(19)
(11)EP 3 929 682 A1

(12)EUROPEAN PATENT APPLICATION

(43)Date of publication:
29.12.2021 Bulletin 2021/52

(21)Application number: 20182392.9

(22)Date of filing:  25.06.2020
(51)International Patent Classification (IPC): 
G05B 23/02(2006.01)
(52)Cooperative Patent Classification (CPC):
G05B 2219/39466; G05B 23/0283
(84)Designated Contracting States:
AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR
Designated Extension States:
BA ME
Designated Validation States:
KH MA MD TN

(71)Applicant: VOLKSWAGEN AG
38440 Wolfsburg (DE)

(72)Inventor:
  • Cortes Mac-Evoy, Carlos
    38442 Wolfsburg (DE)


(56)References cited: : 
  
      


    (54)ESTIMATING A MECHANICAL DEGRADATION OF A MACHINE


    (57) A method for estimating a mechanical degradation of a machine (1) comprises using a drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4) of the machine (1) to move a movable component (2a, 2b, 2c, 2d, 2e) of the machine (1) during an evaluation period. A drive parameter is measured during the evaluation period to set up a set of input data. A cluster analysis is performed by a computing unit (5) on the input data and a degradation value (d) for the machine (1) is determined by the computing unit (5) depending on a result of the cluster analysis.




    Description


    [0001] The present invention relates to a method for estimating a mechanical degradation of a machine, the machine comprising a movable component and drive unit for controlling a motion of the movable component, wherein the drive unit is used to move the movable component during an evaluation period. The invention further relates to a system for estimating a mechanical degradation of a machine and to a computer program product.

    [0002] The phenomenon of degradation of machines is of central importance, for example in condition based monitoring, CBM, or prognostics and health management, PHM, as it provides information in order to determine the current health status of a component or equipment and, therefore, to plan necessary actions for inspection or maintenance, aiming to improve the maintenance schedule, reduce maintenance costs and extend the lifetime of the systems and equipment.

    [0003] Known approaches for the degradation estimation may be grouped in two families: first, the model based approaches, which attempt to reconstruct the dynamics of the deterioration by means of physical models, for example regarding crack growth, fatigue or wear dynamics. Given the mathematical and system's knowledge required to represent the physical degradation, the Wiener process is one of the most popular tools for modeling the deterioration of a system. Generally the physical models have a limited applicability due to the occurrence of multiple and non-linear degradation processes on the operating assets. The second family of approaches are the data-driven methods, which are based upon the system's collected historical data in order to construct a representation of the degradation. These approaches are in general directed to the degradation estimation of single components and focus on the deterioration estimation of systems and components than can be directly measured, for example bearings, turbofans, batteries, pumps and so forth.

    [0004] However, the estimation of the degradation phenomenon of multiple components is much more challenging. In robotic grasping, for example, gripper systems are built using a significant amount of small components, such as spindles, small gearboxes, couplings, pulleys, tendons, and so forth in order to keep a certain equilibrium between performance and size, leading to a very limited amount of sensor data for building a representation of the degradation phenomenon. In order to ensure the execution of grasping tasks over extended periods of time in demanding scenarios without to access to maintenance, such as the outer space, deep sea or hazardous environments, the degradation states of the gripper systems have to be taken into account.

    [0005] It is therefore an object of the present invention to provide an improved concept for estimating a mechanical degradation of a machine, which allows for a reliable degradation estimation also for machines that are subject to complex degradation processes and, in particular, which is not dependent on direct measurements of the degradation status of the machine's components.

    [0006] This object is achieved by the respective subject-matter of the independent claims. Further implementations and preferred embodiments are subject-matter of the dependent claims.

    [0007] The improved concept is based on the idea to perform a cluster analysis based on a set of input data, which is generated depending on a drive parameter measured while the movable component is moved.

    [0008] According to the improved concept, a method for estimating a mechanical degradation of a machine is provided. The machine comprises a movable component and a drive unit for controlling a motion of the movable component. The drive unit is used to move the movable component during an evaluation period. A set of input data is generated, in particular by means of a sensor unit and a computing unit, wherein generating the set of input data comprises measuring a drive parameter of the drive unit, in particular by the sensor unit, during the evaluation period. A cluster analysis is performed by the computing unit by applying a cluster analysis algorithm to the set of input data. A degradation value for the machine, in particular for the movable component, is determined by the computing unit depending on a result of the cluster analysis.

    [0009] The movable component is, in particular, movable by the drive unit with respect to further static or movable components of the machine.

    [0010] The drive unit comprises, in particular, a motor drive, for example an electric motor drive, a pneumatic or hydraulic drive. The motor drive comprises a drive motor, in particular an electric drive motor.

    [0011] Depending on the implementation and operation principle of the drive unit, the drive parameter may correspond to different physical parameters. For example, the drive parameter may be a parameter concerning the electrical drive current in case of an electric motor drive or a work medium, such as a pressure or flow of air or a liquid, in case of a pneumatic or hydraulic drive unit. The drive unit may also comprise a control unit, for example an H-bridge, which is configured to control a position and the electrical drive current.

    [0012] The movement of the movable component during the evaluation period may for example be understood as a movement over at least a part of a predefined moving range of the movable component. Depending on the implementation of the machine and the movable component, the moving range may correspond to a one-dimensional parameter range, for example to a rotation angle or a position, or to a two-dimensional or higher-dimensional parameter space.

    [0013] In particular, the drive parameter may be measured during the evaluation period as a function of time, in particular at a certain sampling rate.

    [0014] The result of the cluster analysis may for example be understood as a description, in particular a mathematical description, of at least two, preferably at least three, clusters of points or vectors corresponding to the input data in a respective parameter space, which may for example include the drive parameter as a degree of freedom. Particularly good results have been achieved with five or six clusters. In particular, the result of the cluster analysis may comprise centroid positions, variances or other statistical quantities of the respective clusters.

    [0015] The degradation value is determined by the computing unit for example depending on respective descriptions of all clusters identified or defined by the cluster analysis. For example, the degradation value may be based on or correspond to a statistical quantity of the parameters describing the individual clusters. For example, a minimum distance between centroids of the clusters may correspond to or form a basis for the degradation value.

    [0016] By means of the improved concept, the cluster analysis based on the drive parameter is used as an indirect means to extract information regarding an actual degradation status of the machine or the movable component, respectively. In particular, the degradation value can be determined independent of several aspects related to the actual application or use of the machine. The inventors have realized that the actual status of the degradation leads to distinctive signatures in the drive parameter, which may be extracted by means of the cluster analysis. In particular, an actual status of the movable component or a part of the movable component, such as individual bearings, gears or connections, do not have to be directly inspected or measured in order to estimate the degradation according to the improved concept. In fact, the improved concept does not rely on the identification of the specific type of degradation or the specific failure type which may have caused the degradation. Therefore, the analysis may be highly automated and may, in particular, also be applied to machines with a high degree of complexity and with a great number of individual movable components, which are subject to degradation processes.

    [0017] According to several implementations of the method for estimating a mechanical degradation according to the improved concept, the cluster analysis algorithm is based on a machine learning model, in particular an unsupervised learning model.

    [0018] According to several implementations, the cluster analysis algorithm comprises a K-means algorithm.

    [0019] In particular, a parameter of the K-means algorithm corresponds to a predefined number of clusters and the K-means algorithm determines, after a certain number of iterations, which points of the input data correspond to which of the clusters and also the centroids of the clusters. The K-means algorithm has proven to be particularly efficient and suitable for extracting the degradation information from the measured values of the drive parameter. Furthermore, the K-means algorithm has the advantage that is does not require labeled training data.

    [0020] According to several implementations, generating the input data comprises generating a set of raw data by the measurement of the drive parameter and generating a point cloud depending on the set of raw data by the computing unit.

    [0021] In particular, the input data may correspond to the point cloud. The point cloud may be a one-dimensional, a two-dimensional, a three-dimensional or an even higher-dimensional point cloud. One degree of freedom of the corresponding parameter space of the input data and the point cloud, respectively, may correspond to the drive parameter. The other degrees of freedom may, if applicable, correspond to one or more further drive parameters of the drive unit and/or to one or more kinematic parameters of the movable component, such as position or orientation of the movable component, or other measurable or derivable parameters.

    [0022] According to several implementations, the set of raw data corresponds to the drive parameter as a function of time, in particular as a discrete function of time, and generating the input data comprises transforming the drive parameter from the time domain into a frequency domain by the computing unit, in particular by applying a discrete Fourier transform. Each point of the point cloud comprises a value of the drive parameter in the time domain and a value of the transformed drive parameter in the frequency domain.

    [0023] In this way, the drive parameter and its frequency content are used as parameters or features for the cluster analysis. It has been shown in tests and experiments, that such sets of parameters are particularly well suited for extracting the degradation information in terms of the cluster analysis.

    [0024] According to several implementations, a kinematic parameter of the movable component is measured during the evaluation period, in particular by using a further sensor unit. Each point of the point cloud comprises a value of the drive parameter and a value of the kinematic parameter.

    [0025] The kinematic parameter may, for example, correspond to a position parameter according to a corresponding degree of freedom of the movable component. In other words, the kinematic parameter may comprise or correspond to a translational position, a rotational position or a generalized position of the movable component or may be derived from them, for example by computing a respective time derivative or another function.

    [0026] Depending on the actual implementation of the machine, it may be particularly simple to measure the kinematic parameter as a function of time. Then, it may be beneficial to use the combination of the drive parameter and the kinematic parameter for the cluster analysis. In this way, the computational step of performing the Fourier transform may be saved. However, also in such implementations the drive parameter and the kinematic parameter may be given in the time domain or in the frequency domain, respectively.

    [0027] According to several implementations, performing the cluster analysis comprises, by the computing unit, determining respective centroids for each of at least three clusters of the point cloud and determining a set of mutual distances between the centroids. The degradation value is determined depending on the set of mutual distances.

    [0028] In particular, as described, the number of clusters may be predefined. The algorithm may determine all points of the point cloud corresponding to the respective clusters automatically and determine the centroids, which can be understood as the n-dimensional coordinates of the centroids in the respective n-dimensional parameter space, automatically.

    [0029] For example, the set of mutual distances comprises a distance between a first cluster and a second cluster of the at least three clusters, a distance between the first cluster and a third cluster of the at least three clusters and a distance between the second and the third cluster. In case more than three clusters are predefined, the set of mutual distances comprises a respective distance for each pair of clusters.

    [0030] The inventors have found out that, in case the degradation of the machine or the movable component is low or zero, the distance between the clusters' centroids is low. In other words, in case the movable component is "healthy", the points of the point cloud may be described well by a single cluster. However, as the movable component degrades over time, additional clusters are formed, whose distances increase over time. This phenomenon is exploited in such implementations by taking into account the set of mutual distances for determining the degradation value. Consequently, the degradation value may be determined with a higher reliability.

    [0031] According to several implementations, performing the cluster analysis comprises, by the computing unit, determining a statistical quantity of the set of mutual distances. The degradation value is determined depending on or as the statistical quantity. In this way, the evaluation and assessment of the degradation becomes particularly simple. In particular, memory requirements for storing the degradation value over time may be decreased.

    [0032] According to several implementations, the statistical quantity comprises a minimum value of the set of mutual distances.

    [0033] It has been found in experiments that the minimum distance between the clusters is particularly suitable for determining the degradation value, since, as described above, the distances between the individual clusters increases with ongoing degradation of the movable component over time. Furthermore, it has been shown that the minimum value features a particularly high monotonicity and a particularly low volatility, which makes it a particularly well-suited descriptor for the degradation.

    [0034] According to several implementations, the statistical quantity comprises a maximum value or a mean value or a variance value or a standard deviation value of the set of mutual distances.

    [0035] Depending on the actual implementation of the machine and the movable component, these quantities are also suitable due to their relatively low volatility and relatively high monotonicity, as has been found in the experiments.

    [0036] According to several implementations, the movement of the movable component during the evaluation period is carried out as a free movement and the drive parameter is measured during the free movement, in particular exclusively during the free movement.

    [0037] The free movement may be understood as a movement of the movable component without any external load been applied to the movable component. For example, if the movable component is designed to grasp an object or lift an object according to an application of the machine, the free movement corresponds to a movement, where the component is actually not lifting or grasping such an object. In other words, the movable component is brought into a load free status and then is moved within its allowed movement range during the evaluation period, such that the movable component is load free during the whole evaluation period.

    [0038] In this way, it may be avoided that the load affects the measured drive parameter of the drive unit during the evaluation period. Therefore, the measured drive parameter remains comparable over time, in particular over several evaluation periods. In consequence, the degradation value and, in particular, its evolution over time may be determined with a higher accuracy and reliability.

    [0039] According to several implementations, the machine comprises a gripping unit and the movement of the movable component is carried out as a gripping movement.

    [0040] In particular, the machine may be implemented as or comprise a robotic hand device, for example an anthropomorphic robotic hand device, which may for example be electrically drivable.

    [0041] The experiments of the inventors have revealed that the improved concept delivers a particularly reliable and meaningful assessment of the degradation for such devices.

    [0042] The movable component may, in case of an anthropomorphic robotic hand device, for example correspond to a finger or a thumb or a part thereof.

    [0043] According to several implementations, the drive parameter corresponds to an electric supply current of an electric motor of the drive unit.

    [0044] According to several implementations, the drive parameter comprises a fluid pressure or a fluid flow rate of a pneumatic or hydraulic unit of the drive unit.

    [0045] The fluid flow rate may be understood as a volume or mass of a respective work fluid, such as air, another gas, water or oil, per time unit and per cross-section area.

    [0046] According to the improved concept, also a system for estimating a mechanical degradation of a machine with a movable component is provided. The system comprises a drive unit, which is configured to control a motion of the movable component and to move the movable component during an evaluation period. The system comprises a sensor unit, which is configured to measure a drive parameter of the drive unit during the evaluation period. The system comprises a computing unit, which is configured to generate a set of input data depending on the drive parameter measured during the evaluation period, perform a cluster analysis by applying a cluster analysis algorithm to the set of input data and to determine a degradation value of the machine depending on a result of the cluster analysis.

    [0047] The machine and/or the movable component are not necessarily parts of the system for estimating the mechanical degradation. However, in several implementations, the system according to the improved concept, comprises the movable component and/or comprises the machine.

    [0048] According to several implementations, the system for estimating mechanical degradation comprises the machine and the machine comprises the drive unit.

    [0049] According to several implementations, the system comprises the machine and the machine comprises a gripping unit.

    [0050] According to several implementations, the drive unit comprises an electric motor and the drive parameter corresponds to an electric supply current of the electric motor.

    [0051] According to several implementations, the drive unit comprises a pneumatic or hydraulic unit and the drive parameter comprises fluid flow rate of the pneumatic or hydraulic unit.

    [0052] Further implementations of the system according to the improved concept follow directly from the various implementations of the method of the improved concept and vice versa. In particular, a system according to the improved concept may be configured to or programmed to carry out a method according to the improved concept or the system carries out such a method.

    [0053] According to the improved concept, also a computer program comprising instructions is provided. When the instructions are executed by a system according to the improved concept, in particular by the computing unit of the system, the instructions cause the system to carry out a method according to the improved concept.

    [0054] According to the improved concept, also a computer readable storage medium storing a computer program according to the improved concept is provided.

    [0055] The computer program and the computer readable storage medium may also be denoted as respective computer program products comprising the instructions.

    [0056] In the following, the invention will be explained in detail with reference to specific exemplary implementations and respective schematic drawings. In the drawings, identical or functionally identical elements may be denoted by the same reference signs. The description of identical or functionally identical elements is not necessarily repeated with respect to different figures.

    [0057] In the figures:
    Fig. 1
    shows schematically an exemplary implementation of a system according to the improved concept;
    Fig. 2
    shows schematically a further exemplary implementation of a system according to the improved concept;
    Fig. 3
    shows schematically a method step of an exemplary implementation of a method according to the improved concept;
    Fig. 4
    shows schematically a further method step of an exemplary implementation of a method according to the improved concept;
    Fig. 5
    shows schematically a further method step of an exemplary implementation of a method according to the improved concept; and
    Fig. 6
    shows schematically a further exemplary method step of an exemplary implementation of a method according to the improved concept.


    [0058] The embodiments described in the following represent preferred embodiments of the improved concept. The components described with respect to the individual embodiments represent individual features of the improved concept, which shall be considered independent of each other and may contribute to further implementations of the improved concept also independent of each other. These features shall be considered individually or in combinations different from the shown combinations as comprised by the improved concept. Furthermore, the embodiments may also be extended by additional features, which have been described above.

    [0059] In Fig. 1, an exemplary implementation of a system 8 according to the improved concept for estimating a mechanical degradation of a machine 1 is shown.

    [0060] The system 8 comprises a computing unit 5 and a sensor unit 7 coupled to the machine 1. The machine 1 comprises a plurality of movable components 2a, 2b, 2c, 2d, 2e and at least one drive unit 3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4 for each of the movable components 2a, 2b, 2c, 2d, 2e. The machine 1 may, in some implementations, comprise additional or less drive units as shown in the figures.

    [0061] In the present example, each of the drive units 3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4 comprises a respective actuator to move the corresponding movable components 2a, 2b, 2c, 2d, 2e or a part of it. The machine 1 has for example one or more electrically driven motors to control the actuators. In other words, each of the drive units 3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4 is associated to a respective electric motor, which is configured to control the motion of the respective actuator. The sensor unit 7 is configured to measure the motor current of the electric motor as a function of time for each of the drive units 3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4. In some implementations, each of the drive units 3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4 comprises an associated electric drive motor and the sensor unit 7 is configured to measure the corresponding electric currents of each motor.

    [0062] The machine 1 may also comprise a further sensor system (not shown) to measure the respective positions of the movable components 2a, 2b, 2c, 2d, 2e. For example, the further sensor system may comprise respective motor encoders for the electric motors to determine the respective positions of the components 2a, 2b, 2c, 2d, 2e.

    [0063] The machine 1 may for example be controlled by the computing unit 5 or by a further computing unit (not shown) to move one or more of the movable components 2a, 2b, 2c, 2d, 2e in a load free status during a respective evaluation period and the sensor unit 7 measures the respective motor currents during the evaluation periods. For one or more of the movable components 2a, 2b, 2c, 2d, 2e, the computing unit 5 generates a respective set of input data depending on the respective motor current measured during the evaluation period. The computing unit 5 performs a cluster analysis by applying a cluster analysis algorithm to each of the sets of input data and determines a degradation value for the machine 1 or, respectively, the corresponding movable component 2a, 2b, 2c, 2d, 2e or a part of it depending on a result of the cluster analysis.

    [0064] In the following, the improved concept is explained with respect to a concrete example for the machine 1 under consideration, namely for the machine 1 corresponding to an anthropomorphic hand 1 as indicated in Fig. 1. Fig. 2 shows the hand 1 grasping a test object 6. It is pointed out that this exemplary choice of the machine 1 does in general not restrict the applicability of the improved concept.

    [0065] The anthropomorphic hand 1 is able to provide sensor data in terms of joint angular positions and motor currents. Such hands 1 are commercially available in June 2020 for example from Schunk GmbH & Co. KG. The hand 1 is driven by nine motors actuating at least nine joints by means of leadscrew mechanisms that convert prismatic motion into rotational motion. For example, each of the motors may be part of one of the drive units 3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4. The movable components 2a, 2b, 2c, 2d, 2e correspond to a thumb 2a, an index finger 2b, a middle finger 2c, a ring finger 2d and a little finger 2e.

    [0066] The hand 1 may be controlled by a position controller (not shown) and a current controller (not shown) working in cascade that are interfaced via a driver (not shown) compatible with the known Robot Operating System ROS. Grasping may be performed by a grasp controller (not shown) actuating in motor space based on the motor torque measurements and therefore, it is possible to determine the enclosure of an object.

    [0067] In the following, an exemplary implementation of a method according to the improved concept is explained with respect to Fig. 3 to Fig. 6, wherein the hand 1 is considered as a non-limiting example for the machine 1.

    [0068] In a method step represented by Fig. 3, the hand 1 is set to perform a free movement on a regular basis, for example once a day during ten minutes. During this time, both the motor current signals of the drive units 3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4 as well as the corresponding positions are recorded, for example using ROS at a frequency of 100 Hz, and afterwards stored on a local database (not shown). The free movement may for example be chosen as a free open-close movement across the mechanical movement range of all the components 2a, 2b, 2c, 2d, 2e of the hand 1 in order to isolate it from external disturbances and loads and have a better picture of the actual system's behavior. Otherwise, the measurements could be influenced by the effects of the loads, disturbances and other phenomena like heat transfer, deformation and so forth.

    [0069] Figure 4 shows schematically the angular position P and its corresponding motor current signal RD for a specific finger as well as the corresponding target position TP. The motor current signal RD as a function of time may be considered as raw data. In an experiment, the aforementioned measurements have been conducted for more than 50 days, leading to a dataset size of about 27 million records.

    [0070] On healthy fingers, the motor current is mostly concentrated on a whole single data unit or cluster across the corresponding mechanical movement range in absence of loads (grasping). However, as the fingers degrade over time, it is observed that their associated motor currents begin to split and form additional clusters, whose distance increases over time. The motors of the fingers of the hand 1 may exhibit the same dynamic behavior during free motion under healthy conditions. When they are traveling to the closed position (fist), the motor currents may be negative, while they may be positive when traveling back. For healthy fingers, the distance between these clusters is small.

    [0071] Based on the previous observation, an unsupervised learning model based on K-means may be trained to analyze the evolution of the clusters over time, as depicted schematically in Fig. 5, which depicts the distribution of the motor current of the drive unit 3b as a function of the corresponding angular position of the respective part of the index finger 2b. The angular positions, motor currents in the time domains and their corresponding values in the frequency domain may be used as features for the models forming a point cloud as input data for the model, since they revealed significant information about the degradation. Using the K-means algorithm, the computing unit 5 may determine the centroids of the clusters in the input data and based on the position of the centroids, computes the mutual Euclidean distances between them. A vector D containing all mutual distances between the cluster centroids across the entire database is set up. D is a real (k x p)-dimensional vector, wherein p = n*(n-1) / 2, n > 2 is a predefined number of clusters and k = 1, 2, ..., 53 corresponds to the respective measurement instance. Several parameters, such as the sum, the minimum, the maximum, the mean value, the standard deviation as well as the area covered by the centroids may be were calculated by the computing unit 5 from vector D.

    [0072] A Spearman and volatility analysis revealed that the minimum distance between the clusters centroids for a given k exhibits a particularly monotonic behavior and very low volatility. Therefore, a degradation value d may be computed by the computing unit 5 for example as minimum value of D for fixed k. Fig. 6 shows the degradation value d as a function of k for the motor of drive unit 3a corresponding to a flexion movement of the thumb 2a as empty circles. The filled circles show d as a function of k for the motor of drive unit 3a' corresponding to an opposition movement of the thumb 2a.

    [0073] As can be seen in Fig. 6, the estimated degradation values d for the thumb 2a were low for most of the time. However, during the course of the experiment, the hand 1 suffered a collision, which caused a crack on a structural screw hole of the hand 1. The degradation values d are consistent with the consequences of this failure as d increased significantly right after the occurrence of the failure event for the thumb opposition 3a'. Maintenance actions such as oiling may partially alleviate degraded situations, as also reflected in Fig. 6.

    [0074] The improved concept allows the online and offline detection of degradation phenomena maintaining a low complexity using real data from a real system. The approach does not necessarily require a large amount of data nor existing datasets for estimating the degradation.

    [0075] The improved concept may be used for anthropomorphic hands as described above but also for other types of robotic hands, gripper systems and generally for machines with one or more moveable components, which are driven by a drive unit with a measurable drive parameter such as the drive current in case of electric motors. In alternative implementations, the air flow in pneumatically driven gripper systems may be used analogously as the motor currents in the description above.

    Reference signs



    [0076] 
    1
    machine, hand
    2a, 2b, 2c, 2d, 2e
    movable components
    3a, 3a', 3b, 3b', 3c, 3d, 3e
    drive units
    4
    drive unit
    5
    computing unit
    6
    test object
    7
    sensor unit
    8
    system
    d
    degradation value
    RD
    raw data
    TP
    target position
    P
    position



    Claims

    1. Method for estimating a mechanical degradation of a machine (1), the machine (1) comprising a movable component (2a, 2b, 2c, 2d, 2e) and a drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4) for controlling a motion of the movable component (2a, 2b, 2c, 2d, 2e), wherein the drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4) is used to move the movable component (2a, 2b, 2c, 2d, 2e) during an evaluation period,
    characterized in that

    - a set of input data is generated, wherein generating the set of input data comprises measuring a drive parameter of the drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4) during the evaluation period;

    - a cluster analysis is performed by a computing unit (5) by applying a cluster analysis algorithm to the set of input data; and

    - a degradation value (d) for the machine (1) is determined by the computing unit (5) depending on a result of the cluster analysis.


     
    2. Method according to claim 1,
    characterized in that
    the cluster analysis algorithm comprises a K-means algorithm.
     
    3. Method according to one of the preceding claims,
    characterized in that
    generating the input data comprises generating a set of raw data (RD) by the measurement of the drive parameter and generating a point cloud depending on the set of raw data (RD) by the computing unit (5).
     
    4. Method according to claim 3,
    characterized in that

    - the set of raw data (RD) corresponds to the drive parameter as a function of time;

    - generating the input data comprises transforming the drive parameter from the time domain into a frequency domain by the computing unit (5); and

    - each point of the point cloud comprises a value of the drive parameter in the time domain and a value of the transformed drive parameter in the frequency domain.


     
    5. Method according to claim 3,
    characterized in that

    - a kinematic parameter of the movable component (2a, 2b, 2c, 2d, 2e) is measured during the evaluation period;

    - each point of the point cloud comprises a value of the drive parameter and a value of the kinematic parameter.


     
    6. Method according to one of claims 3 to 5,
    characterized in that

    - performing the cluster analysis comprises

    - determining respective centroids for at least three clusters of the point cloud; and

    - determining a set of mutual distances between the centroids; and

    - the degradation value (d) is determined depending on the set of mutual distances.


     
    7. Method according to claim 6,
    characterized in that

    - performing the cluster analysis comprises determining a statistical quantity of the set of mutual distances; and

    - the degradation value (d) is determined depending on the statistical quantity.


     
    8. Method according to claim 7,
    characterized in that
    the statistical quantity comprises a minimum value or a maximum value or a mean value or a variance value or a standard deviation value of the set of mutual distances.
     
    9. Method according to one of the preceding claims,
    characterized in that
    the movement of the movable component (2a, 2b, 2c, 2d, 2e) during the evaluation period is carried out as a free movement and the drive parameter is measured during the free movement.
     
    10. Method according to one of the preceding claims,
    characterized in that
    the machine (1) comprises a gripping unit and the movement of the movable component (2a, 2b, 2c, 2d, 2e) is carried out as a gripping movement.
     
    11. Method according to one of the preceding claims,
    characterized in that

    - the drive parameter corresponds to an electric supply current of an electric motor of the drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4); or

    - the drive parameter comprises a fluid flow rate of a pneumatic or hydraulic unit of the drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4).


     
    12. System for estimating a mechanical degradation of a machine (1) with a movable component (2a, 2b, 2c, 2d, 2e), the system comprising a drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4), which is configured to control a motion of the movable component (2a, 2b, 2c, 2d, 2e) and to move the movable component (2a, 2b, 2c, 2d, 2e) during an evaluation period,
    characterized in that

    - the system (8) comprises a sensor unit (7), which is configured to measure a drive parameter of the drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4) during the evaluation period;

    - the system (8) comprises a computing unit (5), which is configured to

    - generate a set of input data depending on the drive parameter measured during the evaluation period;

    - perform a cluster analysis by applying a cluster analysis algorithm to the set of input data; and

    - determine a degradation value (d) for the machine (1) depending on a result of the cluster analysis.


     
    13. System according to claim 12,
    characterized in that
    the system (8) comprises the machine (1) and the machine (1) comprises a gripping unit.
     
    14. System according to one of claims 12 or 13,
    characterized in that

    - the drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4) comprises an electric motor and the drive parameter corresponds to an electric supply current of the electric motor; or

    - the drive unit (3a, 3a', 3b, 3b', 3c, 3c', 3d, 3e, 4) comprises a pneumatic or hydraulic unit and the drive parameter comprises a fluid flow rate of the pneumatic or hydraulic unit.


     
    15. Computer program product comprising instructions, which, when executed by a system (8) according to one of claims 12 to 14, causes the system (8) to carry out a method according to one of claims 1 to 11.
     




    Drawing













    Search report









    Search report