State monitoring device, state abnormality determination method, and state abnormality determination program

Abstract

A state monitoring device monitors a state of a robot that can reproduce a predetermined motion. A time series data acquisition section acquires time series data of a state signal, with a period from a timing of an acquisition start signal to a timing of an acquisition end signal as an object, the acquisition start signal being a signal that shows the start of acquisition of a state signal that reflects the state of the robot, and the acquisition end signal being a signal that shows the end of acquisition of the state signal. The time series data is stored in association with period information that shows the acquisition period and reproduction recognition information that determines the reproduced motion of the robot. A dissimilarity calculation section is configured to calculate a dissimilarity between reference data and comparison data, the reference data being data based on the time series data acquired by at least one run, and the comparison data being data based on the time series data acquired by a run later than the acquisition period of the reference data. A robot state evaluation section uses the dissimilarity as an evaluation quantity for evaluation of the state of the robot.

Description

Technical Field

[0001] This invention relates to a status monitoring device that monitors the status of a robot and supports its maintenance. Background Technology

[0002] If industrial robots are repeatedly used in factories and other similar environments, the various parts of the robot (e.g., mechanical components) will inevitably deteriorate. As the condition deteriorates further, it may eventually lead to robot failure. If a robot malfunctions and causes a prolonged production line shutdown, it will result in significant losses, making it imperative to perform maintenance (upkeep) before the robot fails. On the other hand, frequent maintenance is also difficult from the perspective of maintenance costs.

[0003] In order to enable maintenance at the appropriate time, a device for predicting the remaining lifespan of a robot's reducer and the like has been proposed. Patent Document 1 discloses such a robot maintenance support device.

[0004] The robot maintenance support device of Patent Document 1 is configured to: diagnose the future trend of the current command value based on the data of the current command value of the servo motor constituting the robot drive system, and determine the period before the current command value reaches a preset value based on the diagnosed trend.

[0005] [Existing technical documents]

[0006] [Patent Literature]

[0007] Patent Document 1: Japanese Patent Application Publication No. 2016-117148 Summary of the Invention

[0008] The technical problem that the invention aims to solve

[0009] In the configuration of Patent Document 1, diagnostic items for current command values ​​include an I2 monitor, duty cycle, and peak current. Among these items, some are effective in detecting early signs of robot malfunctions, while others are less effective. Therefore, there is a need for a new indicator that can accurately detect when a robot is approaching a malfunction.

[0010] In view of this, the main objective of the present invention is to effectively detect early signs of robot malfunctions.

[0011] Technical solutions used to solve the problem

[0012] As described above, the problem that this invention aims to solve is now explained below, along with the means used to solve the problem and their effectiveness.

[0013] According to a first aspect of the present invention, a state monitoring device with the following structure is provided. That is, this state monitoring device monitors the state of an industrial robot capable of reproducing predetermined actions. The state monitoring device includes a time-series data acquisition unit, a storage unit, a dissimilarity calculation unit, and a robot state evaluation unit. The time-series data acquisition unit acquires time-series data of the state signal, targeting a period from the timing of an acquisition start signal to the timing of an acquisition end signal. The acquisition start signal is a signal indicating the start of acquisition of the state signal reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signal. The storage unit stores the time-series data acquired by the time-series data acquisition unit, along with period information and reproduction identification information, establishing an association between the period information (information indicating the acquisition period of the time-series data) and the reproduction identification information (information determining the robot's reproduced actions at the time the time-series data was acquired). The dissimilarity calculation unit calculates the dissimilarity between reference data and comparison data. The reference data is data obtained from time-series data acquired by at least one reproduction action, and the comparison data is data obtained from the same reproduction action performed after the acquisition of the time-series data of the reference data. The robot state evaluation unit uses the dissimilarity calculated by the dissimilarity calculation unit as an evaluation metric for robot state evaluation.

[0014] According to a second aspect of the present invention, a state monitoring method with the following structure is provided. That is, in this state monitoring method, the state of an industrial robot capable of reproducing predetermined actions is monitored. This state monitoring method includes: a time-series data acquisition step; a storage step; a dissimilarity calculation step; and a robot state evaluation step. In the time-series data acquisition step, time-series data of the state signal is acquired, taking the period from the timing of an acquisition start signal to the timing of an acquisition end signal as the object. The acquisition start signal is a signal indicating the start of acquisition of the state signal reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signal. In the storage step, the time-series data acquired in the time-series data acquisition step, along with period information and reproduction identification information, are associated and stored. The period information displays information about the acquisition period of the time-series data, and the reproduction identification information determines the robot's reproduced actions at the time the time-series data was acquired. In the dissimilarity calculation step, the dissimilarity between the reference data and the comparison data is obtained. The reference data is based on the time-series data acquired from at least one reproduction action, and the comparison data is based on the time-series data acquired from the same reproduction action performed after the acquisition of the time-series data of the reference data. In the robot state evaluation step, the dissimilarity calculated in the dissimilarity calculation step is used as an evaluation metric to evaluate the robot's state.

[0015] According to a third aspect of the present invention, a state monitoring program with the following structure is provided. That is, in this state monitoring program, the state of an industrial robot capable of reproducing predetermined actions is monitored. In the state monitoring program, a computer executes a time-series data acquisition step, a storage step, a dissimilarity calculation step, and a robot state evaluation step. In the time-series data acquisition step, time-series data of the state signal is acquired, taking the period from the timing of the acquisition start signal to the timing of the acquisition end signal as the object. The acquisition start signal is a signal indicating the start of acquisition of the state signal reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signal. In the storage step, the time-series data acquired in the time-series data acquisition step, along with period information and reproduction identification information, are associated and stored. The period information displays information about the acquisition period of the time-series data, and the reproduction identification information determines the robot's reproduced actions at the time the time-series data was acquired. In the dissimilarity calculation step, the dissimilarity between the reference data and the comparison data is obtained. The reference data is based on the time-series data acquired from at least one reproduction action, and the comparison data is based on the time-series data acquired from the same reproduction action performed after the acquisition of the time-series data of the reference data. In the robot state evaluation step, the dissimilarity calculated in the dissimilarity calculation step is used as an evaluation metric to evaluate the robot's state.

[0016] This makes it easy to detect early signs of robot malfunction. Therefore, robot maintenance can be performed before malfunctions occur.

[0017] The benefits of invention

[0018] According to the present invention, it is possible to effectively detect early signs of robot malfunctions. Attached Figure Description

[0019] Figure 1 This is a perspective view showing the configuration of the robot according to the first embodiment of the present invention;

[0020] Figure 2 This is a block diagram illustrating the electrical structure of the robot and its status monitoring device;

[0021] Figure 3 It is a graph used to illustrate the timing of the trigger signals for acquiring time series data;

[0022] Figure 4 This is a schematic diagram showing the DTW algorithm;

[0023] Figure 5(a) is a graph showing the waveform of the current value without noise, and (b) is a graph showing the waveform of the current value with applied noise.

[0024] Figure 6 It is a graph that displays the shift of time series data based on the assessment results of DTW distance;

[0025] Figure 7 It is a chart that displays the shift of time series data based on the assessment results of PTP;

[0026] Figure 8 It is a chart that shows the shift of time series data based on the root mean square evaluation results;

[0027] Figure 9 It is a chart that displays the baseline data and specific examples of the data used for comparison;

[0028] Figure 10 This is a graph showing an example of a chart displaying DTW distance;

[0029] Figure 11 This is a graph showing an example of how the shift of a+kσ is displayed on a chart, relating to the DTW distance;

[0030] Figure 12 This is a graph showing an example of how the shift in the anomaly of the spatial competition theory is displayed on a chart regarding the DTW distance;

[0031] Figure 13 This is a flowchart used to illustrate the process of acquiring time series data;

[0032] Figure 14 This is a flowchart illustrating the process for calculating the DTW distance;

[0033] Figure 15 It is a chart illustrating the variation of the baseline data and the data used for comparison;

[0034] Figure 16 It is a graph showing the shift of time series data based on the assessment results of the second DTW distance;

[0035] Figure 17 It is a graph showing the shift of the third evaluation metric in time series data;

[0036] Figure 18 It is a graph showing the shift of the fourth evaluation metric in time series data;

[0037] Figure 19 This is a chart displaying an example of a warning message;

[0038] Figure 20 This is a schematic diagram illustrating the calculation of dissimilarity in the second embodiment; and

[0039] Figure 21 This is a schematic diagram illustrating the dissimilarity calculated in the third embodiment.

[0040] Explanation of reference numerals in the attached figures

[0041] 1 robot

[0042] 5. Status monitoring device

[0043] 51. Time Series Data Acquisition Department

[0044] 52 Storage Unit

[0045] 53. Time Series Data Evaluation Department (Dissimilarity Calculation Department)

[0046] 54 Robot Condition Assessment Department Detailed Implementation

[0047] The embodiments of the present invention will now be described with reference to the accompanying drawings. Figure 1 This is a perspective view showing the structure of robot 1 according to an embodiment of the present invention. Figure 2 This is a block diagram showing the electrical structure of robot 1 and status monitoring device 5.

[0048] The status monitoring device 5 of the present invention, for example, is applied to, for example, Figure 1 The robot shown is an industrial robot 1. Robot 1 performs tasks such as coating, cleaning, welding, and transporting workpieces. Robot 1 is implemented, for example, by a vertical articulated robot.

[0049] Below, refer to Figure 1 as well as Figure 2 The structure of robot 1 will be briefly explained.

[0050] Robot 1 has a rotating base 10, a multi-joint arm 11, and a wrist 12. The rotating base 10 is fixed to the ground (e.g., a factory floor). The multi-joint arm 11 has multiple joints. The wrist 12 is mounted on the front end of the multi-joint arm 11. An end effector 13 for working on a workpiece is mounted on the wrist 12.

[0051] like Figure 2 As shown, robot 1 has an arm drive device 21.

[0052] These drive devices may consist, for example, of actuators configured as servo motors and speed reducers. However, the configuration of the drive devices is not limited to the aforementioned configuration. Each actuator is electrically connected to the controller 90. The actuators operate in response to command values ​​input from the controller 90.

[0053] The driving force from each servo motor constituting the arm drive unit 21 is transmitted to each joint of the multi-joint arm 11, the rotating base 10, and the wrist 12 via a reducer. An encoder (not shown) is installed on each servo motor to detect its rotational position.

[0054] Robot 1 operates by reproducing actions recorded in the instruction manual. Controller 90 controls the actuator in a manner that Robot 1 reproduces a series of actions previously taught by the instructor. Teaching Robot 1 can be performed by the instructor operating a teaching device (without illustrations).

[0055] By teaching robot 1, a program is generated to move robot 1. A program is generated each time robot 1 is taught. If the taught actions of robot 1 are different, the program will also be different. By switching between multiple programs, the actions performed by robot 1 can be changed.

[0056] The controller 90 can be, for example, a known computer, which includes a CPU, ROM, RAM, auxiliary storage devices, etc. The auxiliary storage devices can be, for example, HDDs, SSDs, etc. Programs for the mobile robot 1 are stored in the auxiliary storage devices.

[0057] like Figure 1 As shown, the status monitoring device 5 is connected to the controller 90. The status monitoring device 5 acquires the current value and other information flowing in the actuator (servo motor) via the controller 90.

[0058] If the servo motor and its connected reducer malfunction, the current value of the servo motor will fluctuate. Therefore, this current value corresponds to a status signal reflecting the state of robot 1. The shift in the current value can be represented by repeatedly acquiring the current value at short time intervals and arranging multiple current values ​​in chronological order. Hereinafter, the data of the status signal values ​​arranged in a time sequence can also be referred to as time-series data.

[0059] The status monitoring device 5 can determine whether there are any abnormalities in the robot 1 by monitoring the acquired time-series data. In this embodiment, the status monitoring device 5 mainly focuses on the servo motors and reducers of each joint to determine whether there are any abnormalities. Here, "abnormality" refers to a situation where, although the robot has not yet reached a state of malfunction / failure, some kind of precursory condition has already occurred in the servo motor, reducer, or bearing.

[0060] like Figure 2 As shown, the state monitoring device 5 includes a time series data acquisition unit (time series data acquisition unit) 51, a storage unit 52, a time series data evaluation unit (non-similarity calculation unit) 53, a robot state evaluation unit 54, and a display unit 55.

[0061] The status monitoring device 5 is composed of a known computer, which includes a CPU, ROM, RAM, and auxiliary storage devices. The auxiliary storage devices may be, for example, HDDs or SSDs. Programs for evaluating the status of the robot 1 are stored in the auxiliary storage devices. Through the coordinated operation of this hardware and software, the computer can function as a time-series data acquisition unit 51, a storage unit 52, a time-series data evaluation unit 53, a robot status evaluation unit 54, etc.

[0062] The time-series data acquisition unit 51 acquires the time-series data. The time-series data acquisition unit 51 acquires time-series data for all the servo motors included in the arm drive device 21 of the robot 1. Furthermore, it acquires time-series data for each of the multiple servo motors (in other words, multiple reducers) configured in various parts of the robot 1.

[0063] In this embodiment, the status signal is a current value. This current value represents a measured value obtained by a sensor determining the magnitude of the current flowing through the servo motor. The sensor is mounted on a servo driver (not shown) that controls the servo motor. However, the sensor can also be installed separately from the servo driver for monitoring purposes. Alternatively, the current command value provided by the servo driver to the servo motor can be used as the status signal. The servo driver performs feedback control on the servo motor to make the current value close to the current command value. Therefore, from the perspective of detecting abnormalities in the servo motor or reducer, the current value and the current command value are almost identical.

[0064] The torque of a servo motor is proportional to the current. Therefore, torque values ​​or torque command values ​​can be used as status signals.

[0065] As a status signal, the target value of the servo motor's rotational position and the deviation from the actual rotational position obtained through the encoder (rotational position deviation) can also be used. Typically, the servo driver provides this deviation multiplied by a gain as a current command value to the servo motor. Thus, the shift in rotational position deviation shows a similar trend to the shift in the current command value. The actual rotational position of the servo motor can also be used as a status signal.

[0066] Whenever robot 1 reproduces the taught action, time series data acquisition unit 51 acquires time series data for each servo motor. However, it is also possible to acquire time series data only for, for example, the reproduced action once or several times a day, instead of acquiring time series data for all reproduced actions.

[0067] The time-series data acquisition unit 51 acquires time-series data of the current values ​​flowing through each servo motor during the period between the timing of receiving an acquisition start signal and the timing of receiving an acquisition end signal. The acquisition start signal and acquisition end signal are, for example, output by the controller 90.

[0068] Figure 3 The graph shows an example of the current flowing through a servo motor at a certain joint when robot 1 is performing a reproduced motion. For example... Figure 3 As shown, before the program for reproducing the action is executed, the current value of the servo motor is zero. At this time, because the electromagnetic brake (not shown) moves in each joint, the posture of the multi-joint arm 11, etc., can be maintained.

[0069] Next, the program for reproducing the actions of robot 1 begins. Simultaneously, the brake is released, and almost simultaneously, current begins to flow to the servo motor. At this moment, control is applied by stopping the output shaft of the servo motor. After a certain amount of time is allowed to stabilize the angle of the servo motor's output shaft, the servo motor begins to rotate. Thus, the actions of robot 1 essentially begin.

[0070] After the brake is released, and slightly ahead of the start of the servo motor's rotation, the controller 90 will acquire a start signal and output it to the status monitoring device 5 (and, the time series data acquisition unit 51).

[0071] When the robot 1 has completed all its actions, the servo motor is controlled to stop rotating. After the servo motor stops rotating, the controller 90 acquires an end signal and outputs it to the time series data acquisition unit 51 just before the program ends.

[0072] The storage unit 52 is, for example, configured by the auxiliary storage device. The storage unit 52 stores the time-series data acquired by the time-series data acquisition unit 51.

[0073] In this embodiment, the time-series data is data formed by arranging multiple current values ​​obtained through repeated detection at short time intervals in chronological order. Therefore, the current values ​​in the time-series data are sampled values. The time interval for detecting the current values ​​(sampling interval) is, for example, several milliseconds. The sampling interval for the current values ​​can be consistent with or different from the control cycle of robot 1. Figure 3 In the chart, the time series data corresponds to the shift of current values ​​from the timing of acquiring the start signal to the timing of acquiring the end signal.

[0074] Furthermore, the acquisition of start or end signals can be substantially achieved using other existing signals, without the need for special preparation for measurements such as time series data acquisition. For example, the decline of the "reproduced operation" signal, which is mainly used for safety purposes, can be used as the acquisition end signal.

[0075] The storage unit 52 stores period information of the acquisition period of the display data in association with the time series data. The period information can be, for example, a timestamp that displays the date and time of receiving the acquisition start signal.

[0076] Similarly, the storage unit 52 stores reproduction identification information associated with the time-series data. This reproduction identification information shows the actions performed by robot 1 when the time-series data was acquired. The series of actions of robot 1 are defined by a program for reproducing the actions. Therefore, the reproduction identification information can be, for example, a program number or program name, which is assigned to uniquely identify the program for reproducing the actions.

[0077] The following describes the storage of time-series data in detail. Using the file system of the OS of the computer, i.e., the status monitoring device 5, a folder named after the program number or program name is created in the auxiliary storage device. This folder automatically stores files containing time-series data acquired when the program with that number is executed. These files, for example, record time-series data related to the current values ​​of the servo motors of the six joints using comma-separated values ​​(CSV). The file name contains a timestamp string. This allows for the establishment of an association between period information, reproduction identification information, and time-series data. However, this method is only an example; other methods can also be used to establish this association.

[0078] The time series data acquisition unit 51 repeatedly acquires time series data over a long period of time during the operation of the robot 1 in a factory or other similar environment. Each time time series data is acquired, a file is saved.

[0079] Time series data can be divided into baseline data, which is initially collected, and comparison data collected subsequently. The time series data evaluation unit 53 evaluates the comparison data by comparing the baseline data and the comparison data. Based on the evaluation result, the time series data evaluation unit 53 outputs an evaluation quantity for evaluating the state of robot 1.

[0080] As reference data, for example, time-series data acquired by the time-series data acquisition unit 51 is used when the robot 1 is taught an action and then reproduces that action for the first time. However, it is also possible to have the robot 1 perform the reproduced action multiple times, including the initial one, and to calculate the average time-series data by averaging the time-series data from multiple times, and then use the average time-series data as reference data. The multiple times can be consecutive numbers such as the 1st, 2nd, 3rd, etc., or non-consecutive numbers such as the 1st, 4th, 6th, etc.

[0081] The timing of acquiring baseline data does not have to be strictly initial (first time), as long as it is substantially initial (first time). For example, robot 1 can be made to reproduce the action once or multiple times as a break-in run after the action instruction, and then the baseline data can be acquired.

[0082] As baseline data, it is also possible to use data that has been filtered from the initial time series data or from time series data containing the average of multiple initial values.

[0083] This type of data is used as comparison data, derived from time series data that serves as the source of the baseline data. It can be used directly as comparison data, or it can be filtered time series data.

[0084] In the following explanation, the reference data and comparison data may sometimes be referred to as "raw" without filtering and "filtered" after filtering.

[0085] Filtering is performed to remove noise, for example, from time series data. Since data filtering is a well-known technique, detailed explanations are omitted, but any filter, such as a moving average filter or a CR circuit analog filter, can be used.

[0086] The baseline data is prepared for each reproduced action of robot 1 (in other words, each program). Although it is also possible to obtain baseline data for all taught actions, it is also possible to obtain baseline data only for actions that are primarily performed by robot 1 or are simple actions that are not primary actions and are reproduced once a day.

[0087] The time series data evaluation unit 53 uses the DTW algorithm to compare the baseline data with the comparison data. DTW is short for Dynamic Time Warping.

[0088] Here, we will briefly explain the DTW algorithm. The DTW algorithm is used to calculate the similarity between two time series data. A key feature of the DTW algorithm is that it allows for nonlinear scaling of the time series data along the time axis when calculating similarity. Therefore, the DTW algorithm can achieve results that closely approximate human intuition regarding the similarity of time series data.

[0089] The time series data evaluation unit 53 outputs the DTW distance (dissimilarity) as an evaluation metric, which shows the magnitude of the difference between the comparison data and the benchmark data.

[0090] Reference Figure 4 Explain the principle of the DTW algorithm. Multiple (m) current values ​​from the reference data are arranged sequentially along the first horizontal axis in time series. Furthermore, multiple (n) current values ​​from the comparison data are arranged along the second vertical axis.

[0091] Next, m×n cells are defined in a matrix arrangement on a plane defined by the horizontal and vertical axes. Each cell (i, j) displays the correspondence between the i-th current value in the reference data and the j-th current value in the comparison data. Where 1≤i≤m, 1≤j≤n.

[0092] Each cell (i, j) is associated with a value that displays the difference between the i-th current value in the reference data and the j-th current value in the comparison data. In this embodiment, the absolute value of the difference between the i-th current value and the j-th current value is stored in each cell in an associated manner.

[0093] Time series data evaluation unit 53 is used to obtain data from the time series data located at... Figure 4 The twisted path (path) from the starting cell at the bottom left corner of the matrix to the ending cell at the top right corner.

[0094] The starting cell (1, 1) is equivalent to establishing a correspondence between the first current value in the time series (i.e., the 1st) among the m current values ​​in the reference data and the first current value in the time series (i.e., the 1st) among the n current values ​​in the comparison data.

[0095] The termination point cell (m, n) is equivalent to establishing a correspondence between the last current value (i.e., the mth current value) in the time series of the m current values ​​in the reference data and the last current value (i.e., the nth current value) in the time series of the n current values ​​in the comparison data.

[0096] In the m×n matrices constructed as described above, consider the path from the starting cell to the ending cell according to the rules in [1] and [2] below. [1] You can only move to adjacent cells in the vertical, horizontal, or diagonal direction. [2] You cannot move in the direction of returning the reference data or in the direction of returning the comparison data.

[0097] The connection of individual cells in this way is called a path or twist path. A twist path shows how to establish a correspondence between m current values ​​in the reference data and n current values ​​in the comparison data. From another perspective, a twist path shows how to stretch two time series data along the time axis.

[0098] The twisted path from the starting cell to the ending cell can consider multiple possible paths. The time series data evaluation unit 53 seeks the twisted path that has the smallest sum of the numerical values ​​of the differences established with the cells passed through (in this embodiment, the absolute value of the difference between the i-th current value and the j-th current value is established with each cell).

[0099] The twist path can also be referred to as the optimal twist path. Furthermore, the sum of the values ​​of each cell in the optimal twist path is called the DTW distance. Alternatively, the average value obtained by dividing the DTW distance by the number of cells traversed by the optimal twist path can be used to evaluate the comparative data instead of the DTW distance. Alternatively, the value obtained by dividing the DTW distance by the number of features m or n of any time series data can be used to evaluate the comparative data instead of the average value.

[0100] When m and n are large, a large number of twisted paths can be considered. Therefore, assuming that all possible twisted paths are considered, the computational cost of finding the optimal twisted path increases exponentially. To solve this problem, the time series data evaluation unit 53 of this embodiment uses the DP matching method (dynamic programming) to obtain the DTW distance. DP is short for Dynamic Programming.

[0101] Since the DP matching method is a well-known method, it will be explained simply below. Considering the aforementioned rules, then... Figure 4 As indicated by the arrows, when focusing on a specific cell, only three cells from the moving source can move towards that cell. These three cells are the cell adjacent to the cell of interest on the left, the cell adjacent to the cell of interest on the bottom, and the cell adjacent to the cell of interest on the bottom left. The time series data evaluation unit 53 of this embodiment uses these features to calculate the DTW distance.

[0102] The specific explanation is as follows. Time series data evaluation department 53, firstly, regarding... Figure 4In the matrix, for each cell in the bottom row (1,1), (2,1), ..., (m,1), sum the values ​​of all cells contained in the path from the starting cell (1,1) to that cell. This sum can also be called the cell summation. Furthermore, when considering one or more paths from the starting cell (1,1) to a certain cell, the minimum value of the cell summation can also be called the minimum cell summation.

[0103] If we consider the aforementioned rule, the path from the starting cell (1,1) to each cell in the lower row of the matrix can only be a straight line. Therefore, we can say that the sum of all the cells is the minimum sum of all cells. Regarding the cells in the lower row of the matrix, the sum of all cells (the minimum sum of all cells) can be easily calculated by sequentially adding the values ​​of the cells starting from the starting cell (1,1).

[0104] Next, focus on the second line from the bottom.

[0105] First, consider cell (1, 2). As the cell before the path reaches cell (1, 2), it only has the starting cell (1, 1). Therefore, it is easy to find the sum of the values ​​of each cell along the path from the starting cell (1, 1) to cell (1, 2). There is only one path from the starting cell (1, 1) to cell (1, 2). Therefore, the sum obtained is the minimum sum of the values ​​of cells with respect to cell (1, 2).

[0106] Next, consider cell (2,2). As cells before the path reaches cell (2,2), there are three possibilities: starting cell (1,1), cell (2,1), and cell (1,2). The minimum sum of the three cells can be obtained through previous calculations. The time series data evaluation unit 53 selects the cell with the smallest minimum sum from the three cells, adds this value to the value of cell (2,2), and determines the resulting value as the minimum sum of the three cells (2,2). The time series data evaluation unit 53 associates the position of the cell with the smallest minimum sum with the position of cell (2,2) and stores it.

[0107] Next, consider cell (3,2). Cells preceding cell (3,2) can be categorized into three types: cell (2,1), cell (2,2), and cell (3,1). Similar to the case of cell (2,2), the time series data evaluation unit 53 selects the cell with the smallest sum of values ​​from the three cells that can be considered, adds this value to the value of cell (3,2), and determines the resulting value as the minimum sum of values ​​for cell (3,2). The time series data evaluation unit 53 then associates the position of the cell with the smallest sum of values ​​among the three cells with the position of cell (3,2) and stores it.

[0108] The time series data evaluation unit 53 performs the same processing sequentially one by one until the last cell (m, 2) of the second row from the bottom.

[0109] The time series data evaluation unit 53 repeats the aforementioned processing sequentially, row by row, for the 3rd, 4th, ... rows counting from the bottom. Once all cells have been calculated, the minimum sum of values ​​for the terminal cell (m, n) is obtained. This minimum sum of values ​​represents the sum of values ​​for the cells passed through along the path from the starting cell (1, 1) to the terminal cell (m, n) (the optimal twisted path). The time series data evaluation unit 53 outputs the minimum sum of values ​​for the terminal cell (m, n) as the DTW distance.

[0110] In cases where the optimal twist path needs to be determined in addition to the DTW distance, it is sufficient to sequentially trace the positions of the cells that have the smallest sum of individual cells from the terminal cell (m, n) to the starting cell (1, 1). The above description is an example of calculation for each row of the matrix, but it is also possible to calculate for each column.

[0111] Because the dynamic matching method does not fully consider all possible twist paths, its accuracy is not perfect. However, by using the dynamic matching method, the computational cost can be significantly reduced, and the optimal twist path with sufficient practical accuracy can be obtained.

[0112] DTW distance is the degree of dissimilarity between two time series data (in other words, two signal waveforms). As is well known, the DTW algorithm does not need to reflect the differences between the two waveforms in the time axis direction, but it can obtain the dissimilarity (DTW distance) of the two waveforms in a way that can well reflect the differences in waveform amplitude, etc.

[0113] If two waveforms differ only in their periods or phases, then the DTW distance is zero. This means that the two waveforms are evaluated as identical.

[0114] Next, consider Figure 5 The two current waveforms shown are shown. Figure 5 (a) is the basic waveform. Figure 5 (b) A waveform in which noise is intentionally added to the basic waveform. Figure 5 The waveforms shown are for illustrative purposes only and are not actually obtained from a servo motor.

[0115] Figure 5 The basic waveform of (a) is shown by 2cos(ωt), where the amplitude is 2. Figure 5The waveform in (b) is equivalent to... Figure 5 (a) The basic waveform has a current value of -1.732A and a noise of -2.266A applied to the portion where the current value is -1A, and a noise of +2.5A applied to the portion where the current value is -1A.

[0116] If we take the plotted points shown in each chart as time series data, the DTW distance between two time series data points is calculated to be 4.766. This value is equal to the sum of the absolute values ​​of the applied noise, i.e., -2.266A and +2.5A. Thus, the noise is directly reflected in the DTW distance obtained by the DTW algorithm. In addition, both positive and negative noise are reflected in the DTW distance without canceling each other out.

[0117] The time series data evaluation unit 53 calculates the DTW distance between the reference data and the comparison data, and outputs the acquired DTW distance.

[0118] The baseline data and comparison data include time series data of the servo motors for 6 joints. The time series data evaluation unit 53 calculates the DTW distance for each joint (in other words, each servo motor).

[0119] If the actions performed by robot 1 are different, the shift in the current value of the servo motor will naturally be different as well. Taking this factor into account, the DTW distance is calculated by the time series data evaluation unit 53 for each reproduced action. More specifically, the time series data evaluation unit 53 calculates the DTW distance between comparison data and reference data related to the same reproduced action. As a result, time series data can be compared appropriately.

[0120] The DTW distance output by the time series data evaluation unit 53 is stored in the storage unit 52. At this time, various pieces of information are associated with the DTW distance and stored in the storage unit 52. The information stored along with the DTW distance includes period information and reproduction identification information. Period information is information indicating the acquisition period of the comparison data, such as a timestamp. Reproduction identification information is, for example, the program number or program name used to determine the reproduction action of robot 1 at the time the comparison data was acquired.

[0121] It is also possible to use specific signals contained in the program as identification information, in place of program number or program name. For example, if a program only contains signals such as the aforementioned start signal and / or end signal, it can be distinguished from other programs based on the presence or absence of such signals.

[0122] The robot condition assessment unit 54 uses the DTW distance stored in the storage unit 52 to assess the condition of the robot 1 (each part of each part) (robot condition assessment process and robot condition assessment steps).

[0123] The robot state assessment unit 54 uses the DTW distance acquired for each servo motor to assess the state of that servo motor.

[0124] Assuming that over time, the waveform corresponding to the current current value in the time series data gradually deviates from the waveform corresponding to the initial current value in the time series data, the DTW distance can be considered as a quantification of the degree of this deviation. The robot state evaluation unit 54 uses the DTW distance to determine whether there is an abnormality in each servo motor of the six joints. Furthermore, based on the DTW distance over time, it is possible to predict when the servo motor will become malfunctioning / unresponsive in the future.

[0125] Figure 6 This displays an example of the shift in DTW distance associated with a servo motor of robot 1. Each point shows the calculated DTW distance.

[0126] exist Figure 6 In the example, the operation of robot 1 began around August 2019. For about six months after its start of operation, the increasing trend of the DTW distance was slight, but around February 2020, the increasing trend became significantly stronger. Around June 2020, the servo motors became inoperable.

[0127] There were two instances of abnormally high DTW distance values ​​around March and April 2020. Considering that the servo motor subsequently became inoperable within a short period of time, these two points can be considered to indicate precursors to the servo motor's malfunction.

[0128] Although there are many signs that a servo motor or other device may become inoperable, such as noise interference caused by reduced shielding of the control cable, mechanical vibration caused by bearing deterioration, and squeaking noise caused by wear on the gear teeth of the reducer, etc. Figure 6 The two points shown as "precursors of anomalies" exhibited the phenomena described above, affecting the time series data of the current values. This can be considered the cause of the abnormal changes in the DTW distance.

[0129] exist Figure 7 as well as Figure 8 In the comparison example shown in the text, the text is presented in the context of... Figure 6 Under the same conditions, the shift of other indicators that represent changes in time series data.

[0130] Figure 7 The PTP (Peak to Peak) shown is the value obtained by subtracting the low-peak current value from the high-peak current value of the current waveform. This PTP is one type of "peak current" referred to in Patent Document 1. Figure 7As shown, using the PTP method, it is extremely difficult to capture early signs of anomalies in the data from March to April 2020.

[0131] Figure 8 The I2 shown is the root mean square value of the current. This I2 corresponds to "I2" as described in Patent Document 1. Figure 8 As shown, using the I2 method, it is extremely difficult to detect early signs of anomalies in the data from March to April 2020.

[0132] PTP and I2 are values ​​obtained through statistical processing of time series data. Figure 7 as well as Figure 8 The reason why precursors to anomalies cannot be found in the charts can be attributed to the loss of certain features in the time series data that indicate precursors to anomalies during the statistical processing. On the other hand, Figure 6 The DTW distance shown is a value obtained by comparing time series data without performing statistical processing. Therefore, it can be said that the DTW distance is a value contained in time series data that is easily affected by the characteristics of precursors to anomalies.

[0133] Figure 9 The display shows baseline and comparative data for a certain servo motor. The comparative data was obtained approximately 10 months later than the baseline data. It can be observed that some discrepancies arise between the two time series data over time.

[0134] Figure 9 The range shown as A1 in the current waveform corresponds to the action of lowering the arm of robot 1. Within this range, the current value of the comparison data is slightly lower than the reference data. This can be attributed to mechanical wear caused by aging and other factors acting in the direction of the arm's descent, thus assisting the servo motor and causing a reduction in the current flowing through it.

[0135] Figure 9 The range shown as A2 in the current waveform corresponds to the movement of holding the arm of robot 1. The range shown as A3 corresponds to the movement of raising the arm of robot 1. Within the ranges A2 and A3, the current value of the comparative data is slightly greater than the reference data. This can be attributed to mechanical wear caused by aging, etc., acting against the direction of holding or raising the arm, thus increasing the current flowing through the servo motor.

[0136] Thus, mechanical losses, which can be a precursor to faults, can act in the direction that increases or decreases the current value in the time series data, depending on the situation. For example, when using I2, the effects of the two directions are evaluated in a way that partially cancels each other out. However, by using the DTW distance of this embodiment, the dissimilarity of the time series data can be obtained while fully considering the effects of both directions.

[0137] In this embodiment, the time-series data used as the source of both reference data and comparison data is limited to data from... Figure 3 The period shown is from the start signal to the end signal. If the time series data contains, for example, data corresponding to... Figure 3 In the case of waveforms indicating the timing of brake opening, the dissimilarity increases, which may lead to misjudgment as an anomaly. In this embodiment, by appropriately setting the timing of acquiring the start signal and the end signal, it is possible to evaluate only the period that is substantially meaningful from the viewpoint of detecting precursors of faults during the current value transition.

[0138] The robot's state assessment unit 54 can also determine, through appropriate calculations, whether or not the condition exists. Figure 6 Points displayed as "early signs of anomalies" are identified. The calculation method is varied, and can be performed as follows: Focusing on a specific point, calculate the average DTW distance *a* and standard deviation *σ* of the N nearest points to that point. N is 2 or greater. Furthermore, if the DTW distance of the point of interest exceeds *a+kσ*, that point is considered a precursor to an anomaly (in other words, the servo motor or reducer is malfunctioning). Here, *k* is a suitable positive value. Alternatively, a threshold can be set; if the distance exceeds this threshold, it is evaluated as an "early sign of anomalies." Euclidean distance, described later, can also be used instead of DTW distance. Figure 11 The chart shows the shift in the value of a+3σ regarding the DTW distance. The chart also calculates the mean 'a' and standard deviation 'σ' by referring to the 10 most recent points.

[0139] Alternatively, it can be done as follows: Focus on a point, and calculate the average DTW distance *a* and standard deviation *σ* of the N points closest to that point. Then, using the DTW distance *x* of the point of interest, calculate the value of (xa)² / σ², known as the anomaly in Hotelling's Law, and graph it. This value can also be multiplied by a coefficient as needed. If this value exceeds a predetermined threshold, it can also be assessed as a "precursor to anomaly." Hotelling's Law is a method for anomaly detection based on a statistical model. It is known that when the value *x* follows a normal distribution, the value of the anomaly follows a chi-square distribution with 1 degree of freedom. Utilizing this property, in Hotelling's Law, deviation values ​​are detected by comparing the anomaly value with an appropriate threshold. The Euclidean distance, etc., described later, can also be used instead of the DTW distance as *x*. Figure 12 The graph shows the distance to DTW and the shift in the value of (xa)² / σ². If we consider... Figure 12 The chart, if the threshold for anomaly is set to, for example, around 10, can be suitable for detecting early signs of anomalies.

[0140] Whether calculating a+kσ or (xa)2 / σ2, it is possible to treat all points from the beginning of the evaluation as objects instead of only treating the nearest N points as objects.

[0141] DTW distance or Euclidean distance, for example, has the advantage of easily capturing characteristics by taking the distance between corresponding points of two waveforms. By using these distances as objects, standard deviation or outlier can be calculated, for example, providing easily understandable charts for fault monitoring workers.

[0142] The display unit 55 is capable of displaying corresponding... Figure 6 The display unit 55 is composed of a display device such as a liquid crystal display. The operator monitors for any abnormal points on the chart that deviate from the usual trend of the DTW distance. The operator can flexibly utilize this information to appropriately formulate future maintenance plans.

[0143] If the robot condition assessment unit 54 determines that one or more servo motors are malfunctioning, the display unit 55 displays information about the malfunction. For example, a warning message can be displayed on the display unit 55, indicating the cause of the malfunction. This allows the operator to anticipate the situation.

[0144] The robot state assessment unit 54 is able to calculate and obtain a trend line that shows the trend of the DTW distance over time. Figure 10 This shows an example of a trend line. A trend line can be obtained as an approximation of a group of points on a chart. An approximate straight line can be obtained using the well-known least squares method, but the method is not limited.

[0145] The robot condition assessment unit 54 calculates the date and time when the trend line obtained as described above will reach a predetermined lifespan threshold. By calculating the period from now until that date and time, the robot condition assessment unit 54 can predict the remaining lifespan of servo motors, etc.

[0146] On the display unit 55, with Figure 10 The trend lines are displayed in an overlay format. Operators can confirm the trend lines by referring to the screen on display unit 55, enabling them to plan maintenance appropriately.

[0147] When calculating the trend line of DTW distance, it is not necessary to approximate all DTW distances. For example, if robot 1 repeats the same reproduction action multiple times a day, robot state evaluation unit 54 can obtain a representative value of DTW distance each day and obtain a trend line in an approximate representative value manner.

[0148] This representative value can be, for example, the median of multiple DTW distances acquired on a given day. Using a median has the advantage of being less affected even when DTW distances are highly irregular. Alternatively, an average value can be used instead of the median.

[0149] As a representative value, the maximum value among multiple DTW distances acquired on the same day can also be used. In this case, it is easy to reflect the precursors of anomalies on the trend line.

[0150] Next, refer to Figure 13 The flowchart illustrates in detail an example of the time series data acquisition and processing of the time series data acquisition unit 51 in this embodiment.

[0151] like Figure 13 As shown, if the robot 1 begins to reproduce its actions, the controller 90 that controls the actions of the robot 1 will activate the status monitoring device 5 (step S101).

[0152] Next, the controller 90 obtains the program number of the program that causes the robot 1 to perform the reproduced action (step S102). In the program for reproducing the action, the path of the mobile end effector 13 is defined by recording multiple points (1), (2), ..., (E) traversed.

[0153] The controller 90 further obtains the current rotational position of each servo motor (step S103).

[0154] The controller 90 outputs the program number obtained in step S102 to the status monitoring device 5 (step S104).

[0155] Then, after completing the preparations for operation, the controller 90 will receive a start signal and output it to the status monitoring device 5 (step S105). The preparations for operation include... Figure 3 The description of the brake opening is as follows.

[0156] The controller 90 actuates each servo motor in such a way that the end effector 13 moves toward the point (1) (step S106).

[0157] When the end effector 13 reaches point (1), the controller 90 actuates each servo motor in such a way that the end effector 13 moves toward the next point (2) (step S107). The same action is repeated until the end effector 13 is finally moved to the final point (E) (step S108).

[0158] After the end effector 13 reaches point (E), the controller 90 will acquire an end signal and output it to the status monitoring device 5 (step S109), and end the procedure of reproducing the action.

[0159] Next, the processing of the status monitoring device 5 will be explained.

[0160] The controller 90 performs step S101 processing and starts the status monitoring device 5. After starting, the status monitoring device 5 obtains the program number output by the controller 90 of robot 1 in step S104 (step S201).

[0161] Next, the status monitoring device 5 goes into standby mode before receiving the acquisition start signal output by the controller 90 (step S202).

[0162] Upon receiving the start signal, the status monitoring device 5 acquires each current value flowing through the current of each servo motor used to drive the robot 1 (step S203). This step corresponds to the time series data acquisition step (time series data acquisition process).

[0163] The process of step S203 is repeated until the status monitoring device 5 receives the acquisition end signal output by the controller 90 (step S204). The current value is acquired from the servo motor of each of the 6 joints and these current values ​​are stored in the RAM of the status monitoring device 5.

[0164] Upon receiving the acquisition completion signal, the status monitoring device 5 creates a folder in the auxiliary storage device with the program number obtained in step S201 as the name (step S205). If the folder has already been created, the processing of step S205 is skipped.

[0165] The status monitoring device 5 saves the current value data acquired between the timing of receiving the start acquisition signal and the timing of receiving the end acquisition signal as a file in the folder created in step S205 (step S206). The file contains a large amount of current value data arranged in chronological order. The name of the saved file can be set to include a timestamp showing the date and time of receiving the start acquisition signal. Then, the status monitoring device 5 substantially stops processing. Steps S205 and S206 correspond to the storage step (storage process).

[0166] Next, refer to Figure 14 This document provides a detailed example of a flowchart for calculating the DTW distance between the baseline data and the comparison data. The DTW distance calculation shown in this flowchart corresponds to the dissimilarity calculation step (dissimilarity calculation process).

[0167] Figure 14 The process shown is performed once daily, for example, after the factory's operating hours have ended. When starting... Figure 14 During processing, the time series data evaluation unit 53 retrieves the earliest collected time series data from the folder containing the program number corresponding to a specific reproduction action (step S301). The collection date and time can be easily obtained from the timestamp contained in the file name. The time series data retrieved in step S301 corresponds to the reference data.

[0168] Next, the time series data evaluation unit 53 acquires the time series data collected from the same folder based on the reproduced actions of the robot 1 on that day (step S302). The time series data acquired in step S302 corresponds to the comparison data.

[0169] The time series data evaluation unit 53 calculates the DTW distance between the acquired baseline data and the comparison data (step S303).

[0170] Then, the time series data evaluation unit 53 determines whether there is another set of comparison data collected on the same day (step S304). If there is another set of comparison data on the same day, the comparison data is processed in the same way as in steps S302 and S303.

[0171] If the DTW distance is calculated between the reference data and all comparison data collected on the same day, the process proceeds to step S305. In step S305, the time series data evaluation unit 53 obtains the median value of the DTW distance calculated from the comparison data collected on the same day and stores the median value in the storage unit 52.

[0172] Next, the robot state assessment unit 54 generates a graph showing the daily progression of the median value of the DTW distance obtained in step S305 (step S306). This graph is displayed on the display unit 55. Figure 10 The example shown includes a chart. The chart displays a trend line calculated by an approximate straight line from the intermediate value calculated by the robot condition assessment unit 54, and a baseline displaying the lifespan threshold.

[0173] Next, the variation of DTW distance will be explained.

[0174] As mentioned above, the DTW distance is calculated between the original baseline data and the original comparison data. Figure 15 (a) Shows an example with the original baseline data. Figure 15 (b) An example with original comparison data is shown. Hereinafter, the DTW distance in this case can also be referred to as the first DTW distance. The first DTW distance corresponds to the first dissimilarity and the first evaluation metric.

[0175] However, as illustrated below, it is also possible to calculate the DTW distance without using the original baseline data and the original comparison data.

[0176] It can also calculate the DTW distance between the filtered baseline data and the filtered comparison data. Figure 15 (c) Showing an example of filtered baseline data. Figure 15 (d) Shows an example of comparison data after filtering. Figure 15 In the example, a moving average filter is used. Below, the DTW distance in this case can also be referred to as the second DTW distance. The second DTW distance corresponds to the second dissimilarity.

[0177] Also able to Figure 15 (c) shows the filtered baseline data and Figure 15 The original comparison shown in (b) calculates the DTW distance between the data. Hereinafter, the DTW distance in this case can also be referred to as the third DTW distance. The third DTW distance corresponds to the third dissimilarity.

[0178] The time series data evaluation unit 53 can also calculate the difference between the first DTW distance and the second DTW distance obtained as described above, and output it as an evaluation quantity (second evaluation quantity). In addition, it can also calculate the difference between the third DTW distance and the second DTW distance, and output it as an evaluation quantity (third evaluation quantity).

[0179] Figure 16The display shows the shift based on the evaluation results of the second DTW distance. Since high-frequency components are removed by filtering the baseline and comparison data, the second DTW distance can be considered to more clearly reflect changes in gearbox efficiency and mechanical losses caused by aging. Figure 16 As shown, although the value of the second DTW remains basically unchanged along the way, it shows an accelerating trend in the latter half.

[0180] Figure 17 The display shows a shift in the third evaluation metric. As mentioned above, this third evaluation metric is the difference between the third DTW distance and the second DTW distance. The third DTW distance is the DTW distance between the filtered reference data and the original comparison data. Therefore, this third evaluation metric can be considered to more clearly reflect the high-frequency components (e.g., the pulsed portion of the waveform) contained in the comparison data, such as vibrations and squeaking.

[0181] exist Figure 17 The charts clearly show an abnormal increase in the value of the third assessment metric around March and April 2020. This anomaly is highly likely to be used as a precursor to a failure.

[0182] By using filtered data for at least one of the baseline data and the comparison data, it is possible to relatively emphasize some features in the comparison data that indicate the precursors of a fault. This means that using at least one of the second DTW distance, the third DTW distance, the second evaluation quantity, and the third evaluation quantity is effective.

[0183] Although it does not determine dissimilarity between the baseline data and the comparison data, it can still achieve... Figure 15 (c) shows the original comparison data and Figure 15 (d) shows the DTW distance obtained between the filtered comparison data. This DTW distance can also be referred to as the fourth DTW distance. The fourth DTW distance corresponds to the fourth dissimilarity and the fourth evaluation metric.

[0184] Figure 18 The data shows a shift in the fourth assessment measure. Although not very obvious, an unusual increase in the value of the fourth assessment measure can be observed around March to April 2020.

[0185] Since this fourth assessment metric also relatively emphasizes noise components such as squeaking sounds included in the comparative data, depending on the situation, there are cases where it can effectively capture early signs of a fault.

[0186] Whenever the robot state evaluation unit 54 obtains the first DTW distance, for example, Figure 10As shown, a graph plotted for the first DTW distance is displayed on the display unit 55. The trend line obtained by the robot state evaluation unit 54 is also displayed on this graph. Simultaneously, the robot state evaluation unit 54 monitors evaluation quantities other than the first DTW distance (e.g., Figure 17 Does the third evaluation metric (shown) contain outliers? If outliers are found, such as... Figure 19 As shown, the robot's condition assessment unit 54 displays warning messages (alarms) on a screen identical to the chart, alerting the operator. This allows the operator to appropriately assess the situation.

[0187] As explained above, the state monitoring device 5 of this embodiment monitors the state of an industrial robot capable of reproducing predetermined actions. The state monitoring device 5 includes a time-series data acquisition unit 51, a storage unit 52, a time-series data evaluation unit 53, and a robot state evaluation unit 54. The time-series data acquisition unit 51 acquires time-series data of state signals, targeting the period from the timing of an acquisition start signal to the timing of an acquisition end signal. The acquisition start signal is a signal indicating the start of acquisition of a state signal reflecting the state of the robot 1 (specifically, the current value of a servo motor), and the acquisition end signal is a signal indicating the end of acquisition of the state signal. The storage unit 52 stores the time-series data acquired by the time-series data acquisition unit 51, associated with period information and a program number. The period information displays information about the acquisition period of the time-series data, and the program number is used to determine the reproduced action of the robot 1 when the time-series data was acquired. The time-series data evaluation unit 53 calculates the DTW distance, which serves as the dissimilarity between reference data and comparison data. The reference data is data based on time-series data acquired from at least one reproduced action. The comparison data is based on time series data acquired after the acquisition of the reference data, obtained from the same reproducible action. The robot state evaluation unit 54 uses the DTW distance calculated by the time series data evaluation unit 53 as the evaluation quantity (first evaluation quantity) for evaluating the state of robot 1.

[0188] Therefore, it is easy to detect the early signs of failure in robot 1, so that maintenance can be carried out before the failure occurs.

[0189] Furthermore, in the status monitoring device 5 of this embodiment, the time series data evaluation unit 53 performs the following processing. Specifically, it can consider the following: m current values ​​extracted from reference data are sequentially arranged on the horizontal axis according to the time series, and n current values ​​extracted from comparison data are sequentially arranged on the vertical axis according to the time series. The correlation between each current value is represented by a matrix consisting of m×n cells, and the difference between the corresponding sampled values ​​is correlated with each cell. The cell corresponding to the current value in the reference data arranged on the horizontal axis that corresponds to the first timing in the time series, and the cell corresponding to the current value in the comparison data arranged on the vertical axis that corresponds to the first timing in the time series, is designated as the starting cell. The cell corresponding to the current value in the reference data arranged on the horizontal axis that corresponds to the last timing in the time series, and the cell corresponding to the current value in the comparison data arranged on the vertical axis that corresponds to the last timing in the time series, is designated as the ending cell. Under the aforementioned conditions, the time series data evaluation unit 53 determines the path from the starting cell to the ending cell that minimizes the sum of differences corresponding to the cells traversed. The time series data evaluation unit 53 uses the sum (DTW distance) or average of the differences corresponding to each cell traversed by the determined path as the dissimilarity.

[0190] Thus, it is possible to sensitively capture precursors of anomalies in the comparative data while simultaneously determining dissimilarity.

[0191] In addition, in the status monitoring device 5 of this embodiment, reference data can also be obtained by averaging the time series data acquired multiple times.

[0192] In this case, measurement errors can be suppressed.

[0193] In addition, in the status monitoring device 5 of this embodiment, reference data can also be obtained by filtering the time series data.

[0194] In this case, dissimilarity can be obtained by relatively emphasizing the form of appropriate components (e.g., high-frequency components) in the abnormal precursors contained in the comparative data.

[0195] In addition, in the status monitoring device 5 of this embodiment, comparison data can also be obtained by filtering the time series data.

[0196] In this case, dissimilarity can be obtained by relatively emphasizing the form of appropriate components (e.g., low-frequency components) in the abnormal precursors contained in the comparative data.

[0197] Furthermore, in the status monitoring device 5 of this embodiment, the time series data evaluation unit 53 can also be configured to calculate the first DTW distance and the second DTW distance. The first DTW distance is the DTW distance between the reference data without time series data filtering and the comparison data without time series data filtering. The second DTW distance is the DTW distance between the reference data with time series data filtering and the comparison data with time series data filtering. The time series data evaluation unit 53 outputs the difference between the first DTW distance and the second DTW distance as a second evaluation quantity.

[0198] Furthermore, in the status monitoring device 5 of this embodiment, the time series data evaluation unit 53 can also be configured to calculate a third DTW distance and a second DTW distance. The third DTW distance is the DTW distance between the reference data (after filtering the time series data) and the comparison data (after not filtering the time series data). The second DTW distance is the DTW distance between the reference data (after filtering the time series data) and the comparison data (after filtering the time series data). The time series data evaluation unit 53 outputs the difference between the third DTW distance and the second DTW distance as a third evaluation quantity.

[0199] Furthermore, in the status monitoring device 5 of this embodiment, the time series data evaluation unit 53 can also be configured to obtain a fourth DTW distance. The fourth DTW distance is the DTW distance between comparison data that has undergone filtering of the time series data and comparison data that has not undergone filtering of the time series data. The time series data evaluation unit 53 outputs the fourth DTW distance as a fourth evaluation quantity.

[0200] The above assessment metrics can easily capture early signs of anomalies in the comparative data.

[0201] Furthermore, in the status monitoring device 5 of this embodiment, the time series data is data about at least one of the following: current value, current command value, torque value, torque command value, rotational position deviation, and actual rotational position of the servo motor driving the robot 1.

[0202] Therefore, it is easy to assess the status of the servo motor and its surrounding components.

[0203] Furthermore, in the condition monitoring device 5 of this embodiment, the robot condition assessment unit 54 generates trend line data, which shows the trend of the DTW distance obtained by the time series data assessment unit 53 changing over time. The robot condition assessment unit 54 calculates the remaining lifespan based on the trend line.

[0204] Thus, the operator can obtain information to appropriately plan the maintenance time of robot 1.

[0205] Furthermore, the status monitoring device 5 of this embodiment includes a display unit 55 capable of displaying trend lines. For example... Figure 19 As shown, when the evaluation quantity, which is different from the DTW distance, meets the preset conditions, the display unit 55 can display an alarm simultaneously with the trend line.

[0206] As a result, operators can grasp the situations that need attention as early as possible.

[0207] Next, a second embodiment of the present invention will be described. Furthermore, in the description of this embodiment, components that are the same as or similar to those in the previous embodiment can be given the same reference numerals in the drawings, and their descriptions will be omitted.

[0208] The time series data evaluation unit 53 of this embodiment can replace the DTW distance and obtain the dissimilarity of the two data by calculating the Euclidean distance between the reference data and the comparison data.

[0209] The time series data evaluation unit 53 moves one waveform of the waveform of the reference data and the waveform of the comparison data in multiple stages along the time axis, and calculates and obtains the Euclidean distance in each stage.

[0210] Since the Euclidean distance is well known, it is briefly explained below. The Euclidean distance D between time series data a and time series data b is given by setting the length of the time series data to m, and is shown in the following equation (1).

[0211] [Formula 1]

[0212]

[0213] Figure 20 The example shown demonstrates shifting the comparison data by one stage in the negative direction and one stage in the positive direction of the time axis, for a total shift of two stages. The Euclidean distance is calculated in three stages, including no shift. The amount of shift along the time axis for each stage, as well as the number of stages shifted, can be arbitrarily set.

[0214] In calculating the Euclidean distance, the waveform can be substantially shifted by horizontally shifting the correspondence between the m current values ​​in the reference data and the m current values ​​in the comparison data.

[0215] The time series data evaluation unit 53 calculates the minimum value of the multiple Euclidean distances obtained. The time series data evaluation unit 53 outputs this minimum value as the dissimilarity between the baseline data and the comparison data.

[0216] In the first embodiment, the DTW distance substantially eliminates the influence of large time delays or significant distortions along the time axis between two time series data points when assessing dissimilarity. However, assessing the delays and distortions along the time axis of time series data can sometimes be effectively used to detect precursors of anomalies. In this case, when using the irregular Euclidean distance of this embodiment, the permissible time delay is limited to the maximum range of phased lateral shifts of the time series data along the time axis. Therefore, a slight time delay can be tolerated, and a dissimilarity that appropriately reflects a certain degree of significant delays and distortions along the time axis of the time series data can be obtained. In other words, on the one hand, phase deviations as measurement errors can be ignored, and on the other hand, large phase deviations such as those from deterioration of display servo motors can be appropriately detected.

[0217] Furthermore, such as Figure 20 As shown, shifting time series data relative to the time axis results in objectless combinations. If the number of objectless elements (current values) is large, the Euclidean distance decreases as the number of object-containing elements decreases. To address this issue, one could consider dividing the Euclidean distance obtained above by the number of object-containing elements and using the average value as the dissimilarity assessment. Alternatively, a correction could be made by adding elements with values ​​equal to the values ​​obtained at the start or end points as objects to the objectless elements at both ends of the time series data.

[0218] As explained above, the state monitoring device 5 of this embodiment moves at least one waveform of the reference data waveform and the comparison data waveform in multiple stages along the time axis, while calculating the Euclidean distance between the two waveforms. The minimum value of the Euclidean distance is taken as the dissimilarity.

[0219] Therefore, it is possible to obtain a high-precision dissimilarity that reflects distortions such as those along the time axis in time series data.

[0220] Next, a third embodiment of the present invention will be described. Furthermore, in the description of this embodiment, components that are the same as or similar to those in the previous embodiment will be given the same reference numerals in the drawings and their descriptions will be omitted.

[0221] The time series data evaluation unit 53 of this embodiment calculates the difference between the i-th current value among the m current values ​​contained in the reference data and the ip-th to (i+p-th)-th current values ​​among the m current values ​​contained in the comparison data. Figure 21 This is a schematic diagram showing the case where p=2, with the difference in current values ​​displayed as a value with an attached underline.

[0222] Next, the time series data evaluation unit 53 investigates which of the current values ​​from the ipth to the (i+pth)th comparison data is closest to the ith current value of the reference data. Figure 21 In the example, among the current values ​​of the comparison data from the (i-2)th to the (i+2)th, the one with the smallest absolute difference relative to the i-th current value of the reference data is the (i-1)th, which has a difference of -0.2. The i-th current value of the reference data corresponds to the first sample value, while the (i-1)th current value of the comparison data corresponds to the second sample value.

[0223] The time series data evaluation unit 53 calculates the Euclidean distance as the dissimilarity. At this time, the difference between the i-th current value of the reference data and the current value of the closest current value of the reference data among the (i-2)th to (i+2)th current values ​​of the comparison data is regarded as the absolute value of the difference between the i-th current value of the reference data and the i-th current value of the comparison data (|a| in equation (1)). i -b i |).

[0224] The dissimilarity obtained in this way can tolerate slight time delays and distortions, and can appropriately reflect larger delays and distortions in time series data along the time axis. In other words, it corrects for phase deviations as measurement errors and detects larger phase deviations that indicate degradation.

[0225] As explained above, the state monitoring device 5 calculates the Euclidean distance between the waveforms of the reference data (which acquires multiple sample values) and the waveforms of the comparison data. The difference between the i-th current value (the first current value) of one waveform and the second current value (the closest to the first current value among the current values ​​from the ip-th to the (i+p-th)-th (where i and p are integers greater than or equal to 1) current values ​​of the other waveform is considered the difference between the corresponding sample values ​​of the two waveforms. The state monitoring device 5 uses the Euclidean distance as a measure of dissimilarity.

[0226] Therefore, it is possible to obtain a high-precision reflection of the dissimilarity of time series data, such as distortion in the time axis direction.

[0227] The preferred embodiments of the present invention have been described above, but the configuration can be modified as follows, for example.

[0228] The status monitoring device 5 can also acquire time-series data reflecting the status of the robot 1 from the controller 90 of the robot 1 via a communication line such as a network, instead of being directly connected to the robot 1. In this case, the controller 90 acquires and saves the current value in real time during the reproduction of the action, and transmits the time-series data of the current value, along with the program number, the date and time of acquisition of the current value, and the information of the servo motor identification, to the status monitoring device 5 through batch processing or the like.

[0229] The DTW distance can also be obtained without using the DP matching method.

[0230] The status monitoring device 5 can also be integrated into the controller 90 without being separate from it. Alternatively, the status monitoring device 5 can be implemented using the computer of the controller 90 of the robot 1, instead of having a separate computer that functions as a CPU, ROM, RAM, auxiliary storage device, etc. In this case, for example, the display unit 55 can be configured as part of the robot 1's teaching display and can switch between displaying the warning information.

Claims

1. A state monitoring device for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that... The system comprises: a time-series data acquisition unit that acquires time-series data of state signals, targeting a period from the timing of an acquisition start signal to the timing of an acquisition end signal, wherein the acquisition start signal is a signal indicating the start of acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signals; a storage unit that stores the time-series data acquired by the time-series data acquisition unit, along with period information and reproduction identification information, wherein the period information displays information about the acquisition period of the time-series data, and the reproduction identification information determines the robot's reproduction action when the time-series data was acquired; a dissimilarity calculation unit that calculates the dissimilarity between reference data and comparison data, wherein the reference data is data of the time-series data acquired from at least one reproduction action, and the comparison data is data of the time-series data acquired from the same reproduction action performed after the acquisition of the reference data; and a robot state evaluation unit that uses the dissimilarity calculated by the dissimilarity calculation unit as an evaluation metric for robot state evaluation. The dissimilarity calculation unit sequentially arranges m sample values ​​extracted from the reference data on the first axis in a time sequence, and sequentially arranges n sample values ​​extracted from the comparison data on the second axis in a time sequence. The correlation between each sample value is represented by a matrix consisting of m×n cells. Furthermore, when establishing a correlation between the differences between corresponding sample values ​​and each cell, the unit calculates the correlation between the sample value corresponding to the earliest timing in the time sequence from the sample values ​​of the reference data arranged on the first axis, and the sample value of the comparison data arranged on the second axis. From the starting cell corresponding to the sample value corresponding to the earliest timing in the time series, to the ending cell corresponding to the sample value corresponding to the last timing in the time series in the sample values ​​of the reference data configured on the first axis, and the ending cell corresponding to the sample value corresponding to the last timing in the time series in the sample values ​​of the comparison data configured on the second axis, the path with the minimum sum of differences corresponding to the cells passed is obtained, and the sum or average of the differences corresponding to each cell passed by the obtained path is taken as the dissimilarity.

2. A state monitoring device for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that... The system comprises: a time-series data acquisition unit that acquires time-series data of state signals, targeting a period from the timing of an acquisition start signal to the timing of an acquisition end signal, wherein the acquisition start signal is a signal indicating the start of acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signals; a storage unit that stores the time-series data acquired by the time-series data acquisition unit, along with period information and reproduction identification information, wherein the period information displays information about the acquisition period of the time-series data, and the reproduction identification information determines the robot's reproduction action when the time-series data was acquired; a dissimilarity calculation unit that calculates the dissimilarity between reference data and comparison data, wherein the reference data is data of the time-series data acquired from at least one reproduction action, and the comparison data is data of the time-series data acquired from the same reproduction action performed after the acquisition of the reference data; and a robot state evaluation unit that uses the dissimilarity calculated by the dissimilarity calculation unit as an evaluation metric for robot state evaluation. The dissimilarity calculation unit sequentially arranges m sample values ​​extracted from the reference data on the first axis in a time sequence, and sequentially arranges n sample values ​​extracted from the comparison data on the second axis in a time sequence. The correlation between each sample value is represented by a matrix consisting of m×n cells. Furthermore, when establishing a correlation between the differences between corresponding sample values ​​and each cell, the unit calculates the correlation between the sample value corresponding to the earliest timing in the time sequence from the sample values ​​of the reference data arranged on the first axis, and the sample value of the comparison data arranged on the second axis. From the starting cell corresponding to the sampled value corresponding to the earliest timing in the time series, to the ending cell corresponding to the sampled value of the reference data configured on the first axis that corresponds to the last timing in the time series, and the sampled value of the comparison data configured on the second axis that corresponds to the last timing in the time series, the path with the minimum sum of differences corresponding to the cells traversed is obtained, and the sum or average of the differences corresponding to each cell traversed by the obtained path is taken as the dissimilarity. The robot state assessment unit generates trend line data, which shows the trend of the dissimilarity obtained by the dissimilarity calculation unit changing over time, and the robot state assessment unit calculates the remaining lifespan based on the trend line.

3. A state monitoring device for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that... The system comprises: a time-series data acquisition unit that acquires time-series data of state signals, targeting a period from the timing of an acquisition start signal to the timing of an acquisition end signal, wherein the acquisition start signal is a signal indicating the start of acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signals; a storage unit that stores the time-series data acquired by the time-series data acquisition unit, along with period information and reproduction identification information, wherein the period information indicates the period during which the time-series data was acquired, and the reproduction identification information determines the robot's reproduction action when the time-series data was acquired; and a dissimilarity calculation unit that calculates the dissimilarity between reference data and comparison data, wherein the reference data is data of the time-series data acquired from at least one reproduction action, and the comparison data is data of the time-series data acquired from the same reproduction action performed after the acquisition of the reference data. And a robot state evaluation unit, which uses the dissimilarity calculated by the dissimilarity calculation unit as an evaluation metric for robot state evaluation. The dissimilarity calculation unit moves at least one waveform of the waveform of the reference data and the waveform of the comparison data in multiple stages along the time axis, calculates the Euclidean distance between the two waveforms respectively, and takes the minimum value of the Euclidean distance as the dissimilarity.

4. A state monitoring device for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that... The system comprises: a time-series data acquisition unit that acquires time-series data of state signals, targeting a period from the timing of an acquisition start signal to the timing of an acquisition end signal, wherein the acquisition start signal is a signal indicating the start of acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signals; a storage unit that stores the time-series data acquired by the time-series data acquisition unit, along with period information and reproduction identification information, wherein the period information displays information about the acquisition period of the time-series data, and the reproduction identification information determines the robot's reproduction action when the time-series data was acquired; a dissimilarity calculation unit that calculates the dissimilarity between reference data and comparison data, wherein the reference data is data of the time-series data acquired from at least one reproduction action, and the comparison data is data of the time-series data acquired from the same reproduction action performed after the acquisition of the reference data; and a robot state evaluation unit that uses the dissimilarity calculated by the dissimilarity calculation unit as an evaluation metric for robot state evaluation. The dissimilarity calculation unit calculates the Euclidean distance between the waveform of the reference data (which acquires multiple sample values) and the waveform of the comparison data, by taking the difference between the i-th sample value (i.e., the first sample value) of one waveform and the second sample value (from the ip-th to the (i+p-th)-th sample value of another waveform, which is closest to the first sample value) as the difference between the sample values ​​of the two waveforms, and uses the Euclidean distance as the dissimilarity.

5. The status monitoring device according to any one of claims 1 to 4, wherein, The baseline data is obtained by averaging the time series data acquired multiple times.

6. A state monitoring device for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that... The system comprises: a time-series data acquisition unit that acquires time-series data of state signals, targeting a period from the timing of an acquisition start signal to the timing of an acquisition end signal, wherein the acquisition start signal is a signal indicating the start of acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signals; a storage unit that stores the time-series data acquired by the time-series data acquisition unit, along with period information and reproduction identification information, wherein the period information indicates the period during which the time-series data was acquired, and the reproduction identification information determines the robot's reproduction action when the time-series data was acquired; and a dissimilarity calculation unit that calculates the dissimilarity between reference data and comparison data, wherein the reference data is data of the time-series data acquired from at least one reproduction action, and the comparison data is data of the time-series data acquired from the same reproduction action performed after the acquisition of the reference data. And a robot state evaluation unit, which uses the dissimilarity calculated by the dissimilarity calculation unit as an evaluation metric for robot state evaluation. The dissimilarity calculation unit calculates a first dissimilarity and a second dissimilarity, and outputs the difference between the first dissimilarity and the second dissimilarity as a second evaluation metric. The first dissimilarity is the dissimilarity between the reference data and the comparison data without filtering the time series data, and the second dissimilarity is the dissimilarity between the reference data and the comparison data with filtering the time series data.

7. A state monitoring device for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that... The system comprises: a time-series data acquisition unit that acquires time-series data of state signals, targeting a period from the timing of an acquisition start signal to the timing of an acquisition end signal, wherein the acquisition start signal is a signal indicating the start of acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signals; a storage unit that stores the time-series data acquired by the time-series data acquisition unit, along with period information and reproduction identification information, wherein the period information displays information about the acquisition period of the time-series data, and the reproduction identification information determines the robot's reproduction action when the time-series data was acquired; a dissimilarity calculation unit that calculates the dissimilarity between reference data and comparison data, wherein the reference data is data of the time-series data acquired from at least one reproduction action, and the comparison data is data of the time-series data acquired from the same reproduction action performed after the acquisition of the reference data; and a robot state evaluation unit that uses the dissimilarity calculated by the dissimilarity calculation unit as an evaluation metric for robot state evaluation. The dissimilarity calculation unit obtains a third dissimilarity and a second dissimilarity, and outputs the difference between the third dissimilarity and the second dissimilarity as a third evaluation metric. The third dissimilarity is the dissimilarity between the benchmark data after filtering the time series data and the comparison data after not filtering the time series data. The second dissimilarity is the dissimilarity between the benchmark data after filtering the time series data and the comparison data after filtering the time series data.

8. A state monitoring device for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that... The system comprises: a time-series data acquisition unit that acquires time-series data of state signals, targeting a period from the timing of an acquisition start signal to the timing of an acquisition end signal, wherein the acquisition start signal is a signal indicating the start of acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signals; a storage unit that stores the time-series data acquired by the time-series data acquisition unit, along with period information and reproduction identification information, wherein the period information indicates the period during which the time-series data was acquired, and the reproduction identification information determines the robot's reproduction action when the time-series data was acquired; and a dissimilarity calculation unit that calculates the dissimilarity between reference data and comparison data, wherein the reference data is data of the time-series data acquired from at least one reproduction action, and the comparison data is data of the time-series data acquired from the same reproduction action performed after the acquisition of the reference data. And a robot state evaluation unit, which uses the dissimilarity calculated by the dissimilarity calculation unit as an evaluation metric for robot state evaluation. The dissimilarity calculation unit obtains a fourth dissimilarity and outputs the fourth dissimilarity as a fourth evaluation metric. The fourth dissimilarity is the dissimilarity between the comparison data after the time series data has been filtered and the comparison data without the time series data.

9. The status monitoring device according to any one of claims 1 to 4, 6 to 8, wherein, The time-series data is data about at least one of the following: current value, current command value, torque value, torque command value, rotational position deviation, and actual rotational position of the motor driving the robot.

10. The status monitoring device according to any one of claims 3 to 4, 6 to 8, wherein, The robot state assessment unit generates trend line data, which shows the trend of the dissimilarity obtained by the dissimilarity calculation unit changing over time, and the robot state assessment unit calculates the remaining lifespan based on the trend line.

11. The status monitoring device according to claim 10, wherein, The system includes a display unit capable of displaying the trend line. When an evaluation quantity that is different from the dissimilarity meets a preset condition, the display unit can simultaneously display an alarm along with the trend line.

12. A state monitoring method for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that... The system includes: a time-series data acquisition step, which acquires time-series data of state signals from the timing of the acquisition start signal to the timing of the acquisition end signal, wherein the acquisition start signal is a signal that indicates the start of the acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal that indicates the end of the acquisition of state signals; a storage step, which establishes an association between the time-series data acquired in the time-series data acquisition step and period information and reproduction identification information, wherein the period information indicates the acquisition period of the time-series data, and the reproduction identification information determines the robot's reproduction action when the time-series data was acquired; and a dissimilarity calculation step, which calculates the dissimilarity between reference data and comparison data, wherein the reference data is data of the time-series data acquired from at least one reproduction action, and the comparison data is data of the time-series data acquired from the same reproduction action performed after the acquisition of the time-series data of the reference data; And a robot state assessment process, which uses the dissimilarity calculated in the dissimilarity calculation process as an assessment metric to assess the robot's state. In the dissimilarity calculation process, m sample values ​​extracted from the reference data are sequentially arranged in time series on the first axis, and n sample values ​​extracted from the comparison data are sequentially arranged in time series on the second axis. A matrix composed of m×n cells represents the correlation between each sample value. Furthermore, when establishing a correlation between the differences between corresponding sample values ​​and each cell, the sample value corresponding to the earliest timing in the time series from the sample values ​​of the reference data arranged on the first axis, and the sample value from the comparison data arranged on the second axis... The path is calculated from the starting cell corresponding to the sampled value corresponding to the earliest timing in the time series, to the ending cell corresponding to the sampled value of the reference data configured on the first axis corresponding to the last timing in the time series, and the path corresponding to the sampled value of the comparison data configured on the second axis corresponding to the last timing in the time series. The path with the minimum sum of differences corresponding to the cells passed through is then used as the dissimilarity.

13. A state monitoring method for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that... The system includes: a time-series data acquisition step, which acquires time-series data of state signals from the timing of the acquisition start signal to the timing of the acquisition end signal, wherein the acquisition start signal is a signal that indicates the start of the acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal that indicates the end of the acquisition of state signals; a storage step, which establishes an association between the time-series data acquired in the time-series data acquisition step and period information and reproduction identification information, wherein the period information indicates the acquisition period of the time-series data, and the reproduction identification information determines the robot's reproduction action when the time-series data was acquired; and a dissimilarity calculation step, which calculates the dissimilarity between reference data and comparison data, wherein the reference data is data of the time-series data acquired from at least one reproduction action, and the comparison data is data of the time-series data acquired from the same reproduction action performed after the acquisition of the time-series data of the reference data; And a robot state assessment process, which uses the dissimilarity calculated in the dissimilarity calculation process as an assessment metric to assess the robot's state. In the dissimilarity calculation process, m sample values ​​extracted from the reference data are sequentially arranged in time series on the first axis, and n sample values ​​extracted from the comparison data are sequentially arranged in time series on the second axis. A matrix composed of m×n cells represents the correlation between each sample value. Furthermore, when establishing a correlation between the differences between corresponding sample values ​​and each cell, the sample value corresponding to the earliest timing in the time series from the sample values ​​of the reference data arranged on the first axis, and the sample value from the comparison data arranged on the second axis... The path is calculated from the starting cell corresponding to the sampled value of the earliest timing in the time series, to the ending cell corresponding to the sampled value of the reference data configured on the first axis that corresponds to the last timing in the time series, and the path corresponding to the sampled value of the comparison data configured on the second axis that corresponds to the last timing in the time series. The path with the minimum sum of differences corresponding to the cells traversed by this path is then used as the dissimilarity. In the robot condition assessment process, trend line data is generated, which shows the trend of the dissimilarity obtained by the dissimilarity calculation process changing over time. In the robot condition assessment process, the remaining lifespan is calculated based on the trend line.

14. A state monitoring program for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that: The computer executes time-series data acquisition, storage, dissimilarity calculation, and robot state evaluation steps. In the time-series data acquisition step, the period from the timing of the acquisition start signal to the timing of the acquisition end signal is used as the object to acquire time-series data of state signals. The acquisition start signal is a signal indicating the start of acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signals. In the storage step, the time-series data acquired in the time-series data acquisition step is associated with period information and reproduction identification information and stored. The period information is... The display shows information about the acquisition period of the time series data, and the reproduction identification information is information about the robot's reproduction action when the time series data was acquired; in the dissimilarity calculation step, the dissimilarity between the reference data and the comparison data is calculated, where the reference data is data from the time series data acquired by at least one reproduction action, and the comparison data is data from the time series data acquired by the same reproduction action performed after the acquisition of the reference data; and in the robot state evaluation step, the dissimilarity calculated in the dissimilarity calculation step is used as an evaluation metric to evaluate the robot's state. In the dissimilarity calculation step, m sample values ​​extracted from the reference data are sequentially arranged in time series on the first axis, and n sample values ​​extracted from the comparison data are sequentially arranged in time series on the second axis. A matrix consisting of m×n cells represents the correlation between each sample value. Furthermore, when establishing a correlation between the differences between corresponding sample values ​​and each cell, the sample value corresponding to the earliest timing in the time series from the sample values ​​of the reference data arranged on the first axis, and the sample value from the comparison data arranged on the second axis... The path is calculated from the starting cell corresponding to the sampled value corresponding to the earliest timing in the time series, to the ending cell corresponding to the sampled value of the reference data configured on the first axis corresponding to the last timing in the time series, and the path corresponding to the sampled value of the comparison data configured on the second axis corresponding to the last timing in the time series. The path with the minimum sum of differences corresponding to the cells passed through is then used as the dissimilarity.

15. A state monitoring program for monitoring the state of an industrial robot capable of reproducing predetermined actions, characterized in that: The computer executes time-series data acquisition, storage, dissimilarity calculation, and robot state evaluation steps. In the time-series data acquisition step, the period from the timing of the acquisition start signal to the timing of the acquisition end signal is used as the object to acquire time-series data of state signals. The acquisition start signal is a signal indicating the start of acquisition of state signals reflecting the robot's state, and the acquisition end signal is a signal indicating the end of acquisition of the state signals. In the storage step, the time-series data acquired in the time-series data acquisition step is associated with period information and reproduction identification information and stored. The period information is... The display shows information about the acquisition period of the time series data, and the reproduction identification information is information about the robot's reproduction action when the time series data was acquired; in the dissimilarity calculation step, the dissimilarity between the reference data and the comparison data is calculated, where the reference data is data from the time series data acquired by at least one reproduction action, and the comparison data is data from the time series data acquired by the same reproduction action performed after the acquisition of the reference data; and in the robot state evaluation step, the dissimilarity calculated in the dissimilarity calculation step is used as an evaluation metric to evaluate the robot's state. In the dissimilarity calculation step, m sample values ​​extracted from the reference data are sequentially arranged in time series on the first axis, and n sample values ​​extracted from the comparison data are sequentially arranged in time series on the second axis. A matrix consisting of m×n cells represents the correlation between each sample value. Furthermore, when establishing a correlation between the differences between corresponding sample values ​​and each cell, the sample value corresponding to the earliest timing in the time series from the sample values ​​of the reference data arranged on the first axis, and the sample value from the comparison data arranged on the second axis... The path is calculated from the starting cell corresponding to the sampled value of the earliest timing in the time series, to the ending cell corresponding to the sampled value of the reference data configured on the first axis that corresponds to the last timing in the time series, and the path corresponding to the sampled value of the comparison data configured on the second axis that corresponds to the last timing in the time series. The path with the minimum sum of differences corresponding to the cells traversed by this path is then used as the dissimilarity. In the robot state assessment step, trend line data is generated, which shows the trend of the dissimilarity obtained by the dissimilarity calculation step as it changes over time. In the robot state assessment step, the remaining lifespan is calculated based on the trend line.

Images