Method for hard finishing a workpiece with a toothing or a profile on a hard finishing machine

Abstract

The invention relates to a method for hard fine machining of a workpiece on a hard fine machining machine, wherein the hard fine machining machine has a number of machine axes, wherein in the method the workpiece is received on a workpiece spindle and is machined by means of a tool which is received on a tool spindle by engaging the tool with the workpiece.In order to enable improved condition monitoring of the hard fine machining machine, the invention provides that at least in a time segment of the machining phase in which the tool is not in engagement with the workpiece, a test run is carried out on at least one machine axis, wherein the machine axis is driven during the test run, wherein a value is specified for at least one drive parameter for which a target value of the movement of the machine axis is expected, wherein the actual value of the movement of the machine axis is recorded, wherein a monitoring device compares the target value and the actual value and wherein a signal is output by the monitoring device if the comparison between the target value and the actual value has shown that the difference between the target value and the actual value lies outside a permissible range.

Description

[0001] The invention relates to a method for hard fine machining a workpiece with a toothing or a profile on a hard fine machining machine, wherein the hard fine machining machine has a number of machine axes which are actuated in an NC-controlled manner for machining the workpiece, wherein in the method the workpiece is received on a workpiece spindle and is machined by means of a tool which is received on a tool spindle by engaging the tool with the workpiece, wherein the machining phase of the workpiece begins with the introduction of the workpiece into the hard fine machining machine and ends with the removal of the machined workpiece from the hard fine machining machine.

[0002] A generic method is known from EP 3 168 700 A1. In this method, a workpiece is machined in a known manner under numerical control by bringing the tool into engagement with the workpiece to be machined, in particular the workpiece to be ground. In addition to the actual machining time, various non-productive times occur, for example, for loading and unloading workpieces and for dressing a grinding tool.

[0003] Hard finishing machines are subject to wear and tear due to operation, so machine maintenance must be performed at regular intervals and machine components replaced if necessary. The same applies, for example, in the event of an unexpected event (e.g., a crash, overload). It is important to avoid workpiece production while the machine is no longer in proper working order due to wear or damage.

[0004] Therefore, in practice, it is common practice to periodically perform reference runs on the machine to check the status of the machine, and especially to assess wear. These runs involve moving the machine axes (linear axes / rotary axes) outside of the actual machine operation, and recording the response to specified control signals. This allows you to check whether the machine components involved are still in perfect condition or whether changes in the response to specified control signals due to wear have already been detected. In the latter case, appropriate machine maintenance is then required.

[0005] The cyclical reference runs provide a very precise picture of the machine's condition. Accurate results are obtained because the reference runs allow for precise movements under similar conditions, thus minimizing the impact of disruptive factors on the system.

[0006] The disadvantage, however, is that the time-interval for determining changes to the machine depends on the selected period for these reference runs. In practice, one strives to avoid performing reference runs too frequently due to the time required (since no workpieces can be machined during these periods). This then leads to relatively long periods in which the machine is not being tested (long periods between reference runs). Furthermore, longer reference runs can potentially lead to problems with the machine's temperature response.

[0007] The invention is based on the object of developing a method of the type mentioned above in such a way that improved condition monitoring of the hard finishing machine and, in particular, its machine components, in particular the linear and rotary axes, is possible. This should enable the most intensive machine monitoring possible without any loss of machine capacity. Furthermore, the aim is to detect possible causes of wear more precisely.

[0008] This object is achieved by a method for hard fine machining of a workpiece with the features of patent claim 1, according to which it is provided that at least in a time segment of the machining phase in which the tool is not in engagement with the workpiece, a test run is carried out on at least one machine axis, wherein the machine axis is driven during the test run, wherein a value is specified for at least one drive parameter, for which a target value of the movement of the machine axis is expected, wherein the actual value of the movement of the machine axis is recorded, wherein a monitoring device compares the target value and the actual value, and wherein a signal is output by the monitoring device when a) the comparison between the target value and the actual value has shown that the difference between the target value and the actual value is outside a permissible range or b) the comparison between the target value and the actual value has shown that the difference between the target value and the actual value has a value which, compared to previously recorded values ​​of the difference, indicates that a permissible tolerance band has been exceeded.

[0009] During the test run, an operating parameter is specified for at least one machine axis, and its response to the operating parameter is recorded. A signal is issued if a deviation between an expected value (target value) and the measured value (actual value) exceeds a permissible range (case a). However, a trend analysis can also be performed, in which values ​​recorded during previous test runs are incorporated to determine whether the machine axis is about to leave a permissible operating range (case b).

[0010] The at least one time period of the machining phase preferably falls within the time interval during which the workpiece is mounted on the workpiece spindle. However, other time periods of the machining cycle can also be considered here, for example, periods during which balancing or dressing takes place.

[0011] Regarding the nomenclature used here, it should be noted that a "test run" (as already mentioned above) refers to the described monitoring of a machine axis during the regular machining process of a workpiece - even if this only occurs for partial machining processes in which the tool does not engage the workpiece -, whereas a "reference run" is a separate test run of the (complete) machine axes that is carried out outside the machining process of a workpiece.

[0012] The test run preferably takes place during an idle stroke, during which the tool is moved relative to the workpiece without the tool engaging the workpiece. According to another preferred option, the test run can also take place during an acceleration or rotational acceleration process and / or during a braking process of a machine axis. Other sections of the machining cycle are also possible, such as the rotation of a rotary table or dressing.

[0013] The test run is particularly preferably carried out during a dressing phase in which the tool is profiled, and / or during a phase in which the workpiece is aligned for machining in the hard finishing machine, and / or during a phase in which a balancing process is carried out, and / or in a phase in which a loading or unloading process is carried out.

[0014] One advanced development stipulates that the signal is only output after a statistical analysis and / or other evaluation of a number of recorded values ​​has been performed. This is intended to prevent a signal from being output in the case of an "outlier" during the test drive. The aforementioned statistical analysis generally refers to a (further) analysis of the recorded data, which reveals trends that are emerging in the recorded data.

[0015] The proposed method is particularly advantageous when using artificial intelligence (AI). In particular, the assessment of the recorded values ​​can be carried out using an algorithm that takes feedback (or empirical values) into account within the framework of machine learning (ML).

[0016] The detected value is preferably the acceleration detected by an acceleration sensor, and / or the current with which a machine axis is driven, and / or a control difference that occurs during the control of the machine axis, and / or a signal that provides an indication of a deviation between the target value and the actual value.

[0017] In the case of acceleration measurements, correspondingly low values ​​or if a signal deviates from calculated or known curves in the time and / or frequency domain can be used to conclude that defective operation is beginning to occur. If the current used to drive a machine axis becomes noticeably high when a desired movement is specified (in the time and / or frequency domain), this can also be used to conclude that increased wear is occurring. This also applies to the control error, which the controller must compensate for; if increased values ​​of the control error occur here (in the time and / or frequency domain), it can be concluded that wear is making it more difficult for the controller to maintain the specified position. In this respect, it should be noted that the evaluated and monitored signals can be viewed in both the time and frequency domain.

[0018] If the assessment shows that the value is outside the permissible range, the output of the monitoring device signal can trigger a reference run of the hard finishing machine, during which all machine axes are operated without the tool engaging the workpiece. Accordingly, the signal triggers a reference run for the machine if anomalies are detected during one or more test runs.

[0019] The proposed method is preferably used for gear grinding.

[0020] According to the proposed concept, the condition of the machine, and in particular of its components, can be recorded during a single machining cycle of a workpiece, thus enabling close and time-neutral monitoring. The time-consuming and therefore only randomly performed reference runs no longer need to be performed cyclically, but can be performed as needed.

[0021] The monitoring quality increases due to the tight timing, and non-productive times can be advantageously reduced. This results in a reduction in the number of improperly machined workpieces (due to worn or faulty machine components). Furthermore, root cause analysis is significantly simplified and provides a good basis for preventative machine maintenance (“predictive maintenance” concept).

[0022] A further advantage is that causes, especially for sudden wear and tear, can be determined much more easily and assigned much more precisely.

[0023] Another advantage is that the availability of the machine for effective machining of the workpiece can be increased, since non-productive times that are already available are used for test runs and thus unplanned machine downtimes can be reduced or avoided with the procedure described.

[0024] Finally, it is advantageous that critical temperature fluctuations in the machine can be avoided.

[0025] The drawing shows embodiments of the invention. Fig. 1 shows schematically the wear pattern of a component of a gear grinding machine, with a wear index plotted over time, Fig. Figure 2 shows schematically the wear pattern of a component of the machine, with a wear index plotted over a number of measurements, Fig. 3 shows a schematic flow chart of how criteria for carrying out a reference run of the machine are defined, Fig. Figure 4 shows schematically the concept for the data exchange provided according to an embodiment of the proposed method, Fig. 5 shows schematically the course of three machine axes over time, with rectangles marking areas in which test runs are planned, and Fig. 6 schematically shows a flow diagram for the implementation of the proposed concept according to a possible embodiment.

[0026] The following explanations are based on the concept described, according to which, particularly during gear grinding, a test run is carried out on at least one machine axis during a period of the machining phase in which the tool is not in engagement with the workpiece. During this test run, the machine axis is actuated according to at least one predetermined operating parameter, and the reaction of the machine axis to the operating parameter is recorded. The recorded value is assessed by a monitoring device. A signal is output by the monitoring device if the assessment shows that the value lies outside a permissible range. In this case, an irregularity in the machine exists, which can be attributed in particular to corresponding wear.Therefore, a reference run is then carried out in which all axes of the machines are operated and checked according to a defined program.

[0027] In contrast to the previously known approach, all or at least some of the idle strokes and acceleration phases of the axes are now used to evaluate the machine condition. In principle, all movements during the machining process (especially tool dressing, workpiece alignment, etc.) in which the workpiece and tool are not in contact can be used.

[0028] If any anomalies occur in the signals during the idle strokes, a reference run can either be initiated automatically (to determine precise values) or a message can be displayed on the controller indicating that a reference run should be performed. Furthermore, it is possible to initiate a machine stop to prevent subsequent damage.

[0029] This allows additional information on the machine's condition to be collected in a time-neutral manner, allowing monitoring to be virtually seamless and appropriate measures to be initiated very quickly after any unusual values ​​are detected; this can prevent the production of reject workpieces.

[0030] Furthermore, a close-meshed concept for preventive machine maintenance (“predictive maintenance”) can be installed.

[0031] This is particularly advantageous in cases of sudden wear, allowing for timely detection of, for example, the need to replace machine components. If necessary, extrapolation of the parameters can be used to predict the optimal timing for action.

[0032] In phases of machining the workpiece in which the tool does not engage with the workpiece, in particular during dressing of the tool, balancing, swiveling of a rotary table or other machine movements that are necessary for the machining cycle, data from the individual axes are recorded.

[0033] These idle strokes (i.e. machine movements where the tool does not engage the workpiece) can include, in particular, an idle stroke during finish dressing (Y-axis) or rough dressing, the infeed during dressing and grinding (X-axis), an idle stroke during grinding (Z-axis), or moving to the home position for balancing or centrifuging. Furthermore, acceleration or deceleration phases of the machine axes can be taken into consideration. Also suitable are those phases of the machining process which involve balancing, centrifuging, aligning the workpiece, swiveling a turret (a rotary table) for dressing, and changing the workpiece.

[0034] The recorded measurement signals can in particular be signals from an acceleration sensor arranged on a component of the machine, signals internal to the control system (such as currents or control differences) and other signals, if applicable, from existing sensors (for example from a current clamp and a sensor).

[0035] To analyze the signals, characteristic values ​​can be calculated (e.g., a wear index). It is also possible to use the signal curves and evaluate them using neural networks.

[0036] The processing of the signals preferably takes place on an industrial PC.

[0037] The signal curves, evaluations, and analyses can be saved locally in a database on the PC. It is also possible to store and evaluate the data in the cloud.

[0038] The proposed concept can be applied to specific modules, machines, or components, or even across machines. An approach across different machine types is also conceivable.

[0039] Regardless of the scope of the proposed concept, idle strokes will always occur for all machining-relevant axes. These can occur during process preparation, the actual grinding process, dressing, or at the end of the process.

[0040] Individual axes (especially the Y-axis) are usually particularly important for the quality of the grinding process. Monitoring this axis can, for example, be as follows: During the dressing process, the standard dressing technology includes an idle stroke during finish dressing (two or more dressing strokes - oscillation inactive) or an idle stroke during rough dressing (an odd number of dressing strokes). During the idle stroke, the axis moves at a defined speed over almost the entire travel range. Rotary axes (e.g., during the spin-drying process after dressing) can also be relevant. Test runs can then be performed on all of these axes as described. Equivalent to the dressing process, other idle strokes in other parts of the machining process are also helpful for checking various machine components.

[0041] In Fig. 1, a wear index IND is plotted over time t, which resulted in a respective reference index IR during reference runs; the respective determined reference index IR is marked with a cross. Regarding the wear index, it should be noted that this is a value that can be calculated from various values ​​or signals that are relevant or typical for wear.

[0042] The time interval between two reference runs is, for example, one week. The wear index must be below a warning limit W to be considered safe. If it exceeds an error limit F, it is likely that proper workpieces will no longer be produced.

[0043] A wear index for idle strokes (IL) is recorded at much shorter intervals, whereby this is only entered for the times between the last two reference indices (IR) as an example. The respective wear index for idle strokes (IL) is marked with a dot. In comparison to the reference indices (IR), which are only determined at relatively long intervals, a very large number of wear indices are determined for idle strokes (IL). The latter are preferably determined continuously during the machining process.

[0044] It can be seen that there is a sudden increase in wear between the penultimate and last entered reference index IR, since the penultimate reference index is still well below the warning limit W, while the next reference index is already above the error limit F. Therefore, the wear on the affected machine axis could only be detected once a number of workpieces had already been manufactured and would therefore be rejects. However, using the proposed method and the much larger number of wear indices during idle strokes, it can be determined that the wear index IND increases relatively sharply, so that maintenance measures can be initiated before the warning limit W is reached.

[0045] In Fig. Figure 2 illustrates that statistical or trend-based evaluation is also recommended when monitoring a wear index IND. It can be seen how the magnitude of the wear index IND is plotted against the number of measurements for three different machine axes, or their reactions to a predefined operating parameter (represented by dots, crosses, or squares). It can be seen that the machine axes, or their reactions to a predefined operating parameter, behave inconspicuously for the axes marked with a dot and a cross during the idle strokes IL performed, and the wear index IND is at a low level. However, it can be seen that the wear index IND of the axis marked with a square is gradually increasing, resulting in a trend TREF for a reference run.

[0046] Outliers that may occur and are not taken into account in the statistical analysis are marked with OUT.

[0047] However, for the wear index IND of the axis marked with dots, it can be seen from the 64th measurement onwards that there is an accumulation HAUF with a sharply increased value of the wear index, which indicates that there is a sudden sharp increase in wear for this machine axis.

[0048] In this case, it is therefore considered that an algorithm is used that only initiates a reference run under certain circumstances or indicates that one should be performed. Due to various influencing factors, the informative value of the signals from the test run is significantly lower than the informative value of the signals during the reference runs. Therefore, a reference run is not always performed when a limit is exceeded during a test run, but only when a certain accumulation of anomalies is observed or a corresponding trend is recognized. The underlying algorithm for determining whether a reference run is performed is preferably configurable and / or self-learning.

[0049] In Fig. Figure 3 schematically illustrates how the criteria for performing a reference run can be optimized. In the manner described, operating parameters are continuously analyzed during the idle strokes of the machine axes and then checked to determine whether the values ​​are within specified limits ("Criteria met?"). If this is not the case, a reference run is performed and the system reacts accordingly.

[0050] The data sets that can be obtained from the idle strokes, as well as the precise (high-precision) data from the reference runs, are subjected to a "machine learning" concept within the framework of an algorithm in which an optimized training data set is generated through continuous labeling, which in turn can be used to check the data obtained from the idle strokes carried out.

[0051] Thus, another aspect of the proposed approach is that a self-learning algorithm is used which optimizes the correlations between the analysis of the test drive and the results of the reference drive after each reference drive.

[0052] In Fig. Figure 4 schematically illustrates the data processing concept for a possible implementation of the proposed method. The left section of the figure shows a schematic representation of the machine in which the grinding process, warm-up, dressing, and alignment of workpieces take place. During the respective processes, accelerometers can be used to determine whether the machine is in proper working order. A corresponding display can be provided via a human-machine interface (HMI).

[0053] The machine control system controls the processes and drive data and / or vibration data are transmitted to an additional PC via digital-to-analog converter outputs (DAC), traces, sensors or other data from the process.

[0054] The PC contains software with appropriate algorithms to evaluate the received data. For this purpose, the values ​​recorded by the acceleration sensors and / or analog measurement outputs are transmitted to measurement inputs, which are made available to the software. The software is also connected to a database containing stored process data.

[0055] The PC can be connected to a cloud.

[0056] Fig. Figure 5 schematically shows the movements of three axes, plotted over time. Rectangles mark areas where, due to the process control, there is no contact between the tool and the workpiece, thus representing time periods during which test runs can be performed.

[0057] The relevant signals measured here are the current consumption of the axis drive (here: the Y-axis of the machine), the control error in the controller, and the acceleration recorded by the acceleration sensor. These parameters are typically present in the machine during measurement data recording for monitoring the dressing process.

[0058] The idle stroke can be identified via internal control signals (counter dressing stroke, dressing step, etc.) and the assigned position of the axis (via an integrated scale); the corresponding signals can be “cut out” (areas marked with rectangles in Fig. 5).

[0059] The signals are evaluated by calculating statistical parameters during constant speed, as well as by machine learning algorithms during acceleration and deceleration. A machine learning approach can also be advantageously used for evaluating constant speed.

[0060] The calculated wear value (see wear index IND in Fig. 1 and Fig. 2) is stored in a database and compared with previously defined limit values ​​and / or learned limits in order to trigger a reference run if these are exceeded.

[0061] The limit values ​​for this (with automatically learned limits) are determined from previous idle strokes and reference runs.

[0062] It should be noted that, as already mentioned, the signal quality of the idle strokes during a test run is generally poorer than that during a reference run. Therefore, algorithms are preferably used to assess whether a reference run is necessary, particularly through statistical approaches or by using machine learning concepts (particularly through similarity analysis, which can lead to changes in the values).

[0063] In this respect, a reduction to purely static values ​​or limits is not necessary.

[0064] If a reference run is triggered, as already mentioned, the highly accurate results can be correlated with the results of the idle strokes and thus the dynamic limits can be optimized.

[0065] In Fig. Figure 6 shows a flow chart that illustrates this procedure.

Claims

[1] Method for hard finishing a workpiece with a toothing or a profile on a hard finishing machine, wherein the hard finishing machine has a number of machine axes which are NC-controlled for machining the workpiece, wherein in the method the workpiece is received on a workpiece spindle and is machined by means of a tool which is received on a tool spindle by engaging the tool with the workpiece, wherein the machining phase of the workpiece begins with the introduction of the workpiece into the hard finishing machine and ends with the removal of the machined workpiece from the hard finishing machine, characterized by , that at least during a period of the machining phase in which the tool is not in engagement with the workpiece, a test run is carried out on at least one machine axis, wherein the machine axis is driven during the test run, wherein a value is specified for at least one drive parameter for which a target value of the movement of the machine axis is expected, wherein the actual value of the movement of the machine axis is recorded, wherein a monitoring device compares the target value and the actual value, and whereby a signal is issued by the monitoring device when a) the comparison between the target value and the actual value has shown that the difference between the target value and the actual value is outside a permissible range or b) the comparison between the target value and the actual value has shown that the difference between the target value and the actual value has a value which, compared to previously recorded values ​​of the difference, indicates that a permissible tolerance band has been exceeded. [2] Method according to claim 1, characterized bythat at least one time section of the machining phase lies within the time interval in which the workpiece is mounted on the workpiece spindle. [3] Method according to claim 1 or 2, characterized by that the test run takes place during an idle stroke in which the tool is moved relative to the workpiece without the tool engaging with the workpiece. [4] Method according to claim 1 or 2, characterized by that the test run takes place during an acceleration or rotational acceleration process and / or during a braking process of a machine axis. [5] Method according to one of claims 1 to 4, characterized bythat the test run takes place during a dressing phase in which the tool is profiled, and / or during a phase in which the workpiece is aligned for machining in the hard finishing machine, and / or during a phase in which a balancing operation is carried out, and / or in a phase in which a loading or unloading operation is carried out. [6] Method according to one of claims 1 to 5, characterized by that the signal is only output once a statistical evaluation of a number of recorded values ​​has been carried out. [7] Method according to one of claims 1 to 6, characterized by that the assessment of the recorded values ​​is carried out using an algorithm that takes into account empirical values ​​within the framework of machine learning. [8] Method according to one of claims 1 to 7, characterized bythat the recorded value is the acceleration recorded by an accelerometer, and / or the current with which a machine axis is driven, and / or a control difference that occurs when controlling the machine axis, and / or a signal that provides an indication of a deviation between the target value and the actual value. [9] Method according to one of claims 1 to 8, characterized by that the output of the signal from the monitoring device, if the assessment has shown that the value is outside a permissible range, causes the hard finishing machine to carry out a reference run during which all machine axes are actuated without the tool engaging the workpiece. [10] Method according to one of claims 1 to 9, characterized by that the process is a gear grinding process.

Images