METHOD FOR ESTIMATE THE COMPONENT QUALITY OF AN ELECTRON EMISSOR

Description

[0001] The invention relates to a computer-implemented method for estimating the component quality of an electron emitter, a computer-implemented method for training a machine learning method for estimating the component quality of an electron emitter, an associated AI model and an associated computer program product.

[0002] An electron emitter is conventionally used in an X-ray tube. The electron emitter generates free electrons, which can then be accelerated towards an anode, for example. Electron emitters come in various designs. They can typically be divided into two categories: thermionic emitters and cold emitters.

[0003] In thermionic emission of electrons, the electron emitter is at least partially heated until electrons are emitted from the heated emitter region. A conventional thermionic emitter is, for example, a helical emitter. An alternative is the flat emitter. A conventional flat emitter typically consists of an emitter sheet. The emitter sheet is usually structured with incisions.

[0004] A cold emitter, for example, is a field-effect emitter and typically has a large number of emitter needles. An electron current can be extracted from the tips of the emitter needles by applying a gate voltage and utilizing the field effect.

[0005] As a result of the thermal stress during electron emission, deformations can occur, particularly in thermionic electron emitters. These deformations include, for example, C-deformation, S-deformation, or other types of deformation. An unusual deformation, in particular, is an artifact in the electron emitter geometry that can negatively affect component quality and thus be a critical quality parameter. Ideally, the electron emitter geometry is artifact-free. Component quality can correlate with component lifespan.

[0006] To assess component quality, it is essential to correlate the electron emitter geometry with field performance data of comparable electron emitters. This typically requires comprehensive parameterization of the electron emitter geometry. The electron emitter geometry specifically describes the positioning of the electron emitter within a cathode head. Parameterization is performed, for example, using a measuring machine and / or often requires complex automation, which generally results in longer measurement times at atmospheric pressure. For vacuum components like electron emitters, residence times at atmospheric pressure are critical to evaluate. Assessing component quality is complex and regularly exceeds the scope of information that a user of the electron emitter can comprehend.The complexity of this estimate is based in particular on the fact that variances in the micrometer range must be taken into account and / or there may be a strong dependence on the specific use of the electron emitter.

[0007] Currently, there is no suitable tool for the effective and / or standardized measurement of electron emitter geometry and / or for an efficient assessment of component quality, particularly with regard to deformation and / or mispositioning of the electron emitter. The typically isolated measurement of the electron emitter is generally insufficient for this purpose. This is because the selection of measurement points is largely determined by previously known deformations. Conversely, previously unknown deformations potentially remain undetected.

[0008] The determination of electron emitter geometry is conventionally based on the acquisition of a defined series of measurement points. Typically, this is limited to measurement points that characterize the emitter's outer slit. Alternatively or additionally, the height of the electron emitter inserted into the cathode head is measured at specific points relative to the cathode head. Positioning relative to the central rib is generally not considered. Distances between the measurement points can typically be calculated. Electron emitter geometry is usually determined manually, particularly in a manufacturing environment. To minimize the manual effort, only a small number of measurement points are typically recorded. The resulting measurements can therefore only provide a rough indication of the electron emitter geometry.In particular, conventional measurement results do not provide a comprehensive parameterization of the electron emitter.

[0009] Therefore, the various types of deformation on electron emitters, especially thermionic ones, cannot be accurately represented and / or differentiated using this method. In particular, a continuous correlation between the recorded electron emitter geometry and field-achieved performance data does not occur. In other words, the estimation of component lifetime is conventionally triggered individually and / or only when needed.

[0010] JP H07 29494 A deals with the precise determination of the relative positions of the individual components of an electron gun within a short time. An electron gun is placed on a measuring base and photographed with a camera. The camera output is fed to a monitor and a data analysis unit. In the data analysis unit, the profile of each grating of the electron gun is detected, and the center point of each grating is calculated based on this. Using the central axis of the grating that forms a primary lens as a standard, the displacement and inclination of the central axes of other gratings are numerically determined, and these are displayed on a monitor along with the external shapes of the gratings.

[0011] From JP 2002 324495 A, the provision of a high-quality electron gun whose electrode position data can be reliably detected and which is capable of making adjustments that reflect the detection result is known. At least a plurality of electrodes located closer to a cathode are shaped such that, viewed in the beam direction, they have a planar outer shape of different sizes, with the electrode with the smallest outer shape being located at a position closest to the cathode and electrodes with a large outer shape being arranged successively at correspondingly greater distances from the cathode. Positional deviation data from reference data in the XY plane are calculated for each electrode by capturing images of the plurality of electrodes from the cathode side in the Z-axis direction, which is the direction of the beam path, to obtain image data in the XY plane.The obtained image data undergoes image processing, including the extraction of outlines and straight lines, followed by arithmetic processing.

[0012] The invention is based on the objective of providing a computer-implemented method for estimating the component quality of an electron emitter, a computer-implemented method for training a machine learning method for estimating the component quality of an electron emitter, an associated AI model and an associated computer program product with an improved differentiation capability.

[0013] The problem is solved by the features of the independent claims. Advantageous embodiments are described in the dependent claims.

[0014] Regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are included.

[0015] The computer-implemented method according to the invention for estimating the component quality of an electron emitter, which is part of a cathode assembly, wherein the cathode assembly comprises a cathode head and the electron emitter inserted into the cathode head, wherein the electron emitter is in particular a flat emitter, comprises the following steps: Receiving an electron emitter image dataset, wherein image information from the electron emitter image dataset at least partially depicts the electron emitter inserted into the cathode head; receiving an electron emitter geometry model from a storage unit; transforming the received electron emitter geometry model onto the image information from the electron emitter image dataset, wherein an electron emitter geometry information of the electron emitter is calculated as an output parameter of the transformation; determining a degree of similarity of the inserted electron emitter with at least one other electron emitter using the electron emitter geometry information; estimating the component quality as a function of the determined degree of similarity with the at least one other electron emitter.

[0016] The component quality of the electron emitter corresponds in particular to the component lifetime of the electron emitter. Alternatively or additionally, the component quality of the electron emitter can correlate with the absence of a defect pattern and / or the image quality achieved by means of generated X-ray radiation.

[0017] The component lifetime can be specified in units of time and / or load. The load unit can, for example, comprise a sum or time series of X-ray dose, a limiting load, and / or heating current over the lifetime of the electron emitter. Typically, field performance data is specified in units of time and / or load. After the end of its lifetime, the electron emitter can typically no longer be operated within specifications and / or must be replaced. During or until the end of its lifetime, the electron emitter can be operated within specifications. "Within specifications" means that field performance data for the electron emitter falls within the limits defined in the specifications.

[0018] The electron emitter is specifically a thermionic emitter. Alternatively, it is conceivable that the electron emitter is a cold emitter, for example, a field-effect emitter. The thermionic emitter is specifically a flat emitter, which typically consists of an emitter sheet with incisions. The thermionic emitter can alternatively be a helical emitter.

[0019] The cathode assembly typically exhibits a negative high voltage relative to the anode. It is also possible for the cathode assembly to be at ground potential. In this case, the cathode assembly would have a more negative high voltage compared to the anode. A high voltage is typically present between the cathode assembly and the anode. The anode can, in principle, be at ground potential.

[0020] Free electrons generated by the electron emitter are accelerated towards the anode, primarily by the high voltage from the cathode assembly. This high voltage can also be referred to as the accelerating voltage. X-rays are generated when the electrons interact with the anode. Typically, 99% of the kinetic energy of the accelerated electrons is converted into heat upon impact with the anode. The area where the electrons strike the anode is called the focal spot.

[0021] To improve anode cooling, the anode can be rotatable relative to the emitted electrons. If the anode rotates, the focal spot forms part of a circular path. In this case, the anode is, for example, a rotating anode. An X-ray tube with such a rotating anode and cathode assembly is distinguished, for example, by whether the cathode assembly rotates with the anode. In the first case, the X-ray tube is specifically a rotary piston X-ray tube, and in the latter case, it is specifically a rotating anode X-ray tube. If the anode is stationary, the X-ray tube is specifically a stationary anode X-ray tube. Alternatively, the cathode assembly can be located within a linear accelerator unit.

[0022] Electrons accelerated by the accelerating voltage have an energy of up to 150 keV, particularly suitable for diagnostic imaging and / or materials testing. Electrons accelerated by the linear accelerator unit can have an energy of up to 20 MeV, suitable for non-destructive imaging, materials testing, and / or therapy. Depending on the electron energy, the energy of the X-rays, for example, has a maximum energy of 150 keV or 20 MeV.

[0023] The cathode assembly can consist of the cathode head and the electron emitter inserted into the cathode head. The cathode head can be configured to block or deflect the free electrons. The electrical potential of the cathode assembly is applied, in particular, to the cathode head. The cathode head can be configured, in particular, to cool the electron emitter. The cathode head and the electron emitter can be galvanically isolated and / or electrically insulated from each other.

[0024] The fact that the electron emitter is inserted into the cathode head can, in particular, mean that the electron emitter is finally mounted in the cathode head. "Finally" means that the assembly process of the cathode assembly, especially of the electron emitter in the cathode head, has already been completed. Typically, the position of the electron emitter relative to the cathode head is kept constant, for example, by means of suitable fasteners, and / or is no longer changed. For example, the electron emitter can be secured relative to the cathode head by means of fasteners such as a screw, a spring, a bolt, a weld, and / or a solder point.

[0025] The electron emitter image dataset can be acquired, for example, using a sensor, in particular a camera and / or a scanner and / or a microscope and / or a white light interferometer. The sensor can be a 2D or 3D sensor. The electron emitter image dataset can, in particular, include 2D image information, especially for describing lateral extent, and optionally additional depth information.

[0026] The electron emitter image dataset can be a photograph containing the image information. It can also be a series of photographs and / or a film. Additionally, the electron emitter image dataset can contain metadata describing the image information. The image information can include, in particular, grayscale, black and white, or color. The electron emitter image dataset is preferably stored in a digital file, preferably in an image format. The electron emitter image dataset can, in particular, correspond to a digitized version of an analog photograph. The image information can be transformed into an image space or a frequency space. The image information can be, for example, raw image information and / or final image information, especially as processed by the sensor.

[0027] The image information shows the electron emitter at least partially, preferably completely. In particular, the image information preferably depicts at least part of the electron emitter and the cathode head. It is conceivable that the image information depicts neither the entire electron emitter nor the entire cathode head. In principle, it is conceivable that the image information does not depict the cathode head at all.

[0028] The image information preferably represents the electron emitter one-to-one. Advantageously, the electron emitter therefore appears undistorted in the image information. The image information preferably represents the electron emitter in a rectified and / or undistorted form. It is conceivable that the image information no longer contains any distortion caused by optics during image acquisition, but is corrected in this respect. For example, the post-processed final image information can be corrected in this way.

[0029] The electron emitter image data set can be received, in particular, by the sensor and / or a storage unit. Receiving the data set can include the sensor and / or storage unit providing it. For example, the electron emitter image data set is retrieved by a processing unit and then received by the processing unit. Receiving the electron emitter image data set can also include transmitting it via an interface, which may be a network interface.

[0030] Receiving the electron emitter image dataset can include reconstructing, filtering, and / or annotating the image information. Reconstruction can specifically include applying an (inverse) Fourier transform to the image information. Filtering can specifically include region growth, edge smoothing, edge enhancement, image segmentation, and / or contrast enhancement. Annotation can specifically include location marker recognition and / or pattern recognition. Reconstruction, filtering, and / or annotation can be performed, for example, in the processing unit, the sensor, and / or the storage unit. The storage unit can be part of the processing unit and / or the sensor.

[0031] Receiving the electron emitter geometry model can involve querying the electron emitter geometry model from the storage unit. The electron emitter geometry model can be transferred, for example, via the interface.

[0032] The electron emitter geometry model corresponds, for example, to a 1:1 representation of the electron emitter(s) before insertion into the cathode head. In this case, the electron emitter geometry model specifically depicts the electron emitter(s) without distortion. The electron emitter geometry model shows the electron emitter(s) schematically and / or abstractly.

[0033] Alternatively or additionally, it is conceivable that the electron emitter geometry model, based on a 1:1 mapping, is adapted to acceptable deformations of the electron emitter that may occur during insertion into the cathode head. These acceptable deformations may fall within the specified parameters and / or be mandatory due to the fastening elements. The acceptable deformations are typically harmless, non-critical, and / or insignificant for the quality and thus the component quality and / or service life of the electron emitter. The adaptation of the 1:1 mapping can be achieved, for example, by modeling and / or averaging electron emitter geometry information from other electron emitters that exhibit comparatively high component quality and / or long service lives.Based on this modeling and / or averaging, the electron emitter geometry model can preferably be adapted or adapted.

[0034] Typically, the electron emitter geometry model is annotated with numerous measurement points describing the electron emitter geometry. This annotation can be performed automatically and / or by a user. The measurement points can, in particular, identify an edge, a centroid, and / or a center point.

[0035] Transforming the received electron emitter geometry model onto the image information includes, in particular, shifting at least one measurement point of the electron emitter geometry model to a measurement point mapped in the image information. If the electron emitter geometry model is 100% geometrically identical to the image information, or if the electron emitter geometry model is geometrically identical to the image information, the output parameter is, for example, a one-to-one mapping. In this case, the transformation does not include shifting at least one measurement point.

[0036] The transformation is, in particular, an image transformation. The transformation is preferably performed while minimizing a complex correlation factor. The transformation can, in particular, comprise a rigid and / or a non-rigid transformation. It is conceivable that the image information and / or the electron emitter geometry model are subdivided into different regions with a rigid or a non-rigid transformation.

[0037] The transformation and / or calculation of the output parameter takes place, for example, in the processing unit, such as a CPU or GPU. The output parameter is usually calculated during the transformation process.

[0038] The electron emitter geometry information contains, in particular, a relation between the image information and the electron emitter geometry model. The electron emitter geometry information specifically characterizes a difference between the image information and the electron emitter geometry model. The electron emitter geometry information describes, in particular, the displacement and / or non-displacement of at least one measurement point. The electron emitter geometry information can, in particular, describe a spatial displacement of at least one measurement point and / or a relative distance between two measurement points. The electron emitter geometry information is, for example, a transformation matrix. In its simplest form, the transformation matrix corresponds to a one-to-one mapping and / or is equal to 1. Additionally, the electron emitter geometry information can contain the image information.

[0039] The electron emitter geometry information can include a classification of the image information into a relative category such as undistorted, in particular 1:1 mapping, slightly distorted, severely distorted or not transformable, for example because it is too severely distorted, and / or a quantitative category such as X distortions of type Y.

[0040] Determining the degree of similarity can involve identifying the most similar electron emitter among the other electron emitters and / or defining the degree of similarity based on the similarity between the most similar electron emitter and the other electron emitters. Alternatively or additionally, the determined degree of similarity can be used, for example, to identify and / or select the most similar electron emitter among the other electron emitters. The determination of the degree of similarity is performed primarily in the processing unit.

[0041] Determining the degree of similarity using electron emitter geometry information specifically involves comparing the electron emitter geometry information with the electron emitter geometry information of at least one other electron emitter. Determining the degree of similarity may include retrieving, receiving, and / or providing the electron emitter geometry information of the at least one other electron emitter.

[0042] Determining the degree of similarity can be performed using multiple electron emitters or all electron emitters. It is conceivable that the degree of similarity is determined multiple times for the same electron emitter. Determining the degree of similarity can include ranking and / or filtering the results. Furthermore, it can involve classifying the determined degrees of similarity and selecting a class containing the classified degrees of similarity, or selecting a single degree of similarity from a class.

[0043] The degree of similarity is, in particular, a measure of similarity. Specifically, it describes a similarity with respect to electron emitter geometry information. Typically, the degree of similarity is higher the smaller the differences are between the electron emitter or its electron emitter geometry information and at least one other electron emitter or its electron emitter geometry information.

[0044] The degree of similarity can be a correlation factor and / or a geometric mapping. For example, the degree of similarity can be expressed in categories such as "dissimilar" or "similar." Alternatively or additionally, the degree of similarity can be a value between 0 and 100%. In this case, for example, a threshold can define a boundary that separates similar and dissimilar electron emitters.

[0045] Estimating component quality can, in particular, correspond to estimating component service life. Estimating component quality can involve calculating and / or determining the component quality. Estimating component quality can also include transmitting the estimated component quality to the storage unit.

[0046] The component quality can be estimated, for example, by assigning a component quality of zero if the first similarity degree is lower than the second similarity degree. In this case, for instance, the electron emitter inserted into the cathode head exhibits extraordinary deformation that is outside the specification and / or unacceptable. This extraordinary deformation can be, in particular, severe. The inserted electron emitter can then be assigned a value corresponding to "deformed" and / or "defective."

[0047] Alternatively, for the second degree of similarity, the component quality can be estimated based on a starting value that is influenced by at least one other electron emitter with the highest degree of similarity. This starting value can, for example, be based on or dependent on field-obtained performance data for this at least one other electron emitter. Alternatively or additionally, the component quality can be estimated by selecting the class with the at least one other electron emitter and using the field-obtained performance data of those electron emitters in the selected class to determine the starting value. For example, the field-obtained performance data can be averaged and / or modeled to determine the starting value.Alternatively, the performance data obtained in the field can be used to extract the component quality of at least one other electron emitter and provided as an estimated component quality.

[0048] Generally, the higher the estimated component quality and / or the longer the component lifetime, the better the expected performance data in the field. In other words, the less likely it is that the electron emitter will fail relatively early, exhibit a fault pattern, or display an artifact, the higher the estimated component quality and / or the longer the estimated component lifetime. Conversely, the lower the expected performance data of the electron emitter in the field, the lower the estimated component quality and / or lifetime.

[0049] One embodiment provides that a single electron emitter image data set is received and the component quality of the electron emitter is estimated using this single image data set. This embodiment is particularly advantageous because it greatly simplifies the estimation of the component quality. This means that not multiple electron emitter image data sets are received, but only the single electron emitter image data set.

[0050] One embodiment provides that the estimated component quality of the electron emitter is stored in the storage unit. This embodiment advantageously allows access to the estimated component quality, for example, during operation of the electron emitter.

[0051] One embodiment provides that determining the degree of similarity involves inputting the electron emitter geometry information into an AI model trained using a machine learning method and providing the degree of similarity at an output of the AI ​​model. Advantageously, the AI ​​model can be used to determine the degree of similarity if it has been trained according to the training method of the invention. In this case, the AI ​​model typically links component quality and / or field-achieved performance data with electron emitter geometry information. The linkage in the AI ​​model can be validated after training. In particular, the AI ​​model can be configured to filter the image information and / or annotate the image information, as previously described, as an optional step when receiving the electron emitter image dataset.The AI ​​model can be adapted, or adapted, specifically to the electron emitter geometry model. Determining the degree of similarity can be achieved by inputting the electron emitter geometry information into the AI ​​model. Estimating component quality can, in principle, be done using the AI ​​model.

[0052] One embodiment provides that the input electron emitter geometry information is dimensionally reduced using the AI ​​model, and the degree of similarity is determined using the dimensionally reduced electron emitter geometry information. In particular, a vector can be calculated for each electron emitter using the AI ​​model, wherein the dimension of the vector is smaller than the number of power data points obtained in the field, and wherein the dimension of the vector is preferably less than or equal to five, and more preferably less than or equal to three. The dimension of the vector can, in particular, be equal to two and / or greater than one.

[0053] The AI ​​model generates, in particular, a dimensionally reduced information density distribution. This dimensionally reduced information density distribution is, in particular, an output parameter of the AI ​​model and / or a so-called "latent space," "latent feature space," or "embedding space." When generating the dimensionally reduced information density distribution, the input data is combined in such a way that input data irrelevant to the differentiation is compressed and / or removed compared to input data relevant to the differentiation. Input data relevant to the differentiation enables, in particular, a delimitation and / or results in a greater abstract distance between the electron emitters. Input data irrelevant to the differentiation can be a label for the electron emitter and / or generally applicable technical information.Input data relevant for differentiation can include, in particular, the electron emitter geometry information assigned to the selected component and the electron emitter geometry information of other components, which can be linked to the performance data obtained in the field.

[0054] It is conceivable that the AI ​​model processes only the input data relevant for differentiation and filters out and / or discards the input data that is irrelevant. A non-binary, for example, linear weighting of the input data between relevant and irrelevant information can preferably be performed using the AI ​​model. In particular, the application of the AI ​​model allows for the linking of electron emitter geometry information and / or field-achieved performance data from different electron emitters.

[0055] The AI ​​model enables, in particular, the extrapolation or interpolation of especially scattered input data. When applying the AI ​​model, the number of input data points is typically reduced so that the dimensionally reduced information density distribution is, for example, two-dimensional or three-dimensional and / or can be visualized. This allows a user to determine and / or perceive differences in electron emitter geometry as (previously so-called abstract) distances between the dimensionally reduced input data points. It is conceivable that the distances between the dimensionally reduced input data points could be determined automatically. The AI ​​model is specifically trained so that the closer such electron emitters are to each other in the information density distribution the more similar their distinguishing input data points are.

[0056] One embodiment provides that the AI ​​model is used to differentiate a first class with electron emitter geometries containing an artifact from a second class with electron emitter geometries without artifacts, and that the degree of similarity is determined by assigning the electron emitter geometry information to the first or second class. This embodiment specifically describes the classification process for determining the degree of similarity. In principle, it is conceivable that the AI ​​model distinguishes the classes according to the number and / or severity of deformations. In particular, more than two classes are possible. Possible classes could be, corresponding to the classification, "not deformed," "acceptable deformations," and "extraordinary deformations."

[0057] The computer-implemented method according to the invention for providing a trained AI model for use in the computer-implemented method according to one of the preceding claims comprises the following steps: Receiving field-acquired power data from additional electron emitters as input data, applying a neural network comprising an encoder and a decoder to the input data, calculating an output vector, where the encoder maps a first number of input values ​​to a second number of output values, and where the decoder maps a second number of input values ​​to a first number of output values, the second number being smaller than the first number, adjusting a parameter of the neural network based on a comparison of the output vector with the input data, and outputting the decoder as a trained AI model.

[0058] One embodiment provides that additional electron emitter geometry information from the other electron emitters is received as input data. This embodiment is particularly advantageous because it improves the estimation of component quality.

[0059] An AI model according to the invention is configured for use in the computer-implemented method according to the invention. The AI ​​model is, in particular, trained according to the computer-implemented method for providing the trained AI model.

[0060] A cathode arrangement advantageously has a cathode head and an electron emitter inserted into the cathode head, wherein the component quality of the electron emitter is estimated or is estimated using a computer-implemented method according to the invention for estimating the component quality of the electron emitter.

[0061] One embodiment provides that the electron emitter is a flat emitter. This embodiment is particularly advantageous because, due to its rigidity, the flat emitter is more prone to deformation, and thus the component quality of the flat emitter can be estimated according to the invention.

[0062] The computer program product can be a computer program or comprise a computer program. The computer program product particularly includes the program code means that implement the process steps according to the invention. This allows the process according to the invention to be defined and executed repeatably, and enables control over the transfer of the process according to the invention. The computer program product is preferably configured such that the computing unit can execute the process steps according to the invention by means of the computer program product. The program code means can, in particular, be loaded into a memory of the computing unit and typically executed by means of a processor of the computing unit with access to the memory.When the computer program product, in particular the program code, is executed in the processing unit, all embodiments of the described method according to the invention can typically be carried out. The computer program product is, for example, stored on a physical, computer-readable medium and / or digitally stored as a data packet in a computer network. The computer program product can represent the physical, computer-readable medium and / or the data packet in the computer network. Thus, the invention can also start from the physical, computer-readable medium and / or the data packet in the computer network. The physical, computer-readable medium is usually directly connectable to the processing unit, for example, by inserting the physical, computer-readable medium into a DVD drive or plugging it into a USB port, thereby allowing the processing unit to access the physical, computer-readable medium, particularly for reading.The data packet can preferably be retrieved from the computer network. The computer network can contain the computing unit itself or be indirectly connected to the computing unit via a wide-area network (WAN) or a (wireless) local area network (WLAN or LAN) connection. For example, the computer program product can be stored digitally on a cloud server at a storage location within the computer network and transferred to the computing unit via the WAN over the internet and / or via WLAN or LAN, particularly by accessing a download link that points to the storage location of the computer program product.

[0063] Features, advantages, or alternative embodiments mentioned in the description of the device are also transferable to the method, and vice versa. In other words, claims relating to the method can be further developed with features of the device, and vice versa. In particular, the device according to the invention can be used in the method.

[0064] The invention will now be described and explained in more detail with reference to the exemplary embodiments shown in the figures. In principle, structures and units that remain essentially the same in the following figure descriptions will be named with the same reference numeral as when the respective structure or unit first appears.

[0065] They show: Fig. 1 a computer-implemented method according to the invention for estimating the component quality of an electron emitter, Fig. 2 the method in a first embodiment, Fig. 3 the method in a second embodiment, Fig. 4 a computer-implemented method according to the invention for providing a trained AI model and Fig. 5 a cathode device according to the invention.

[0066] Fig. 1 Figure 1 shows the computer-implemented method according to the invention for estimating the component quality of an electron emitter, which is part of a cathode device, wherein the cathode device comprises a cathode head and the electron emitter inserted into the cathode head, wherein the electron emitter is in particular a flat emitter, in a flowchart with steps S100 to S104.

[0067] Process step S100 characterizes the reception of an electron emitter image data set, wherein image information of the electron emitter image data set at least partially depicts the electron emitter inserted into the cathode head.

[0068] Procedure step S101 characterizes the reception of an electron emitter geometry model from a storage unit. The electron emitter geometry model can be annotated with a multitude of measurement points describing the electron emitter geometry.

[0069] Process step S102 involves transforming the received electron emitter geometry model into the image information of the electron emitter image dataset, where an electron emitter geometry information is calculated as an output parameter of the transformation. The transformation is performed, in particular, by minimizing a complex correlation factor. The electron emitter geometry information can describe a spatial displacement of at least one measurement point and / or a relative distance between two measurement points.

[0070] Procedure step S103 characterizes the determination of a degree of similarity between the electron emitter used and at least one other electron emitter using the electron emitter geometry information.

[0071] Procedure step S104 characterizes an estimation of the component quality depending on the determined degree of similarity with at least one other electron emitter.

[0072] Fig. 2 A first embodiment of the method is shown in a flowchart.

[0073] Procedure step S100.A indicates that a single electron emitter image data set is received.

[0074] Procedure step S104.A indicates that the component quality of the electron emitter is estimated using the individual electron emitter image data set of the electron emitter.

[0075] Process step S105 indicates that the estimated component quality of the electron emitter is stored in a storage unit.

[0076] Fig. 3 A second embodiment of the method is shown in a flowchart.

[0077] Procedure step S103.A indicates that determining the degree of similarity involves inputting the electron emitter geometry information into an AI model trained using a machine learning method and providing the degree of similarity at an output of the AI ​​model.

[0078] Procedure step S103.B indicates that the input electron emitter geometry information is dimensionally reduced using the AI ​​model and the degree of similarity is determined using the dimensionally reduced electron emitter geometry information.

[0079] Procedure step S103.C indicates that, using the AI ​​model, a first class with electron emitter geometries containing an artifact is distinguished from a second class with electron emitter geometries without an artifact, and that the degree of similarity is determined by assigning the electron emitter geometry information to the first class or the second class.

[0080] Fig. 4 Figure 1 shows a computer-implemented method according to the invention for providing a trained AI model in a flowchart comprising steps S110 to S113.

[0081] Process step S110 involves receiving power data from additional electron emitters obtained in the field as input data. It is conceivable that electron emitter geometry information from these additional emitters could also be received as input data.

[0082] Procedure step S111 characterizes the application of a neural network comprising an encoder and a decoder to the input data, where an output vector is calculated, wherein the encoder maps a first number of input values ​​to a second number of output values, and wherein the decoder maps a second number of input values ​​to a first number of output values, where the second number is smaller than the first number.

[0083] Procedure step S112 characterizes an adjustment of a parameter of the neural network based on a comparison of the output vector with the input data.

[0084] Procedure step S113 indicates the output of the decoder as a trained AI model.

[0085] Fig. 5 shows an advantageous cathode arrangement 10 in a schematic cross-section.

[0086] The cathode assembly 10 comprises a cathode head 11 and an electron emitter 12 inserted into the cathode head 11. The electron emitter 12 is a flat emitter. The quality factor of the electron emitter 12 is estimated using the method according to the invention.

[0087] The cathode assembly 10 can, in particular, be part of an evacuated X-ray tube. In this case, an anode is typically arranged opposite the cathode assembly 10, onto which the electrons generated by the electron emitter 12 strike, provided an accelerating voltage is applied between the cathode assembly 10 and the anode. The interaction of the electrons on the anode generates the X-rays. The X-rays can be used, in particular, in diagnostic imaging, e.g., computed tomography, mammography, and / or angiography, or in materials testing, where attenuated X-ray profiles can be reconstructed to form an image.

[0088] Although the invention has been illustrated and described in detail by the preferred embodiments, the invention is nevertheless not limited by the disclosed examples and other variations can be derived by the person skilled in the art without leaving the scope of protection of the invention.

Claims

1. Computer-implemented method for assessing a component quality of an electron emitter, which is part of a cathode facility, wherein the cathode facility comprises a cathode head and the electron emitter inserted in the cathode head, wherein the electron emitter is, in particular, a flat emitter, with the following steps: - receiving an electron emitter image dataset, wherein items of image information of the electron emitter image dataset at least partially map the electron emitter inserted in the cathode head, - receiving an electron emitter geometry model from a memory unit, - transforming the received electron emitter geometry model to the items of image information of the electron emitter image dataset, wherein an item of electron emitter geometry information of the electron emitter is calculated as an output parameter of the transformation, - ascertaining a degree of similarity of the inserted electron emitter with at least one further electron emitter by using the electron emitter geometry information, - assessing the component quality as a function of the ascertained degree of similarity with the at least one further electron emitter.

2. Method according to claim 1, wherein a single electron emitter image dataset is received and the component quality of the electron emitter is assessed by means of the single electron emitter image dataset of the electron emitter.

3. Method according to one of the preceding claims, wherein the electron emitter geometry model is annotated with a large number of measuring points describing the electron emitter geometry.

4. Method according to one of the preceding claims, wherein transforming takes place while minimising a complex correlation factor.

5. Method according to one of the preceding claims, wherein the electron emitter geometry information describes a spatial shift of at least one measuring point and / or a relative distance between two measuring points.

6. Method according to one of the preceding claims, wherein ascertaining the degree of similarity comprises inputting the electron emitter geometry information into a AI model trained by means of a machine learning method and providing the degree of similarity at an output of the AI model.

7. Method according to claim 6, wherein by means of the AI model, the input electron emitter geometry information is dimensionally reduced and the degree of similarity is ascertained by means of the dimensionally reduced electron emitter geometry information.

8. Method according to one of claims 6 or 7, wherein by means of the AI model, a first category with electron emitter geometries with an artifact is distanced from a second category with electron emitter geometries without artifact and wherein the degree of similarity is ascertained by means of an allocation of the electron emitter geometry information to the first category or the second category.

9. Method according to one of the preceding claims, wherein the assessed component quality of the electron emitter is stored in a memory unit.

10. Computer-implemented method for providing a trained AI model for use for a computer-implemented method according to one of the preceding claims, comprising the following steps: - receiving performance data, obtained in the field, of further electron emitters as input data, - applying a neural network, which comprises an encoder and a decoder, to the input data, wherein an output vector is calculated, wherein the encoder maps a first number of input values to a second number of output values, and wherein the decoder maps a second number of input values to a first number of output values, wherein the second number is smaller than the first number, - adapting a parameter of the neural network on the basis of a comparison of the output vector with the input data, - outputting the decoder as a trained AI model.

11. Method according to claim 10, wherein items of electron emitter geometry information of the further electron emitters are additionally received as input data.

12. AI model for use for a method according to one of claims 1 to 9, wherein the AI model is generated according to one of claims 10 or 11.

13. Computer program product, which can be loaded directly into a memory of a computing unit, with program code means in order to carry out a method according to one of claims 1 to 11 when the computer program product is executed in the computing unit.