Method and system for determining the dynamic characteristics of a machine

Abstract

The present invention relates to a method for determining the dynamic characteristics of a machine (11) having a plurality of axes (12, 13), the method comprising the following method steps: a. performing a measurement stroke for each axis (12, 13) over its entire working range; b. detecting and recording data associated with the measurement stroke; c. determining a time-frequency representation of the recorded data or system characteristics by means of a data processing unit (20); d. analyzing the time-frequency representation or a related representation by means of an image processing algorithm.

Description

Field of the Invention

[0001] The present invention relates to a method for determining the dynamic characteristics of a machine having a plurality of axes. Furthermore, the present invention relates to a corresponding system. Background Art

[0002] So far, measurements for evaluating the dynamic properties of a machine (especially a machine tool), especially complex frequency measurements, have been carried out individually for each axis according to position or posture. It is only very limitedly possible to achieve an explanation of the dependence of the dynamic properties of an axis on the position of the individual axis and the positions of other axes and the identification of complex causal relationships.

[0003] In particular, in the prior art, measurements of dynamic properties are only carried out at a few selected positions in the workspace. This greatly limits the validity of the explanation regarding position dependence. If the dependence on the positions of other axes is to be checked, this can only be achieved by pre-configuring several starting positions. Even so, only discrete explanations of position are allowed. Summary of the Invention

[0004] The object of the present invention is to determine the dynamic characteristics of a machine having a plurality of axes within the entire workspace.

[0005] According to the present invention, this object is achieved by a method for determining the dynamic characteristics of a machine having at least one axis, the method comprising the following method steps:

[0006] a. Performing a measurement stroke for each axis within its entire working area;

[0007] b. Detecting and recording data associated with the measurement stroke;

[0008] c. Using a data processing unit to determine the time-frequency representation of the recorded data;

[0009] d. Analyzing the time-frequency representation or a representation related to the time-frequency representation, especially in an image area, by means of an image processing algorithm.

[0010] In a machine tool, a processing unit (such as a tool holder for holding a tool or a laser processing head) moves along a specific axial direction through a drive device and possibly mechanical components (such as a transmission or a gantry) connected therebetween, and the processing unit can be mounted on these mechanical components. This is usually simply referred to as an axis. Therefore, the axial position refers to the position of the processing unit in the axial direction, which is caused by the relevant axis. The working area of an axis is the area in which the processing unit can move through the corresponding axis. The workspace is generated by the superposition of the working areas of the axes of the machine tool. Therefore, the workspace is the spatial area that the processing unit can reach. The posture or spatial attitude is a combination of the position and orientation of an object (such as the processing unit).

[0011] With the method according to the invention, the measuring travel is preferably carried out in such a way that the entire working space is covered. The measuring travel for each axis is carried out within the entire working area of the axis to be inspected. In particular, multiple measuring travels can be carried out for each axis.

[0012] During the measuring travel, data from different data sources can be detected and recorded. For example, sensors can be provided on the machine as data sources for carrying out the measurement. The controllers and drive units of the machine can also be data sources, and the input and / or output signals of these controllers and drive units can be detected as data.

[0013] The data detected and recorded during the measuring travel can be recorded in a time-synchronized manner or synchronized later. The recording duration may depend on the axis length of the axis to be observed. Preferably, a short sampling time (sampling theorem) suitable from the perspective of the system characteristics to be observed is selected for data detection. The method according to the invention provides a comprehensive description of the position dependence of the dynamic properties of the machine with time-(space) frequency resolution, which is only limited by the feed rate used for measurement and the sampling frequency of data detection. In particular, the description of the dynamic properties of the axis can be based on the position of a single axis and the positions of other axes. Thus, a global description of the complete position dependence can be achieved. Causal relationships can be identified in an automated manner.

[0014] The measuring travel of the axis can be carried out in segments. In particular, different segments can be detected at different times. However, the aim is to detect the entire axis (i.e., the entire working area of the axis).

[0015] The measuring travel can be carried out during auxiliary time or during operation of the machine. Here, the auxiliary time is the time when the machine does not perform workpiece machining. However, it is also conceivable to detect at least some segments of the working area of the axis in parallel with production, for example, during positioning between contours, i.e., when the machining unit moves from one machining point to another machining point. However, data recording can also be implemented as an on-line measurement during equipment operation.

[0016] Alternatively, the measuring travel can be carried out in a separate measurement program. For example, the measurement program can be carried out before the machine is actually started or during the auxiliary time of the machine.

[0017] System characteristics can be determined based on the time-frequency representation of the recorded data (in particular by a data processing unit). By means of time-indexed measurements, the spectra of the excitation signal and the output signal can be determined. The ratio of these spectra gives the representation of the dynamic characteristics of the system in the frequency domain.

[0018] In particular, the analysis of the time-frequency representation can be carried out in an automated manner.

[0019] The measuring stroke can be carried out with a pre-given input signal (in particular the input signal of a drive device), which includes a pre-given signal superimposed with an excitation signal. For example, the pre-given pre-given signal can be a constant speed superimposed with an excitation signal. For example, the excitation signal can be a technical approximation of white noise. Depending on the measuring stroke or type of measurement to be carried out, a specific suitable speed curve can be pre-given. For measurement types used to evaluate dynamic properties (diagnosis of mechanical properties, or the interaction between control settings and mechanical properties) (such as the reference frequency characteristics of a position control loop, the reference frequency characteristics of a rotational speed / speed control loop, the noise frequency characteristics of a rotational speed / speed control loop, the rotational speed / speed controller section, the mechanical frequency characteristics), a constant speed superimposed with a suitable excitation signal can be selected as the input signal.

[0020] Components involved in the movement and present in the drive train of each axis (such as drive devices, racks, measuring systems) can be diagnosed by zeroing the superimposed excitation signal. Multiple single measuring strokes can be selected with different feed rates (i.e., speeds) or other suitable speed curves (for example, the speed linearly increases with the axial position).

[0021] A multi-stage approach can be envisaged to check the dependence on the working space position. To take into account the dependence of dynamic properties on the working space position of axes other than the axis to be checked, the said axis can be pre-positioned at a specific position and the measurement / measuring stroke can be carried out for each posture. Thus, axes other than the axis to be checked can be pre-positioned at different positions, and the measuring stroke for the axis to be checked can be carried out for each position of the other axes.

[0022] The measuring stroke can be carried out on at least one axis during the movement of another axis. In particular, the simultaneous movement of multiple axes can be combined with the data recording of the axis to be checked. In particular, when multiple axes move simultaneously, their mutual influence can be checked by appropriately superimposing the excitation signal and recording the input signal and output signal, and thus a depiction of all moving (mechanical) components of the entire system (machine (machine tool)) can also be generated. Various forms of this depiction can be envisaged. The characterization can be carried out as frequency characteristics in the frequency domain. In addition, system identification methods in the time domain can be used, such as NARMAX (nonlinear autoregressive moving average method with exogenous input), subspace system identification, etc. In particular, a coupling matrix can be generated for linear time-invariant systems.

[0023] According to the present invention, when measuring a single axis, the specified input signal and output signal are recorded as data within the entire operating range of the axis. The measured output signal can be, for example, current (especially the current drawn by the drive of the axis), feed rate, rotational speed (the rotational speed of the drive, or if there is a gear, the output rotational speed), acceleration, etc. To check the coupling of multiple axes, the number of data items to be recorded increases according to the transfer characteristics to be checked.

[0024] The data can be detected, for example, by a data detection unit that is capable of recording data in a time-synchronized manner throughout the entire measurement (especially the measurement stroke) period. The data detection unit can be part of the (machine's) controller and can interact with a storage unit. Alternatively, the data detection unit can be an external component, such as an IPC (industrial PC), or a data detection unit with real-time capabilities, having at least one physical communication interface and especially having a data storage function, which is connected to the data source so that data can be recorded and appropriately stored.

[0025] The time-frequency representation can be determined, for example, by a transformation, especially a Fourier transform, square transform, or Wigner-Ville distribution. For example, a short-time Fourier transform can be performed. The time reference of at least one axis can be converted to a position reference. The representation can be analyzed or evaluated in the image area by an image processing algorithm. This can be performed, for example, based on a specified metric, such as based on the statistical value of a single feature, such as a threshold with / without a position reference. In addition, the analysis can be performed based on the representation itself as a highly correlated tensor to be further processed in a data processing unit. For example, the data of the Bode plot can be provided in two matrices, one matrix involving amplitude information and the other matrix involving phase information, and these matrices are respectively converted to normalized gray values based on position. Thus, the data can be checked using an image processing algorithm, and a neural network can be trained to replace the previous series correlation checking method that only had local validity with a comprehensive method. This leads to a significant simplification of the diagnostic process.

[0026] When analyzing the data, other parameters can be considered, such as the life and wear of components and service interventions. This analysis can be performed for a population or for the time characteristics of the same or similar machines for each feature, which also allows for the automatic identification of relevant features. A population can be a group of systems (=machines (machine tools)) having nominally the same characteristics (=motor, transmission, machine body, etc.). The "features" extracted from the measurement can be, for example, natural frequency, attenuation, hysteresis, etc. These features have a statistical distribution. The characteristics of these distributions (such as mean value, standard deviation) may be relevant in themselves for the characterization of the population.

[0027] For example, if the standard deviation of the natural frequencies of the machines from production line A is greater than the standard deviation of the natural frequencies of the machines from production line B, this indicates a difference in the (nominally identical) production processes.

[0028] The image processing algorithm can include machine learning. In particular, the algorithm can establish a statistical model based on training data (i.e., the detected data). In particular, patterns and principles can be identified in the detected data. Thus, unknown data can also be evaluated after the learning process is completed. A possible application learning variant using artificial neural networks is so-called deep learning, which can be applied according to the present invention. This can include, for example, convolutional neural networks, deep autoencoders, generative adversarial networks.

[0029] In particular, it can be provided that the neural network is trained by feeding the analysis of the representation in the image region into the artificial neural network.

[0030] Metadata can be provided to the neural network. In particular, information from other data sources (such as data markings based on service intervention history (e.g., information about actions performed or parts replaced)) can be fed into the neural network. These metadata can be considered during the analysis in order to automatically classify the measurement data. Thus, complex causal relationships can be automatically identified.

[0031] At least some of these method steps can be performed on a spatially distributed system. For example, the image processing algorithm can be performed on a machine, on a data detection unit with real-time capabilities, having at least one physical communication interface and in particular having a data storage function, on a computer or in a cloud architecture.

[0032] Furthermore, within the framework of the present invention, there is also included a system for determining the dynamic characteristics of a machine having a plurality of axes, the system having a machine with a plurality of axes (in particular a machine tool), at least one data source connected to a data detection unit, a data storage unit for storing the detected data, and a data processing unit configured to determine the time-frequency representation of the recorded data, and an image processing device for processing the time-frequency representation. The data source can be an internal machine data source or an external machine data source. For example, an internal machine data source can provide current, acceleration, speed, actual position, and / or setpoint. In addition, the internal machine data source can come from a controller, such as a programmable logic controller (PLC) or a numerical controller (NC). In addition, the drive device of the axis can be a data source. Furthermore, the data source can provide laser power, gas pressure, scattered light, etc. as data.

[0033] External data sources of the machine can be, for example, microphones, microelectromechanical system (MEMS) sensors, or cameras. The data detection unit can be part of the controller. Alternatively, the data detection unit can be an external component for data detection, such as a data detection unit with real-time capabilities, having at least one physical communication interface and in particular having a data storage function, or an IPC.

[0034] The image processing device can be in the form of a neural network or can include a neural network. Such an image processing device allows the use of self-learning image processing algorithms, in particular machine learning and so-called deep learning. In addition, data can be classified based on additional information, such as based on service intervention reports, spare part replacements, etc. Thus, a comprehensive description of the position dependence of the dynamic properties of a machine tool with a time-(space)-frequency resolution that is only limited by the feed rate used for measurement and the sampling frequency of data detection can be made. Local differences in the system under inspection can be detected. In addition, a global description of the position dependence can be made in a shorter time than before. Complex causal relationships can be identified in an automatic manner.

[0035] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention, reference to the drawings showing important details of the present invention, and from the claims. The features shown here should be understood as not necessarily being drawn to scale and being presented in a manner such that the characteristic features according to the present invention can be clearly visible. In variants of the present invention, the various features can be implemented individually in each case or in any desired combination as multiple implementations. Description of the Drawings

[0036] Exemplary embodiments of the present invention are shown in schematic diagrams and are explained in the following description.

[0037] In the drawings:

[0038] Figure 1 A system for determining the dynamic properties of a machine according to the present invention is shown;

[0039] Figure 2 A position-frequency representation of data is shown;

[0040] Figure 3 A flowchart of a method according to the present invention is shown. Detailed Description

[0041] Figure 1A system 10 for determining the dynamic characteristics of a machine 11 having a plurality of axes 12, 13 is shown. The machine 11 also includes data sources 14, 15, 16, and the data sources 14, 15 are associated with the axes 12, 13. In particular, this can be an internal sensor system. In addition, the data sources 14, 15 can be controllers. For example, the data that can be output is current or rotational speed. The machine 11 can include additional data sources 16 in the form of sensors, for example. For example, this enables the acceleration of the processing unit to be measured. An external data source 17 can be additionally provided. For example, the data source 17 can be in the form of a microphone or a camera. Using additional data sources (such as acceleration sensors, rotational speed sensors, microphones, cameras, etc.) enables the characteristics of the system properties to be determined.

[0042] The data detected from the data sources 14, 15, 16, 17 is fed to a data detection unit 18. The data from the data sources 14, 15, 16, 17 can be recorded in a time-synchronized manner; in particular, the data sources 14 - 17 can be synchronized. Alternatively, the data can be synchronized subsequently, for example, in the data detection unit 18. In particular, the data detection unit 18 can be configured to record data in a time-synchronized manner throughout the measurement period. The data detection unit 18 can be part of a controller. In the exemplary embodiment shown, it can be in the form of an external data detection unit with real-time capabilities, which has at least one physical communication interface and in particular has a data storage function.

[0043] During data detection, the axes 12, 13 or the machine 11 move within its entire working area. A possible input signal is a constant speed of a single axis of the system (a pre-given constant speed of moving the processing unit by the axis), which is superimposed on an excitation signal. For example, the output signal (i.e., the data provided by the data sources 14 - 17) can correspond to current, feed rate, rotational speed, or acceleration.

[0044] To check the coupling of the plurality of axes 12, 13, the number of data items to be recorded is increased. The recorded data is stored in a data storage unit 19. After that, data processing or data preprocessing is performed in a data processing unit 20. In particular, the time-frequency representation of the recorded data is determined here. Alternatively or additionally, the system characteristics of the machine 11 can be determined as, for example, a transfer function under the assumption of a linear time-invariant system, or determined as a representation using the NARMAX method. Optionally, the time-frequency representation can subsequently be converted into a position-frequency representation.

[0045] Subsequently, the data is analyzed in the image processing device 21. In particular, the representation generated by the time-frequency transformation can be evaluated in the image area based on a specified metric. For example, statistical values can be determined for individual features, such as thresholds with and without position references. The representation itself can be further processed as a highly correlated tensor. For example, one method is to represent the information of the Bode plot in two matrices, which are used to convert the amplitude information in one matrix and the phase information in the other matrix into normalized gray values based on position. Therefore, modern image processing algorithms, especially machine learning, can be used to examine the data and train a suitable neural network to replace the previous series of correlation inspection methods that only had local validity with a comprehensive method. This leads to a significant simplification of the diagnostic process, including determining the attributes related to the inspection.

[0046] The so-called metadata can be fed to the image processing device 21 by another memory 22. Therefore, additional parameters can be considered in the data analysis. Additional parameters are, for example, the life, wear, service intervention, etc. of the machine.

[0047] Figure 2 A position-frequency curve graph is shown. This is determined by measuring the frequency characteristics of the x-axis of the machine while moving the y-axis. The color or gray level shows the amplitude of the frequency characteristics. This representation can be used as a basis for determining the frequency characteristics of the x-axis at any time and any point. Therefore, not only can point-by-point data be obtained as before, but the frequency characteristics can generally be determined over the entire working area of the x-axis.

[0048] Figure 3 A flowchart of the method according to the present invention is shown. In step 100, at least one measurement run is performed for each axis over its entire working area.

[0049] In step 101, the data associated with the measurement run is detected and recorded. In step 102, a data processing unit is used to determine the time-frequency representation of the recorded data. In step 103, the time-frequency representation or a related representation, such as a position-frequency representation, is analyzed by an image processing algorithm.

Claims

1. A method for determining the dynamic characteristics of a machine (11) having at least one axis (12, 13), the method comprising the following method steps: a. Performing a measurement stroke for each axis (12, 13) within its entire working area; b. Detecting and recording data associated with the measurement stroke; c. Determining the time-frequency representation of the recorded data by means of a data processing unit (20); d. Evaluating the time-frequency representation or the associated position-frequency representation based on a specified metric by means of an image processing algorithm, Among them, In step a, during data detection, each axis moves within its entire working area.

2. The method according to claim 1, characterized in that, The measurement stroke for the axis (12, 13) is performed in segments; and / or In step b, data from different data sources are detected and recorded during the measurement stroke.

3. The method according to claim 1 or 2, characterized in that, The measurement stroke is performed during auxiliary time or while the machine (11) is running.

4. The method according to claim 1 or 2, characterized in that, The measurement stroke is performed with a pre-given input signal, where the input signal includes a pre-given signal superimposed on an excitation signal.

5. The method according to claim 1 or 2, characterized in that Multiple measurement strokes are performed for at least one axis (12, 13) with different input signals.

6. The method according to claim 1 or 2, characterized in that, A measurement stroke is performed for at least one axis (12, 13) during the movement of another axis.

7. The method according to claim 1 or 2, characterized in that, Another axis (12, 13) is pre-positioned at different positions, and a measurement stroke is performed for each position of the other axis (12, 13).

8. The method according to claim 1 or 2, characterized in that, The time-frequency representation is determined by means of a transformation; The transformation is a Fourier transform, a square transform or a Wigner-Ville distribution.

9. The method according to claim 1 or 2, characterized in that, For at least one axis (12, 13), the time reference is converted to a position reference.

10. The method according to claim 1 or 2, characterized in that, The time-frequency representation or the position-frequency representation is evaluated in an image area by means of an image processing algorithm.

11. The method according to claim 1 or 2, characterized in that, The neural network is trained by feeding the analysis of the time-frequency representation or the position-frequency representation in the image area into the artificial neural network.

12. The method according to claim 11, characterized in that, Metadata is provided to the neural network.

13. The method according to claim 1 or 2, characterized in that, At least some of the method steps are performed on a spatially distributed system.

14. A system (10) for determining the dynamic characteristics of a machine (11) having a plurality of axes (12, 13), the system having a machine (11) including the plurality of axes (12, 13), at least one data source (14 - 17) connected to a data detection unit (18), a data storage unit (19) for storing the detected data, and a data processing unit (20) configured to determine a time-frequency representation of the recorded data, and an image processing device (21) for processing the time-frequency representation, wherein, The system (10) is arranged to perform the method according to one of claims 1 to 13.

15. The system according to claim 14, wherein The image processing device (21) is configured as a neural network or includes a neural network.

Images