Method for grinding teeth

Abstract

The invention relates to a method comprising the method steps of: grinding a toothing (2) of a component (4) by means of a gear grinding machine (6), detecting a plurality of component-specific machine data, such as machining parameters, spindle currents, adjustment deviations, and the like, during the grinding of the component (4); according to the invention, one or more results of a computer-implemented rolling test (cWP) of the toothing (2) of the component (4) are determined by transmitting component-specific machine data or parameters derived therefrom as input data to a data model, the data model has a correlation between a result of the test-bed-based rolling test (pWP) and component-specific machine data associated with the result of the test-bed-based rolling test (pWP), and wherein the machine data is associated with the test-bed-based rolling test (pWP). Output data of the data model determined by means of the input data corresponds to the one result or the plurality of results to be determined of the computer-implemented rolling test (cWP).

Description

Technical Field

[0001] The invention relates to a method for grinding a toothing of a component by means of a gear grinding machine, wherein a plurality of component-specific machine data, such as machining parameters, spindle currents, control deviations etc., are detected during the grinding of the component. Background Art

[0002] Toothed parts are often manufactured in the process chain "pre-toothing", "hardening" and "hard finishing". For example, hard finishing can be carried out by grinding the toothed part.

[0003] In order to ensure that the required tolerances are met for toothed parts produced, for example in series production, quality control loops are used according to the prior art. Here, the produced toothed parts are measured and / or tested in order to determine the deviations of the produced toothed parts from predetermined target values ​​for the toothed parts. On the one hand, the components can be classified as good or bad parts and, if necessary, post-processed with the aid of the deviations, and on the other hand, corrections can be derived for the corresponding production steps for producing the toothed parts with the aid of the deviations.

[0004] The basis of such a quality control loop is therefore the deviations determined on the relevant toothing of the component, so that, for example, the above-mentioned corrections to the grinding process are determined on the basis of the component deviations.

[0005] Toothed parts (which are used, for example, in the transmission system of electrically operated vehicles) are often tested in so-called rolling tests for their noise behavior. Since the transmission noise is no longer masked by the engine noise or is masked by the engine noise to a lesser extent in purely electrically operated vehicles, it is important not to deliver toothed parts with significant noise and not to install them in such purely electrically operated vehicles. Therefore, in many cases, 100% of all toothed parts produced for electric vehicles in mass production are tested. Obviously, this results in considerable test costs. The rolling test is not an end-of-line test on an end-of-line test bench (EOL test bench). This means that the rolling test of the component is not carried out in the final installation in the gearbox, but by rolling with the main wheel or the corresponding mating wheel.

[0006] It is also known to monitor the axes and drives of a gear grinding machine during machine operation in order to identify defects or maintenance requirements of the gear grinding machine in question and / or component deviations. DE102018122041A1 of the applicant describes a method for calculating a "virtual" twin of a produced component with the aid of machine data detected during the production process in a component-specific manner. This virtual twin is analyzed computer-aided to identify noise significance and / or machine defects. A disadvantage here is that the calculation and analysis of the virtual twin requires high computing costs. In addition, the accuracy of the calculation of the virtual twin depends on precise knowledge of the tool geometry and the deviations of the machine axes and machine drives.

[0007] The classic quality control loop is based purely on measured test bench data, the method according to DE 10 2018 122 041 A1 is purely data-driven, both of which are time-consuming processes. Summary of the invention

[0008] Against this background, the invention is based on the technical problem of specifying a method that makes it possible to reduce the test expenditure for rolling tests. Furthermore, a method for producing a plurality of components is to be specified.

[0009] The technical problems described above are solved by the features of the independent claims. Advantageous embodiments of the invention are revealed from the dependent claims and the following description.

[0010] According to a first aspect, the invention relates to a method comprising the following method steps: grinding a toothing of a component by means of a gear grinding machine, detecting a plurality of component-specific machine data, such as machining parameters, spindle currents, control deviations, etc., during the grinding of the component; characterized in that a result or a plurality of results of a computer-implemented rolling test on the toothing of the component is determined by transmitting the component-specific machine data or parameters derived therefrom as input data to a data model, wherein the data model has a correlation between the results of the test bench-based rolling test and the component-specific machine data associated with the results of the test bench-based rolling test, and wherein output data of the data model determined as a function of the input data correspond to the result or the plurality of results to be determined of the computer-implemented rolling test.

[0011] According to the invention, the component-specific machine data are therefore evaluated with the aid of a data model based on real test bench data. In this way, the advantages of a conventional quality control loop based on measured deviations can be combined with the advantages of a data-driven approach. Thus, no complex and computationally intensive simulations are required to predict the expected results of the rolling test for the detected data sets of the component-specific machine data.

[0012] The rolling test is not an end-of-line test on an end-of-line test bench (EOL test bench). This means that the rolling test of the component is not carried out in the final installation state in the gearbox, but by rolling with the main wheel or the corresponding mating wheel.

[0013] In this context, the feature "test bench-based rolling test" means that this is a real, practical experiment that was carried out on a rolling test bench on a ground gear. The term "test bench-based" can therefore also be used synonymously with the terms "real" or "actual", since a test bench-based rolling test is a real or actual rolling test that is carried out on a rolling test bench.

[0014] If the term "computer-implemented rolling test" is used here, it means that this is a virtual rolling test or a computer-assisted rolling test. A computer-implemented rolling test is therefore a data processing with a computing device on which a software program product is implemented, which includes a data model or accesses a data model in order to determine the results of the virtual or computer-assisted rolling test as output data from component-specific machine data as input data.

[0015] The data model may have, for example, at least one AI model, the abbreviation “AI” standing for “artificial intelligence” as is known. The AI ​​model may have been trained using training data, which includes the results of a test bench-based rolling test of a plurality of components and component-specific machine data associated with the plurality of components.

[0016] The results of the test bench-based rolling test can include the results of the test bench experiments, i.e. the results of the single-flank rolling test and / or the double-flank rolling test. It can be provided that a test bench for the single-flank rolling test has been used for determining the training data. Alternatively or additionally, it can be provided that a test bench for the double-flank rolling test has been used for determining the training data.

[0017] It can be provided that the training data and the AI ​​model trained with the training data are machine-specific and that the AI ​​model is therefore assigned to a specific gear grinding machine. Alternatively, it can be provided that the training data are generated for a plurality of gear grinding machines across gear grinding machines, so that the training data are determined based on components that have been manufactured using different gear grinding machines.

[0018] According to one embodiment of the method, it can be provided that one or more of the following test characteristics are output as quantitative and / or qualitative results of a computer-implemented rolling test with the aid of the data model: center spacing, radial runout, rolling runout, rolling deviation, double ball size, radial runout error, long-wave and / or short-wave inter-tooth amplitude, maximum rolling deviation, rotation error and dynamic torsional play, noise characteristics, surface error. In particular, it can be provided that quantitative and / or qualitative results of a computer-implemented rolling test are output, which include the indexing error for the toothing in question.

[0019] It can be provided that the data model has an AI model for each of the corresponding output test features. It can therefore be provided that an AI model is generated for each individual test feature and trained with the aid of training data. The data model can therefore have a plurality of AI models, in particular a separate AI model for each individual test feature.

[0020] According to one embodiment of the method, it can be provided that the AI ​​model assigned to the respective test features is an artificial neural network or a classification model.

[0021] It may be provided that a frequency analysis, such as an FFT analysis or the like, is performed based on the one result of the computer-implemented rolling test or the multiple results of the computer-implemented rolling test in order to determine the dominant frequency when the toothing rolls.

[0022] Alternatively or additionally, provision can be made for the one result of the computer-implemented rolling test to be a dominant frequency when the toothing rolls.

[0023] Alternatively or additionally, it may be provided that the plurality of results of the computer-implemented rolling test are a plurality of main frequencies when the toothing rolls.

[0024] According to one embodiment of the method, provision can be made for a machine correction for the gear grinding machine to be determined as a function of the results of the computer-implemented rolling test on the toothing, in particular by means of a multi-objective optimization.

[0025] It can be provided that the machine correction comprises a change in process parameters for the gear grinding machine, wherein the process parameters are located within an n-dimensional process window, wherein the process window is limited by process limitations, such as, for example, maximum permissible axis speed and / or drive speed and / or axis acceleration and / or drive acceleration, conflicting structures within the machine area, maximum permissible tool temperature and / or workpiece temperature, grinding burns, etc.

[0026] The process parameters may be, for example, known process parameters of the grinding process, including feed rate, tool rotation speed, cutting distance per stroke, displacement distance per stroke, removal per unit time, feed amount, cutting depth, etc.

[0027] According to one embodiment of the method, it can be provided that an extrapolable model, such as a linear regression model, an AI model, etc., is provided for at least one process parameter, wherein the extrapolable model reflects the correlation between the process parameter and the process limit, and wherein the extrapolable model can verify compliance with the process limit for a predetermined change of the process parameter through the machine correction.

[0028] Such an extrapolable model can therefore have, for example, the influence of the feed rate and / or the tool rotation speed and / or the amount removed per unit time on the formation of grinding burn, so that it can be checked whether grinding burn forms for a proposed machine adjustment and the resulting changed process parameters. Another example is checking the mechanical or thermal load of the grinding tool for changed process parameters, wherein separate extrapolable models can be provided for the mechanical and thermal loads, respectively.

[0029] Therefore, one or more extrapolable models are provided for testing the corrections for process parameters derived from the computer-implemented rolling experiment with respect to their practical suitability.

[0030] It can be provided that the process parameters resulting from the machine correction are checked with respect to a stability criterion, in particular that the machine parameters are robust with respect to process fluctuations and / or that the machine correction represents a steady state of the gear grinding machine.

[0031] Certain combinations of process parameters can lead to the grinding process becoming unstable with only slight deviations from these process parameters, ie a satisfactory component quality can no longer be achieved. The stability criterion is therefore used to check how sensitive the changed process parameters resulting from the machine correction are to parameter fluctuations.

[0032] For this purpose, the data model can also be used, which contains the connection between the component-specific machine data and the results of the rolling test. Thus, the expected results of the rolling test and thus the component quality can be determined with the aid of the data model for changes in the process parameter in the direct vicinity of the process parameter. In this way, it is possible to predict to what extent the component quality is still achieved in the case of minor process fluctuations.

[0033] Furthermore, the stability criterion can be used to check to what extent the machine correction enables a steady-state state or steady-state operation of the gear grinding machine. This means checking to what extent the process now being carried out with the process parameters changed by the machine correction is a stable process, for example with regard to tool wear. Here again, an extrapolable model can be provided which indicates the connection between tool wear and one or more process parameters. Such a stability criterion can also be used as a process limit or process boundary as already described above, so that a changed process parameter or a machine correction is discarded if this would result in excessive tool wear.

[0034] The dedicated AI model can be provided with separate training data both for stability criteria and for process limitations, which are based on actual measurements and experiments.

[0035] According to one embodiment of the method, it can be provided that the machine data include axis motions and / or axis accelerations and / or vibration data of a machine axis or multiple controlled machine axes of the gear grinding machine, wherein the results of the computer-implemented rolling test contain periodic deviations of the actual geometry of the toothing from a target geometry of the toothing, and wherein the data model in particular reflects the correlation between the periodic deviations of the actual geometry of the toothing and the axis motions and / or axis accelerations and / or vibration data of the gear grinding machine.

[0036] The controlled machine axis has vibration characteristics for a specific motion curve or acceleration curve, in particular vibration characteristics in the range of its natural frequency, which can be transmitted to the tooth surface of the toothing during grinding. The vibration of the machine axis is generally not converted in a 1:1 manner into a ripple to be measured on the tooth surface or a deviation to be measured on the toothing, but due to the machine kinematics and the corresponding transmission, the following transfer function is generated, with which the specific vibration of the individual machine axis is reflected on the toothing. The connection between this vibration excitation of the machine axis and the corresponding deviation to be measured on the toothing can be represented by a transfer function. It can be provided that the data model thus reflects one or more transfer functions for one or more machine axes, so that the periodic deviation of the toothing can be determined as output data with the axis motion or acceleration curve as input data for the data model.

[0037] According to a further aspect, the invention relates to a method comprising the following method steps: carrying out the method according to any one of the preceding claims on a plurality of components; carrying out a test bench-based rolling test on one or more of the components in order to verify and / or improve the data model.

[0038] The data model can thus be improved and / or verified step by step. For example, it can be provided that actual experiments, i.e., rig-based rolling tests, are carried out on those components that were classified as defective by means of the computer-implemented rolling test in order to confirm the results of the computer-implemented rolling test. The results of the rig-based rolling tests can each be used as additional training data for the computer-implemented rolling test, provided that this is an AI model trained by means of training data.

[0039] If a gear grinding method for a new gear geometry is intended for series production, it can be provided, for example, that all manufactured components are first subjected to a test bench-based rolling test. The results of the test bench-based rolling test and the component-specific machine data associated therewith are used to train the AI ​​model for such a long time that the AI ​​model reflects the results of the test bench-based rolling test with a predetermined accuracy. In other words, the test bench-based rolling test is carried out for such a long time until the AI ​​model is successfully verified.

[0040] In other words, the validation cycle can be performed in the above manner for such a long time until a predetermined model quality is achieved. For example, it can be provided that the predetermined model quality for the validation is achieved when the coefficient of determination is greater than 0.8, in particular greater than 0.9, in particular greater than 0.95.

[0041] As soon as the predefined model quality is achieved, it is possible to switch from the test rig-based rolling test to the computer-implemented rolling test in series production, so that the test rig-based rolling test is only still used when necessary.

[0042] If a sufficient database is already available for a specific gear grinding method, the performance of a test bench-based rolling test can be dispensed with completely, even when the gear grinding method is already being set up. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The invention is described in more detail below with the aid of the accompanying drawings showing exemplary embodiments. In the drawings schematically:

[0044] Figure 1 A flow chart of the method according to the invention is shown. DETAILED DESCRIPTION

[0045] The method according to the invention has a method step (A) which comprises grinding the toothing 2 of a component 4 by means of a gear grinding machine 6. Such a gear grinding machine 6 is shown by way of example for method step (I) and will be described in more detail below.

[0046] During the grinding of the toothing 2 of the component 4 in method step (A), a plurality of component-specific machine data are detected, such as machining parameters, spindle current, setting deviations of the gear grinding machine 6, etc. The machining parameters are, for example, the feed or rotation speed of the grinding tool 8, the cutting depth, the amount removed per unit time, the stroke speed, the displacement path per stroke, etc., i.e. the machining parameters are the parameters required for defining the grinding process on the gear grinding machine 6. In the present case, the grinding process involves continuous generating grinding by means of the grinding tool 8 designed as a grinding worm.

[0047] After the toothing 2 of the component 4 has been ground, in method step (B) one result or several results of a computer-implemented rolling test of the toothing 2 of the component 4 are determined.

[0048] The computer-implemented rolling test cWP is carried out in that component-specific machine data or parameters derived therefrom are transmitted as input data to a data model, the data model having a correlation between the results of the test bench-based rolling test pWP and the component-specific machine data associated with the results of the test bench-based rolling test pWP, and wherein output data of the data model determined based on the input data correspond to a result to be determined or a plurality of results to be determined of the computer-implemented rolling test cWP.

[0049] With regard to method step (B), the results of a computer-implemented rolling test cWP are shown by way of example, wherein the rotation error F (in mm) is plotted by way of example and schematically over the rotation U of the component 4, i.e. the results of a single tooth flank rolling test of the toothed component 4 are shown. From this, values ​​for the first-order radial runout error Fr', the inter-tooth amplitude fi' and the maximum rolling deviation Fi' can be determined in a known manner, for example. According to an alternative embodiment of the invention, it can be provided that in method step (B), the pitch error of the toothing 2 is determined by means of a computer-implemented rolling test.

[0050] The data model cWP may have an AI model, wherein the AI ​​model has been trained with the aid of training data. The training and validation of the AI ​​model is described below with the aid of method steps (I), (II) and (III).

[0051] First, in method step (I), a plurality of toothings 2 on a plurality of components 4 are ground successively by means of a gear grinding machine 6, wherein each of the components 4 is subjected to a test bench-based rolling test pWP after grinding, which is carried out in method step (II). For this purpose, a test bench 10 for a single-flank rolling test and a test bench 12 for a double-flank rolling test are shown as examples and schematically with respect to method step (II).

[0052] The results of the test bench-based rolling tests pWP and the associated component-specific machine data are used to train the AI ​​model.

[0053] Method steps (I), (II) and (III) are repeated until the coefficient of determination is greater than 95%, that is, until the AI ​​model reflects the results of the actual rolling test pWP with high accuracy.

[0054] Method steps (I), (II) and (III) may also be referred to as a validation loop or training loop for the AI ​​model.

[0055] One or more of the test characteristics listed below can be output as quantitative and / or qualitative results of a rolling test implemented by a computer with the aid of a data model: axis spacing, radial runout, rolling runout, rolling deviation, double ball size, radial runout error, (long-wave and / or short-wave) inter-tooth amplitude, maximum rolling deviation, rotational error and dynamic torsional play, noise characteristics, surface error, indexing error.

[0056] It can be provided that for each individual test feature a separate AI model is trained and verified, so that there is a separate AI model for each test feature. Therefore, the data model D can have multiple AI models K1, K2...Kn, each of which is trained and verified separately.

[0057] Based on the results of the computer-implemented rolling test cWP of the gear teeth, a machine correction MK can be determined for the gear grinding machine 6, in particular by means of a multi-objective optimization. In this case, corrected values ​​are given for the feed or rotational speed of the grinding tool 8, the cutting depth, the removal per unit time, the stroke speed, the displacement path per stroke, etc. (i.e. the parameters required for defining the grinding process on the gear grinding machine 6) in order to improve the grinding result or the quality of the gear teeth with respect to the test characteristics of the rolling test.

[0058] After the correction MK of the production parameters of the gear grinding machine has been carried out, the AI ​​model or the plurality of AI models K1 , K2 . . . Kn can be revalidated, more precisely by means of method steps (I), (II) and (III).

[0059] The machine correction MK can relate both to the tooth grinding performed by means of the gear grinding machine 6 and to the dressing of the grinding tool 8 performed by means of the gear grinding machine 6 .

[0060] Before the correction MK is applied to the gear grinding machine 6, it can be checked in method step (B1) whether the correction MK complies with process limits and / or the correction can be checked with respect to its stability. It can therefore be provided that an extrapolable model MS, such as a linear regression model, an AI model, etc., is provided for at least one process parameter, wherein the extrapolable model MS reflects the correlation between the process parameter and the process limits, and wherein the extrapolable model for the change of the process parameter predetermined by the machine correction MK makes it possible to check whether the process limits are complied with.

[0061] If method step ( B1 ) shows that the process limits and / or stability criteria are not being met, then the adjusted correction can be determined with the aid of the multi-objective optimization and rechecked in step ( B1 ).

[0062] Furthermore, provision can be made for the toothed parts 2 which have been identified as defective by means of the computer-implemented rolling test cWP to be fed to a test bench-based rolling test pWP in order to verify the results of the computer-implemented rolling test cWP. The results of the test bench-based rolling test pWP can be used together with the associated component-specific machine data as training data to improve the AI ​​model or the plurality of AI models K1, K2 .. Kn.

[0063] It can be provided that during series production, the toothing 2 is tested in a decentralized manner by means of a test bench-based rolling test pWP not only for components 4 declared as good parts by means of a computer-implemented rolling test cWP, but also for components declared as defective parts, in order to generate further training data for improving the AI ​​model or the multiple AI models K1, K2 .. Kn.

[0064] According to one embodiment of the present invention, the machine data may include axis motion and / or axis acceleration and / or vibration data of a machine axis or multiple controlled machine axes X, Y, Z, A, B, C, B2, C3 of the gear grinding machine 6, wherein the result of the computer-implemented rolling test cWP contains periodic deviations of the actual geometric structure of the tooth portion 2 from the target geometric structure of the tooth portion 2, and wherein the data model D particularly reflects the correlation between the said periodic deviations of the actual geometric structure of the tooth portion 2 and the axis motion and / or axis acceleration and / or vibration data of the gear grinding machine 6.

[0065] Reference numerals list

[0066] 2 teeth

[0067] 4 Components

[0068] 6 Gear grinding machine

[0069] 8 Grinding worm

[0070] 10Test bench for single tooth flank rolling test

[0071] 12Test bench for double tooth flank rolling test

[0072] (A) Methods and steps

[0073] (B) Methods and steps

[0074] (B1) Methods and steps

[0075] (I) Methods and steps

[0076] (II) Methods and steps

[0077] (III) Methods and steps

[0078] DData Model

[0079] K1AI Model

[0080] K2AI Model

[0081] KnAI Model

[0082] MK machine correction

[0083] MS extrapolated models

Claims

1. A method comprising the following steps: - grinding a toothing (2) of a component (4) by means of a gear grinding machine (6), during which a plurality of component-specific machine data, such as machining parameters, spindle current, control deviations, etc., are detected; It is characterized in that - determining one or more results of a computer-implemented rolling test (cWP) on a toothing (2) of a component (4) by: -- component-specific machine data or parameters derived therefrom are transferred as input data to the data model, - wherein the data model has a correlation between the results of the test rig-based rolling test (pWP) and component-specific machine data assigned to the results of the test rig-based rolling test (pWP), and - wherein the output data of the data model determined as a function of the input data correspond to the result to be determined or the results to be determined of the computer-implemented rolling test (cWP).

2. The method according to claim 1, characterized in that - the data model has at least one AI model, - the AI ​​model has been trained using training data, and The training data include results of a test bench-based rolling test (PWP) of a plurality of components and component-specific machine data assigned to the plurality of components.

3. The method according to claim 2, characterized in that The results of the test bench based rolling test (pWP) include the results of the test bench experiments, ie the results of the single flank rolling test and / or the double flank rolling test.

4. The method according to claim 3, characterized in that One or more of the following test characteristics are output as quantitative and / or qualitative results of the computer-implemented rolling test (cWP) using the data model: Axis spacing, radial runout, rolling runout, rolling deviation, double ball size, radial runout error, long-wave and / or short-wave inter-tooth amplitude, maximum rolling deviation, rotational error and dynamic torsional backlash, noise characteristics, surface error, indexing error.

5. The method according to claim 4, characterized in that The data model has in each case one AI model for the corresponding output test features.

6. The method according to claim 5, characterized in that The AI ​​model matched with the corresponding test characteristics is an artificial neural network or a classification model.

7. The method according to any one of the preceding claims, characterized in that Performing a frequency analysis, such as an FFT analysis, based on the one result of the computer-implemented rolling test (cWP) or the multiple results of the computer-implemented rolling test (cWP) to determine the main frequency when the tooth (2) rolls; and / or One result of the computer-implemented rolling test is a main frequency when the tooth portion (2) rolls; and / or The plurality of results of the computer-implemented rolling test are a plurality of main frequencies when the tooth portion (2) rolls.

8. The method according to any one of the preceding claims, characterized in that Based on the results of the computer-implemented rolling test (cWP) on the toothing (2), a machine correction (K) for the gear grinding machine is determined, in particular by means of a multi-objective optimization.

9. The method according to claim 8, characterized in that - the machine correction (K) comprises a change in process parameters of the gear grinding machine (6), - in, The process parameters are located within an n-dimensional process window, - wherein the process window is limited by process limitations, such as, for example, maximum permissible axis speed and / or drive speed and / or axis acceleration and / or drive acceleration, interfering structures within the machine area, maximum permissible tool temperature and / or workpiece temperature, grinding burns, etc.

10. The method according to claim 9, characterized in that - Providing an extrapolable model for at least one process parameter, such as a linear regression model, an AI model, etc., - wherein the extrapolable model reflects the interrelationships between the process parameters and process limits, and - wherein the extrapolable model enables a check of compliance with the process limits for a change of the process parameter which is predetermined by the machine correction.

11. The method according to claim 9 or 10, characterized in that: - checking the process parameters resulting from the machine correction with respect to stability criteria, in particular: - The machine parameters are robust to process fluctuations, and / or - The machine correction represents the steady state condition of the gear grinding machine.

12. The method according to any one of the preceding claims, characterized in that - the machine data comprise axis motions and / or axis accelerations and / or vibration data of a machine axis or of a plurality of controlled machine axes of the gear grinding machine, - wherein the results of the computer-implemented rolling test include periodic deviations of the actual toothing geometry from a target toothing geometry, and wherein the data model reflects in particular the correlation between the periodic deviations in the actual geometry of the toothing and the axis movements and / or axis accelerations and / or vibration data of the gear grinding machine.

13. A method comprising the following method steps: - carrying out a method according to any of the preceding claims on a plurality of components (4), - performing a test rig based rolling test on one or more of the components to verify and / or improve the data model.

Images