with a comparatively poor specification or in measurement tasks with comparatively large tolerances. Nevertheless, this deterioration can be identified, and an initial warning can be issued or countermeasures can be initiated.

[0033] According to another example, the method may further comprise the following step, if the condition characteristic indicates that a fault condition of the device may exist: determining measurement data of a reference object using the device; and analyzing the measurement data of the reference object to determine information as to whether the fault condition of the device exists.

[0034] In this example, a reference object is measured using the device for object inspection to obtain measurement data from the reference object when a faulty state of the device is suspected. One option is to interrupt the ongoing measurement to measure a reference object. Alternatively, the measurement can be waited until the completion of any ongoing measurement before measuring a reference object. The obtained measurement data from the reference object is used to perform an analysis to determine whether the device is in a faulty state.

[0035] This means that if it turns out that an error condition could exist and it cannot be determined from the current data whether an error condition actually exists or what the cause of the error condition is, a type of self-diagnosis can be requested or carried out.

[0036] For this purpose, a suitable reference object, known as a master part or standard, ideally with a known geometry, is measured under suitable acquisition parameters to generate measurement data that can be analyzed in a targeted manner to obtain more information about the causes of the defect. The selection of which master part or standard should be measured and under which acquisition parameters in order to systematically generate as much new information as possible, or to obtain the missing information, can be made within the process. For example, if several possible causes of the defect are still possible based on the available data, a measurement can be selected that can differentiate between precisely these causes.

[0037] If, for example, increased blur is detected in the volume data and it cannot be clearly determined whether this is due to an enlarged focal spot or incorrect geometric calibration, a single radiography of a small object, such as a small sphere or a wire, can be automatically performed at high geometric magnification. Reconstruction is not necessary. If the blur of this projection is not increased, the focal spot can be ruled out as the cause. In this case, incorrect geometric calibration is the likely cause, as it does not affect projection blur.

[0038] In another example, if it is unclear whether incorrect geometric calibration or drift effects are the cause of an error, a fast scan of an object (e.g., with a small number of projections) and a slow scan (e.g., with a large number of projections) can be automatically performed with otherwise identical acquisition parameters. If the error only occurs during the slow scan, the cause is probably a drift effect. If the problem still occurs during the fast scan, the cause is probably not a drift effect, but incorrect geometric calibration. Through further tests, e.g., at different magnification levels, the procedure can narrow this down further if necessary.

[0039] Ideally, these measurements can be performed automatically by changing parts. Alternatively, or additionally, a user can be instructed to perform this measurement. Furthermore, ideally, various master parts or standards are available for this measurement.

[0040] These self-analysis steps can be performed as soon as there is a suspicion of an impending error condition.

[0041] In another example, the method may further comprise the following step: determining a fault cause parameter at least from the condition characteristic, wherein the fault cause parameter indicates a possible fault cause for a fault condition.

[0042] The error cause parameter can be used to determine a possible cause for an error condition. To determine the error cause parameter, the following step can preferably be performed: comparing the operating data and the at least one quality parameter with predefined comparison values ​​for the operating data and the at least one quality parameter to determine the possible cause of the error. This allows the operating data and quality parameters responsible for the error condition to be determined from the comparison of the operating data and the at least one quality parameter with corresponding predefined comparison values. In this way, a possible cause of the error condition can be determined. This limits the search for causes of the error condition for a user and saves time and money.

[0043] Therefore, in the event of an error condition, not only is the information represented by the condition characteristic value output, indicating "that something is wrong," which is helpful information in itself, as the user now at least knows that troubleshooting must be initiated, but it also provides a hint as to where the troubleshooting should begin. Ideally, it can be predicted before an error condition occurs that a fault condition will occur in the near future. For example, in a computed tomography device, the X-ray source begins to exhibit fluctuations in power, which is detected by a larger variation in the grayscale level in the background of the projection data. Based on this example, it can therefore be assumed that major problems with the X-ray source are to be expected in the near future.

[0044] According to an example, it is possible to conclude from the monitored variables whether an error condition exists and, if so, what the cause might be, for example by using a table or decision tree.

[0045] An example of this could be specifying certain limit values ​​for the variables to be monitored. This check determines whether these limit values ​​are exceeded or not met. This provides an initial indication of a potential error condition. Based on the information about which variables do not comply with the limit values ​​and in which direction and to what extent the limit values ​​are exceeded, possible causes of errors can be derived based on prior knowledge.

[0046] For example, it may be determined that the volume data is unusually blurred, i.e. the point spread function is unusually wide. At the same time, however, it was found that no unobtrusively large blur can be detected in the projection data. Thus, it is unlikely that an enlarged spot, e.g. due to defocusing of the electron beam, is the cause, as this would also be reflected in the projection data. One possible cause could be system misalignment, i.e. the position and orientation of the source, detector, and turntable in space is not accurately known when reconstructing the volume data. This results in blurred volume data. This analysis result can be passed on to the user. This is an example of the use of a decision tree to determine the cause of an error.

[0047] For each monitored variable, possible error causes could be specified that could lead to a violation of the limit values. Similarly, if two or more variables violate limit values ​​simultaneously, corresponding possible error causes could be specified. The other way around, it is also possible to specify for each error cause which variables would typically cause a violation of the limit values. These are examples of the use of corresponding tables to determine the cause of an error.

[0048] These specifications could be entered manually. However, the corresponding values ​​for past error conditions, possibly with a known cause, that were outside the limit intervals can also be stored, thus building up a database over time, possibly automatically.

[0049] In another example, no fixed thresholds need to be defined; instead, empirical values ​​that can be automatically adjusted can be used to identify anomalies. Anomalies could be, for example, if the temperature in the measurement volume is 3°C above the average for the last week, or if the noise in the background of the projections is two standard deviations above the average for the last month.

[0050] In summary, this approach is most suitable for detecting those relationships between monitored variables and error states (including cause, if applicable) that are comparatively obvious and describable for an expert.

[0051] The method may, for example, further comprise the following step if the error cause parameter could not be determined: determining measurement data of a reference object by means of the device; and analyzing the measurement data of the reference object to determine information for determining the error cause parameter.

[0052] This step can be performed alternatively or in addition. If no error cause parameter could be determined, a reference object is measured using the device in order to determine reference measurement data. This reference measurement data is analyzed to determine the cause of the error. The cause of the error can be determined, for example, from the deviations between the results of the measurement data of the reference object and the measurement data of the object to be measured compared to known reference values. The operating data can also be analyzed in a similar way. The advantage of using a reference object is that more or more precise reference values ​​are typically available. This means that even if the error cause parameter could not be determined in previous steps, the computer-implemented method can use this step to determine a fault cause parameter and make troubleshooting easier for the user.

[0053] Further, for example, if in the step of determining a fault cause parameter a fault cause parameter is determined which indicates that a geometric calibration of the device has an accuracy value which is located outside a predefined accuracy value interval, the method may further comprise the following step: calibrating the device.

[0054] According to this example, a calibration of the device is performed if the error cause parameter indicates that the geometric calibration of the device is faulty, i.e., has an accuracy value that lies outside a predefined accuracy interval. This means that if the computer-implemented method determines that the error cause is the faulty calibration of the device, the computer-implemented method automatically calibrates the device to increase the accuracy value for the geometric calibration of the device so that the accuracy is again within the predefined accuracy value interval. The accuracy value interval can, for example, be between 95% and 100%, with 100% representing the maximum accuracy of the geometric calibration. However, any interval between 0% and 100% can be selected.

[0055] If the error condition indicates that the geometric calibration is compromised, increased measurement uncertainty can be expected. The accuracy interval can be defined with respect to general requirements or a specific measurement task. For dimensional measurements, the uncertainty of the measurement itself can be estimated, e.g., based on the voxel size error. For pore analysis, limits for data uncertainty can be defined.

[0056] If it turns out in this way that the functional status of the computed tomography system is inadequate, a so-called calibration process for geometric calibration can be requested or carried out automatically. This can mean that a user is prompted to carry out such a calibration process. If possible, this can be carried out by the procedure. A calibration process usually means that a special standard, which usually consists of one or more spheres, is positioned in the beam path, possibly in different positions, and is x-rayed. The geometry of the computed tomography system can be recorded on the basis of the radiographic image, e.g. based on the depicted positions of the spheres. For axial CT, this can generally be modeled using nine degrees of freedom: the position and orientation of the detector and the position of the X-ray source, each with respect to the rotary table.For free trajectories, this geometric calibration can be significantly more complex, involving a larger number of degrees of freedom. The acquisition of this geometry is referred to as geometric calibration. Reconstruction can be performed or parameterized on this geometry in such a way that the most error-free and sharp volume data can be calculated.

[0057] This calibration process can be carried out completely automatically, e.g., by positioning the standard used in the measuring volume or on the rotary table using a part changer controlled by the measuring system.

[0058] For example, calibration can also be performed prophylactically if a fault condition is indicated by the device. This means that even if the cause of the fault is unknown, i.e., if the fault-causing parameter could not be determined, a calibration of the device can be performed as a precautionary measure. With a preventative calibration of the device, it can at least be ruled out that the geometric calibration of the device is the cause of the fault if the newly determined condition parameter after calibration continues to indicate a fault condition.

[0059] However, a geometric calibration can also be performed based on the projection data acquired from a normal measurement, completely without a real standard. This allows, for example, the geometric calibration that best fits the data to be determined based on the available measurement data or projections. This can also be performed retrospectively for measurements that have already been performed.

[0060] Furthermore, the device can, for example, have an automatic object changing unit, wherein the method further comprises the following step: exchanging a reference object for the object into the device for examining objects in order to determine measurement data about the reference object and / or exchanging a calibration object for the object into the device for examining objects in order to calibrate the device for examining objects.

[0061] In one example, the object changing unit has a robot arm that can grasp an object, possibly including a holder for the object, and place it in the beam path or on the turntable.

[0062] In another example, the object changing unit comprises at least one turntable that can approach a battery of objects or holders from below and, if necessary with a suitable holder, can pick up the objects. In none of the examples is any user action required.

[0063] According to an example, the method may further comprise the following steps: performing at least the step of determining measurement data of the object with a first object, wherein the measurement is a transmission measurement; performing at least the step of determining measurement data of the object with a second object, wherein the measurement is a transmission measurement; determining recording parameter sets of the device that are identical for determining measurement data for the first object and for determining measurement data for the second object; determining first projection representations from the measurement data for the first object using the determined recording parameter sets and determining second projection representations from the measurement data for the second object using the determined recording parameter sets;and analyzing at least one first quality parameter associated with one of the first projection representations and at least one second quality parameter associated with at least one of the second projection representations for differences. In an alternative or additional example, the method may further comprise the following steps if a recording parameter set of a first object and a recording parameter set of a second object are at least partially not identical: defining recording parameter sets of the device for determining measurement data; performing at least the step of determining measurement data of the object with the first object using the defined recording parameter sets to determine first projection representations, wherein the measurement is a transmission measurement;Performing at least the step of determining measurement data of the object with a second object, using the defined recording parameter sets to determine second projection representations, wherein the measurement is a radiative measurement; and analyzing at least one first quality parameter associated with one of the first projection representations and at least one second quality parameter associated with at least one of the second projection representations for differences; wherein a geometry of the second object deviates from a geometry of the first object within a predefined tolerance interval.

[0064] Radiographic measurements can be, for example, X-ray measurements, e.g., radiography or computed tomography.

[0065] Acquisition parameters of a projection can be the transmission geometry of the projection and / or setting options that can be adjusted when transmitting radiation through an object, for example the current and voltage of the X-ray source, the exposure time or the selected filter.

[0066] A transmission geometry describes the direction in which the object is irradiated during the transmission measurement, the position of the irradiated area, and the magnification. In an X-ray measurement or transmission, this geometry is therefore described by the position of the X-ray source and detector relative to the measurement object, resulting in nine geometric degrees of freedom. Translational freedom results in three degrees of freedom for the X-ray source and detector, and rotational freedom results in three additional degrees of freedom for the detector. A transmission geometry can be defined with respect to the object or with respect to the device used to examine objects.

[0067] In the step of analyzing at least one first quality parameter associated with one of the first projection representations and at least one second quality parameter associated with at least one of the second projection representations for differences, not only quality parameters of the projection representations themselves but also, for example, quality parameters of volume data reconstructed from the projection representations can be analyzed.

[0068] This allows monitoring measurements in which at least partially different acquisition parameters are used, for example, partially different radiographic geometries or trajectories for measurement objects of the same or similar geometry. In a first alternative, a set of projection images is identified for the analysis of the functional status between at least two measurements for which the radiographic geometry and the remaining acquisition parameters were identical. These are compared and evaluated to determine any difference in the quality parameters of the two measurements. In the second alternative of the example, projection images with the same radiographic geometry and the same settings are specifically acquired and analyzed.

[0069] This allows a reference or a common database to be established between different measurements, especially adaptive measurements.

[0070] In the second alternative, comparability between a large number of measurements is achieved by deliberately creating identical projections. This requires additional time. Conventional monitoring methods that analyze this data and, for example, check for drifts can be used for this purpose. This can involve either directly comparing identical projections or determining and analyzing volume data based on a reconstruction from identical projections. The methods already mentioned above can be used to analyze this data for monitoring purposes.

[0071] In the first alternative, such a comparison is possible between at least two measurements, whereby these measurements can be "adjacent" in time. A comparison is therefore initially only possible for a comparatively small number of measurements. If, for example, a comparison of a quality parameter is known between measurements A and B, as well as between B and C, a comparison of A and C can be performed, at least qualitatively. For this purpose, a quality parameter should be used that exhibits a certain transitivity in this respect. In this way, many different measurements can be compared or monitored "in a chain," thus enabling long-term monitoring, even if each individual measurement can only be directly compared with a comparatively small number of measurements. This analysis can also be performed based on projections and / or reconstructed volume data.

[0072] Both alternatives require that the measurement objects of the individual measurements have at least a similar geometry, e.g., the same target geometry.

[0073] A combination of these two alternatives would be to define certain standard projections during the measurement, from which the required ones can be flexibly selected for each measurement. Since this limits the selection of "permitted" projections, if enough standard projections are defined without severely restricting flexibility, the probability that two measurements can be compared increases, or the average number of projections that are comparable between two measurements increases. These standard projections can also be measured separately or specifically during a calibration process to increase accuracy. This is particularly advantageous for robotic CT, where the positioning accuracy is comparatively low, but the "selection" of radiographic directions is extremely large.

[0074] In a further example, the method may comprise the following steps: performing at least the step of determining measurement data of the object with a first object, wherein the measurement is a transmission measurement; performing at least the step of determining measurement data of the object with a second object, wherein the measurement is a transmission measurement, wherein a different transmission geometry is used than for the first object; determining projection representations from the measurement data of the first object; determining projection representations from the measurement data of the second object; comparing at least one blur between the projection representations of the first object and the projection representations of the second object; and analyzing the at least one blur to determine the condition characteristic value.

[0075] In this example, measurements are monitored using different radiographic geometries or trajectories, as may be the case with adaptive measurements, for example. The image blur is determined from the projection data and analyzed to monitor the functional status.

[0076] For this purpose, it is advantageous to compare projection images with the same setting options, although the radiographic geometry may differ. Regarding the radiographic geometry, it can be advantageous to compare projections with the same or similar geometric magnification, although the other geometric parameters may differ. This ensures, for example, that in X-ray images, the blur caused by the expansion of the X-ray spot is reflected similarly in the measurement data, enabling a more reliable comparison.

[0077] The advantage of analyzing blur in projection image data is that it can be compared across multiple measurements, regardless of the geometry of the object being irradiated or the direction from which it is irradiated. These remaining influencing factors that vary between measurements, such as the acquisition parameters, especially the resulting focal spot size, and the geometric magnification, can be relatively easily quantified, and their influence on the blur of the projections can be calculated back.

[0078] However, it is not trivial to determine the blur caused by the imaging system or the blur caused by the point spread function in the projection data. The problem is that the radiation does not form a sharp edge from light to dark in the projections, since the gray value depends on the radiation length. If an object is irradiated, the radiation length usually increases continuously. For example, if a sphere is irradiated, the gray value profile exhibits a gradient from the edges of the sphere to its center. This gradient suggests blur, which is caused by the geometry of the object being irradiated and not by the blur of the computed tomography system itself, which is what we want to determine. A real gray value curve results as a first approximation from the theoretical gray value curve caused by the object, which is convolved with the point spread function of the imaging system.However, this point spread function is extremely small compared to the grayscale gradients caused by the object. This makes it difficult to separate these two effects, especially when the geometry of the object being illuminated is not precisely known. However, if the theoretical grayscale gradient caused by the object forms a sharp edge, the blur caused by the point spread function can be more easily determined. This case rarely occurs, however, because many surfaces are either not sufficiently flat and exhibit certain deviations, which in turn cause false blur, and it is also not known exactly in which, if any, projections are correspondingly "grazing" through the radiation. Identifying the areas to be evaluated is therefore not trivial.

[0079] If, in an example, the geometry of the measurement object is known in principle, e.g., through prior knowledge such as a CAD representation of the object, a specific direction of transmission can be targeted in which the sharpest possible edge forms in the transmission images, which can then be analyzed for blurriness. Prior knowledge of the expected blurriness caused by the geometry can also be used, and the influence or blurriness of the measurement system can be separated from the measured blurriness.

[0080] In another example, if the geometry is unknown in advance, an evaluation of the possibly preliminary measurement data can be performed, e.g., a reconstruction and surface determination, to determine the geometry. In this way, suitable regions in the projection data that exhibit suitable edge transitions can be identified.

[0081] In another example, the spatial frequencies occurring, including their respective amplitude and contrast, can be analyzed or statistically evaluated in all projections. In a histogram, for example, the peak at the highest spatial frequency, which is clearly visible, could be a measure of blur. This could be, for example, the transitions that coincidentally meet a transition with which the point spread function can be adequately determined. High-frequency components caused by noise may need to be identified and excluded from the analysis.

[0082] In another example, a suitable, ideally known, geometry can be captured in addition to the measurement object, e.g., a small sphere or a wire. Accordingly, the projection data contains a suitable edge transition that can be evaluated.

[0083] According to another example, the method may further comprise the following step: determining an estimated value of an uncertainty of a measured variable of the object determined from the measured data by means of the operating data and the at least one quality parameter.

[0084] To estimate the measurement uncertainty, information can be compiled from a variety of sources, e.g., empirical values, an analysis of current data quality, or statistical analyses of the stability of a measurement result. Monitoring the functional status of the computed tomography system in use can provide additional helpful information that can be incorporated into the uncertainty assessment.

[0085] For example, monitoring may reveal that the geometric calibration already exhibits a certain degree of uncertainty, which in turn negatively impacts the measurement uncertainty of the measurements to be performed. Accordingly, an increased "baseline level" of measurement uncertainty can be considered when determining the measurement uncertainty in order to determine more realistic uncertainty values. This can also take into account the fact that an uncertainty or slight disturbance in the geometric calibration has different effects on different measurands, for example, a stronger effect on longer measurands.

[0086] In another example, monitoring may detect an increased noise level in the data, e.g., due to a lower X-ray tube power. This increases the uncertainty of the measured values. The sensitivity of the individual measured values ​​to the error condition can also be taken into account: for example, the measurement of a shape deviation reacts significantly more strongly to an increased noise level than the measurement of the distance between sphere centers.

[0087] During the step of determining measurement data of the object by means of the device, the following steps can be performed, for example: determining preliminary measurement data and / or at least one preliminary quality parameter from the measurement data; adapting the step of determining measurement data of the object by means of the device taking into account the preliminary measurement data and / or the at least one preliminary quality parameter from the measurement data.

[0088] In this example, the information obtained during a measurement can be used during the measurement to optimize the measurement or the recording parameters to be used.

[0089] According to a further example, the method may further comprise the following step: determining whether a measured variable of the object determined from the measured data lies within a predefined tolerance range, preferably taking into account an uncertainty of the measured variable.

[0090] In this case, a reliable statement means that the measurement result, taking into account the uncertainty, is either safely within or safely outside the tolerance range.

[0091] This allows you to identify those areas where a reliable conclusion cannot yet be drawn. Additional information can then be collected primarily for this area during the measurement. This is not necessary for areas where a reliable conclusion can already be drawn. This saves time when recording the measurement data.

[0092] Furthermore, the measuring process can be terminated, if necessary prematurely, if a reliable statement is possible for all sizes and / or for a critical size it is possible to reliably state that the size is outside the tolerance range.

[0093] The determination of uncertainty can use information from monitoring.

[0094] The information from the monitoring, i.e., the current state of the device used to examine objects, e.g., a CT system, can be fed into a digital twin of the device to keep it up-to-date and as identical as possible to the real device. This has advantages when using the digital twin, for example, for simulative measurement uncertainty determination.

[0095] 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.

[0096] 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.

[0097] 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. They show: Fig. 1 shows a flowchart of an example of the computer-implemented method with optional steps; Fig. 2 shows a flowchart of further optional steps of the method; Fig. 3a shows a flowchart of further examples of the method with further optional steps; and Fig. 4 shows a flowchart of a further example with further optional steps.

[0098] In the following, the computer-implemented method for condition monitoring of a device for examining objects is referenced in its entirety by the reference numeral 100. The examination of the object by the device comprises determining measurement data by measuring the object and, during the examination of the object, determining operating data of the device.

[0099] Figure 1shows an example of the computer-implemented method 100. In a first step 102, measurement data of the object are determined using the device for examining objects. The measurement can be, for example, a radiographic measurement using a computer tomography scanner. The measurement data can thus be, for example, two-dimensional projection image data. Alternatively or additionally, the measurement data can be volume image data reconstructed from the projection image dataset or surface data determined based on the volume image data. If the device uses a different measurement method or a different sensor, correspondingly different data is defined as measurement data.

[0100] During step 102, step 104 is carried out simultaneously, in which operating data of the device are determined.

[0101] Operating data of the device can include, for example, the dark current and bright current of a detector in the device, i.e., pixel-resolved information used for light-dark adjustment or flat-field correction. Furthermore, this can include, for example, the number and distribution of defective pixels in the detector or the measuring camera. These are usually identified using certain methods and stored as a mask.

[0102] In relation to a rotary table, these can be operating variables such as voltage and current. Furthermore, these can be geometric parameters such as tilt and concentricity. External sensors, e.g., laser interferometers, can also be used. "Internal sensors" can also be used by irradiating a suitable or dedicated object at various angular positions and evaluating the image on the detector. A spiral arrangement of spheres, for example, is well suited for this purpose, as the spheres on the detector do not shadow each other, and their position in the beam path can therefore be accurately calculated. Such a measurement is often performed during a regularly performed calibration process. Thus, corresponding properties or variables / parameters of the most recently performed calibration process can be monitored.

[0103] With respect to an X-ray tube / source of the device, the operating data may be the current curve, the voltage curve, the current of focusing and centering, or the difference between tube and target current.

[0104] Operating data can also include device information, for example, such as temperature, where various measurement points are conceivable, e.g., outside the device, inside a radiation protection cabin, on a detector, a tube, a turntable, on the axes, or on the measurement object. Other operating data include the time of day, the length of time since the last extended period in which the measuring device or its components were deactivated, the length of time since the last light-dark balance, the time since the last maintenance, or the time since the last calibration process.

[0105] Furthermore, basic information about the configuration, manufacturer, and design of the computed tomography system used, e.g., cone beam, fan beam, or robotic CT, can be used as operating data. Furthermore, the acquisition parameters selected during the measurement, e.g., the scan duration, exposure time per projection, image averaging per projection, nominal voltage, nominal current, pre-filtering, geometric magnification, or nominal voxel size, can be considered as operating data.

[0106] Furthermore, meta information about the object to be measured, e.g. the target geometry or the material, can be operating data.

[0107] In iterative reconstruction, the convergence speed of the reconstruction can be used as operating data.

[0108] In step 106, at least one quality parameter is determined from the measured data. This quality parameter can be, for example, a possibly local match between the gray values ​​of the projections and a forward projection of reconstructed volume data. This can be a measure of the consistency of the volume data with the projection data. If the match is low, artifacts or image errors are to be expected in the volume data.

[0109] Further analysis of the evenly illuminated background in projections can be performed with regard to homogeneity, mean gray value, signal-to-noise ratio, or noise power spectrum. This can also include blur or the point spread function in the projections.

[0110] Quality parameters can be derived from measurement data available, for example, as volume data, as an analysis of gray values ​​of the volume data with regard to their quality in the vicinity of the surface, in particular the focus can be on the blur or the point spread function and the deviation from an ideal model gray value transition, since the quality parameters determined in this way are determined precisely on the basis of the data that influence the result of the surface determination, whereby the surface in turn is important for many evaluations, the significance of these quality parameters is particularly high.

[0111] Furthermore, an analysis of homogeneous regions of the volume data, such as the background or the homogeneous material, can be performed with regard to signal-to-noise ratio and homogeneity. This analysis is facilitated by the existence of prior knowledge about these image regions through a model, ideally with constant gray values ​​in the material and background. Deviations from this are then presumably due to data quality and not to changes in the object.

[0112] Furthermore, an analysis of a gray value histogram of measurement data available as volume data can be carried out, e.g. with regard to the distance and width of the individual material peaks.

[0113] Once defects have been detected, the volume data in these areas can be analyzed using knowledge about the existing defects, e.g., regarding blur, point spread function, or contrast.

[0114] Furthermore, an analysis of the surface determined from the volume data can be carried out, e.g. with regard to surface properties such as shape deviations and parameters for waviness or roughness.

[0115] The acquisition of the above-mentioned variables, both operating data and quality parameters, is generally possible with different time resolutions, although the options may vary depending on the variables to be monitored. Sampling can be continuous, quasi-continuous with a high sampling rate, or at a fixed temporal frequency. Furthermore, one measured variable can be acquired per recorded projection or for each predefined number of recorded projections. Furthermore, a measured variable can be acquired during specific events, such as the start and end of a measurement, the opening of the protective cabin of the computed tomography system, or after performing a light-dark balance.

[0116] In a further step 108, the operating data and the at least one quality parameter are analyzed. The analysis determines whether a fault condition exists in the device for examining an object.

[0117] For this purpose, in step 110, a condition characteristic value is determined based on the analysis of the operating data and the at least one quality parameter. The condition characteristic value indicates the condition of the device. This can be a faulty condition or a normal condition. Furthermore, the condition characteristic value can also indicate one or more conditions between the faulty condition and the normal condition. These conditions between the faulty condition and the normal condition describe ambiguous conditions in which it is unclear whether they are a faulty condition or a normal condition. The closer an intermediate condition is to the normal condition or the faulty condition, the higher the probability that a normal condition or a faulty condition exists.

[0118] Step 110 can optionally include a sub-step 116, in which it is determined whether the status characteristic indicates a fault condition of the device. This means that when determining the status characteristic, the status characteristic is not initially evaluated further. Only after determining in sub-step 116 which status characteristic is now present is the status characteristic evaluated.

[0119] If it has been determined that an intermediate state has been determined using the characteristic state value, measurement data of a reference object can optionally be determined using the device in step 118. For this purpose, in a further optional step 126, a reference object for the object can be exchanged into the device for examining objects in advance in order to determine measurement data about the reference object. The device has an automatic object exchange unit for performing this step 126.

[0120] In a further optional step 120, the measured data of the reference object and, if necessary, the operating data during the measurement of the reference object can be analyzed. This allows information to be determined as to whether a fault condition exists in the device. This allows the information provided by the condition characteristic, which merely indicates an intermediate state indicating that a fault condition may exist, to be improved.

[0121] In a further optional step 122, a fault cause parameter can be determined from the status characteristic. The fault cause parameter indicates a possible cause of the fault condition. However, in many cases, the fault cause parameter does not provide precise information about the cause of the fault condition.

[0122] If, in step 122, a fault-cause parameter is determined that indicates that a geometric calibration of the device is incorrect, i.e., if the geometric calibration of the device has an accuracy value that lies outside a predefined accuracy value interval, the further step 124 is performed, in which the device is calibrated. For this purpose, step 126 can optionally be performed, in which case the device has an automatic object exchange unit. In step 126, which is performed for step 124, a calibration object for the object is exchanged into the device for examining objects.

[0123] The calibration object has a known geometry so that the device for examining objects can be calibrated accordingly.

[0124] In a further optional step 158, an estimated value can be determined that indicates an uncertainty of a measured variable of the object determined from the measurement data. The estimated value is determined based on the operating data and the at least one quality parameter.

[0125] In a further optional step 160, preliminary measurement data are determined during step 102. Alternatively or additionally, at least one preliminary quality characteristic is determined from the measurement data obtained in step 102.

[0126] In a further optional step 162, step 102 is adapted using the preliminary measurement data or the preliminary quality parameter. This performs an adaptive measurement whose measurement parameters can be changed during the measurement process based on the results obtained so far, which can be based on preliminary measurement data or preliminary quality parameters.

[0127] Furthermore, after steps 160 and 162, the optional step 164 can be performed, in which it is determined whether a measured variable of the object determined from the measurement data lies within a predefined tolerance range. A measured variable of the object can be, for example, the length of a side of the object. Furthermore, this can also be, for example, a thickness or density at a specific position or a specific area within the object. If the measured variable lies within a predefined tolerance range, whereby the predefined tolerance range can be specified by a target geometry, the adaptation of the measurement or step 102 is considered successful.

[0128] Furthermore, step 164 can be performed taking into account an uncertainty of the object's measured variable. This means that the adaptation is only considered successful if, even taking into account the uncertainty of the measured variable, this measured variable lies within the predefined tolerance range.

[0129] Figure 2 shows a flowchart of step 110 with optional substeps 112 and 114. The two substeps 112 and 114 can each be provided independently of each other.

[0130] Substep 112 involves a comparison of the operating parameters with predefined comparison values ​​for the operating data and a comparison of the at least one quality parameter with predefined comparison values ​​for the at least one quality parameter. The comparison serves to determine the condition characteristic value.

[0131] The computer-implemented method 100 can use the comparison to determine whether there are deviations between the operating parameters and their comparison values, or whether there are deviations between the at least one quality parameter and its comparison value. If deviations exist, this may indicate a deviation from the normal state of the device. The condition characteristic value is therefore set to indicate either an intermediate state or a fault state.

[0132] Substep 114 involves a comparison of the operating data with pre-determined sample data for the operating data derived from training data, and a comparison of the at least one quality parameter with pre-determined sample data for the at least one quality parameter derived from training data. This comparison also serves to determine the condition characteristic value.

[0133] Here, the computer-implemented method 100 can use the comparison from sub-step 114 to determine whether there are deviations between the operating parameters and the sample data for the operating parameters, or whether there are deviations between the at least one quality parameter and its sample data. The sample data is based on training data used to determine the sample data prior to sub-step 114. Deviations between the operating data and their sample data, or between the at least one quality parameter and its sample data, indicate a deviation from the normal state of the device. The condition characteristic value is therefore set to indicate either an intermediate state or a fault state.

[0134] The sample data can be determined, for example, from a simulation that uses the training data as input. Simulation tools can be used for this purpose, which can realistically recreate the entire measurement or radiographic process, including reconstruction and further data analysis. In this way, a large amount of training data can be generated with little effort and in a targeted manner. This data contains defined error causes and their effects on the measurement data. Any number of different error causes can be simulated, both separately and in various combinations.

[0135] The advantage is that the specific properties of the device used to examine the object, e.g., a CT system, can also be simulated. This ensures that the generated training data is more closely matched to the monitored CT system. Measurement-task-specific training is also possible by irradiating the same objects that are to be measured in real-world operation during the simulated measurements.

[0136] Alternatively, a reference object is first examined to determine whether a fault condition exists. The geometry of the reference object may differ from the geometry of the actual object to be measured. Training data can be obtained directly from this.

[0137] Furthermore, additional measurements from other objects to be measured in the temporal vicinity, i.e., shortly before and / or shortly after, the measurement of the reference object to be measured can also be used. The geometry(s) of the other objects to be measured may in turn differ from the geometry of the reference object and the object to be measured. Furthermore, the geometry of the other objects does not need to be known, nor does it need to be available for reference measurement results, to be able to estimate whether an error condition exists for these measurements. This information is taken from the "temporally close" measurement of the object to be measured, i.e., it is assumed that the state of the device has not changed significantly in this short period of time.

[0138] In this way, additional training data can be generated during normal measurement operations without additional reference measurements.

[0139] In another example, training data can be generated using information from maintenance. If it is determined during maintenance that a certain error occurred, the characteristics of the monitored data before that time can be assigned to this error, even retrospectively. Furthermore, an analysis of the differences in the characteristics of the data before and after the maintenance, during which the error was corrected, can be performed in order to identify the relevant characteristics. This makes it possible to detect the corresponding error earlier in the future. For example, a manufacturer of CT systems could build up corresponding databases of training data for monitoring its CT systems over many maintenance cycles, possibly divided by design, and make these available to users.

[0140] Furthermore, error states of a CT system can also be deliberately induced in order to analyze their effects on data quality and include them in the training data.

[0141] The Figures 3a and 3b relate to further embodiments of the computer-implemented method 100, in which a first and a second object are measured to determine quality parameters. The geometry of the second object deviates only slightly from the geometry of the first object, ie, the geometry of the second object lies within a predefined tolerance interval of the geometry of the first object.

[0142] According to Figure 3a The method 100 may include step 128, which includes at least step 102 and is performed with the first object. Step 102 is performed by means of a radiographic measurement of the first object. However, further steps of the method 100 may be provided in step 128.

[0143] In this example, the method further includes step 130, which includes at least step 102 and is performed with the second object. Step 102 is performed by means of a radiographic measurement of the second object. However, further steps of method 100 can be provided in step 130.

[0144] In a further step 132, recording parameter sets of the device are determined. To determine the recording parameter sets, the recording parameters of the measurement data of the first object, summarized as recording parameter sets, and the recording parameters of the measurement data of the second object, summarized as recording parameter sets, are compared with one another. Recording parameter sets that are identical for the measurement data of both objects are selected as the recording parameter sets to be determined. This means that if measurement data of the first object and measurement data of the second object were determined with identical recording parameters, the recording parameter sets underlying these measurement data are determined by step 132. If, for example, the same position of a detector and a radiation source, i.e. the same transmission geometry, with the same voltages, currents, etc., is required for both objects.was used to record measurement data, identical recording parameters are given.

[0145] In a further step 134, the measurement data of the first and second object associated with the determined acquisition parameter sets are used to generate first projection representations of the first object and second projection representations of the second object. In a further step 136, quality parameters are determined for the first projection representations and the second projection representations and analyzed for differences. First quality parameters are determined for the first projection representations, and second quality parameters are determined for the second projection representation.

[0146] In an alternative or additional example of the method 100 according to Figure 3b In a step 138, recording parameter sets of the device can be defined for determining the measurement data.

[0147] These acquisition parameter sets are used in step 140 to acquire measurement data from the first object in order to determine initial projection representations. Step 140 includes at least step 102, which is performed with the first object. The underlying measurement is a transmission measurement. However, further steps of the above-mentioned method 100 can be used.

[0148] In step 142, these acquisition parameter sets are used to acquire measurement data from the second object in order to determine a second projection representation. Step 142 includes at least step 102, which is performed with the second object. The underlying measurement is a transmission measurement. However, further steps of the above-mentioned method 100 can be used.

[0149] Furthermore, at least one first quality parameter of the first projection representations and at least one second quality parameter of the second projection representation are determined. In step 144, the first quality parameters and second quality parameters are analyzed for differences.

[0150] By comparing the first quality parameter and the second quality parameter, the functional status of the device for examining objects can be analyzed. Since quality parameters based on the same recording parameter sets are compared, the quality parameters of the measurement data for both objects should be at least similar. If there are differences, this indicates that the functional status of the device for examining objects has changed between the measurement of the first object and the second object. The measurements of the first object and the second object can be separated by a large time period. This means that a change in the functional status over a large period of time can be examined.

[0151] Figure 4shows a further alternative or additional embodiment of the computer-implemented method 100 in which a first and a second object are measured in order to analyze the functional state of the device for examining objects.

[0152] In a step 146, this example includes at least step 102, which is performed with the first object. However, further steps of method 100 may also be provided. The measurement used is a transmission measurement.

[0153] In a step 148, this example includes at least step 102, which is performed with the second object. However, further steps of method 100 can also be provided. The measurement used is a transmission measurement. For the second object, a different transmission geometry is used in step 102 than for the first object.

[0154] In step 150, projection representations are determined from the measurement data of the first object. Furthermore, in step 152, projection representations are also determined from the measurement data of the second object.

[0155] The projection representations of the first object and the projection representation of the second object are analyzed, and a blurriness of the projection representations is determined. In step 154, the blurriness of the projection representations of the first object and the blurriness of the production representation of the second object are compared.

[0156] In a step 156, the blur of the first projection image and the blur of the second projection image are analyzed to determine the condition characteristic. Since the blur of projection images remains comparable across multiple measurements, regardless of the geometry of the object being irradiated, a change in the blur between the two projection images can be inferred from a change in the remaining influencing variables that vary between the measurements. In this way, a fault condition can be detected.

[0157] 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.

[0158] The invention is not limited to any of the embodiments described above, but is defined by the appended set of claims.

Claims

1. Computer-implemented method for monitoring the state of a device for examining objects, wherein the examination of an object comprises determining measurement data by way of a measurement of the object and operating data of the device are determined during the examination of the object, wherein the method (100) comprises the following steps: - determining (102) measurement data of the object by way of the device; - determining (104) operating data of the device during the step of determining (102) measurement data of the object; - determining (106) at least one quality parameter from the measurement data; - analysing (108) the operating data and the at least one quality parameter; and - determining (110) a state characteristic value based on the analysis of the operating data and the at least one quality parameter for monitoring the state of the device, wherein the state characteristic value indicates a state of the device, wherein the method (100) further comprises the following steps: - carrying out (146) at least the step of determining (102) measurement data of the object with a first object, wherein the measurement is a radiographic measurement; - carrying out (148) at least the step of determining (102) measurement data of the object with a second object, wherein the measurement is a radiographic measurement, wherein a radiographic geometry used is different from that used for the first object; - determining (150) projection representations from the measurement data of the first object; - determining (152) projection representations from the measurement data of the second object; - comparing (154) at least one blur between the projection representations of the first object and the projection representations of the second object; and - analysing (156) the at least one blur in order to determine the state characteristic value.

2. Method according to Claim 1, characterized in that the step of determining (110) a state characteristic value comprises the following sub-step: - comparing (112) the operating data and the at least one quality parameter with predefined comparison values for the operating data and the at least one quality parameter for determining a state characteristic value.

3. Method according to either of Claims 1 and 2, characterized in that the step of determining (110) a state characteristic value further comprises the following sub-step: - comparing (114) the operating data and the at least one quality parameter with predetermined sample data derived from training data for the operating data and the at least one quality parameter for determining a state characteristic value.

4. Method according to Claim 3, characterized in that the training data are generated from measurement data which are determined by way of the device by way of at least one measurement within a predefined time interval before and / or after determining (102) measurement data of a reference object, wherein the reference object has a known geometry and / or a known property.

5. Method according to any of Claims 1 to 4, characterized in that the step of determining (110) a state characteristic value further comprises the following sub-step: - determining (116) whether the state characteristic value indicates a fault state of the device.

6. Method according to Claim 5, characterized in that the method (100) further comprises the following step if the state characteristic value indicates that a fault state of the device might be present: - determining (118) measurement data of a reference object by way of the device; and - analysing (120) the measurement data of the reference object in order to determine whether the fault state of the device is present.

7. Method according to any of Claims 1 to 6, characterized in that the method (100) further comprises the following step: - determining (122) a fault cause parameter at least from the state characteristic value, wherein the fault cause parameter indicates a possible fault cause for a fault state.

8. Method according to Claim 7, characterized in that, if, in the step of determining (122) a fault cause parameter, a fault cause parameter is determined which indicates that a geometric calibration of the device has an accuracy value which is disposed outside a predefined accuracy value interval, the method (100) further comprises the following step: - calibrating (124) the device.

9. Method according to Claim 6 or 8, characterized in that the device comprises an automatic object changing unit, wherein the method (100) further comprises the following step: - substituting (126) a reference object for the object into the device for examining objects in order to determine measurement data about the reference object and / or substituting a calibration object for the object into the device for examining objects in order to calibrate the device for examining objects.

10. Method according to any of Claims 1 to 9, characterized in that the method (100) further comprises the following steps: - carrying out (128) at least the step of determining (102) measurement data of the object with a first object, wherein the measurement is a radiographic measurement; - carrying out (130) at least the step of determining (102) measurement data of the object with a second object, wherein the measurement is a radiographic measurement; - determining (132) recording parameter sets of the device which are identical for the determination of measurement data for the first object and for the determination of measurement data for the second object; - determining (134) first projection representations from the measurement data for the first object by way of the determined recording parameter sets and determining second projection representations from the measurement data for the second object by way of the determined recording parameter sets; and - analysing (136) at least one first quality parameter assigned to one of the first projection representations and at least one second quality parameter assigned to at least one of the second projection representations, for differences; and / or in that the method (100) further comprises the following steps if a recording parameter set of a first object and a recording parameter set of a second object are at least partly not identical: - defining (138) recording parameter sets of the device for the determination of measurement data; - carrying out (140) at least the step of determining (102) measurement data of the object with the first object with the defined recording parameter sets for determining first projection representations, wherein the measurement is a radiographic measurement; - carrying out (142) at least the step of determining (102) measurement data of the object with a second object with the defined recording parameter sets for determining second projection representations, wherein the measurement is a radiographic measurement; and - analysing (144) at least one first quality parameter assigned to one of the first projection representations and at least one second quality parameter assigned to at least one of the second projection representations, for differences; wherein a geometry of the second object deviates from a geometry of the first object within a predefined tolerance interval.

11. Method according to any of Claims 1 to 10, characterized in that the method (100) further comprises the following step: - determining (158) an estimated value of an uncertainty of a measurement variable of the object determined from the measured data by way of the operating data and the at least one quality parameter.

12. Method according to any of Claims 1 to 11, characterized in that, during the step of determining (102) measurement data of the object by way of the device, the following steps are carried out: - determining (160) preliminary measurement data and / or at least one preliminary quality parameter from the measurement data; - adapting (162) the step of determining (102) measurement data of the object by way of the device taking into account the preliminary measurement data and / or the at least one preliminary quality parameter from the measurement data.

13. Method according to Claim 12, characterized in that the method (100) further comprises the following step: - determining (164) whether a measurement variable of the object determined from the measurement data, preferably taking into account an uncertainty of the measurement variable, lies within a predefined tolerance range.

14. Computer program product comprising instructions which are executable on a computer and which, when executed on a computer, cause the computer and a device for examining objects to carry out the method according to any of the preceding claims.