Gap detection device and gap detection method for robot joints

Abstract

Provided are a device and method for determining the gap between kinematic pairs in a driven joint of a robot having a driving link and a driven link. The gap detection device comprises: a measuring unit that measures the driving torque or current value of a motor when the robot is actually moving along an arbitrary motion trajectory; a simulation unit that sets an arbitrary second gap amount between the kinematic pairs of a plurality of kinematic pairs, simulates the robot moving along the same motion trajectory as the arbitrary motion trajectory, and estimates the driving torque or current value of the motor; a feature quantity calculation unit that calculates a first feature quantity representing fluctuations in a value related to the measured driving torque or current value, and a second feature quantity representing fluctuations in a value related to the estimated driving torque or current value; and a gap calculation unit that calculates an index related to the first gap amount based on the first feature quantity, the second feature quantity, and the second gap amount.

Description

Technical Field

[0001] The present invention relates to a device and a method for detecting gaps at joints of a robot. Background Art

[0002] As an example of a robot with a linkage mechanism, a parallel linkage robot is known that has a triangular parallel linkage mechanism. This triangular parallel linkage mechanism has the function of three-dimensionally positioning a movable part including an end effector. A triangular parallel linkage robot comprises a base, a movable part, and a driving link and a driven link connecting the base and the movable part. In most cases, three pairs of driving links and driven links are provided. By controlling the operation of each pair separately, the movable part can be moved with three degrees of freedom (X, Y, and Z).

[0003] Typically, the driven link and the driving link, and the driven link and the movable part, are connected via a three-degree-of-freedom ball joint. Ball joints can be categorized as either separate or integrated. As an example of prior art for detecting disengagement in separate ball joints, a parallel link robot is provided with a sensor for detecting the tilt of the end plate, the final output of the parallel link robot. Based on the output value of this sensor, it is detected that the connection between the links has been disconnected at at least one of the multiple connection points of the links connected by the ball joint (see, for example, Patent Document 1).

[0004] Furthermore, a detection device is known in which an internal passage is formed that opens on the surface of the ball head of a ball joint and determines whether the connection of the ball joint is disconnected based on a pressure detection value of the internal passage (see, for example, Patent Document 2).

[0005] On the other hand, an integrated ball joint, for example, has a connecting rod ball joint structure in which the ball head and the housing are integrated, preventing the housing and the ball head from easily separating. As a prior art example of detecting the gap between the housing and the ball head in an integrated ball joint, the following device and method are known. These devices generate a robot motion trajectory in which the ball head collides with the housing in a target joint and slides with the housing in another joint. Based on the magnitude of the torque fluctuation when the robot is driven along this motion trajectory, the device determines whether there is an abnormal gap in the target joint (see, for example, Patent Documents 3 and 4).

[0006] Prior art literature

[0007] Patent Literature

[0008] Patent Document 1: Japanese Patent Application Laid-Open No. 2017-056507

[0009] Patent Document 2: Japanese Patent Application Laid-Open No. 2017-013160

[0010] Patent Document 3: Japanese Patent Application Publication No. 2019-136838

[0011] Patent Document 4: Japanese Patent Application Laid-Open No. 2020-142353 Summary of the Invention

[0012] Problems to be solved by the invention

[0013] In a split-type ball joint, when an unexpectedly high-speed movement or collision occurs, the restraining force for attracting the ball head to the housing at the joint portion of the driven link is insufficient, and there is a risk that the joint portion will disintegrate.

[0014] On the other hand, in an integrated ball joint, it is understood that the housing and the ball head will not easily separate even in the event of a collision, due to the mechanical connection. However, when using a connecting rod ball head structure, if the ball head or the housing wears out due to use, a gap will be generated between the ball head and the housing, which may cause a deterioration in the positioning accuracy of the robot's movable parts and an increase in vibration. Due to the deterioration in positioning accuracy and the increase in vibration, it is difficult for the robot to perform normal operations and assembly operations, and sometimes it also leads to major problems such as reduced production efficiency and stopped production processes. Therefore, if there is an abnormality in the gap of the connecting rod ball head, it is desirable to know this abnormal state in advance.

[0015] Existing methods for detecting disengagement in ball joints have difficulty detecting widening gaps between the ball head and the housing (socket) in ball joints with structures that are difficult to disengage. Typically, even a gap of 0.1 to 0.2 mm is considered abnormal, but detecting such minute gaps using sensors is extremely difficult.

[0016] Furthermore, the robot motion required to identify joints with abnormal clearances sometimes requires specialized motions that differ from standard production operations. Such specialized robot motions can be difficult to implement due to interference between the robot and peripheral equipment, or due to layout constraints involving the robot itself. Therefore, a method is desired that can detect abnormal clearances even during standard production operations.

[0017] Means for solving problems

[0018] One embodiment of the present disclosure is a gap detection device for detecting a first gap between kinematic pair elements of a kinematic pair connected to a driven link in a robot, wherein the robot includes: a driving link driven by a motor; a plurality of driven links driven by movement of the driving link; and a plurality of kinematic pairs respectively connected to the plurality of driven links. The gap detection device includes: a measuring unit that measures a driving torque or a current value of the motor when the robot is actually moved along an arbitrary motion trajectory; a simulation unit that sets an arbitrary second gap between the kinematic pair elements of the plurality of kinematic pairs, performs a simulation of moving the robot along the same motion trajectory, and estimates the driving torque or current value of the motor; a feature value calculation unit that calculates a first feature value representing a change in a value related to the driving torque or current value measured by the measuring unit and a second feature value representing a change in a value related to the driving torque or current value estimated by the simulation unit; and a gap calculation unit that calculates an index related to the first gap value based on the first feature value, the second feature value, and the second gap value.

[0019] Another embodiment of the present disclosure is a gap detection method, in which a first gap amount between kinematic pair elements of a kinematic pair connected to a driven link is detected in a robot, wherein the robot has: a driving link driven by a motor; a plurality of driven links driven as the driving link moves; and a plurality of kinematic pairs respectively connected to the plurality of driven links, the gap detection method including the following processing: measuring the driving torque or current value of the motor when the robot is actually moved along an arbitrary motion track; setting an arbitrary second gap amount between the kinematic pair elements of the plurality of kinematic pairs, performing a simulation of moving the robot along the same motion track as the motion track, and estimating the driving torque or current value of the motor; calculating a first characteristic quantity for representing a change in a value related to the measured driving torque or current value, and a second characteristic quantity for representing a change in a value related to the estimated driving torque or current value; and calculating an indicator related to the first gap amount based on the first characteristic quantity, the second characteristic quantity, and the second gap amount.

[0020] Another embodiment of the present disclosure is a gap detection device, which detects a first gap amount between kinematic pair elements of a kinematic pair connected to a second link in a robot, wherein the robot has: a first link that can be driven by a motor; a second link connected to the first link; and one or more kinematic pairs connected to the second link, and the gap detection device has: a measuring unit that measures a first output value of the motor when the robot is actually moved along an arbitrary motion track; a simulation unit that sets an arbitrary second gap amount between the kinematic pair elements of the multiple kinematic pairs, performs a simulation of moving the robot based on the motion track, and estimates the second output value of the motor; and a gap calculation unit that calculates the first gap amount based on changes in the first output value, changes in the second output value, and the second gap amount.

[0021] Effects of the Invention

[0022] According to the present disclosure, it is possible to easily and accurately determine the clearances between the various kinematic pairs (joints) of a robot and determine whether the clearances are abnormal. In addition, such determination and judgment can be performed during any movement of the robot and are not restricted by the robot's layout. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The gap detection device according to the preferred embodiment is shown together with a delta-type parallel link robot as an example of an application target of the device.

[0024] Figure 2 Yes Figure 1 A partial enlarged view of the structure of each ball joint of the parallel linkage robot.

[0025] Figure 3 Represents the structural model of the parallel link robot.

[0026] Figure 4 This is a flowchart showing an example of the gap detection method according to the first embodiment.

[0027] Figure 5 This is a graph showing an example of temporal changes in the driving torque of the electric motor.

[0028] Figure 6 This is a graph showing an example of temporal changes in the difference between the actual measured driving torque and the normal condition.

[0029] Figure 7 This is a flowchart showing an example of the gap detection method according to the second embodiment.

[0030] Figure 8 This is a flowchart showing an example of the steps of kinetic analysis.

[0031] Figure 9 An example of a mass model for a parallel link robot mechanism is shown.

[0032] Figure 10 Explain the constraints of the kinematic pair taking into account the clearance.

[0033] Figure 11 This is a block diagram showing an example of a method for calculating torque.

[0034] Figure 12 The table shows the combinations of kinematic clearances obtained using dynamic analysis and actual machine measurements.

[0035] Figure 13 The coordinates of the start and end points of the generated multiple tracks are expressed in a table format.

[0036] Figure 14 This is a graph showing an example of an acceleration waveform in the motion direction of an output joint of a robot.

[0037] Figure 15a This is a data scatter plot obtained by performing principal component analysis on the dynamic analysis results of a certain kinematic pair.

[0038] Figure 15b This is a data scatter plot obtained by performing principal component analysis on the dynamic analysis results of a certain kinematic pair.

[0039] Figure 15c This is a data scatter plot obtained by performing principal component analysis on the dynamic analysis results of a certain kinematic pair.

[0040] Figure 15d This is a data scatter plot obtained by performing principal component analysis on the dynamic analysis results of a certain kinematic pair.

[0041] Figure 15e This is a data scatter plot obtained by performing principal component analysis on the dynamic analysis results of a certain kinematic pair.

[0042] Figure 15f This is a data scatter plot obtained by performing principal component analysis on the dynamic analysis results of a certain kinematic pair.

[0043] Figure 16a This is a data scatter plot obtained by performing principal component analysis on the actual measurement results of a certain kinematic pair.

[0044] Figure 16b This is a data scatter plot obtained by performing principal component analysis on the actual measurement results of a certain kinematic pair.

[0045] Figure 16c This is a data scatter plot obtained by performing principal component analysis on the actual measurement results of a certain kinematic pair.

[0046] Figure 16d This is a data scatter plot obtained by performing principal component analysis on the actual measurement results of a certain kinematic pair.

[0047] Figure 16e This is a data scatter plot obtained by performing principal component analysis on the actual measurement results of a certain kinematic pair.

[0048] Figure 16f This is a data scatter plot obtained by performing principal component analysis on the actual measurement results of a certain kinematic pair.

[0049] Figure 17 This is a flowchart showing an example of the verification procedure.

[0050] Figure 18 It means according to Figure 17 A chart of the verification results performed in the order shown.

[0051] Figure 19 Another configuration example to which this embodiment is applicable is schematically shown.

[0052] Figure 20 Still another configuration example to which this embodiment is applicable is schematically shown. DETAILED DESCRIPTION

[0053] Figure 1 The present disclosure shows a schematic structure of a gap detection device according to a preferred embodiment of the present disclosure and a triangular parallel link robot as an example of a structure to which the gap detection device can be applied. The parallel link robot (hereinafter, also referred to as the robot) 10 includes a base 12, a movable portion 14 disposed separately from the base 12 (typically below the base 12), two or more (three in the illustrated example) link portions 16a to 16c connecting the base 12 and the movable portion 14 and each having one degree of freedom relative to the base 12, and a plurality of (typically the same number as the link portions, three in the illustrated example) servo motors or other motors 18a to 18c that drive the link portions 16a to 16c, respectively. An end effector such as a manipulator can be mounted on the movable portion 14.

[0054] The connecting rod portion 16a is composed of a driving link 20a connected to the base portion 12, and a pair (two) of driven links (distal links) 22a that connect the driving link (base link) 20a to the movable portion 14 and extend parallel to each other. The driving link (first link) 20a and the driven link (second link) 22a are connected by a pair (two) first joints 24a. In addition, the movable portion 14 and the driven link 22a are connected by a pair (two) second joints 26a. In addition, in this embodiment, the first joint and the second joint (kinematic pair) are both configured as ball joints (spherical bearings).

[0055] Figure 2 This is a partially enlarged view showing the structure (link ball joint structure) of each ball joint (here, ball joint 24a or 26a) of the robot 10. The ball joint 24a or 26a has a ball head (convex portion) 28 and a housing (concave portion) 30 that accommodates the ball head 28, which serve as a kinematic sub-element (joint element) described later. A lining 32 is disposed between the ball head 28 and the housing 30. Figure 1 As shown, the robot 10 has a restraining plate 34a provided on the driving link side (upper side) of the driven link to be connected between the housings of the first ball joint 24a in order to restrain the rotation of the two parallel driven links 22a around their respective axes.

[0056] The other connecting rod portions 16b and 16c can also have the same structure as the connecting rod portion 16a, so the corresponding components are given reference figure marks with only the last digit changed (for example, the elements corresponding to the driven connecting rod 22a are given reference figure marks 22b or 22c), and detailed descriptions are omitted.

[0057] like Figure 1 As schematically shown, the parallel link robot 10 is connected to a control device 36 that controls the motion of the robot 10. Furthermore, the gap detection device 38 that detects the actual gap (first gap amount) of the ball joint includes: a measuring unit 44, such as a torque sensor or ammeter, that measures the driving torque or current value of the motors 18a-18c when the robot 10 is actually moved along an arbitrary motion trajectory; a simulation unit 40 that sets an arbitrary second gap amount between the kinematic pair elements of a plurality of kinematic pairs, simulates the movement of the robot 10 along the same motion trajectory, and estimates the driving torque or current value of the motors 18a-18c; a feature quantity calculation unit 46 that calculates a first feature quantity representing fluctuations in a value related to the driving torque or current value measured by the measuring unit 44, and a second feature quantity representing fluctuations in a value related to the driving torque or current value estimated by the simulation unit 40; and a gap calculation unit 42 that generates a mathematical model (described later) that relates the first feature quantity, the second feature quantity, the first gap amount, and the second gap amount, and calculates an index related to the first gap amount based on the mathematical model. In addition, the gap detection device 38 may also arbitrarily have a judgment unit 48, which judges that the gap of a moving pair whose indicator related to the first gap amount among multiple moving pairs exceeds a predetermined reference value is abnormal. In this case, the gap detection device 38 functions as an abnormality detection device for detecting whether each moving pair has an abnormal (excessive) gap.

[0058] The robot control device 36 is configured to generate motion instructions for moving the robot 10 and control each axis (motor) of the robot 10 based on the motion instructions. Furthermore, the gap detection device 38 may also include: a storage unit 50, such as a memory, which stores measurement data, feature quantities, etc.; an output unit 52, which outputs the above-mentioned simulation results, judgment results, etc. in a manner that can be recognized by the operator; and an input unit 53, such as a keyboard or touch panel, which is used by the operator to perform various settings, input data, etc. Specific examples of the output unit 52 include a display that can display simulation results and judgment results, a speaker that outputs simulation results and judgment results as sound, and a vibrator that can be carried by the operator and vibrates when it is determined that the friction state of the joint is abnormal (abnormal gap exists). The operator receives the output from the output unit 52 and can repair or replace the abnormal joint.

[0059] The gap detection device 38 can be realized as a calculation processing device such as a personal computer (PC) having a processor and a memory connected to the robot control device 36. Figure 1 In the figure, the gap detection device 38 is shown as a device separate from the robot control device 36, but it can also be incorporated into the control device 36 in the form of a processor and memory. In addition, a device such as a PC can also be responsible for part of the gap detection function, and the robot control device 36 can be responsible for other functions.

[0060] Figure 3 express Figure 1 The structural model of the parallel link robot 10. The parallel link robot 10 has a closed-loop link structure including three rotation drive units (motors) and 12 driven kinematic pairs (here, ball joints).

[0061] Reference Figure 2 As the robot 10 moves, the ball head 28 slides against the inner lining 32 in each joint. To minimize frictional resistance during this process, the inner lining 32 is often made of a low-friction material such as resin. However, as the inner lining 32 wears due to the robot's repeated movements, a gap (air gap) forms between the ball head 28 and the inner lining 32 (housing 30). If this gap exceeds a certain value, it may cause problems such as deterioration in the positioning accuracy of the robot 10 and increased vibration associated with the robot's movement. Therefore, in this embodiment, dynamic analysis is performed to determine the gap in each joint for any robot movement and determine whether the gap is an abnormal value.

[0062] (First embodiment)

[0063] Figure 4This is a flowchart showing an example of the processing in the gap detection device 38 of the first embodiment. First, in step S1, the actual robot 10 for which the (abnormal) gap is to be detected is driven along an arbitrary motion trajectory. Specifically, one or more robot motion trajectories for determining the kinematic gap are provided. Here, the trajectory is represented by the symbol m. ’ In addition, the so-called arbitrary motion trajectory can be, for example, a production motion in a factory.

[0064] Next, in step S2, the driving torque waveform of the motor driving each axis of the robot 10 is measured. Specifically, the driving torque τ during operation is measured. mes，m′ (t). The measured τ meas，m′ (t) is time series data representing the change in torque. However, a parallel link robot generally has three actuators, so the torque waveform corresponds to a vector that combines the waveforms of the three actuators. Figure 5 An example of the torque waveform at this time is shown in FIG. Measured data is stored in the storage unit 50, for example, once to several times a day.

[0065] Next, in step S3, the difference between the measured torque waveform and the torque waveform in the initial state is calculated. Specifically, the following formula (1) is used to obtain the ideal waveform τ of the driving torque obtained in advance under the condition that the clearance of all moving pairs is normal using the same action track as that used in the measurement: meas，idea，m′ (t) and the torque waveform τ measured in S2 meas，m′ (t) The waveform of the difference Δτ meas，m′ (t). Figure 6 An example of the torque waveform at this time is shown.

[0066] [Formula 1]

[0067] Δτ meas，m′ (t) = τ meas，m′ (t)-τ meas，ideal，m′ (t) (1)

[0068] On the other hand, the simulation (dynamic analysis) described later is performed using the same track as that used in the actual machine measurement (S5) to estimate the driving torque waveform τ sim，m′，n′ (t)(S6). The simulation results are stored in the storage unit 50, etc. In order to compare the results of the dynamic analysis with the results of the actual machine measurement in the subsequent processing to implement the determination of the kinematic pair clearance, the dynamic analysis is performed under various kinematic pair clearance conditions c sim，n′ The condition of the kinematic clearance is determined by the symbol n'. As for the result of the dynamic analysis, the waveform Δτ, which is the difference from the ideal waveform, is calculated using the following formula (2) in the same way as step S3. sim，m′，n′ (t)(S7).

[0069] [Formula 2]

[0070] Δτ sim，m′，n′ (t) = τ sim，m′，n′ (t)-τ sim，ideal，m′ (t) (2)

[0071] Here, Δτ in equations (1) and (2) meas，m′ (t) and Δτ sim，m′，n′ (t) is time series data and multidimensional data, so calculations using it tend to be complex and time-consuming. Therefore, in order to reduce the amount of calculation for gap estimation described later, feature selection is used to reduce the dimensionality of the time series data.

[0072] Specifically, when comparing drive torque data obtained using the same track but under different kinematic pair clearance conditions, it is believed that the differences (deviations) in these waveforms are caused by changes in the kinematic pair clearance conditions. Therefore, by resetting the orthogonal coordinates in descending order of the deviations, and extracting the first few coordinates as new variables, it is possible to reduce the dimensionality of the data while removing the influence of the kinematic pair clearance. Therefore, feature selection based on principal component analysis is implemented here. By reducing the data dimensionality through feature selection, the new characteristic value Δτ of the drive torque is obtained. meas，FS (first feature quantity), Δτ sim，FS (Second feature quantity) (S4, S8).

[0073] Furthermore, while this embodiment demonstrates a feature selection method using only principal component analysis, it is expected that the accuracy of the mathematical model can be improved by combining signal processing methods such as Fourier transforms and wavelet transforms, as well as various feature selection methods within the field of machine learning. For example, by using Fourier transforms or wavelet transforms to extract only specific frequency regions that are susceptible to the effects of kinematic pair clearance, other sources of error can be eliminated, thereby improving the accuracy of the model. Furthermore, extracting specific frequency regions also reduces data dimensionality. By performing the principal component analysis described above on the extracted data in the specific frequency regions, data dimensionality is further reduced.

[0074] Next, in step S9 , a mathematical model is constructed that relates the characteristic amount of the driving torque obtained through dynamic analysis and actual machine measurement to the size of the kinematic clearance, as shown in the following equation (3).

[0075] [Formula 3]

[0076]

[0077] Here, C sim Indicates the size of the clearance between the kinematic pairs (second clearance) after dynamic analysis, C estIndicates that the actual size of the kinematic pair clearance to be determined (the first clearance amount) is an unknown. The mathematical model is represented by a probabilistic model. The first equation in equation (3) is a model related to actual machine measurement, and the second equation is a model related to simulation (dynamic analysis). Error term ∈ 1，m′ and ∈2 are probability variables that follow a certain probability density function. Based on the understanding that increasing the clearance between the moving parts increases the torque fluctuation, a model that relates these variables to a linear equation related to the clearance is adopted as the simplest model that satisfies this condition.

[0078] ∈ 1，m′ It is believed that the error is caused by the modeling in the linear equation. Since the dynamics of a parallel link robot with kinematic clearance is nonlinear, there are nonlinear terms. It is also believed that the more the number of kinematic pairs with clearance increases, the greater the influence of the nonlinear terms. In order to consider this influence, the error term ∈ 1，m′ The error term ∈2 is an error term that takes into account the difference between the mathematical model of the dynamic analysis and the actual machine measurement. The main reasons for this difference are the modeling error in the dynamic analysis and the measurement error of the actual machine measurement. In this embodiment, each error term ∈ 1，m′ ,∈2 is modeled according to the (multivariate) normal distribution using the following formula (4).

[0079] [Formula 4]

[0080]

[0081] N(μ, ∑) represents a multivariate normal distribution with mean μ and variance-covariance matrix ∑. I is the identity matrix. 1，m′ , considering the correlation between feature quantities in the same orbit and modeling them according to a multivariate normal distribution. Regarding the nonlinear terms mentioned above, it is assumed that the value of the feature quantity that appears due to the gap in one ball kinematic pair varies due to the gap in another ball kinematic pair. The correlation between feature quantities in the same orbit is considered because this variation appears as error correlation. Meanwhile, ∈2 assumes that all elements are modeled according to the same one-dimensional normal distribution.

[0082] Next, in step S10, the constructed mathematical model is solved, more specifically, the maximum likelihood estimation solution of the constructed mathematical model is obtained, thereby determining the kinematic pair clearance (S10). The mathematical model constructed in this embodiment contains two error terms, so it is difficult to calculate the exact solution as in the matrix calculation of the least squares method. Therefore, as an example, consider deriving an approximate value of the maximum likelihood estimation solution by using Bayesian inference using the MCMC method, which is a type of Monte Carlo method. In Bayesian inference, the parameters of the probability model are estimated as probability variables based on the consideration method of Bayesian statistics. The most frequent value of the probability distribution related to the kinematic pair clearance derived by the MCMC method is obtained, thereby obtaining the maximum likelihood estimation solution related to the kinematic pair clearance. In addition, since the clearance amount is obtained from the probability distribution, the accuracy of the estimated clearance amount can also be known. In this embodiment, the maximum likelihood estimation solution obtained in this way is set as the estimated value of the kinematic pair clearance (an indicator related to the first clearance amount).

[0083] Finally, in step S11, gap abnormalities are detected based on the estimated gap size. Specifically, the estimated gap value is compared with a predetermined reference value considered abnormal. If the estimated value exceeds the reference value, the gap is determined to be abnormal. Thus, in addition to being able to quantitatively estimate the gap amount of each kinematic pair, this embodiment can also automatically identify abnormal kinematic pairs (those with excessive gaps).

[0084] (Second embodiment)

[0085] Figure 7 1 is a flowchart showing an example of processing in the gap detection device 38 of the second embodiment. Here, only the points different from the first embodiment will be described, and the description of the points similar to the first embodiment will be omitted.

[0086] In the first embodiment, the first and second equations of equation (3) are solved simultaneously by Bayesian inference, but in the second embodiment, the coefficient X is obtained in advance based on the results of dynamic analysis. 0，m ′ and X 1，m '. That is, in the second embodiment, according to Δτ sim，FS，m′，n′ and c sim，n The value of 'predetermines the coefficient X 0，m ′ and X 1，m '(S9"), according to the Δτ obtained by measurement meas，FS，m ′ and the determined X 0，m ′ and X 1，m ′, and estimate the kinematic clearance C of the actual robot est (S9').

[0087] In either of the first and second embodiments, by following the processes (a) to (d) described below, it is possible to extend the scope to a case where the motion trajectory of the robot used for gap estimation changes.

[0088] (a) The gap calculation unit 42 and the like divide the time interval before and after the time when the motion trajectory changes.

[0089] (b) The measuring unit 44 and the like measure torque using the same motion trajectory in each divided time interval.

[0090] (c) τ of formula (1) meas，ideal，m ′(t) is applied to the torque waveform initially measured in each time interval to calculate Δτ meas，m′ (t).

[0091] (d) The gap calculation unit 42 and the like use the Δτ obtained in (c) in all the divided time intervals. meas，m′ (t), using the method of the above embodiment, the gap in each time interval is determined. The determined gap is equivalent to the change in the gap in each time interval. Therefore, by adding these changes, the kinematic pair gap after the motion trajectory changes can be determined.

[0092] (simulation)

[0093] Reference Figure 8 , the details of step S5, that is, the steps of simulating (dynamic analysis) the parallel linkage robot 10 with a gap in a specific ball kinematic pair when it moves along a certain track, are described. Here, the analysis is performed considering the gaps of some of the 12 ball kinematic pairs possessed by the parallel linkage robot 10. The ball kinematic pairs other than the rotating kinematic pair and the object ball kinematic pair whose gap is considered are assumed to move ideally, so the gaps of the ball kinematic pairs other than the rotating kinematic pair and the object ball kinematic pair are not considered. In addition to the viscoelasticity of the contact between the kinematic pair elements, it is assumed that all the links are rigid bodies, and Figure 9 The mass model of the parallel link robot mechanism is shown.

[0094] like Figure 8 As shown, first set the output node that makes the robot move (for example Figure 1 The target trajectory of the movable plate 14 shown in FIG5 is determined (S51). Next, the ball kinematic pair for which an arbitrary gap is to be set and the size of the gap between the ball kinematic pairs are set (S52). Next, the equation of motion is derived for the combination of ball kinematic pairs with the set gaps (S53). Finally, the equation of motion is solved using an ordinary differential equation solver or the like to obtain a numerical solution for the driving torque (S54). This enables dynamic analysis to determine the driving torque.

[0095] A specific example of the method of deriving the motion equation in step S53 will be described. First, the following equation (5) represents the equation using the generalized coordinates q l equation of motion.

[0096] [Formula 5]

[0097]

[0098] In formula (5), the Einstein convention is applied. When a subscript appears twice in a term, the sum of the subscript is calculated. l represents generalized coordinates, F l represents the generalized force corresponding to the generalized coordinates (known quantity), (M l，m ) represents the vector q=(q l ) corresponds to the mass matrix. 1, m, and n all represent the numbers of the generalized coordinates, and N represents the number of generalized coordinates (the dimension of the mechanical system). C l，m，n is the first K-symbol, defined by the following formula (6).

[0099] [Formula 6]

[0100]

[0101] Substituting equation (6) into equation (5) and rearranging it, we obtain the following equation (7). In this case, the following equation (8) holds true.

[0102] [Formula 7]

[0103]

[0104] [Formula 8]

[0105]

[0106]

[0107] Here, is the Kronecker function, which is defined by the following equation (9).

[0108] [Formula 9]

[0109]

[0110] (M l，m ) is the mass matrix M GC According to formula (7), if the mass matrix and the partial differentials related to the generalized coordinates of the mass matrix are given, the force F lThe motion (generalized acceleration) caused by the forward dynamics analysis can be predicted by numerically integrating the obtained generalized acceleration. The system's degrees of freedom are roughly divided into the degrees of freedom of the mechanism when the clearance between all kinematic pairs is zero and the degrees of freedom of the kinematic pairs due to the clearance between the kinematic pairs. Therefore, it is considered to set the generalized coordinates as shown in the following equation (10).

[0111] [Formula 10]

[0112]

[0113] In formula (10), θ i represents the displacement of the three rotating motion pairs (actuators). J′ is the total number of ball motion pairs taking into account the clearance, is the first spherical kinematic pair considering the gap. Figure 10 As shown, Represents the spherical kinematic pair with clearance in the stationary coordinates The relative position vector between the centers of the kinematic pair elements (ball head 28 and cup 30) is called the kinematic pair error vector. By setting the generalized coordinates in this way, the motion of the mechanical system can be expressed non-redundantly. By solving the kinematics of the robot for the set generalized coordinates, the coefficients M of the motion equation in equation (5) can be derived. l，m and C l，m，n .

[0114] About the mass matrix M GC , using the matrix M that combines the mass of the connecting rod and the inertia tensor of the connecting rod's center of gravity coordinate system link , expressed by the following formula (11).

[0115] [Formula 11]

[0116] M GC = t A t J all (θ) -1 M link J all (θ) -1 A (11)

[0117] in, The vector qall (=(θ, δ all Here, it is assumed that the kinematic pair error is small enough relative to the mechanism constants such as the connecting rod length, and the Jacobian matrix J allIt depends only on the actuator displacement θ. A is a matrix determined by the location of the gap and is defined by the following equation (12).

[0118] [Formula 12]

[0119]

[0120] I is the identity matrix, J′ represents the number of kinematic pairs considering clearance, s l Indicates the number of the kinematic pair considering the clearance. link and A is a matrix that does not depend on q. In this case, qall=Aqq. Substituting equation (11) into equation (8), we get the following equation (13). Among them, equation (14) holds true. t Represents the transpose of a matrix.

[0121] [Formula 13]

[0122]

[0123] [Formula 14]

[0124]

[0125] Next, find the J in the parallel linkage robot. all (θ), The derived formula of . Figure 9 As shown, for connecting rod A i 、B i，j Set up a local orthogonal coordinate system and use the symbols in the figure as its basis vectors. The positions of the joints and the center of gravity of the connecting rods are defined by the symbols in the figure. Assuming that the mechanism constants are large enough relative to the size of the gap, Figure 9 The result of displacement analysis when all the symbols in the use gaps are all zero. Therefore, Figure 9 The signs of all depend only on θ. i The movement of only considers the rotational motion pair R i The axis e A，i，y The rotation of the link B is 1 degree of freedom. i，j The motion of the center of gravity is considered to remove the three degrees of freedom of translation and the remaining degrees of freedom (around e B，i，j，z The mass matrix M has two degrees of freedom except the rotation of link It is composed of the mass and inertia tensor of the above motion. When the mechanism is not in a specific posture, the Jacobian matrix J all (θ) becomes normalized and is expressed by the following equation (15) using the symbols in the figure.

[0126] [Formula 15]

[0127]

[0128] Among them, the following equations (16) to (18) hold true.

[0129] [Formula 16]

[0130]

[0131] [Formula 17]

[0132]

[0133] [Formula 18]

[0134]

[0135] In formulas (15) to (18), I n represents the n-th degree identity matrix, O m，n represents the zero matrix column of m rows and n columns, and [*]x represents the antisymmetric matrix generated by the vector *. Partial differential of the Jacobian matrix By performing partial differentiation on equations (15) to (18), the following equation (19) is obtained, where equations (20) and (21) hold.

[0136] [Formula 19]

[0137]

[0138] [Formula 20]

[0139]

[0140] [Formula 21]

[0141]

[0142] According to the above, if the external force F is determined l The derived formula of can make the equation of motion valid and can be used for forward dynamic analysis. Consider the contact force F of the kinematic pair joint , driving torque F based on the control law of the actuator actuator , and the gravity F acting on each connecting rod gravity , the external force F is given by the following formula (22): l .

[0143] [Formula 22]

[0144] (F l )=F joint +F actuator +F gravity (twenty two)

[0145] In this disclosure, the Lankarani model, which incorporates energy loss into Hertz's elastic contact theory, is used as a contact model for describing the separation, collision, and sliding of kinematic pair components. It is preferable to use a model appropriate for the kinematic pair's conditions and materials. The actuator control law utilizes the calculated torque method, frequently used in industrial robots.

[0146] Figure 11 This is a block diagram showing an example of the torque calculation method. Kp and Kd represent the P gain and D gain of the torque calculation method, respectively, s represents the differential operator, M(θ) represents the mass matrix, and V(θ, sθ) represents the generalized force obtained by combining gravity and gyroscopic torque. In the derivation of M(θ) and V(θ, sθ), all ball kinematic pairs use ideal mechanism models. Figure 11 By doing so, the driving torque τ generated in the actuator can be calculated. l，m 、C l，m，n , and thus the equations of motion can be solved.

[0147] (Based on experimental verification)

[0148] To verify the validity of the kinematic clearance determination method disclosed in this paper, a driving torque measurement experiment was conducted using a real robot. Here, actual machine measurements and dynamic analysis were performed under various kinematic clearance and track conditions, and the resulting data was combined for use.

[0149] Figure 12 (Table 1) shows the combination of kinematic pair clearances that were subjected to dynamic analysis and actual machine measurement. The sizes of the radial clearances of the kinematic pairs used are 0.00mm, 0.14mm, and 0.15mm (measured values). In the table, 0 indicates a radial clearance of 0.00mm, A indicates a radial clearance of 0.14mm, and B indicates a radial clearance of 0.15mm. In Table 1, the conditions that the number of ball kinematic pairs with excessive clearances (0.14mm, 0.15mm) is zero (case 1), one (cases 2-13), and two (cases 14-43) are set. In addition, in order to suppress the scale of the experiment, only the 6 ball kinematic pairs located at the output joint among the 12 ball kinematic pairs (in Figure 1 Equivalent to the excessive gap in the ball joints 26a~26c).

[0150] In Table 1, the ball kinematic pair indicated as 0 was measured using an ideal ball kinematic pair with a measured clearance of 0.00 mm in the actual machine measurement, and numerical calculations were performed under the conditions of the ideal ball kinematic pair in the dynamic analysis. Numerical calculations were performed using these values for the 0.14 mm and 0.15 mm gaps in the dynamic analysis. The track used in the measurement and analysis was a linear track that moved the output node a certain distance l0 (here, l0 = 200 mm). The starting and ending points of the track were randomly selected from the operating area to generate multiple linear tracks (set to 100 in this study). Figure 13 Table 2 summarizes the coordinates of the start and end points of multiple trajectories actually generated. In all trajectories, the acceleration waveform in the direction of motion of the output node uses the same deformed trapezoidal curve. Figure 14 Acceleration waveform indicating the direction of motion of the output node.

[0151] In this way, drive torque waveform data was obtained from actual machine measurements and dynamic analysis under various kinematic clearance and track conditions. Below, the subscript n' is used to identify kinematic clearance conditions, and the subscript m' is used to identify track conditions. By combining these acquired data and repeatedly performing kinematic clearance determination, verification of this disclosure was achieved.

[0152] The result of feature selection when using track 1 (m'=1) in Table 2 will be described. First, the waveform Δτ of the difference between the driving torque of each actuator and the waveform when the gap of all ball kinematic pairs is sufficiently small is obtained. sim，1，n ′(t), Δτ meas，1 (t). Since the dynamic analysis results are performed under multiple gap conditions, the results Δτ sim，1，n ′(t) performs principal component analysis. Figures 15a to 15f A scatter plot showing the first principal component score (PC1) and the second principal component score (PC2). Figures 15a to 15f This is a scatter plot of the same data. The ideal ball motion pair is represented by a black circle mark, and the ball motion pair with excessive clearance is represented by a white circle mark.

[0153] Next, the same graph is made for the actual machine measurement results. The size of the positive projection of the actual machine measurement results (the difference waveforms for all actuators are summarized) for each principal component vector obtained by principal component analysis of the dynamic analysis results is used as the principal component score of the actual machine measurement results data. Figures 16a-16f The scatter plot shows the first principal component score (PC1) and the second principal component score (PC2) of the actual machine measurement results under the conditions where all the kinematic clearances were measured. Figures 15a to 16fIt can be seen that although the scale of the vertical axis (PC2) is different in the dynamic analysis results and the actual machine measurement results, the appearance of the black and white circle patterns is similar. In particular, the ball motion pair S 2，2，2 The effect of the gap is significantly seen in the second principal component score ( Figure 16d ). In this way, the influence of the kinematic pair clearance can be extracted by principal component analysis, so by combining multiple types of orbital data, the kinematic pair clearance can be determined with high accuracy by constructing and solving a mathematical model using the principal component scores.

[0154] Next, the kinematic clearance was determined, and the estimated clearance value obtained was compared with the actual clearance size, thereby verifying the validity of the present disclosure. Figure 17 This is a flowchart showing an example of the verification procedure. In this verification, the kinematic clearance is repeatedly determined while randomly changing the combination of data used.

[0155] First, the orbital group used in determining the gap is determined. Here, n orbitals are randomly selected from 100 orbitals. traj Regarding the data of dynamic analysis, for n traj For the analysis results of each track, all the combinations (43 types) of data shown in Table 1 were selected. For the actual machine measurement results, one of the combinations (43 types) shown in Table 1 was randomly selected as the data for determining the kinematic pair clearance conditions. Figure 4 The gap determination process shown in FIG. 1 is used to obtain an estimated value of the gap size for each ball kinematic pair. Since the gap size of the ball kinematic pair used in the actual machine measurement is measured in advance, a pair of estimated values and measured values of the kinematic pair gap can be obtained. By repeatedly performing this operation, a pair of measured values and estimated values of the kinematic pair gap size can be obtained, and the estimation performance can be evaluated. In this verification, let n be the number of pairs. traj =10, Figure 17 The cycle shown is repeated 100 times, performing the determination of the kinematic joint clearance.

[0156] Here, the result of one of the 100 executions of kinematic pair clearance determination is described. The selected track numbers (numbers shown in Table 2) are (64, 60, 58, 80, 70, 55, 67, 44, 35, 61), and the number indicating the clearance condition to be determined (number shown in Table 1) is 15. The measured value of the clearance S 1，1，2 0.14mm, S 2，1，2 is 0.15mm, and the other ball motion pairs are 0mm. The estimated value of the motion pair clearance at this time is as follows Figure 18In this figure, the 95% confidence interval of the Bayesian inference is represented by error bars. This figure shows that, although the estimated gap values vary, large values can be estimated for ball kinematic pairs that actually have excessive gaps, confirming the usefulness of the present disclosure.

[0157] In addition, in the above-mentioned embodiment, the determination of the driving torque of the motor is performed as the output value of the motor, but the time differential value of the driving torque may be used instead. In addition, instead of using a value related to the driving torque (here, the driving torque or its time differential value), a value related to the current value of the motor (for example, the current value or its time differential value) may be used as the output value of the motor. Generally, since the driving torque is proportional to the current value, the same process as described above can be applied even when a value related to the current value is used. Furthermore, in the above-mentioned embodiment, a characteristic value representing the variation of the output value of the motor is used to determine the first gap amount, but the variation data itself may be used instead of the characteristic value.

[0158] In addition, in this embodiment, a parallel link robot is described as a robot to which the gap detection device and gap detection method disclosed herein can be applied, but the application is not limited thereto. As other preferred examples to which the gap detection device and gap detection method disclosed herein can be applied, a six-axis vertical multi-joint robot without a closed-loop linkage mechanism, such as a Figure 19 or Figure 20 A robot as schematically shown has at least partially a closed-loop linkage mechanism.

[0159] Figure 19 A robot 84 having a planar linkage mechanism is shown, wherein the planar linkage mechanism comprises a driving joint 80 and three driven joints 82, and a load is applied to the front end. Figure 20 A robot 90 having a five-link mechanism is shown, wherein the five-link mechanism comprises two driving joints 86 and three driven joints 88, which can be used for positioning devices, etc. Figure 1 The parallel linkage robot shown similarly has a driving link driven by an electric motor, multiple follower links that follow the movement of the driving link, and multiple kinematic pairs respectively connected to the multiple follower links. Therefore, it is possible to identify and detect joints (driven kinematic pairs) with abnormal gaps in the same way as described above.

[0160] In addition, in this embodiment, a spherical joint (ball joint) is described as a kinematic pair (joint) to which the gap detection device and gap detection method disclosed herein can be applied, but the application is not limited to this. The gap detection device and gap detection method can also be applied, for example, to a hinge structure (rotating joint) with one degree of freedom. In this case, the rotating joint (hinge structure) has a generally cylindrical member (convex portion) as a kinematic pair element and a generally cylindrical member (concave portion) that engages with the cylindrical member. Even in such a hinge structure, an abnormal radial gap may occur between the cylindrical member and the cylindrical member due to aging (wear of the cylindrical member or cylindrical member, or wear of the lining between the cylindrical member and the cylindrical member), so the gap detection device and gap detection method disclosed herein can also be applied.

[0161] In the present disclosure, a program for causing the gap detection device to perform the above-described processing can be stored in the device's storage unit or other storage device. Alternatively, the program can be provided as a computer-readable, non-transitory recording medium (CD-ROM, USB memory, etc.) containing the program.

[0162] Description of Reference Numerals

[0163] 10 parallel linkage robot

[0164] 12 Basic Department

[0165] 14 movable parts

[0166] 16a connecting rod

[0167] 18A electric motor

[0168] 20a driving connecting rod

[0169] 22a driven connecting rod

[0170] 24a, 26a ball joints (spherical bearings)

[0171] 28 ball head

[0172] 30 shell

[0173] 32 lining

[0174] 34a Constraint plate

[0175] 36 control devices

[0176] 38 gap detection device

[0177] 40 Simulation Department

[0178] 42 Gap calculation unit

[0179] 44 Measurement Department

[0180] 46 Feature calculation unit

[0181] 48 Judgment Department

[0182] 50 Storage Department

[0183] 52 output unit

[0184] 53 Input unit

[0185] 80, 86 drive joints

[0186] 82, 88 follower joints

[0187] 84 parallel linkage robot

[0188] 90 five-section linkage robot.

Claims

1. A gap detection device for detecting a first gap between kinematic pair elements of a kinematic pair connected to a driven link in a robot, wherein: The robot comprises: a driving link driven by a motor; a plurality of driven links driven by the movement of the driving link; and a plurality of kinematic pairs connected to the plurality of driven links, It is characterized in that The gap detection device comprises: a measuring unit configured to measure a driving torque or a current value of the motor when the robot is actually moved along an arbitrary motion trajectory; a simulation unit configured to set an arbitrary second gap between the kinematic pair elements of the plurality of kinematic pairs, execute a simulation for causing the robot to move along a motion trajectory identical to the motion trajectory, and estimate a driving torque or current value of the motor; a feature quantity calculation unit that calculates a first feature quantity indicating a change in a value related to the driving torque or current value measured by the measurement unit and a second feature quantity indicating a change in a value related to the driving torque or current value estimated by the simulation unit; and A gap calculation unit calculates an index related to the first gap amount based on the first feature amount, the second feature amount, and the second gap amount.

2. The gap detection device according to claim 1, characterized in that: The gap calculation unit generates a mathematical model that relates the first feature quantity, the second feature quantity, the first gap quantity, and the second gap quantity, and calculates an index related to the first gap quantity based on the mathematical model.

3. The gap detection device according to claim 2, characterized in that: The mathematical model is a probabilistic model.

4. The gap detection device according to claim 3, characterized in that: The gap calculation unit calculates an index related to the first gap amount based on a probability distribution related to the probability model.

5. The gap detection device according to any one of claims 1 to 4, characterized in that: The feature quantity calculation unit calculates the first feature quantity and the second feature quantity through principal component analysis.

6. The gap detection device according to any one of claims 1 to 5, characterized in that: The gap calculation unit divides the time interval before and after the moment when the robot's motion trajectory changes. The measuring unit uses the same motion trajectory to measure the driving torque or current value in each divided time interval. The gap calculation unit determines the change in the gap in all the divided time intervals, and calculates the indicator related to the first gap value by adding these change amounts.

7. The gap detection device according to any one of claims 1 to 6, characterized in that: The gap detection device further includes a determination unit that determines that the gap of a kinematic pair, among the plurality of kinematic pairs, for which the index related to the first gap amount is equal to or greater than a predetermined reference value, is abnormal.

8. The gap detection device according to any one of claims 1 to 7, characterized in that: The driving link and the driven link constitute at least one closed-loop link.

9. A gap detection method for detecting a first gap between kinematic pair elements of a kinematic pair connected to a driven link in a robot, wherein: The robot comprises: a driving link driven by a motor; a plurality of driven links driven by the movement of the driving link; and a plurality of kinematic pairs connected to the plurality of driven links, It is characterized in that The gap detection method comprises: measuring the driving torque or current value of the motor when the robot is actually moved along an arbitrary motion trajectory; Setting an arbitrary second gap between the kinematic pair elements of the plurality of kinematic pairs, performing a simulation in which the robot moves along a motion trajectory identical to the motion trajectory, and estimating a driving torque or current value of the motor; calculating a first characteristic value indicating a change in a value related to the measured driving torque or current value, and a second characteristic value indicating a change in a value related to the estimated driving torque or current value; and An index related to the first gap amount is calculated based on the first feature amount, the second feature amount, and the second gap amount.

10. A gap detection device for detecting a first gap between kinematic pair elements of a kinematic pair connected to a second link in a robot, wherein: The robot comprises: a first link that can be driven by a motor; a second link connected to the first link; and one or more kinematic pairs connected to the second link. It is characterized in that The gap detection device comprises: a measuring unit configured to measure a first output value of the motor when the robot is actually moved along an arbitrary motion trajectory; a simulation unit that sets an arbitrary second gap amount between the kinematic pair elements of the plurality of kinematic pairs, executes a simulation of causing the robot to move based on the motion trajectory, and estimates a second output value of the motor; and A gap calculation unit calculates the first gap amount based on the change in the first output value, the change in the second output value, and the second gap amount.

Images