DE102021209951A1 - Method and device for checking a cast component - Google Patents

Abstract

The invention relates to a method for checking a cast component, with volume data (10) of the cast component being recorded and generated by computer tomography being received, with three-dimensional target model data (20) of the cast component being received, with predetermined partial areas (21-x) being received in the target model data (20). or are defined, with a global registration of the volume data (10) of the cast component with the three-dimensional target model data (20) of the cast component being carried out, with the specified sub-areas (21-x) independently of one another locally with corresponding sub-areas (11-x) in each case are compared with the volume data (10), and wherein a verification result (30) is generated and output based on the respective comparison results of the partial areas (11-x, 21-x). The invention also relates to a device (1) for checking a cast component.

Description

The invention relates to a method and a device for checking a cast component.

In industrial manufacturing, the casting of metals and alloys is a widespread manufacturing process for components. Here, a melt of a liquid metal or an alloy is filled into a hollow mold in which the melt then solidifies. A common casting process is gravity die casting. Here, a cast component is formed in a mold and by means of sand cores inserted therein. The sand cores or parts thereof can be used here in particular to represent internal contours, in particular cavities in the subsequent cast component. The sand cores are destroyed and removed after casting.

A position of the sand cores is subject to process fluctuations in the manufacturing process, such as positional tolerances when inserting core packages into the mold or during casting, influences from melt flow (flow pressure) and thermal stress on different core areas. In order to be able to check the quality of a cast component, it is necessary to know the actual position of a sand core or parts thereof during casting and to be able to check it after casting. A computer tomographic method, for example, can be used for the three-dimensional recording and visualization of components.

From W.E. Lorensen and H.E. Cline, Marching Cubes: A high resolution 3D surface construction algorithm, Computer Graphics, Vol. 21, No. 4, July 1987, pp. 163-169, Association for Computing Machinery, New York, NY, USA, https://doi.org/10.1145/37402.37422, a method for extracting surfaces from volume data is known.

From P.J. Besl and N.D. McKay, A Method for Registration of 3-D Shapes, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 14, No. 2, February 1992, is a method for registering three-dimensional shapes, especially free-form curves and surfaces , known.

It is known to adapt (register) a CAD model to three-dimensional volume data that has been recorded by means of a computer tomograph. An alignment (estimation of a translation and a rotation) of the volume data to the CAD model is determined, for example, by minimizing a quadratic error of a CAD surface to a surface extracted from the three-dimensional volume data. This allows the CAD model and the three-dimensional volume data to be registered. However, the aforementioned process fluctuations can only be mapped insufficiently using this method. For the 100% inspection of castings in the process line, there is only a limited time available for capturing image data (especially radiographic images). Accordingly, an image quality that can be achieved is impaired by the long transmission lengths of solid components (e.g. crankcases or electric motor housings). In particular, the surface of solid components cannot be reconstructed without errors at all points in the recorded volume. This makes it difficult to measure wall thicknesses in relevant areas.

The object of the invention is to improve a method and a device for checking a cast component.

The object is achieved according to the invention by a method having the features of patent claim 1 and a device having the features of patent claim 10 . Advantageous configurations of the invention result from the dependent claims.

It is one of the basic ideas of the invention to separately consider and separately consider predetermined partial areas in three-dimensional reference model data (for example CAD data) of a cast component during checking. For this purpose, the specified sub-areas are each compared separately with the respective corresponding sub-areas in recorded and generated volume data of the cast component. In order in particular to obtain a common reference system, a global registration of the volume data of the cast component with the three-dimensional target model data of the cast component is first carried out. This can be done in particular starting from an outer shell of the cast component. For example, a global registration of the three-dimensional target model data with the volume data is carried out exclusively with regard to the outer shell of the cast component. After this initial global registration, local sub-areas within the cast component are considered. These local subareas are or will be specified or defined. In particular, these subareas are or will be defined in such a way that they coincide with functional subareas and/or functional modules (in internal combustion engines, for example, a water jacket, an inlet or outlet channel or oil channels, etc.). In particular, the partial areas are or will be specified or defined in such a way that they coincide with the sand cores or parts thereof used in the production process len. The sub-areas are each individually compared with one another. A check result is generated and output based on the respective comparison results of the partial areas. In other words, the cast component is checked on the basis of individual comparisons of the specified sub-areas. A comparison result includes in particular a deviation between the partial areas in the volume data and the three-dimensional target model data, in particular a shift and/or a difference between a position of the partial areas in the volume data and the three-dimensional target model data relative to one another. In this way, the checking can be improved, since the checking is not based on a comparison of the entire cast component with rigid three-dimensional target model data, but rather on (individual) comparisons of a plurality of partial areas within the cast component. As a result, an inner structure, in particular an inner surface structure, which is produced by sand cores or parts thereof, can be mapped and checked in an improved manner. Overall, the cast component can be checked in more detail and thus improved.

In particular, a method for checking a cast component is provided, with volume data of the cast component recorded and generated by computer tomography being received, with three-dimensional target model data of the cast component being received, with predetermined partial areas being received or defined in the target model data, with a global registration of the volume data of the cast component is carried out with the three-dimensional target model data of the cast component, with the specified sub-areas being locally compared independently of one another with respective corresponding sub-areas in the volume data, and with a check result being generated and output on the basis of the respective comparison results of the sub-areas.

Furthermore, in particular a device for checking a cast component is created, comprising an input device, a computing device, and an output device, wherein the input device is set up to receive computer tomographically recorded and generated volume data of the cast component, three-dimensional target model data of the cast component and predetermined partial areas in the target model data. wherein the computing device is set up to carry out a global registration of the volume data of the cast component with the three-dimensional target model data of the cast component, to compare the specified sub-areas independently of one another locally with the corresponding sub-areas in the volume data, and to provide a verification result based on the respective comparison results of the sub-areas generate, and wherein the output device is set up to output the generated verification result.

Three-dimensional target model data are, in particular, three-dimensional Computer Aided Design (CAD) data.

A computer tomograph is in particular a computer tomograph from industrial measurement technology. A cast component is in particular a workpiece that is produced using a casting process.

The volume data recorded and generated by computer tomography are recorded in particular by means of an industrial computer tomograph. For this purpose, the cast component is irradiated in particular by means of X-rays from a plurality of irradiation directions, the associated radiographic images being recorded in each case. The acquired radiographic images are reconstructed into the three-dimensional volume data (also referred to as volume model or object volume). Volume data refer in particular to raw data from voxel values that are obtained or were obtained by reconstruction of transmission images captured in different transmission directions. In particular, the volume data is recorded and generated as part of an inline check of the cast component. Provision can be made for the method to include the measure of acquiring and generating the three-dimensional volume data.

A global registration is intended to mean in particular that the three-dimensional target model data as a whole, that is to say not subdivided into partial areas, are registered with the three-dimensional volume data in a rigid, that is to say rigid, form. This can be done, for example, by minimizing a sum of a squared error between surfaces in the three-dimensional target model data and the three-dimensional volume data. In particular, this can be done starting from an outer shell of the cast component, that is, only the outer shell of the cast component is considered for global registration, internal structures are not taken into account and are only rigidly displaced together with the outer shell. In particular, it is provided that a coordinate system of the target model data serves as a reference coordinate system after the global registration, which is referred to in particular when comparing the partial areas.

A partial area is in particular a subset within the cast component. A partial area is in particular a three-dimensional section from the cast component, which can be defined, for example, by associated positions and/or surface contours. A sub-area corresponds in particular to a sand core or a part thereof.

Parts of the device, in particular the input device, the computing device and the output device, can be designed individually or combined as a combination of hardware and software, for example as program code that is executed on a microcontroller or microprocessor. However, it can also be provided that parts are designed individually or combined as an application-specific integrated circuit (ASIC) and/or field-programmable gate array (FPGA). In particular, a memory device is assigned to the computing device.

The verification result is generated in particular by combining and/or aggregating the individual comparison results. For example, the comparison results can be weighted. If, for example, deviations (e.g. in translation and rotation) between the volume data and the target model data are determined as comparison results for the individual partial areas, tolerance values can be specified which the comparison results must not exceed or fall below. Provision can be made here, for example, for the comparison results for all sub-areas to be within the respective tolerance range. If at least one comparison result is not within the respective tolerance range, this is displayed and/or encoded accordingly in the verification result.

The result of the check is made available, in particular output, in particular as an analog or digital signal, for example as a data packet.

In one embodiment it is provided that the result of the check includes a rejection decision. The rejection decision can be generated and output, for example, in the form of an analog or digital control signal. This allows direct intervention in a manufacturing process or such can be controlled directly. For example, it can be provided that a cast component is divided into the categories “OK”/“Not OK” depending on the test result. Accordingly, the rejection decision or a corresponding control signal then causes the checked cast component to be carried on or sorted out. In principle, a classification into different (graded) quality classes can also be provided in order, for example, to select and/or parameterize a subsequent process step.

In one embodiment, it is provided that the comparison takes place in each case by adapting a three-dimensional surface contour of the specified sub-areas derived from the three-dimensional target model data to a surface contour of a corresponding sub-area in the volume data, with transformation parameters of a transformation between the respective sub-area into the three-dimensional Target model data and the volume data are estimated. The surface contour in the volume data is determined, for example, by means of the Lorensen and Cline method cited at the outset. The adjustment, which can also be referred to as local registration, takes place separately for all specified sub-areas, that is, the adjustment takes place locally and independently of one another. For example, a sum of a mean squared error between points on the surface contours can be determined and minimized for fitting. This allows the transformation parameters to be determined, since the (rigid or rigid) target model data for the sub-area are shifted (translation) and/or rotated (rotation) within the initially determined coordinate system until the sum of the squared errors is minimized. A comparison with the original position of the sub-area within the target model data provides the transformation parameters. The transformation parameters, in particular for a translation (expressed, for example, by differences in x, y, z) and a rotation (expressed, for example, by angles alpha, beta and gamma), are in particular a measure of an error in the sub-area under consideration. If the sub-area under consideration coincides with a sand core or a part thereof in the casting mold used, a location (position and orientation) of the sand core or part thereof can be determined and checked during the casting process that has already taken place. In principle, other robust estimation methods such as RANSAC, M-Estimator (MSAC) or Hough Voting can also be used here, which in particular allow as many surface points as possible to be taken into account during the adjustment. In this way, in particular inconsistent surface points (which cannot be extracted robustly or consistently with the surface contour) can be discarded when determining or estimating the transformation parameters. These methods usually still work with an outlier share of up to 50-80% (depending on a statistical distribution of the outliers).

In one embodiment, it is provided that at least one distance between specified reference points on surface contours of the sub-areas in the volume data that correspond to the specified sub-areas is determined, with the at least one determined distance being compared with a corresponding target distance between corresponding reference points in the three-dimensional target model data, and wherein the verification result is generated taking into account a comparison result for the at least one distance determined. In this way, a position of the partial areas relative to one another can be checked. The reference points between which a distance is determined lie, in particular, on surface contours that belong to different partial areas.

In this case, a distance can be determined in particular by means of the Euclidean distance measure.

Provision can be made for a production quality to be assessed. For this purpose, it is provided in particular that a plurality of distances between respectively specified reference points are determined. A predetermined minimum distance (tolerance) is assigned to each distance. The determined distances are compared with the minimum distance specified in each case. If all of the specified distances exceed the assigned minimum distances, the cast component is evaluated as “good” in the inspection signal. However, if or if at least a minimum distance is exceeded, the cast component is evaluated as "bad" in the inspection signal. Based on this, a rejection decision can be generated and made available as part of the verification result.

In one embodiment, at least one wall thickness between adjacent partial areas is determined based on surface contours of the partial areas corresponding to the specified partial areas in the volume data, the determined at least one wall thickness being taken into account when generating the verification result. The wall thickness is in particular a minimum distance between two cavities within the cast component, which were created by sand cores or parts thereof. In this case, a distance can be determined in particular by means of the Euclidean distance measure. The at least one wall thickness is determined in particular at at least two reference points, which are permanently specified with respect to a surface contour of the respective partial areas. Alternatively, provision can also be made to determine a distance in pairs between all points on the surface contours of the two partial areas and to use the smallest distance as an estimated value for the wall thickness. Efficient methods such as OctTrees can be used here. A position of the reference points and points in the volume data is determined in particular via the (in each case locally determined) transformation.

In one embodiment, it is provided that a set with at least one comparison region is or will be defined for each predetermined sub-region, the at least one comparison region containing part of a surface contour of the respective predetermined sub-region, with target voxel values derived from the three-dimensional target model data within the at least one comparison region compared to voxel values in the volume data. The individual comparison results are then taken into account when generating the verification result. In particular, this embodiment makes it possible to check the individual sub-areas and to compare them with the associated target model data, even if it is not possible to adapt surface contours of the sub-areas to one another, for example because a surface contour cannot be determined with sufficient quality. In particular, several comparison regions are or will be defined for each sub-area. The comparison regions within a set are, in particular, evenly distributed over a surface contour of a partial area. The comparison regions are in particular three-dimensional areas. A comparison of the voxel values therefore also takes place in particular for a three-dimensional volume in each case.

In one embodiment, it is provided that, in order to compare the voxel values, the at least one comparison region is adapted to a region in the three-dimensional target model data that corresponds to the at least one comparison region, with transformation parameters of a transformation between the at least one comparison region and the corresponding region in the three-dimensional target model data being estimated become. The adjustment, which can also be referred to as local registration, takes place in particular taking into account all voxel values within a comparison region under consideration. The transformation parameters for each of the comparison regions are determined in principle in the same way as has already been described above with reference to the partial regions.

One embodiment provides for the adaptation to be carried out using a trained neural network. This can do that Adjusting can be done in an efficient and reliable way. The input data of the trained neural network are the voxel values within the comparison region derived from the target model data and the voxel values of the volume data within the comparison region. As an output, the trained neural network estimates transformation parameters for a transformation that describes a shift (in particular translation and rotation) of the comparison region under consideration between the target model data and the volume data. Training data for training such a neural network can be collected empirically, for example. In this case, a target function can be approximated, for example, by adapting surface contours to one another, as has already been described above. Alternatively, the training data can also be generated by simulating computer tomographic volume data in which displacements of partial areas and/or comparison regions are generated in a targeted manner. The neural network is trained using methods known per se. The neural network is in particular a deep neural network. The neural network is a convolutional network, for example.

In one embodiment it is provided that, in order to generate the verification result, a statistical evaluation of the comparison results determined for the sub-areas and/or the associated comparison regions is carried out at least for sub-areas. In particular, an aggregation of the individual comparison results for the comparison regions and/or the sub-areas can be carried out for this purpose. Robust estimation methods can be used for this, such as RANSAC, M-Estimator (MSAC) and/or Hough Voting etc.

Further features for the configuration of the device result from the description of configurations of the method. The advantages of the device are in each case the same as in the embodiments of the method.

Furthermore, in particular, a computer tomography system for checking a cast part is also created, comprising a computer tomograph and a device according to one of the described embodiments. For example, the system can be used for inline inspection of cast components in a production line. In particular, a subsequent manufacturing step can be controlled on the basis of the generated and provided verification result, which includes, for example, a rejection decision for a verified cast component. The cast component is either sorted out or further processed based on the inspection result provided or the rejection decision.

• 1 eine schematische Darstellung einer Ausführungsform der Vorrichtung zum Überprüfen eines Gussbauteils;
• 2a-2f schematische Darstellungen zur Verdeutlichung einer Ausführungsform der Vorrichtung und des Verfahrens zum Überprüfen eines Gussbauteils;
• 3 eine schematische Darstellung zur Verdeutlichung einer Ausführungsform der Vorrichtung und des Verfahrens zum Überprüfen eines Gussbauteils;
• 4 ein schematisches Ablaufdiagramm einer Ausführungsform des Verfahrens zum Überprüfen eines Gussbauteils.

The invention is explained in more detail below on the basis of preferred exemplary embodiments with reference to the figures. Here show:

• 1 a schematic representation of an embodiment of the device for checking a cast component;
• 2a-2f schematic representations to illustrate an embodiment of the device and the method for checking a cast component;
• 3 a schematic representation to clarify an embodiment of the device and the method for checking a cast component;
• 4 a schematic flowchart of an embodiment of the method for checking a cast component.

In 1 a schematic representation of an embodiment of the device 1 for checking a cast component is shown. The device 1 comprises an input device 2, a computing device 3 (for example a microprocessor), a memory device 4 and an output device 5. The input device 2 and the output device 5 are designed as interfaces, for example. The arithmetic unit 3 can access data stored in the storage unit 4 and carry out arithmetic operations on it. In particular, the device 1 performs the method described in this disclosure for inspecting a cast component.

One application situation is, for example, the following: cast components produced on a production line are computer tomographically recorded using a computer tomography system 50 and volume data 10 are generated from recorded radiographs. The device 1 can also be part of the computed tomography system 50 .

The volume data 10 of the cast component, recorded and generated by computer tomography, are supplied to the input device 2, which receives the volume data 10 and supplies them to the computing device 3, which stores the volume data 10 in the storage device 4 for processing. In particular, it is provided that voxels (grey values that correspond to a material-dependent attenuation of X-ray radiation) of the cast component and of air have already been identified in the volume data 10 so that their respective positions in the volume data 10 are known. The identification is for example by means methods known per se using the computed tomography system 50.

In addition, the input device 2 is supplied with three-dimensional target model data 20 of the cast component and specified partial areas 21 in the target model data 20 . The three-dimensional target model data 20 are CAD data, for example, and the specified partial areas 21-x are subsets within the CAD data, which are defined, for example, by position information and/or by functional units within the CAD data (e.g. a water jacket, channels for media, etc .). The input device 2 receives the target model data 20 and the specified sub-areas 21-x and feeds them to the computing device 3, which stores the target model data 20 and the specified sub-areas 21-x in the storage device 4 for processing.

The computing device 2 performs a global registration of the volume data 10 of the cast component with the three-dimensional target model data 20 of the cast component. This is done in particular starting from an outer shell of the cast component, i.e. only the outer shell of the cast component is considered in the global registration, so that the outer shell in the target model data 20 is registered with the outer shell in the volume data 10 (the remaining internal structures are in particular moved rigidly with the outer shell). Methods known per se can be used here, such as the method according to Besl and McKay cited at the outset. After the global registration, in particular a coordinate system of the target model data 20 registered in this way with the outer shell is used as a reference coordinate system in order to carry out subsequent comparisons.

The computing device 2 compares the specified partial areas 21-x independently of one another locally with the corresponding partial areas 11-x (cf. e.g 2 B or 3 ) in the volume data 10.

Based on the respective comparison results of the partial areas, the computing device 2 generates a verification result 30 which is output by the output device 5 . The verification result 30 can include a rejection decision 31, for example, which is output, for example, in the form of a rejection decision signal.

The test result 30, in particular the rejection decision 31, can then be fed to a production controller 51, for example, which uses a corresponding selection device (not shown) to separate out the respectively checked cast component or return it to the production line for further processing.

Provision can be made for the comparison to take place in each case by adapting a three-dimensional surface contour of the specified partial regions 21-x derived from the three-dimensional target model data 20 to a surface contour of a corresponding partial region in the volume data 10, with transformation parameters of a transformation between the respective Section 21 can be estimated in the three-dimensional target model data 20 and the volume data 10 . The surface contour of a corresponding partial area in the volume data 10 can be extracted, for example, by means of the Lorensen and Cline method cited at the outset. In the examples shown, the surface contours coincide with a respective outer contour of the subregions 11-x, 21-x, but in principle they can also be designed differently.

This is shown schematically in the 2a until 2f shown. The 2a shows a schematic representation of three-dimensional target model data 20 (eg CAD data), which can also be referred to as target geometry. In the example shown, the three-dimensional target model data 20 includes an outer shell 20-0 and three sub-areas 21-1, 21-2, 21-3, which in particular coincide with internal cavity structures that are formed during the casting process by sand cores or parts thereof placed in a mold be formed.

The 2 B shows the volume data 10 of the cast component in addition to the three-dimensional target model data 20 . Since errors have occurred during production, for example because the sand cores used to form the cavities or parts thereof have floated, the partial areas 11-1, 11-2, 11-3 not at intended target positions, as is shown schematically and by way of example in FIG 2 B is indicated. The displacements of the partial areas 11-1, 11-2, 11-3 differ from one another here.

In the 2c the global registration is clarified. For this purpose, the outer shell 20-0 of the three-dimensional target model data 20 is registered with an outer shell 10-0 of the volume data 10 using methods known per se. After the global registration, the outer shells 10-0, 20-0 are on top of each other. In particular, starting from the outer shells 10-0, 20-0, a transformation P0 of the volume data 10 to the target model data 20 is determined as part of the global registration. This is particularly necessary to avoid global effects, such as For example, to compensate for a positioning inaccuracy of a holder of the cast component when loading the computer tomograph.

After the global registration, the specified partial areas 21-1, 21-2, 21-3 are compared locally, independently of one another, with corresponding partial areas 11-1, 11-2, 11-3 in the volume data 10.

For this purpose, three-dimensional surface contours of the predetermined partial areas 21-1, 21-2, 21-3 derived from the three-dimensional target model data 20 are adapted to a surface contour of a corresponding partial area 11-1, 11-2, 11-3 in the volume data 10. In other words, a local registration of the surface contours takes place in each case. Here, transformation parameters of transformations P1, P2. P3 is estimated between the respective partial area 21-1, 21-2, 21-3 in the three-dimensional target model data 20 and the volume data 10. This is shown schematically in the 2d , 2e and 2f clarified. The adaptation or local registration of the surface contours takes place individually for each of the predetermined partial areas 21-1, 21-2, 21-3, that is, in particular a translation and rotation takes place during the adaptation or local registration of the respective partial area 11-1 , 11-2, 11-3 of the volume data 10 with the predetermined partial areas 21-1, 21-2, 21-3 of the target model data 20 detached from the respective other partial areas 11-1, 11-2, 11-3 and remaining areas of Volume data 10 of the cast component. It can be provided here that the predetermined partial areas 21-1, 21-2, 21-3 of the target model data 20 are shifted and, alternatively, that the respective partial areas 11-1, 11-2, 11-3 of the volume data 10, the two approaches being basically equivalent to one another. The translation parameters for the translation and the rotation of the respective transformations P1, P2, P3 are determined from the respective displacement.

Based on the transformation parameters of the individual transformations P0, P1, P2 and P3, the verification result is generated and made available. For example, the transformation parameters can be compared to predefined threshold values (tolerances) for the translation and/or the rotation. Depending on the individual comparison results, the test result includes, in particular, a rejection decision, for example a qualitative assessment of the cast component.

It can be provided that at least one distance D1, D2 ( 2 B ) between specified reference points x0', x1', x2', x3` ( 2 B ) on surface contours of the sub-areas 11-1, 11-2, 11-3 corresponding to the predetermined sub-areas 21-1, 21-2, 21-3 in the volume data 10 is determined, the at least one specific distance D1, D2 with a herewith corresponding target distance S1, S2 ( 2a ) between corresponding reference points x0, x1, x2, x3 ( 2a ) is compared in the three-dimensional target model data 20, and the check result 30 ( 1 ) is generated taking into account a comparison result for the determined at least one distance D1, D2.

$$\text{x1}\text{'}=\text{P2}*\text{P}0*\text{x1}$$

$$\text{x2}\text{'}=\text{P2}*\text{P}0*\text{x2}$$

$$\text{x3}\text{'}=\text{P3}*\text{P}0*\text{x3}$$

With the specific transformations P0, P1, P2 and P3 (which in particular each correspond to a linear mapping), the reference points x0`, x1', x2', x3` in the volume data 10 can be derived directly from the reference points x0, x1, x2, x3 of the Target model data 20 are determined: $$\text{x}0\text{'}=\text{<figure-callout id="1" label="P" filenames="DE102021209951A1\_0005.png,DE102021209951A1\_0006.png" state="{{state}}">P</figure-callout>}1*\text{P}0*\text{x}0$$

$$\text{x1}\text{'}=\text{p2}*\text{P}0*\text{x1}$$

$$\text{x2}\text{'}=\text{p2}*\text{P}0*\text{x2}$$

$$\text{x3}\text{'}=\text{P3}*\text{P}0*\text{x3}$$

In particular, it can be provided that the determined distances D1, D2 may only deviate from the target distances S1, S2 within a specified tolerance range. If at least one of the distances D1, D2 is outside of such a tolerance range, a corresponding check result 30 ( 1 ) generated. For example, the result of the check then includes a reject decision 31 that marks the cast component as reject. Otherwise, the verification result 30 includes, for example, a rejection decision 31 that marks the cast component as free of defects.

It can be provided that, based on surface contours of the sub-areas 11-1, 11-2, 11-3 corresponding to the predetermined sub-areas 21-1, 21-2, 21-3 in the volume data 10, at least one wall thickness between adjacent sub-areas 11- 1, 11-2, 11-3 is determined, wherein the determined at least one wall thickness when generating the verification result 30 ( 1 ) is taken into account. In principle, this can be done in the same way as has already been described for the distances D1, D2. In order to determine the wall thicknesses, in particular distances between all points on the surface contours of two of the partial areas 11-1, 11-2, 11-3 are determined. The smallest of the distances determined in this way is then taken as a measure for the (minimum) wall thickness. To generate the verification result 30 ( 1 ) It can be provided that the at least one wall strength is compared with at least one predetermined threshold. If at least one of the wall thicknesses falls below the specified at least one threshold value, a corresponding check result 30 ( 1 ) generated. For example, the verification result 30 then includes a scrap decision 31 that marks the cast component as scrap. Otherwise, the verification result 30 includes, for example, a rejection decision 31 that marks the cast component as free of defects.

In the 3 a schematic representation is shown to clarify a further embodiment. The device 1 is basically like that in FIG 1 embodiment shown formed. In the further embodiment, it is alternatively or additionally provided that a set with at least one comparison region 22-1, 22-2, 22-3 is or will be defined for each predetermined partial area 21-1, 21-2, 21-3. wherein the at least one comparison region 22-1, 22-2, 22-3 contains a part of a surface contour of the respective predetermined partial area 21-1, 21-2, 21-3, wherein the computing device 2 ( 1 ) target voxel values derived from the three-dimensional target model data within the at least one comparison region 22-1, 22-2, 22-3 are compared with voxel values in the volume data 10. This embodiment is particularly advantageous when surface contours of the corresponding partial areas 11-1, 11-2, 11-3 in the volume data 10 cannot be determined with sufficient quality. In this case, a comparison can be made based on the individual comparison regions 22-1, 22-2, 22-3.

The comparison regions 22-1, 22-2, 22-3 are in this case defined in particular with reference to the target model data 20 and correspond to regions in the volume data 10. The corresponding regions in the volume data 10 are determined here starting from the global registration by starting From this, the regions in the volume data 10 that correspond to the comparison regions 22-1, 22-2, 22-3 are determined without an adjustment between the partial regions 11-1, 11-2, 11-3, 21-1, 21 -2, 21-3 must be done. The target voxel values are determined on the basis of the target model data 20 of the cast component, which, in addition to geometric information, in particular also includes material parameters from which the values of the individual voxels can be determined.

To compare the comparison regions 22-1, 22-2, 22-3, it can be provided, for example, that the respective parts of the surface contours within the respective comparison region 22-1, 22-2, 22-3 are adapted to one another, in particular brought into registration with one another become. Based on this, transformation parameters can be determined for each of the comparison regions 22-1, 22-2, 22-3. A shift (translation and rotation) of a subarea 21-1, 21-2, 21-3 can be determined by aggregating the individual results of the respective comparison regions 22-1, 22-2, 22-3, with robust estimation methods being able to be used for this purpose.

In a further development, it can be provided that the adaptation is carried out using a trained neural network. For this purpose, the derived setpoint voxel values of a comparison region 22-1, 22-2, 22-3 and the voxel values of a region of the volume data 10 corresponding to the comparison region 22-1, 22-2, 22-3 are supplied as input data to the trained neural network. The neural network is trained, for example, to identify or estimate a shift (translation and rotation) based on this (e.g. values for a translation in the x, y, z direction and a rotation around the angles alpha, beta, gamma ). For the individual comparison regions 22-1, 22-2, 22-3, this takes place, in particular, independently of one another (in parallel or sequentially). A shift or transformation parameters of a transformation (translation and rotation) of a subregion 21-1, 21-2, 21-3 can be determined by aggregating the individual results of the respective comparison regions 22-1, 22-2, 22-3, with robust estimation methods can be used.

It can be provided that, in order to generate the verification result, at least in some areas using the computing device 2 ( 2 ) a statistical evaluation of the comparison results determined for the partial areas 21-1, 21-2, 21-3 and/or the associated comparison regions 22-1, 22-2, 22-3 is carried out. This is done, for example, using robust estimation methods such as RANSAC, M-Estimator (MSAC) or Hough Voting.

The 4 shows a schematic flowchart of an embodiment of the method for checking a cast component. The method is, for example, using a device according to in the 1 embodiment shown executed.

In a measure 100, computer tomographically recorded and generated volume data of the cast component are received. Provision can also be made for the acquisition and generation of the volume data by means of a computer tomography system or computer tomograph to be part of measure 100 .

In measure 101, three-dimensional target model data, in particular CAD data, of the cast component are received.

In measure 102, predetermined partial areas are received or defined in the target model data. In this case, the predetermined partial areas correspond in particular to functional structures in the cast component, in particular to structures that are produced during casting using sand cores or parts thereof.

In measure 103 a global registration of the volume data of the cast component with the three-dimensional target model data of the cast component is carried out. This is done in particular starting from an outer shell of the cast component, neglecting structures located within the outer shell.

In measure 104, the predetermined sub-areas are compared locally, independently of one another, with respective sub-areas in the volume data that correspond to them.

In step 105, a check result is generated based on the respective comparison results of the partial areas and output in step 106, for example as an analog or digital signal, for example in the form of a digital data packet.

The verification result can include a rejection decision, for example, which is output in the form of a rejection decision signal, for example. To generate the elimination signal, provision can be made in a measure 105a for the comparison results to be compared with specified criteria, for example in order to carry out a threshold value comparison and/or to check whether deviations lie within a specified tolerance. Depending on whether the specified criteria are met or not, a corresponding reject decision signal results.

In a measure 106, the test result, in particular the reject decision, can be fed to a production controller, for example, which uses a selection device set up accordingly for this purpose to separate out the respectively checked cast component in a measure 106a or to return it to the production line in a measure 106b for further processing.

Measure 104 can provide for the comparison to be carried out by adapting a three-dimensional surface contour of the specified sub-areas derived from the three-dimensional target model data to a surface contour of a corresponding sub-area in the volume data, with transformation parameters of a transformation between the respective sub-area being used in each case can be estimated in the three-dimensional target model data and the volume data.

Measure 104 can alternatively or additionally provide for a set with at least one comparison region to be defined or to be defined for each specified sub-area, with the at least one comparison region containing a part of a surface contour of the respective specified sub-area, with data derived from the three-dimensional target model data Target voxel values within the at least one comparison region are compared with voxel values in the volume data.

Further embodiments of the method have already been described above with reference to the other figures.

Reference List

1

contraption

2

entrance facility

3

computing device

4

storage device

5

exit facility

10

volume data

10-0

outer shell

11-x

corresponding section

20

target model data

20-0

outer shell

21-x

subarea

22-x

comparison region

30

verification result

31

committee decision

50

computed tomography system

51

production control

100-106

Measures

D1

Distance

D2

Distance

P0

transformation

P1

transformation

p2

transformation

P3

transformation

S1

target distance

S2

target distance

x0

Reference point (nominal model data)

x1

Reference point (nominal model data)

x2

Reference point (nominal model data)

x3

Reference point (nominal model data)

x0'

Reference point (volume data)

x1'

Reference point (volume data)

x2'

Reference point (volume data)

x3`

Reference point (volume data)

Claims (11)

Procedure for checking a cast component, wherein computer tomographically recorded and generated volume data (10) of the cast component are received, wherein three-dimensional target model data (20) of the cast component is received, wherein predetermined partial areas (21-x) are received or defined in the target model data (20), wherein a global registration of the volume data (10) of the cast component with the three-dimensional target model data (20) of the cast component is carried out, wherein the predetermined partial areas (21-x) independently of one another locally with corresponding partial areas (11-x) in the volume data (10 ) are compared, and a check result (30) is generated and output based on the respective comparison results of the partial areas (11-x, 21-x). procedure after claim 1 , characterized in that the verification result (30) includes a rejection decision (31). procedure after claim 1 or 2 , characterized in that the comparison takes place in each case by adapting a three-dimensional surface contour of the specified partial areas (21-x) derived from the three-dimensional target model data (20) to a surface contour of a corresponding partial area (11-x) in the volume data (10). , wherein in this case transformation parameters of a transformation between the respective partial area (11-x, 21-x) in the three-dimensional target model data (20) and the volume data (10) are estimated. Procedure according to one of Claims 1 until 3 , characterized in that at least one distance (D1, D2) between specified reference points (x0'-x3') on surface contours of the sub-areas (11-x) corresponding to the specified sub-areas in the volume data (10) is determined, the at least one specific distance (D1, D2) is compared with a corresponding target distance (S1, S2) between corresponding reference points (x0-x3) in the three-dimensional target model data (20), and the verification result (30) taking into account a comparison result for the specific at least a distance (D1,D2) is generated. Procedure according to one of Claims 1 until 4 , characterized in that based on surface contours of the sub-areas (11-x) corresponding to the specified sub-areas (21-x) in the volume data (10), at least one wall thickness between adjacent sub-areas (11-x) is determined, the determined at least one Wall thickness is taken into account when generating the verification result (30). Method according to one of the preceding claims, characterized in that for each predetermined sub-area (21-x) a set with at least one comparison region (22-x) is or will be defined, the at least one comparison region (22-x) being part of a Contains surface contour of the respective predetermined partial area (21-x), wherein target voxel values derived from the three-dimensional target model data (20) within the at least one comparison region (22-x) are compared with voxel values in the volume data (10). procedure after claim 6 , characterized in that to compare the voxel values, the at least one comparison region (22-x) is adapted to a region in the three-dimensional target model data (20) that corresponds to the at least one comparison region (22-x), with transformation parameters of a transformation between the at least a comparison region (22-x) and the corresponding region in the three-dimensional target model data (20) are estimated. procedure after claim 7 , characterized in that the adaptation is carried out by means of a trained neural network. Method according to one of the preceding claims, characterized in that to generate the verification result (30) a statistical evaluation of the respective comparison results determined for the sub-areas and/or the associated comparison regions (22-x) is carried out at least in part-areas. Device (1) for checking a cast component, comprising: an input device (2), a computing device (3), and an output device (5), wherein the input device (2) is set up to convert computer tomographically recorded and generated volume data (10) of the cast component, three-dimensional target model data (20) of the cast component and predetermined partial areas (21-x) in to receive the target model data (20), the computing device (3) being set up to perform a global registration of the volume data (10) of the cast component with the three-dimensional target model data (20) of the cast component, the specified partial areas (21-x) independently of one another locally to be compared with corresponding sub-areas (11-x) in the volume data (10), and to generate a verification result (30) based on the respective comparison results of the sub-areas (11-x, 21-x), and wherein the output device (5 ) is set up to output the verification result (30) generated. Computer tomography system (51) for checking a casting, comprising: a computer tomograph and a device (1). claim 10 .

Images