How to monitor the condition of your gear cutting machine

Abstract

In a method for monitoring the condition of a machine tool 1 having multiple machine axes, at least some of the machine axes are systematically tested in a test cycle and associated condition data are ascertained by performing measurements. On this basis, EOL data are predicted. This data is correlated with the noise behavior of a gear train including a workpiece machined using a gear cutter. The reverse direction is also disclosed, in which condition data are predicted from the EOL data.

Description

[Technical field]

[0001] The present invention relates to a method for monitoring the condition of a gear cutting machine, in particular a gear grinding machine, for machining toothed workpieces. [Background technology]

[0002] During hardening of toothed workpieces in gear cutting machines, especially external or internal gears, manufacturing deviations naturally occur, which manifest themselves as tolerances of the actual shape of the actually manufactured workpiece relative to the specified nominal shape. Manufacturing deviations can be caused, among other things, by failures or wear of various components of the gear cutting machine or by improper assembly of the components. For example, manufacturing deviations can be caused by the drive moving the slide of the gear cutting machine to a position other than the nominal position specified by the machine controller, by worn bearings in the spindle, or by machine parts not being properly connected to each other so that vibrations are sufficiently damped, etc.

[0003] Hardening finish is usually the last step in the machining of the workpiece. After this machining step, the workpiece is assembled into a gear train. After assembly, a final test of the gear train ("EOL test") is performed on an EOL test stand (EOL: End of Line). In particular, the noise behavior of the gear train is tested. Manufacturing deviations of the workpiece often result in undesirable noise excitation. Even the smallest manufacturing deviations that may occur during machining on a defect-free machine can result in noise. Especially in electric drives, the gear train components can sometimes be the main noise source, and noise generated in the gear train is of particular concern.

[0004] To avoid costly disassembly operations that would be necessary if the gear train is found to be prone to noise during EOL testing, it is desirable to be able to predict the noise behavior caused by the workpiece before it is mounted on the gear train. It is also desirable to be able to draw direct conclusions about specific causes in the gear cutting machine from the measured gear train noise behavior.

[0005] EP 1 239 633 A1 discloses a method for hardening the tooth flanks with correction and / or modification on a gear cutting machine, in which gear pairs which mesh with one another in a transmission or test device are machined taking into account their respective mating tooth flanks, and the tooth flanks of the associated workpieces are subjected to a periodic waviness correction or modification. The extent of the rotational error is determined by means of a gear measuring device and / or an error measurement of the rotational distance of the gear pair in the transmission. The measurement results serve as input values ​​for defining the amplitude, frequency and phase position of the periodic waviness correction for the tooth flanks of the gear pairs produced on the gear cutting machine. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] US Patent Application Publication No. 20140256223 Summary of the Invention

[0007] [First aspect: prediction of noise excitation] In a first aspect, it is an object of the present invention to provide a method for monitoring the condition of a gear cutting machine, which predicts at an early stage the occurrence of disturbance noise in a gear train comprising a workpiece machined by means of the gear cutting machine.

[0008] This object is achieved by a method according to claim 1. Further embodiments are set forth in the dependent claims.

[0009] Thus, a method for monitoring the condition of a gear cutting machine having several machine axes is provided, comprising the following steps: a) performing a test cycle in which at least a portion of the machine axes are systematically actuated and associated machine measurement data is obtained; b) performing a spectral analysis of the mechanical measurement data and calculating mechanical spectrum data from the mechanical measurement data; c) determining predicted EOL spectrum data based on the machine spectrum data, indicative of orders of excitations to be expected in the EOL spectrum when the workpiece machined using the gear cutting machine is mounted in a gear train and rotates in mesh with a mating gear within the gear train; A method is disclosed that includes:

[0010] The method further comprises: d) outputting the predicted EOL spectral data or at least one quantity derived from the predicted EOL spectral data; may include.

[0011] For the proposed method, machine measurement data are determined as described above in a test cycle while the machine axes are selectively in motion. The test cycle is performed while the gear cutting machine is stopped working, i.e. during the test cycle the gear cutting machine machining tool is not in working engagement with the workpiece. The machine measurement data can in particular be acceleration values ​​determined by means of an acceleration sensor, position values ​​determined by means of a position sensor and / or current values ​​determined by means of a current sensor. While the machine axes are in motion, machine spectral data (spectral state data of the machine) are calculated from the machine measurement data by means of a spectral analysis, in particular an order analysis, in order to determine at which frequencies or orders periodic excitations of the machine components occur.

[0012] Based on the machine spectrum data, the EOL spectrum data are predicted. This prediction is based on the following considerations: excitations of the machine components are propagated to the manufactured workpiece during machining. When a machining tool (machine tool) is dressed using the machine axes, these excitations are also propagated to the so-dressed tool and from there further to the workpiece during machining. Using the known kinematic linkages between the individual machine components and taking into account the selected machining parameters, it is possible to calculate how such excitations affect the manufactured workpiece. In particular, such excitations can lead to periodic deviations (waviness) in the tooth flanks of the workpiece. The expressions "kinematic linkage" or "kinematic chain" are used to indicate how the movement of one component is transferred to another during the machining process. In the machining process, the kinematic linkage between the tool spindle and the workpiece spindle, i.e. the correlation between the movement of the tool and the movement of the workpiece, plays a central role. In the case of a gear cutting machine being a machine that processes a workpiece by a generating process, such as a generating grinder or a skiving machine, this linkage is characterized by a rolling mesh between the workpiece and the tool and is determined by the geometry of the workpiece and the geometry of the tool. It is easy to calculate how the periodic excitation of the tool is propagated by this linkage as a waviness to the workpiece tooth surface. In particular, if the order of the excitation of the tool that causes it is known based on the rotation of the tool, it is easily possible to calculate what order, based on the rotation of the workpiece, the excitation in the gear train caused by the waviness will be. The ratio between these two orders is referred to below as the "propagation coefficient". It is also possible to calculate, for all other components to which the motion is propagated, how the vibration of one of these components affects the other components, in particular to calculate the corresponding propagation coefficient between the perturbation order of these components and the resulting perturbation order in the EOL spectrum.

[0013] Thus, calculating the predicted EOL spectral data can include applying a propagation factor to the machine spectral data, where the propagation factor depends on the kinematic linkage between the machine axis and the workpiece for which the machine spectral data was determined.

[0014] Thus, predicted EOL spectrum data are obtained by a specified method. These predicted EOL spectrum data are based on testing of the actual machine state and therefore include sources of perturbations in the machine that may not be known in advance and, as a result, cannot be included in the calculation or simulation of EOL spectrum data based only on the known design of the machine. As a result, the method allows a reliable prediction of in what order noise excitations will occur in the finished gear train. In this way, noise excitations can be predicted even before the workpieces are actually mounted in the gear train. Affected workpieces can be excluded and, if necessary, inspected in more detail, thus avoiding cumbersome disassembly of the fully assembled gear train. In addition, measures can be taken to identify and eliminate the sources of errors that lead to the expected noise occurrence.

[0015] It is important to note that obtaining predicted EOL spectral data is not necessarily a quantitative prediction of the noise power at various orders, but rather a qualitative indication of which orders are "perturbation orders" in the first place, i.e., which orders are expected to have significant noise power in the first place. In particular, the predicted EOL spectral data can include perturbation orders and associated intensity indicators, which represent estimates (which may only be very rough estimates) of the expected perturbation strength at the perturbation orders.

[0016] In particular, the predicted EOL spectrum data can be determined separately for each actuated machine axis, i.e. for each actuated machine axis, individual machine measurement data is determined, from which individual machine measurement data individual machine spectrum data is calculated, and on this basis individual EOL spectrum data is predicted for each actuated machine axis. In this way, it is possible to predict which machine axis may cause which perturbation order in the EOL spectrum data.

[0017] Preferably, steps a) to c) are repeated several times, The workpiece is machined with the gear cutting machine between test cycles, which are performed during a pause in the process when the machining tool is not engaged with the workpiece. The evolution of the predicted EOL spectrum data as a function of test cycles or time is then visualized and / or analyzed. This is based on the idea that the predicted EOL spectrum data based on a single test cycle may be of little value. On the other hand, wear or failure of gear cutting machine components may occur during the process, which will result in significant changes in the predicted EOL spectrum data. It is therefore proposed to take into account the evolution of the predicted EOL spectrum data over time. By appropriately visualizing the evolution of the predicted EOL spectrum data over time, it is possible to visually easily identify the perturbation orders at which significant changes in the behavior of the noise excitation are expected to occur over time. The evolution of the calculated EOL spectrum data can also be analyzed numerically. For example, the numerical analysis may include performing a regression analysis of the predicted perturbation strengths of selected perturbation orders or all perturbation orders using a suitable regression function, such as a polynomial of at least the second degree. For example, if at least one perturbation order is predicted to have a gradient of perturbation strength that meets a particular warning criterion, then based on this, a warning indicator may be determined and output.

[0018] If reference machine spectrum data for many reference machines required in many different reference test cycles are available, the reference EOL spectrum data can be predicted from each of these reference machine spectrum data. In that case, by comparing the predicted EOL spectrum data of the machine to be evaluated with the predicted reference EOL spectrum data of the reference machines or quantities derived from them, it is possible to automatically evaluate the predicted EOL spectrum data of the machine to be evaluated, and thus to automatically infer expected noise problems without requiring special knowledge and without requiring measured EOL spectra as the basis for the evaluation. In particular, a statistical analysis of the predicted reference EOL spectrum data can be performed for this purpose. For the ideas underlying this approach and further embodiments, reference is made to the patent application entitled "Method for Monitoring the Condition of a Machine Tool" filed on the same date as the present application and filed by the same applicant, the contents of which are incorporated herein by reference in their entirety.

[0019] [Second aspect: Identifying components causing noise problems] The reverse is also possible: the EOL measurements can be measured on the EOL test stand while the workpiece rotates in mesh with the mating gear in the gear train, a spectral analysis of the EOL measurements can be performed to obtain measured EOL spectrum data, and from the EOL spectrum data it can be concluded whether any of the individual components of the gear cutting machine conditions are causing perturbation orders in the EOL spectrum data due to their conditions. The EOL measurements can be determined by any suitable sensors on the EOL test stand, in particular acceleration sensors and sensors for determining rotation errors.

[0020] Thus, a method for monitoring the condition of a gear cutting machine having several machine axes is provided, comprising the following steps: a) performing an EOL test on a gear train having a workpiece machined by a gear cutting machine, the workpiece meshing with a mating gear within the gear train and rotating, and associated EOL measurement data is obtained; b) performing a spectral analysis of the EOL measurement data and calculating EOL spectrum data from the EOL measurement data; c) determining predicted condition data based on the EOL spectrum data, the predicted condition data indicating which orders of the at least one machine axis correspond to the calculated EOL spectrum data; A method is disclosed, including:

[0021] The method comprises: d) outputting the predicted state data or at least one quantity derived from the predicted state data; may include.

[0022] Using this method, it is possible to reach a conclusion as to which machine axes and, if applicable, which components of these machine axes are responsible for the perturbation noises that actually occur after the workpiece machined by means of the gear cutting machine is mounted in the gear train. The method also makes use of the fact that, if the kinematic linkages between the components of the gear cutting machine are known and the selected machining parameters are taken into account, the order of the perturbation noises associated with the rotation of the workpiece in the gear train can be easily calculated from the order of the measurement data determined for the components of the gear cutting machine.

[0023] In particular, the method can be carried out without the need for condition measurements on the gear cutting machine itself. However, particular advantages are obtained when the method is combined with condition measurements on the gear cutting machine. In this respect, the method comprises: e) performing a test cycle during which at least some of the machine axes are systematically actuated and associated machine measurement data is obtained; f) performing a spectral analysis in which machine spectrum data is calculated from the machine measurement data; g) determining predicted EOL spectrum data based on the machine spectrum data, indicative of at what orders excitations are expected in the EOL spectrum when the workpiece machined by the gear cutting machine is mounted in a gear train and rotates in mesh with a mating gear within the gear train; determining the predicted condition data includes comparing EOL spectrum data calculated from the EOL measurement data to the predicted EOL spectrum data.

[0024] By comparing the EOL spectrum data determined by measurements during the commissioning of the gear train with the EOL spectrum data predicted from the measurement data of the gear cutting machine, the origin of the perturbation noise can be determined particularly reliably.

[0025] [Third aspect: Use of machine learning algorithms] The methods discussed so far require knowledge of the kinematic linkages between the components of the gear cutting machine. In another aspect, the invention provides a method that allows prediction of the noise behavior of a gear train based on machine state measurements or drawing conclusions about the state of the gear train from the measured noise behavior of the gear train, even without knowledge of the kinematic linkages. The method uses a trained machine learning algorithm, the input variables of which are state data of the gear cutting machine and the output variables are predicted EOL data characteristic of the expected noise behavior of the gear cutting machine, or the input variables are EOL data and the output variables are predicted state data characteristic of the expected state of the machine.

[0026] The machine learning algorithm follows the steps below: a) performing a test cycle in which at least some of the machine axes are systematically actuated and associated status data are determined by measurements; b) machining at least one workpiece with the gear cutting machine while the gear cutting machine is in a state corresponding to the state data; c) mounting the machined workpiece onto a gear train; d) performing an EOL test on the gear train in which the workpiece rotates in mesh with a mating gear and associated EOL data is determined; e) storing the state data and corresponding EOL data in a training dataset; f) repeating steps a) to e) for a number of test cycles on workpieces having the same nominal shape and machined under the same processing conditions; g) training a machine learning algorithm with the training dataset; and is trained using

[0027] Thus, the training data set includes multiple condition data for multiple workpieces having the same nominal shape, machined under the same processing conditions, and mounted in the same type of gear train, along with corresponding EOL data.

[0028] The nominal shape includes in particular quantities such as the normal module, the number of teeth and the tooth helix angle, but may also include further quantities such as specified tooth flank modifications. Machining conditions are considered to be the same in particular if the machine axes are moved in the same way during the machining operation. For example, in case generating grinding is used as the machining process, if the workpieces are machined with the same radial feed, the same axial feed rate, the same shift speed, the rotational speed of the tool is the same for all workpieces, and the grinding worms used have the same number of threads and the same pitch height for all workpieces, so that the rotational speed of the workpieces is also the same. The processing conditions are the same. If the grinding worm is a dressable grinding worm that is dressed using a rotating disk-shaped dressing tool, the conditions during dressing, in particular the rotational speed of the tool spindle and the rotational speed of the dressing tool during the dressing process, are also part of the processing parameters.

[0029] A machine learning algorithm is trained using the condition data and the corresponding EOL data. As a result, the machine learning algorithm can predict EOL data based on condition data or vice versa without requiring knowledge of the kinematic linkages between the components of the gear cutting machine.

[0030] Many different types of machine learning algorithms are known that can be used in this regard, and the structure of the training data set may vary accordingly. In particular, classification algorithms are suitable for practical implementation. For this purpose, the output variables can be reduced to a limited number of classes. For example, if the input variables are EOL data and the output variables are predicted state data, the predicted state data can consist of, for example, one real value per machine axis. Each value can then indicate, for example, the probability that the corresponding machine axis is responsible for the observed EOL data. The training data would then include state data representing a single real state value per machine axis and the associated EOL data. For example, if the input variables are state data and the output variables are predicted EOL data, the predicted EOL data can consist of one real value per order for a relatively small number of orders (orders that are particularly important in practice). Each value can then indicate, for example, the predicted relative spectral intensity of the corresponding order. The training data would then include the corresponding EOL data. Of course, completely different and even more complex output quantities can also be envisaged. For practical implementations, for example, an artificial neural network (ANN) or a support vector machine (SVM) are suitable. In a particularly simple example, the input variables can be state data and the output variable is a single real value characterizing the global noise behavior of the entire gear train at the EOL test stand. For example, a random forest is suitable as a machine learning algorithm to predict such a value. With such a value, for example, expected problematic noise behaviors can be easily detected and measures can be taken to prevent affected workpieces from being installed in the gear train.

[0031] The condition data may generally include various types of data correlating with the condition of the machine shaft with respect to the vibration behavior of the machine shaft, and in particular the condition data may include the machine spectrum data defined with respect to the first and second aspects.

[0032] The EOL data may also include various types of data that correlate with noise characteristics of the gear train. In particular, the EOL data may include EOL spectrum data as defined in relation to the first and second aspects.

[0033] The training data can be stored in a database. This database can be located remotely from the machine being monitored. This database can also be implemented in the cloud, for example in the form of a computer resource shared by multiple users as a service. An evaluation computer can access the database to train the machine learning algorithm. This evaluation computer is preferably spatially separate from the machine tool. This evaluation computer is connected to the machine tool by a network connection. This evaluation computer also does not have to be a single physical device, but can be implemented in the cloud.

[0034] The invention further provides an apparatus for monitoring the state of a gear cutting machine having multiple machine axes, comprising a processor and a storage medium on which a computer program is stored. The computer program, when executed in the processor, performs at least a part of the method steps of one of the methods described above. The invention further provides a corresponding computer program. This computer program can be stored on a non-volatile storage medium.

[0035] Preferred embodiments of the present invention will now be described with reference to the drawings, which are intended to illustrate, but not to limit, these preferred embodiments of the present invention. [Brief description of the drawings]

[0036] [Figure 1] FIG. 1 is a schematic diagram of a generating grinding machine. [Diagram 2]FIG. 2 is a diagram illustrating the spectrum data acquired in a measurement cycle. [Diagram 3] FIG. 3 is a schematic diagram with an EOL test stand. [Figure 4A] FIG. 4A shows an excerpt from a table for converting machine component orders to workpiece orders that are attached to the gear train at the EOL test stand. [Figure 4B] FIG. 4B shows an excerpt from a table for converting machine component orders to workpiece orders that are attached to the gear train at the EOL test stand. [Diagram 5] FIG. 5 is a diagram illustrating the spectrum prediction data. [Figure 6] FIG. 6 is a schematic diagram of the time progression of perturbation strength in the perturbation order. [Figure 7] FIG. 7 is a schematic diagram illustrating a machine learning algorithm. [Figure 8] FIG. 8 shows an excerpt of exemplary training data for a machine learning algorithm. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0037] [Example structure of a generating grinding machine] FIG. 1 shows a generating grinding machine 1 as an example of a gear cutting machine, which will be referred to as the "machine" for short. The machine 1 has a machine bed 11, on which a tool carrier 12 is guided so as to be displaceable along a radial feed direction X. The tool carrier 12 holds an axial slide 13, which is guided so as to be displaceable along an axial feed direction Z relative to the tool carrier 12. A grinding head 14 is attached to the axial slide 13 and can be pivoted about a pivot axis extending parallel to the X direction (so-called A axis) to match the helix angle of the gear to be machined. The grinding head 14 also holds a shift slide that can shift a tool spindle 15 along a shift direction Y relative to the grinding head 14. A worm-shaped toothed grinding wheel (grinding worm) 16 is clamped to the tool spindle 15. The grinding worm 16 is driven by the tool spindle 15 to rotate about a tool axis B.

[0038] The machine bed 11 is also fitted with a swiveling workpiece carrier 20 in the form of a turret which can be swiveled between at least three positions about the swiveling axis C3. Two identical workpiece spindles are mounted diametrically opposite each other on the workpiece carrier 20. Of these workpiece spindles, only one workpiece spindle 21 with an associated tailstock 22 is visible in FIG. 1. A workpiece is clamped on each of the workpiece spindles and can be driven to rotate about the workpiece axis C1 or C2. The workpiece spindle 21 visible in FIG. 1 is located in a machining position in which a workpiece 23 clamped on it can be machined with a grinding worm 16. The other workpiece spindle, offset by 180° and not visible in FIG. 1, is located in a workpiece exchange position in which a finished workpiece can be removed from it and a new raw part can be clamped on it. A dressing device 30 is mounted offset by 90° to the workpiece spindles.

[0039] The machine 1 thus has a large number of movable components, such as slides or spindles, which can be moved under the control of corresponding drives, which are often referred to in the technical world as "NC axes" or "machine axes", or simply as "axes". In some cases, this designation also includes the components that are driven by drives, such as slides or spindles.

[0040] The machine 1 also has a number of sensors. By way of example, only two sensors 18, 19 are shown diagrammatically in FIG. 1. The sensor 18 is an acceleration sensor (vibration sensor) which detects vibrations of the housing of the grinding spindle 15. The sensor 19 is a position sensor which detects the position of the axial slide 13 along the Z direction relative to the tool carrier 12. However, in addition to these, the machine 1 also comprises a number of further sensors. These sensors include, in particular, a further position sensor which detects in each case the actual position of one linear axis, a rotation angle sensor which detects in each case the rotational position of one rotary axis, a current sensor which detects in each case the drive current of one axis and a further vibration sensor which detects in each case the vibrations of one driven component.

[0041] All driven axes of the machine 1 are digitally controlled by a machine controller 40. The machine controller 40 comprises several axis modules 41, a control computer 42 and a control panel 43. The control computer 42 receives operator commands from the control panel 43 and sensor signals from various sensors of the machine 1 and calculates therefrom control commands for the axis modules 41. The control computer 42 also outputs operating parameters for display on the control panel 43. The axis modules 41 provide at their outputs the control signals for one machine axis each.

[0042] A monitoring device 44 is connected to the control computer 42 .

[0043] The monitoring device 44 can be a separate hardware unit associated with the machine 1. The monitoring device 44 can be connected to the control computer 42 via an interface known per se, for example the known Profinet standard, or it can be connected to the control computer 42 via a network, for example the Internet. The monitoring device 44 can be spatially part of the machine 1 or it can be located spatially remote from the machine 1.

[0044] The monitoring device 44 receives various different measurement data from the control computer 42 during operation of the machine. Among the measurement data received from the control computer are sensor data obtained directly by the control computer 42 and data read out by the control computer 42 from the axis modules 41, including for example data indicating the target positions of the various machine axes and the target current consumption in the axis modules.

[0045] The monitoring device 44 may optionally have its own analog and / or digital sensor inputs for directly receiving sensor data as measurement data from further sensors, which are typically sensors not directly required for the control of the actual machining process, for example acceleration sensors for detecting vibrations or temperature sensors.

[0046] The monitoring device 44 can alternatively be implemented as a software component of the machine controller 40, for example running in a processor of the control computer 42, or can be designed as a software component of a service server 45, which will be described in more detail below. The service server 45 comprises a processor 451, which is only shown diagrammatically, and a storage medium 452.

[0047] The monitoring device 44 communicates with a service server 45 either directly or via the internet and a web server 47. The service server 45 in turn communicates with a database server 46 having a database DB. These servers may be located remotely from the machine 1. The server does not have to be a single physical device. In particular, the server may be implemented as a virtual unit in the so-called "cloud".

[0048] The service server 45 communicates with a terminal 48 via a web server 47. The terminal 48 is capable of running, among other things, a web browser for visualizing the received data and their evaluation. The terminal does not require any particular computing power. For example, the end device can be a desktop computer, a notebook computer, a tablet computer, a mobile phone, etc.

[0049] [Workpiece lot processing] In the following, it will be explained how a workpiece is machined using the machine 1.

[0050] To process a workpiece (unmachined part) that has not yet been machined, the workpiece is clamped by an automatic workpiece changer to a workpiece spindle in a workpiece exchange position. The workpiece exchange takes place in parallel with the processing of another workpiece in another workpiece spindle in a processing position. Once the new workpiece to be processed has been clamped and the processing of the other workpiece has been completed, the workpiece carrier 20 is swiveled 180° about the C3 axis so that the spindle with the new workpiece to be processed moves to the processing position. Before and / or during the swiveling process, a meshing operation is carried out with the associated meshing probe. For this purpose, the workpiece spindle 21 is rotated and the position of the tooth space of the workpiece 23 is measured with the meshing probe 24. The rolling angle is determined on this basis.

[0051] When the workpiece spindle holding the workpiece 23 to be machined reaches the machining position, the workpiece 23 is brought into collision-free meshing with the grinding worm 16 by moving the tool carrier 12 along the X-axis. The workpiece 23 is then machined by the rollingly meshed grinding worm 16. During machining, the workpiece is continuously advanced along the Z-axis with a constant feed in the radial direction X. In addition, the tool spindle 15 is slowly and continuously moved along the shift axis Y (so-called shift movement) in order to continuously use unused areas of the grinding worm 16 for machining.

[0052] In parallel with the workpiece machining, the finished workpiece is removed from the other workpiece spindle and another blank is clamped onto this spindle.

[0053] If, after machining a certain number of workpieces, the grinding worm 16 has become too dull and / or the tooth flank shape too inaccurate due to continued use, the grinding worm is dressed. For this purpose, the workpiece carrier 20 is swiveled ±90° so that the dressing device 30 reaches a position opposite the grinding worm 16. The grinding worm 16 is then dressed with the aid of a dressing tool 33, which here is a rotating dressing wheel.

[0054] [Gear cutting machine test cycle] While machining is halted, the monitoring system 44, in conjunction with the machine controller 42, executes test cycles in which the condition of individual or all components of the machine 1 is checked. During such test cycles, selected some or all of the machine axes are systematically actuated and measurements are made on the machine.

[0055] For example, each linearly movable component moves with an associated machine axis, and the current instantaneous position of that component is continuously determined using the position sensors mentioned above. From this process, the position deviation between the specification (nominal position) and the measurement (actual position) is continuously determined and transmitted to the monitoring device 44. The same can be done for a rotationally driven spindle, in which case a rotation angle sensor is used to determine the position deviation.

[0056] The vibration behavior of selected machine axes is also determined while the corresponding machine axes are in operation. Acceleration sensors (vibration sensors) connected to these components are used for this purpose. The results of the vibration measurements are also transmitted to a monitoring device 44.

[0057] Furthermore, the power consumption of the drive motors of the machine axes is determined continuously while they are in operation. Current sensors integrated in the axis modules 41 can be used for this purpose, for example. The results of the current measurements are also transmitted to the monitoring device 44.

[0058] All this can be done while one machine axis is actuated alone. However, it is also possible to actuate two or more machine axes in combination so that the behavior of the machine is recorded when two or more machine axes are actuated simultaneously. In this case, for example, amplified vibrations may occur that are greater than would be expected based only on the vibration behavior when a single machine axis is actuated, or controller errors may be detected that can only be determined when two machine axes are actuated synchronously.

[0059] [Status Data] The monitoring device 44 determines various status data from the received measurement data. These status data make it possible to draw direct or indirect conclusions about the state of the machine or its individual components. In particular, the status data comprises spectral data obtained from the measurement data by means of a spectral analysis. The complete spectrum can be determined or only the spectral intensity at selected discrete excitation frequencies can be determined.

[0060] Figure 2 shows an example of a spectrum that can be obtained by filtering and FFT calculation from the time signal of an acceleration sensor, position sensor or current sensor recorded during the operation of a machine axis (here the B axis, i.e. the tool spindle). The spectrum of Figure 2 contains several clearly visible peaks at integer and non-integer multiples (orders) of the rotation frequency of the machine axis in question.

[0061] For example, strong peaks in the tool rotation speed and its multiples may indicate a concentricity error in the tool spindle. Peaks at higher multiples of the tool rotation speed may indicate bearing damage in the tool spindle, and the bearing order can be inferred from the multiple. Once the bearing order is known, the bearing causing the peak can be identified.

[0062] The monitoring device 44 transmits the status data thus acquired to the service server 45 .

[0063] [EOL Test] The finished workpieces are then each mounted into a gear train that is tested on an EOL test stand prior to shipment, as will be described in more detail with reference to FIG.

[0064] 3 shows very diagrammatically the machine 1 together with the various servers 45-47 already mentioned above and a terminal device 48. Also shown very diagrammatically is an EOL test bed 2 which communicates with a service server 45 via a web server 47.

[0065] As previously mentioned, the machine 1 has a number of sensors including an acceleration sensor (vibration sensor) 51, a position sensor 52, and a current sensor 53. As also previously mentioned, the machine uses these sensors to collect measurement data and transmits status data derived from these measurement data to the service server 45.

[0066] The EOL test stand also has various sensors, including an acceleration sensor 54 that measures acoustic signals when a workpiece attached to a gear train rotates in mesh with a mating gear, a rotation angle sensor, etc. The EOL test stand calculates EOL data from these by spectrum analysis, and also transmits the EOL data to the service server 45.

[0067] [Service Server] The service server 45 processes the received data, calculates further quantities from the received data if necessary, and stores the received data and the calculated further quantities if necessary in a database DB. In particular, the service server stores the following data: status data having associated status identification data that are uniquely identifiable for the machine for which the status data is sought, the associated operation in the test cycle (in particular the machine axis actuated), and the time of the test cycle; · process data of the machining process for each workpiece accompanied by workpiece identification data enabling a unique identification of the workpiece; · EOL data from the EOL test stand with associated workpiece identification data.

[0068] The service server can read and merge data from the database, in particular, the service server can merge the EOL data for a particular workpiece with those machine condition data that best characterize the machine condition for the associated process data and processing conditions to form respective data sets.

[0069] [Prediction of perturbative noise excitations in EOL spectra based on kinematic linkage] The service server can perform a qualitative prediction of the intensity of the perturbation orders at the EOL test stand. For this purpose, the service server calculates the corresponding expected excitation spectrum at the EOL test stand (EOL spectrum) from the spectrum of FIG.

[0070] In this calculation, the service server makes use of the known kinematic linkages between the components of the machine 1. This is explained in more detail with reference to Figures 4A and 4B.

[0071] FIG. 4A shows an example of a section of a table where the known possible perturbation orders of the B axis (i.e. the tool spindle) and the corresponding expected perturbation orders in the EOL spectrum are entered. In this example, these perturbation orders are given by the kinematic linkage between the tool spindle and the workpiece, i.e. by the rolling mesh between the tool and the workpiece, and are in a fixed ratio of 3.45 determined by the geometry of the workpiece and the grinding worm. Figuratively speaking, this ratio describes how the vibrations of the B axis are propagated to the waviness on the tooth surface of the workpiece. This ratio can be calculated by taking into account the contact conditions between the grinding worm and the workpiece. This ratio is referred to below as the "propagation coefficient". The possible perturbation orders of the B axis can be calculated in advance if the orders of the components of the B axis are known, e.g. the bearing order and the motor order. The actual perturbation orders of the B axis can be determined by measurement.

[0072] FIG. 4B shows an example of a section of a table in which the possible perturbation orders of the Y-axis (i.e. the shift axis) and the corresponding expected perturbation orders in the EOL spectrum are entered. The table distinguishes, on the one hand, between the different components of the Y-axis that can be responsible for these perturbation orders, e.g. the ball screw drive BSD and the drive motor, and, on the other hand, between the perturbation orders in the EOL spectrum that are due to vibrations during workpiece machining (grinding) and dressing. Vibrations during grinding lead directly to tooth flank ripples (waviness) on the tooth flank of the workpiece. Vibrations during dressing first lead to tooth flank ripples on the grinding worm and from there are also transformed into tooth flank ripples on the workpiece tooth flank during grinding. The corresponding propagation coefficients between the possible perturbation orders of the Y-axis and the resulting perturbation orders in the EOL spectrum can also be easily calculated if the kinematic linkage and the machining parameters during grinding and dressing are known. The possible perturbation orders of the Y-axis can again be measured or calculated.

[0073] This type of analysis of possible perturbation orders of the machine axes and the resulting perturbation orders in the EOL spectrum can be performed for each machine axis involved in the grinding process.

[0074] A prediction of the EOL spectrum is then made based on the spectrum determined in the test cycle on the machine and the known propagation coefficients between the perturbation orders of the machine axes and the associated perturbation orders in the EOL spectrum. This will be explained in more detail with reference to FIG. 5. FIG. 5 shows the predicted EOL spectrum expected when the test of the B axis in the test cycle resulted in the spectrum of FIG. 2. This predicted EOL spectrum is essentially the same as the spectrum of FIG. 2, but stretched along the horizontal axis by the propagation coefficient of 3.45 already mentioned above as an example. However, the absolute signal value of this EOL spectrum should be considered with caution. That is, after all, how strong the EOL signal of a given perturbation order actually is depends not only on the strength of the corresponding perturbation order of the causative machine axis, but also on a number of other factors during the machining of the workpiece and the mounting of the workpiece in the gear train. In this respect, the spectrum of FIG. 5 does not allow a quantitative statement of the expected signal strength. However, the spectrum of Fig. 5 allows prediction of which perturbation orders are actually present in the EOL spectrum due to the perturbation orders present in the spectrum of the relevant machine axis, as well as a qualitative estimation of the expected signal strength at these perturbation orders. For example, the spectrum of Fig. 5 allows a rough estimation of the signal strength at specific perturbation orders of interest that cause noise that is perceived as particularly annoying. Such perturbation orders are marked by circles in Fig. 5 as an example.

[0075] Overall, a good prediction can be made as to what perturbation orders, due to what causes and with what signal strength, are expected in the EOL spectrum.

[0076] For example, worn bearings in the tool spindle can cause vibrations in the tool spindle, and the order of these vibrations (associated with the tool rotation) depends on the order of the bearing. The order of the bearing results from the bearing design and can often be obtained from the bearing manufacturer. Thus, the direct cause of the vibrations measured in the test cycle can be found in the worn bearing. These vibrations can be measured, for example, by an acceleration sensor on the housing of the workpiece spindle. The vibrations are propagated to the workpiece by the machining process and appear on the workpiece as periodic deviations (ripples / waviness) on the tooth flank. After mounting in the gear train, these ripples appear as noise excitations when the workpiece rotates in tooth mesh with the mating gear. The order of these noise excitations related to the rotation of the workpiece in the gear train can be easily calculated based on the above considerations. In this way, it is possible to calculate how worn bearings affect the noise spectrum of the gear train.

[0077] The calculated spectrum of FIG. 5 does not provide a quantitative description of the expected signal strength. However, by observing how this expected spectrum changes from test cycle to test cycle, it is possible to make a very good estimation of which perturbation order changes and how. This is illustrated below with reference to FIG. 6. FIG. 6 shows a diagram in which the expected spectrum strength I at the EOL test stand for a certain order (here the order is 52) is plotted as a function of the number of workpieces machined with the machine. It can be seen that the expected noise strength increases significantly over time. By fitting a suitable regression function (here a quadratic regression function), this increase can be quantified and, depending on the regression parameters found, appropriate actions can be triggered, for example a warning signal can be issued. The time course of the expected signal strength at the different EOL orders can also be visualized in a suitable way. This allows even an inexperienced user to unravel the noise problem.

[0078] [Identification of perturbation sources based on kinematic linkages] The reverse is also possible: if an EOL spectrum is determined by measurements on an EOL test stand, the above considerations can be used to deduce which machine axis, and if necessary which component of the machine axis, caused the perturbation order in the measured EOL spectrum. For this, it is back-calculated which order of the machine axis corresponds to the perturbation order in the measured EOL spectrum, and the expected perturbation order in the spectrum of the machine axis is searched for as the component corresponding to the order thus back-calculated. This process can be easily automated.

[0079] [Procedure without knowledge of kinematic linkage] If the kinematic linkage of the drive train is not known or should not be used in the calculations for other reasons, machine learning algorithms can be used to predict the signal strength at specific EOL perturbation orders and to identify the perturbation source.

[0080] This is explained below with reference to FIG. Figure 7 is a highly simplified schematic diagram of an artificial neural network (ANN). In this example, the network has only three inputs, two outputs, and one hidden layer. In reality, ANNs usually have many more inputs, outputs, and hidden layers.

[0081] In this example, condition data characterizing the vibration tendencies of one of the machine's axes B, Y, and Z are fed to the input of the ANN. From this data, the ANN calculates predicted EOL spectral data as expected spectral intensities at two particular EOL orders, here orders 52 and 59.

[0082] The ANN is pre-trained using training data, an example of a section of such training data is shown in Figure 8. FIG. 8 shows a table into which, on the one hand, state data determined from many test cycles of the machine are input. On the other hand, EOL data are input as spectral intensities in orders 52 and 59 obtained by EOL measurements on gear trains which were fitted with workpieces machined with the machine in the state in which the state data were obtained (i.e. immediately before and / or after the respective test cycle). The table contains a large number of rows of this type. The table can be taken from the database DB of FIG. 1 and FIG. 3. The ANN has been trained with this data in a manner known per se. The ANN is therefore able to reliably predict which EOL intensities in the abovementioned orders a state of the machine (represented by state data) will result.

[0083] The reverse is also possible: the input variables of the corresponding ANN can be EOL data and the output data can be predicted state data.

[0084] Of course, the above example is highly simplified, but it demonstrates the basic approach. Instead of an ANN, other types of ML algorithms can be used, in particular other known classifiers.

[0085] [Result output and visualization] The visualization of the results of these analyses can be done platform-independently on any client computer via a web browser. Other evaluation measures can be performed accordingly platform-independently. This makes the analysis easier, even remotely. In particular, the status of any machine can be viewed in detail from any mobile device via the cloud.

[0086] In addition, when a situation requires action, a corresponding message may be sent automatically via SMS, push message or email.

Claims

1. 1. A method for monitoring the condition of a gear cutting machine having multiple machine axes, comprising: a) performing a test cycle in which at least some of the machine axes are systematically actuated and associated machine measurement data is obtained; b) performing a spectral analysis of the mechanical measurement data and calculating mechanical spectral data from the mechanical measurement data; c) based on the machine spectrum data, determining predicted EOL spectrum data indicating at what orders excitations are expected in the EOL spectrum when the workpiece machined using the gear cutting machine is attached to a gear train and rotates in mesh with a mating gear within the gear train; The method comprising:

2. d) outputting the predicted EOL spectral data or at least one quantity derived from the predicted EOL spectral data; The method of claim 1 , comprising:

3. 3. The method of claim 1, wherein determining the predicted end-of-life spectrum data includes applying a propagation factor to the machine spectrum data, the propagation factor being dependent on a kinematic linkage between the workpiece and the machine axis for which the machine spectrum data was determined.

4. The method of claim 1 or 2, wherein the predicted EOL spectral data is determined separately for each actuated machine axis.

5. a) to c) are repeated several times, a workpiece is machined using the gear cutting machine between the test cycles, the test cycles being performed during a machining pause when the machining tool is not in machining engagement with the workpiece; 3. The method according to claim 1 or 2, wherein the progression of the predicted EOL spectrum data as a function of the test cycles performed, the workpieces processed or time is visualized and / or analyzed, in particular by regression analysis.

6. 3. The method of claim 1 or 2, Reference machine spectral data can be obtained for a plurality of reference machines, the reference machine spectral data being determined by a plurality of reference test cycles performed on the reference machines; From the reference machine spectral data, predicted reference EOL spectral data is determined; The predicted EOL spectral data determined based on the machine spectral data of the gear cutting machine being monitored is compared with the predicted reference EOL spectral data or a quantity derived from the predicted reference EOL spectral data; The method optionally comprising statistical analysis of the predicted reference EOL spectral data.

7. 1. A method for monitoring the condition of a gear cutting machine having multiple machine axes, comprising: a) performing an EOL test on a gear train including a workpiece machined by the gear cutting machine, in which the workpiece rotates in mesh with a mating gear within the gear train, and associated EOL measurement data is determined; b) performing a spectral analysis of the EOL measurement data and calculating EOL spectrum data from the EOL measurement data; c) determining predicted condition data based on the EOL spectrum data, the predicted condition data indicating which orders of at least one machine axis the predicted condition data of that machine axis match the calculated EOL spectrum data; and The method comprising:

8. d) outputting said predicted state data or at least one quantity derived from said predicted state data; The method of claim 7, comprising:

9. e) performing a test cycle in which at least some of the machine axes are systematically actuated and associated machine measurement data is obtained; f) performing a spectral analysis of the mechanical measurement data, from which mechanical spectral data is calculated; g) determining, based on the machine spectrum data, predicted EOL spectrum data indicating at what orders excitations are expected in the EOL spectrum when the workpiece machined by the gear cutting machine is attached to a gear train and rotates in mesh with a mating gear within the gear train; Further comprising:

9. The method of claim 7 or 8, wherein determining the predicted condition data comprises comparing the EOL spectrum data calculated from the EOL measurement data with the predicted EOL spectrum data.

10. 1. A method for creating a training dataset for a machine learning algorithm, comprising: a) performing a test cycle in which at least some machine axes are systematically actuated and associated condition data are determined by measurements; b) machining at least one workpiece using the gear cutting machine while the gear cutting machine is in a state corresponding to the state data; c) mounting the machined workpiece in a gear train; d) performing an EOL test on the gear train, wherein the workpiece rotates in mesh with a mating gear within the gear train and associated EOL data is determined; e) storing the status data and the corresponding end-of-life data in the training data set; f) repeating steps a) through e) above for multiple test cycles on workpieces having the same nominal shape and processed under the same processing conditions; The method comprising:

11. 11. A method of training a machine learning algorithm, wherein the machine learning algorithm is trained using the training dataset of claim 10.

12. 1. A method for monitoring the condition of a gear cutting machine having multiple machine axes, comprising:

11. The method, characterized in that a machine learning algorithm trained with the training data set of claim 10 is used in the method.

13. The machine learning algorithm has the gear cutting machine state data as an input variable and has predicted EOL data as an output variable; The method comprises: a) performing a test cycle in which at least some of the machine axes are systematically actuated and associated condition data are determined by measurements; b) determining predicted end-of-life data based on the state data by providing the state data as input variables to the trained ML algorithm; c) optionally outputting the predicted EOL spectral data or at least one quantity derived from the predicted EOL spectral data; 13. The method of claim 12, comprising:

14. The machine learning algorithm has EOL data as an input variable and predicted state data of the gear cutting machine as an output variable; The method comprises: a) performing an EOL test on a gear train in which the workpiece rotates in mesh with a mating gear within the gear train and associated EOL data is determined; b) determining predicted state data based on the EOL data by providing the EOL data as input variables to the trained ML algorithm; c) optionally outputting said predicted state data or at least one quantity derived from said predicted state data; 13. The method of claim 12, comprising:

15. The method according to any one of claims 10 to 14, wherein the machine learning algorithm is a classification algorithm, in particular an artificial neural network or a support vector machine, or a random forest.

16. the condition data correlates with the condition of the machine shaft with regard to the vibration behavior of the machine shaft and in particular comprises machine spectral data calculated by spectral analysis of machine measurement data; and / or The method according to any one of claims 10 to 14, wherein the EOL data correlates with the noise behavior of the gear train and in particular comprises EOL spectral data calculated by spectral analysis of EOL measurement data.

17. An apparatus for monitoring the state of a gear cutting machine having multiple machine axes, A processor (451) and a storage medium (452) in which a computer program is stored, The computer program, when executed on the processor, performs the following steps: receiving status data determined by a test cycle for the gear cutting machine, in which at least some of the machine axes are systematically operated and associated status data is determined by measurement; determining predicted end-of-life data based on the condition data, the predicted end-of-life data being correlated with noise behavior of a gear train having a workpiece machined using the gear cutting machine; The device performs the above.

18. An apparatus for monitoring the state of a gear cutting machine having multiple machine axes, A processor (451) and a storage medium (452) in which a computer program is stored, The computer program, when executed on the processor, performs the following steps: receiving EOL data determined by the EOL test for a gear train including a workpiece machined by the gear cutting machine, the workpiece in the gear train rotating in mesh with a mating gear and associated EOL data being determined; determining predicted condition data based on the EOL data, the predicted condition data correlating with a condition of at least one machine axis with respect to a vibration behavior of the at least one machine axis; The device performs the above.

Images