METHOD AND SYSTEM FOR SYNCHRONIZING SIGNALS

Description

[0001] The invention relates to a method and a system for synchronizing signals related to a technical system, in particular to a machine and / or a machining process. The method comprises the following steps: a) recording data from a first data source to obtain a first signal track; b) recording data from at least a second data source independent of the first data source to obtain at least a second signal track.

[0002] High-precision, temporal mapping of machine signals from internal and external sensors is currently not possible. Common machine control systems, as well as the machines themselves, lack an interface to enable higher-level signal processing with external signals.

[0003] From EP 2 434 360 A1 a method for motion control is known, wherein a first motion controller is connected to a second motion controller via a data bus, wherein first trace data of the first motion controller have a time stamp dependent on a global time and wherein second trace data of the second motion controller have a time stamp dependent on the global time, wherein the different trace data are linked via the time stamp.

[0004] EP 3 051 374 A1 discloses an analysis method for comparative analysis taking into account a time offset.

[0005] EP 3 521 792 A1 discloses a method for temporal synchronization of two data time series.

[0006] The state of the art therefore provides for data to be recorded in a time-synchronized manner so that they can be assigned to one another later.

[0007] The object of the present invention is to provide a method and a system with which signal traces of different data sources, which are independent of each other and not synchronized in time, can be brought into temporal alignment.

[0008] This object is achieved according to the invention by a method according to claim 1 and a system according to claim 10. A method for synchronizing signals related to a technical system, in particular a machine and / or a machining process, is disclosed, comprising the method steps: a) recording data from a first data source to obtain a first signal trace, b) recording data from at least a second data source independent of the first data source to obtain at least a second signal trace, c) analyzing the signal traces using previously known domain knowledge, d) temporally linking the signal traces.

[0009] Data sources within the meaning of the invention can be, for example, measurement sources, sensors, controllers, etc. The recorded data can be measured data. Furthermore, the data can be input variables or output variables from controllers. Furthermore, drives of a machine can represent data sources. The data can accordingly be data from a drive. The data is recorded in a temporal relationship. When the data is transmitted, it is transmitted as signals. The temporal progression of a signal is referred to as a signal trace.

[0010] Domain knowledge globally describes the relationship between vibration excitation by machine components, axis dynamics, the absolute position of the kinematic chain, possibly depending on the workspace, actuators, e.g., valves, the operating state of a machining unit, and noise emissions (sound waves). Examples of machining units that can be considered are lasers, punching devices, presses, milling heads, saws, drills, and water jets. In machine tools, the machining units are moved in a specific axial direction via drives and possibly interposed mechanical components, such as gears or gantries. This is often abbreviated to "axis." All components, especially axes, that contribute to the movement of a machining unit are called the kinematic chain.Furthermore, domain knowledge includes the relationship between individual components, in particular the infrastructure, movement trajectories, processing processes and properties of all components involved.

[0011] By linking the signal traces chronologically, they can be synchronized. In particular, the signal traces can be assigned to a common time axis. This allows for early detection of faulty machine states, especially slowly progressing defects. Furthermore, noises not originating from the machine or the axes can be suppressed. This improves the signal-to-noise ratio.

[0012] Even with two different, but sensibly selected signal traces, temporal synchronization can be clearly determined from the aforementioned domain knowledge. The more different signal traces available, the more reliable the analysis. With the method according to the invention, real-time synchronization is conceivable, enabling real-time evaluations based on different data sources. This opens up new possibilities for error detection, error diagnosis, condition monitoring, and predictive maintenance of entire systems. In particular, real-time error diagnosis of entire systems with differently running clocks is possible. Sources of interference can be suppressed using known and expected signal patterns. Processing quality can be improved. False alarms and incorrect error interpretations can be reduced.The method according to the invention can be implemented with minimal effort and cost-effectively, as no additional effort is required for time synchronization. Furthermore, the method according to the invention is scalable, as it can be used for two or more different data sources. Cross-system use through cascading is also possible. This allows for the diagnosis of entire production plants or factory halls.

[0013] The analysis of signal traces can be model-based. In particular, domain knowledge can be used to automatically assign data values to time from signal traces from different data sources. Deviations, such as those in sound pressure, atomic numbers, or mechanical resonances, lead to rapid error detection, precise error identification, and efficient troubleshooting.

[0014] The analysis of signal traces can be performed, in particular, using pattern recognition based on reference patterns. The reference patterns are known from domain knowledge. Using pattern recognition and pattern comparison, it is possible to temporally overlap signal traces from different data sources, especially measurement sources. For example, the kinematic chain generates a known excitation pattern corresponding to the motion profile of the actuators / axes. This excitation pattern is translated into various data recordings, especially measurement recordings. In addition, mechanical resonance points of a machine can be excited, which also manifest themselves in known vibration phenomena.

[0015] Particular advantages arise when the signal traces are recorded without temporal synchronization. Thus, it is not necessary to time-stamp the signal traces, as is the case with the state of the art.

[0016] At least one signal track from an internal machine data source and at least one signal track from an external machine data source can be used. An internal machine data source can be, for example, an internal machine controller or drive. An external machine data source can be, for example, a camera or microphone used to observe the process being performed on the machine. Using both internal and external machine data sources can improve and facilitate system diagnostics, and in particular, fault detection.

[0017] Preferably, the time periods in which the data from the data sources are recorded overlap. This makes it possible to align the recorded signal traces in time and, in particular, to synchronize them after analysis.

[0018] After synchronizing the signal traces, a time-frequency transformation, such as a Fourier transform, can be performed. This can facilitate analysis and troubleshooting.

[0019] To improve the analysis result, each signal trace is intended to contain at least a predetermined number of data points. This number can depend on the frequency at which the data points are acquired. This can vary over many orders of magnitude. An NC controller controls in the millisecond range, the interpolation of the NC controller even faster. This is the frequency at which, for example, drives are addressed, e.g. the motor current is adjusted during control to achieve a target speed. This would be around 1 kHz. Optical or acoustic sensors can measure in a wide frequency band. A camera, for example, in the order of 10 or 100 Hz, and possibly even higher for special applications. Photodiodes measure in the MHz or even GHz range. Acoustic sensors, for example, have a resolution in the audible range, i.e. in the kHz range, but there are also sensors in the MHz or GHz range.Two traces can thus be linked if a characteristic signal can be resolved by the data sources involved and a corresponding number of measuring points (depending on the measuring frequency or recording frequency) has been recorded.

[0020] The technical system, in particular the first or second data source, and thus the recorded data, are deliberately manipulated, thereby deliberately exciting mechanical resonance points. Input variables, for example, from control systems, can be deliberately manipulated. If input variables are deliberately manipulated, a specific result or behavior can be expected from the recorded data. It can then be analyzed to determine whether the recorded signal exhibits the expected behavior. Based on this analysis, possible error sources can be identified.

[0021] Each data recording can be individually time-normalized and meet the Nyquist criterion. This can improve the reliability of the analysis.

[0022] Based on the analyzed signal traces, fault detection, fault diagnosis, condition monitoring and / or predictive maintenance are carried out.

[0023] Applications for the method according to the invention include, for example, machine diagnostics, such as axis diagnostics, process diagnostics, and diagnostics of other external causes. The invention allows for aggregated and correlated evaluation of data sources for the early detection of impending errors.

[0024] Also disclosed is a system for synchronizing signals, comprising a first data source providing a first signal trace and a second data source providing a second signal trace, an analysis device to which the signal traces are fed and which is connected to or has a storage device in which domain knowledge is stored, wherein the analysis device is configured to temporally link the signal traces using the stored domain knowledge. With such a system, signal traces originating from different data sources and not having a time stamp can be synchronized. This makes it possible to analyze the system and, if necessary, detect errors. Preferably, at least one data source is arranged internally on the machine and at least one data source is arranged externally on the machine.

[0025] Further features and advantages of the invention will become apparent from the following description of exemplary embodiments of the invention, based on the figures of the drawing, which illustrate details essential to the invention, and from the claims. The features shown therein are not necessarily to scale and are presented in such a way that the special features of the invention can be clearly seen. The various features can be implemented individually or in combinations in variants of the invention.

[0026] The schematic drawing shows embodiments of the invention and explains them in more detail in the following description.

[0027] They show: Fig. 1 shows a schematic representation of a system; Figs. 2a to 2c show diagrams for explaining signal synchronization; Figs. 3a to 3c show diagrams for explaining error identification; Fig. 4 shows a further diagram for explaining error identification; Fig. 5 shows a flowchart for explaining the method according to the invention.

[0028] The Fig. 1 shows a system 1 for signal synchronization. A machining process is being performed on a machine 2. The machine 2 has a first data source 3. The first data source 3 can be, for example, a control system of the machine 2. In particular, it can be a machine-internal data source. Data from the data source 3 is recorded and transmitted as a signal trace to an analysis device 4. The analysis device 4 can be located internally or externally to the machine.

[0029] In the illustrated embodiment, a further, second data source 5 is located external to the machine. For example, the second data source 5 can be a microphone or a camera. The data from data source 5 is also recorded and transmitted as a signal trace to the analysis device 4. The data from data sources 3 and 5 are recorded independently of one another. In particular, it occurs without a synchronized time stamp between the data sources 3 and 5 and the analysis device 4 and each other.

[0030] So-called domain knowledge is stored in memory 6. This can include previously recorded measurement data, simulation results, historical data from machine 2 itself, data from other machines, etc. The analysis device 4 can access the domain knowledge. Based on the domain knowledge, the signal traces from data sources 3, 5 are analyzed and temporally correlated. The result can be displayed on a display device 7.

[0031] The Figur 2a shows the signal trace 8, which corresponds to the recorded data of data source 3. The Figur 2b shows the signal trace 9, which corresponds to the recorded data of the data source 5. In the domain knowledge, it is known that the signal trace 8 is seen in response to a certain excitation signal. Furthermore, it is known in the domain knowledge that the signal trace 9 is to be expected in response to the same excitation signal. Furthermore, it is known in which temporal relationship the signal traces 8 and 9 stand to the excitation signal. Based on this knowledge, the signal traces 8 and 9 can be related in time, which is shown in the Figur 2c Signal tracks 8 and 9 are shown here as being assigned to a common time axis.

[0032] Based on the Fig. 3a bis 3c The method according to the invention will be explained. An external data source, in particular a microphone, recorded a noise from a machine. Fig. 3a The spectral analysis of the noise is shown, with the amplitude plotted against the frequency. The spectral analysis was created after the noise signal was first placed in temporal relation to other signal traces and a Fourier transformation was performed. At a frequency of 566.4 Hz, the first harmonic 10 is displayed. At a frequency of 1132.9 Hz, the second harmonic 11 is displayed.

[0033] In the Fig. 3b is the reference frequency response of a speed control loop of the Z-axis of a machine, with the amplitude plotted against the frequency. Curve 12 represents the reference frequency response of a first commissioning. Curve 13 represents the reference frequency response of a second commissioning, for example at a customer's. Curves 12 and 13 have a similar shape and show no abnormalities. Curve 14 corresponds to a reference frequency response that was recorded as a second signal trace when the noise appeared. Curve 14 was determined by first synchronizing the second signal trace with the noise and then Fourier transforming it. A peak 15 can be seen at a frequency of 566.4 Hz. This means that an abnormality was detected in the Z-axis at a frequency that corresponds to the first harmonic 10 of the noise.The second harmonic 11 does not correlate with the frequency response of the Z-axis and therefore has a different cause.

[0034] In the Fig. 3c The torque-generating current (the current responsible for generating the drive's torque) is plotted against frequency. The torque-generating current of the Z-axis also exhibits a peak of 16 at a frequency of 566.4 Hz. This confirms a fault in the Z-axis.

[0035] By first relating the signal traces to each other in time, it was possible to reconcile the spectra determined from the signal traces. No peak at the frequency of 566.4 Hz was detected in the moment-generating currents of the other axes. This ruled out the possibility that the error causing the noise was caused by one of the other axes. Domain knowledge allows us to identify the origin of peaks at specific frequencies. Thus, based on the recorded signal traces, it is possible to determine where an error exists and correct it specifically.

[0036] In the diagram of the Fig. 4 The spectrum of the torque-generating current is shown along an x-axis. In the plane, the velocity v is plotted against the frequency f. The vertical axis shows the amplitude of the torque-generating current. Along line 20, the excitation (harmonics) from the pinion-rack meshing is shown. Along line 21, the excitation from a motor is shown. Line 22 shows superpositions of sound levels. Sharp peaks can be seen at a constant frequency of approximately 550 Hz. They indicate mechanical resonance points. By superimposing sound levels, a statement about the severity and extent of the vibrations is possible.

[0037] Here, it can be seen that the signal traces resulting from excitations by the motor and from excitations due to pinion-to-rack meshing, as well as signal traces recorded via a microphone, were temporally correlated to obtain information about the machine's behavior. It can be seen that resonances occur at 550 Hz for different X-axis speeds. This suggests that the resonances are due to a structural element of the machine and not to the drive (motor).

[0038] In the flowchart of the Fig. 5100 denotes the step of recording data from a first data source to obtain a first signal trace. In step 101, data from a further data source is recorded, wherein the further data source is independent of the first data source. This produces a second signal trace. In step 102, the signal traces are analyzed using known domain knowledge. In step 103, the signal traces are temporally related to one another. A further step may follow in which the temporally synchronized signal traces are transformed into the frequency domain and the result is (automatically) analyzed. This analysis can also be carried out with the aid of or supported by domain knowledge.

Claims

1. A computer-implemented method for the synchronization of signal tracks (8, 9) which are related to a technical system, in particular a machine and / or a processing process, comprising the procedural steps: a. recording data from a first data source to obtain a first signal track (8), b. recording data from at least a second data source that is independent of the first data source and not synchronized in time, in order to obtain at least a second signal track (9), c. analyzing the signal tracks (8, 9) based on prior domain knowledge, d. bringing into temporal agreement and synchronizing the signal tracks (8, 9) based on domain knowledge, wherein error detection, error diagnosis, status monitoring and / or predictive maintenance are carried out on the basis of the analyzed and synchronized signal tracks (8, 9), characterized in that the technical system, in particular the first or second data source, and thus the recorded data, are specifically manipulated so that mechanical resonance points are specifically excited.

2. The method according to claim 1, characterized in that the signal tracks (8, 9) are assigned to a common time axis.

3. The method according to claim 1 or 2, characterized in that the analysis of the signal tracks (8, 9) is model-based.

4. The method according to one of the preceding claims, characterized in that the analysis of the signal tracks (8, 9) is carried out by means of pattern recognition and on the basis of reference patterns.

5. The method according to one of the preceding claims, characterized in that the recording of the signal tracks (8, 9) is carried out without synchronizing the recordings in time.

6. The method according to one of the preceding claims, characterized in that at least one signal track (8) of a data source (3) inside the machine and at least one signal track (9) of a data source (5) outside the machine are used.

7. The method according to one of the preceding claims, characterized in that the time periods during which the data from the data sources (3, 3) are recorded overlap.

8. The method according to one of the preceding claims, characterized in that each signal track (8, 9) comprises at least a predetermined number of data points.

9. The method according to one of the preceding claims, characterized in that each data recording is time-normalized and complies with the Nyquist criterion.

10. A system (1) for synchronizing signal tracks (8, 9), comprising a first data source (3) which supplies a first signal track (8) and a second data source (5) which is independent of the first data source and is not synchronized in time and which supplies a second signal track (9), an analysis device (4) to which the signal tracks (8, 9) are fed and which is connected to or has a memory device (6) in which domain knowledge is stored, wherein the analysis device (4) is configured to bring the signal tracks (8, 9) into temporal agreement and to synchronize them on the basis of the stored domain knowledge, and, on the basis of the analyzed and synchronized signal tracks (8, 9), to perform error detection, error diagnosis, status monitoring and / or predictive maintenance, characterized in that the analysis device (4) is designed and configured to specifically manipulate the technical system (1), in particular the first or second data source (3, 5), and thus the recorded data, so that mechanical resonance points can be specifically excited.

11. The system according to claim 10, characterized in that at least one data source (3) is arranged inside the machine and at least one data source (5) is arranged outside the machine.