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

Abstract

To collect data for predicting remaining life in a process where a robot is operated by executing a program for actual work.SOLUTION: An operating state of a robot is determined, and an operating-state time being a time during which the operating state continues is measured. A basic-data acquisition count is calculated by dividing the operating-state time by a basic-data upper-limit length. From position time-series data of each axis corresponding to the operating-state time, basic data are acquired by the basic-data acquisition count. When the position time-series data of each axis that are repeatedly reproduced match the basic data, time-series data of state signals of each axis reflecting a state of the robot in a matched time section are acquired. An evaluation value is calculated from the time-series data. A trend line representing a tendency in which the evaluation value changes over time is created, and a timing at which the trend line reaches a life threshold is obtained as a predicted life timing.SELECTED DRAWING: Figure 6

Description

[Technical Field]

[0001] The present disclosure relates primarily to lifespan prediction for robots. [Background technology]

[0002] When industrial robots are repeatedly operated in factories, it is inevitable that the various parts of the robot (for example, mechanical components) will deteriorate. As this situation progresses, the robot will eventually break down. If a robot breaks down and the line is stopped for an extended period of time, this will result in significant losses, so there is a strong demand for carrying out maintenance (preservation) before a robot breaks down. However, carrying out frequent maintenance is difficult due to factors such as maintenance costs.

[0003] In order to perform maintenance at an appropriate time, a device has been proposed for predicting the remaining life of a robot's reducer, etc. Patent Document 1 discloses this type of robot maintenance support device.

[0004] The robot maintenance support device of Patent Document 1 is configured to diagnose the future change trend of the current command value based on data on the current command value of the servo motor that constitutes the robot drive system, and determine the period until the current command value reaches a preset value based on the diagnosed change trend. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2016-117148 A Summary of the Invention [Problem to be solved by the invention]

[0006] In order to predict the remaining life using a device such as that disclosed in Patent Document 1, it is necessary to actually operate the robot and collect data for evaluation (for example, the aforementioned current command value).

[0007] One approach is to pre-program the robot to perform specific operations for data collection. The robot executes the program periodically or irregularly, allowing it to reproduce a specific motion pattern for evaluation. However, this approach requires the robot to perform special operations that differ from the actual work, which inevitably reduces operational efficiency.

[0008] The present disclosure has been made in consideration of the above circumstances, and its main purpose is to collect data for predicting remaining lifespan in the process of operating a robot by executing a program for actual work. [Means for solving the problem]

[0009] The problem to be solved by the present disclosure is as described above. Next, the means for solving this problem and the effects thereof will be described.

[0010] According to an aspect of the present disclosure, a status monitoring device having the following configuration is provided. Specifically, this status monitoring device monitors the status of a robot capable of reproducing predetermined actions. The status monitoring device includes a motion state time measurement unit, a basic data processing unit, a search processing unit, a time-series data acquisition unit, an evaluation value calculation unit, and a lifespan estimation unit. The motion state time measurement unit determines the motion state of the robot and measures the motion state time, which is the time during which the motion state of the robot continues. The basic data processing unit calculates the number of basic data acquisitions by dividing the motion state time by a preset basic data upper limit length, acquires the number of basic data acquisitions from position time-series data of each axis of the motion state of the robot corresponding to the motion state time, and stores at least one basic data. The search processing unit acquires the position time-series data of each axis that is repeatedly reproduced and determines whether or not it matches the stored basic data. The time-series data acquisition unit acquires time-series data of a state signal of each axis that reflects the state of the robot during a time period that matches the basic data. The evaluation value calculation unit acquires an evaluation value for evaluating the state of the robot based on the time-series data acquired by the time-series data acquisition unit. The lifespan estimation unit creates a trend line that represents the tendency of the evaluation value to change over time, and determines the timing at which the trend line reaches a predetermined lifespan threshold as a predicted lifespan timing.

[0011] This makes it possible to appropriately obtain basic data for predicting the remaining life from position time-series data when the robot is actually operating by executing a program for an actual task. [Effects of the Invention]

[0012] According to the present disclosure, data for predicting the remaining life can be collected in the process of running a program for an actual task and operating a robot. [Brief explanation of the drawings]

[0013] [Figure 1] FIG. 1 is a perspective view showing a configuration of a robot according to an embodiment of the present disclosure. [Figure 2] FIG. 2 is a block diagram showing an outline of the electrical configuration of the robot and the state monitoring device. [Figure 3] 10 is a graph showing an example of time-series data of current values. [Figure 4] 10 is a graph illustrating the transition of evaluation values ​​and trend lines obtained from time-series data of current values. [Figure 5] 10 is a flowchart illustrating a process of acquiring a reference pattern from position time-series data. [Figure 6] 10 is a graph illustrating a process of extracting a time interval from position time-series data to obtain a reference pattern. [Figure 7] 10 is a flowchart mainly illustrating a pattern matching process for searching for a reference pattern from position time-series data. [Figure 8] FIG. 4 is a schematic diagram illustrating a pattern matching process. [Figure 9] 10A and 10B are schematic diagrams illustrating the average error and maximum error of position time series data in a time interval in which a reference pattern is matched. [Figure 10] 10 is a graph showing an example of time-series data of current values ​​that change as the motion pattern of the robot is changed. [Figure 11] 6 is a graph illustrating a first method for consistently processing evaluation values ​​before and after a change in the robot's motion pattern. [Figure 12] 10 is a graph illustrating processing when the robot's motion pattern is changed twice in the first method. [Figure 13] Graph illustrating the second method. [Figure 14] Graph illustrating the third method. [Figure 15] Graph illustrating the fourth method. [Figure 16] Graph illustrating the fifth method. [Figure 17] Graph illustrating the sixth method. [Figure 18] Graph illustrating the seventh method. [Figure 19] Graph illustrating the eighth method. [Figure 20] Graph illustrating the ninth method. [Figure 21] Graph illustrating the tenth method. [Figure 22] Graph illustrating the eleventh method. [Figure 23] Graph illustrating the twelfth method. DETAILED DESCRIPTION OF THE INVENTION

[0014] Next, an embodiment of the present disclosure will be described with reference to the drawings. Fig. 1 is a perspective view showing the configuration of a robot 1 according to an embodiment of the present disclosure. Fig. 2 is a block diagram showing the electrical configuration of the robot 1 and a status monitoring device 5. Fig. 3 is a graph showing an example of time-series data of current values. Fig. 4 is a graph illustrating the transition and trend line of an evaluation value obtained from the time-series data of current values.

[0015] A condition monitoring device 5 according to the present disclosure is applied to, for example, a robot (industrial robot) 1 as shown in Fig. 1. The robot 1 performs work such as painting, cleaning, welding, and transporting on a workpiece. The robot 1 is realized by, for example, a vertical articulated robot.

[0016] The configuration of the robot 1 will be briefly described below with reference to FIGS. 1 and 2.

[0017] The robot 1 includes a swivel base 10, an articulated arm 11, and a wrist 12. The swivel base 10 is fixed to the ground (for example, the floor of a factory). The articulated arm 11 has multiple joints. The wrist 12 is attached to the tip of the articulated arm 11. An end effector 13 is attached to the wrist 12 to perform work on a workpiece.

[0018] As shown in FIG. 2, the robot 1 includes an arm driving device 21.

[0019] These driving devices are composed of actuators configured as servo motors, reducers, etc. However, the configuration of the driving devices is not limited to the above. Each actuator is electrically connected to a controller 90. The actuators operate to reflect command values ​​input from the controller 90.

[0020] The driving force from each servo motor constituting the arm driving device 21 is transmitted via a reducer to each joint of the articulated arm 11, the swivel base 10, and the wrist 12. Each servo motor is fitted with an encoder (not shown) that detects its rotational position.

[0021] The robot 1 performs tasks by playing back the motions recorded through instruction. The controller 90 controls the actuators so that the robot 1 reproduces the series of motions previously taught by the instructor. Hereinafter, this series of motions may be referred to as a motion pattern. The robot 1 can be taught by the instructor operating a teaching pendant (not shown).

[0022] A program for moving the robot 1 is generated by teaching the robot 1. A program is generated each time teaching is performed on the robot 1. If the movement pattern taught to the robot 1 is different, the program will also be different. By switching between multiple programs to be executed, the movement pattern performed by the robot 1 can be changed.

[0023] The controller 90 is configured as a known computer including, for example, a CPU, a ROM, a RAM, an auxiliary storage device, etc. The auxiliary storage device is configured as, for example, an HDD, an SSD, etc. The auxiliary storage device stores programs for operating the robot 1, etc.

[0024] 1, the state monitoring device 5 is connected to the controller 90. The state monitoring device 5 acquires, via the controller 90, the transition of the current value of the current flowing through the actuator (servomotor), etc.

[0025] If an abnormality occurs in the servo motor or the reducer connected to it, the servo motor's current value is thought to fluctuate as a result. Therefore, this current value corresponds to a status signal that reflects the state of the robot 1. The transition of the current value can be expressed by repeatedly obtaining the current value at short time intervals and arranging many current values ​​in a time series. Hereinafter, data in which the values ​​of the status signal are arranged in a time series may be referred to as status time series data.

[0026] The status monitoring device 5 can determine whether or not there is an abnormality in the robot 1 by monitoring the acquired status time-series data. In this embodiment, the status monitoring device 5 determines whether or not there is an abnormality, mainly targeting the servo motors and reducers of each joint. Here, "abnormality" includes cases where some condition that does not lead to malfunction / inability to operate but is a precursor to such an abnormality occurs in the servo motor, reducer, or bearing.

[0027] As shown in FIG. 2, the condition monitoring device 5 includes a time-series data acquisition unit 51, a storage unit 52, a time-series data evaluation unit 53, a life estimation unit 54, and a display unit 55.

[0028] The condition monitoring device 5 is configured as a known computer including a CPU, ROM, RAM, auxiliary storage device, etc. The auxiliary storage device is configured as, for example, an HDD, SSD, etc. Various programs, such as a state monitoring program for the robot 1, are stored in the auxiliary storage device. This state monitoring program causes the computer to execute a time-series data acquisition step, a storage step, a lifespan estimation step, etc. This realizes the condition monitoring method of the present disclosure. Cooperation of these hardware and software allows the computer to operate as a time-series data acquisition unit 51, a storage unit 52, a time-series data evaluation unit 53, a lifespan estimation unit 54, etc.

[0029] The time-series data acquiring unit 51 acquires the above-mentioned state time-series data. The time-series data acquiring unit 51 acquires state time-series data for all servo motors provided in the arm driving device 21 of the robot 1. The state time-series data is acquired for each of the multiple servo motors (in other words, multiple reducers) arranged in each part of the robot 1.

[0030] In this embodiment, the status signal is a current value. Here, the current value refers to a measurement value obtained by measuring the magnitude of the current flowing through the servo motor using a sensor. The sensor is provided in a servo driver (not shown) that controls the servo motor. However, a monitoring sensor may be provided separately from the servo driver. Alternatively, the current command value given to the servo motor by the servo driver may be used as the status signal. The servo driver performs feedback control on the servo motor so that the current current value approaches the current command value. Therefore, for the purpose of detecting abnormalities in the servo motor or reducer, there is almost no difference between the current value and the current command value.

[0031] The magnitude of the torque of the servo motor is proportional to the magnitude of the current, so the torque value or the torque command value may be used as the status signal.

[0032] The deviation (rotational position deviation) between the target value for the rotational position of the servo motor and the actual rotational position obtained by the encoder may be used as the status signal. Normally, the servo driver multiplies this deviation by a gain and provides the result to the servo motor as a current command value. Therefore, the transition of the rotational position deviation shows a similar trend to the transition of the current command value.

[0033] The time-series data acquiring unit 51 acquires the status time-series data for each servo motor every time the robot 1 reproduces a taught motion pattern. However, instead of acquiring the status time-series data for all reproduced motions, the time-series data may be acquired for only one or several reproduced motions per day, for example.

[0034] In this embodiment, the movement patterns for which the time-series data acquiring unit 51 acquires state time-series data are movement patterns taught to the robot 1 to perform actual work. In other words, the robot 1 does not operate according to any special movement pattern defined for state monitoring.

[0035] The time-series data acquisition unit 51 acquires time-series data of the current values ​​flowing through each servo motor between the timing at which the acquisition start signal is received and the timing at which the acquisition end signal is received. The acquisition start signal and the acquisition end signal are output by, for example, the controller 90.

[0036] The graph in Figure 3 shows an example of the current value flowing through the servo motor of a certain joint when the robot 1 reproduces a taught movement pattern. In the graph in Figure 3, the vertical axis represents the current value and the horizontal axis represents time. As shown in Figure 3, before the program for the reproduced movement is executed, the current value of the servo motor is zero. At this time, electromagnetic brakes (not shown) are operating in each joint, so the posture of the articulated arm 11, etc. is maintained.

[0037] Next, the playback operation program for the robot 1 is started. This releases the brake, and at almost the same time, current begins to flow to the servo motor. At this point, the output shaft of the servo motor is controlled to stop. After a certain amount of time has passed, which is necessary for the angle of the output shaft of the servo motor to stabilize, the servo motor starts to rotate. This essentially starts the operation of the robot 1.

[0038] After the brake is released, the controller 90 outputs an acquisition start signal to the state monitor 5 (and thus to the time-series data acquisition unit 51) shortly before the servo motor starts to rotate.

[0039] When the series of operations included in the operation pattern are all completed, the servo motor is controlled to stop rotating. After the servo motor stops rotating, but before the program ends, the controller 90 outputs an acquisition end signal to the time-series data acquisition unit 51.

[0040] The storage unit 52 is configured by, for example, the above-mentioned auxiliary storage device. The storage unit 52 stores the time-series data acquired by the time-series data acquisition unit 51, etc.

[0041] In this embodiment, the time series data is a chronological arrangement of a large number of current values ​​obtained by repeatedly detecting them at short, fixed time intervals. Therefore, the current values ​​in the time series data are sampled values. The time interval (sampling interval) for detecting the current values ​​is, for example, several milliseconds. The time series data corresponds to the transition of the current value from the timing of the acquisition start signal to the timing of the acquisition end signal in the graph of FIG. 3.

[0042] Time information indicating the time when the data was acquired is stored in association with the time-series data in the storage unit 52. The time information may be, for example, a timestamp indicating the date and time when the acquisition start signal was received.

[0043] Similarly, the storage unit 52 stores playback identification information indicating the movement pattern being performed by the robot 1 when the time-series data of the current value was acquired, in association with the time-series data. The movement pattern of the robot 1 is defined by a playback movement program. Therefore, the playback identification information can be, for example, a program number or program name assigned to uniquely identify the playback movement program.

[0044] The storage of time-series data will now be described in detail. Using the file system of the OS provided in the status monitoring device 5, which is a computer, a folder named after the program number or program name is created in the auxiliary storage device described above. A file of status time-series data obtained when the program with that number is executed is automatically saved in this folder. In this file, time-series data regarding the current values ​​of the servo motors of the six joints is written, for example, in comma-separated value format (CSV). The file name includes a timestamp string. In this way, the timing information for the status time-series data is associated with the reproduction identification information. However, the above is an example, and the association may be achieved in other ways.

[0045] In this embodiment, the time series data acquisition unit 51 acquires time series data on the rotational positions of the servo motors of the six joints when acquiring time series data on the current values ​​of the servo motors of the six joints. This rotational position may be the target value described above or may be the actual rotational position obtained by an encoder. The time series data on the rotational position is used for generating a reference pattern, which will be described later. Hereinafter, the time series data on the rotational positions of the servo motors acquired in correspondence with the state time series data may be referred to as position time series data. Like the state time series data, the position time series data can be written in, for example, CSV format. The state time series data and the position time series data may be saved in a single file or in separate files.

[0046] The time-series data evaluation unit 53 evaluates the state time-series data acquired by the time-series data acquisition unit 51. The state time-series data to be evaluated is, for example, time-series data of current values ​​from the acquisition start signal to the acquisition end signal, as shown in FIG. 3. The time-series data evaluation unit 53 outputs an evaluation value for evaluating the state of the robot 1 based on the evaluation result. In this embodiment, I2 and the frequency analysis integrated value are used as the evaluation value. Details of the evaluation will be described later.

[0047] I2 is the root mean square value of the current. This I2 corresponds to "I2" in Patent Document 1, so a detailed explanation will be omitted.

[0048] The frequency analysis integrated value is obtained by performing frequency analysis on time-series data of current values ​​and integrating components from 0 Hz to several tens of Hz, for example, for the amplitude spectrum obtained. Because 0 Hz represents a DC component, the integrated value may be obtained by integrating components from several Hz to several tens of Hz, excluding DC. Frequency analysis can be performed using a known method, typically a fast Fourier transform. It may be preferable to use the frequency analysis integrated value to determine whether or not there are signs of a malfunction, such as whether the robot 1 is experiencing a tendency to vibrate. There are various possible causes for a tendency to vibrate, but one example is increased lost motion due to wear in the reducer. Instead of the amplitude spectrum, a power spectrum or power spectral density may be used for integration. The amplitude spectrum, power spectrum, and power spectral density are all types of frequency spectrum.

[0049] Assuming that the natural frequency of the robot 1 system is 8 Hz, the 8 Hz component in the amplitude spectrum becomes large due to resonance. In this regard, in the frequency analysis integrated value of this embodiment, a value obtained by integrating components from 0 Hz to several tens of Hz, for example, is used for evaluation. Over time, in addition to the vibration of the original main mode, vibrations due to the deterioration of various parts often occur. By observing a wide range of frequencies, the detection range for signs of failure can be expanded.

[0050] The evaluation value is not limited to I2 and the frequency analysis integrated value. For example, peak, PTP (Peak to Peak), etc. can be used as the evaluation value. A peak is the peak value of a current waveform. A PTP is a value obtained by subtracting the current value of a low peak from the current value of a high peak of a current waveform. Both peak and PTP are types of "peak current" as defined in Patent Document 1. The evaluation value can also be the dissimilarity obtained by comparing the current waveform obtained this time with the current waveform at the start of condition monitoring. The similarity can be calculated by any known method, and for example, the sum or average of Euclidean distances can be used. The DTW distance, etc. can also be used as the similarity.

[0051] The lifespan estimation unit 54 can predict the timing at which each joint of the robot 1 will malfunction / become unable to operate in the future, based on the transition of I2 and the frequency analysis integrated value over time up to the present. Hereinafter, this timing may be referred to as lifespan timing.

[0052] The graph in Figure 4 shows an example of the state of the robot 1 approximately 70 days after it began monitoring the state of the robot 1. This graph focuses on one of the six joints of the robot 1 and shows the progress of I2 as an evaluation value. I2 values ​​have been acquired at appropriate times from the start of monitoring to the present and plotted on the graph.

[0053] The lifespan estimation unit 54 creates a trend line TL1 that represents the increasing trend of I2 over time. This trend line TL1 can be obtained, for example, by the well-known least squares method. A lifespan threshold TH1 is set in advance for I2. The point at which the obtained trend line TL1 reaches the lifespan threshold TH1 indicates the timing at which the joint is predicted to malfunction or become inoperable. Hereinafter, this predicted lifespan timing may be referred to as the predicted lifespan timing PL1. Similarly, the lifespan timing can be predicted for the frequency analysis integrated value by creating a trend line and determining the timing at which the frequency analysis integrated value reaches the lifespan threshold. After determining the predicted lifespan timing for both I2 and the frequency analysis integrated value, the lifespan estimation unit 54 determines whichever timing occurs first as the overall predicted lifespan timing.

[0054] The display unit 55 can display the changes in I2 and the frequency analysis integrated value over time, for example, in the form of a graph. The display unit 55 is configured with a display device such as a liquid crystal display. The display unit 55 can also display the trend line TL1, life threshold TH1, and predicted life timing PL1 mentioned above on the graph.

[0055] The operator monitors the graphs for I2 and frequency analysis integrated values ​​to see if there are any unusual points that deviate from the usual trends. The operator can use this information to appropriately plan future maintenance.

[0056] Next, the detailed configuration of the time-series data evaluation unit 53 will be described.

[0057] The time series data evaluation unit 53 determines an evaluation value representing the state of the robot 1 by focusing on only a partial time interval, rather than on all of the time series data of the current value. Hereinafter, the time interval to be evaluated may be referred to as the evaluation interval. As a prerequisite for determining the evaluation interval, the time series data evaluation unit 53 determines and stores one or more reference patterns (basic data) corresponding to a portion extracted from the position time series data. The evaluation interval of the state time series data is determined by a pattern matching process that searches for the reference data from the position time series data.

[0058] As shown in FIG. 2, the time-series data evaluation unit 53 includes a reference pattern creation unit 61, a search processing unit 62, an evaluation value calculation unit 63, and an operation change monitoring unit 64.

[0059] The reference pattern creation unit 61 creates a predetermined reference pattern. The reference pattern can be obtained by extracting an appropriate time interval from the position time-series data acquired when the status monitoring device 5 first starts monitoring.

[0060] In some teaching programs, the last teaching point of the previously played program is always aligned with the first teaching point of the next program, for example by defining a home position. In other teaching programs, the last teaching point of the previously played program is allowed to differ from the first teaching point of the next program. In such cases, the next program is played back with a movement between these two points. Because it may not be appropriate to include this positioning operation in the reference pattern, the reference pattern may be created from only the portion of the position time-series data after a predetermined time has elapsed since the start of program execution. The process of extracting the reference pattern will be described in detail later.

[0061] To properly perform pattern matching, a lower limit and an upper limit of the length of the reference pattern are predetermined, for example, but not limited to, 10 seconds and 60 seconds.

[0062] After creating the reference pattern, the reference pattern creating unit 61 adds identification information that uniquely identifies the reference pattern and stores it in the storage unit 52.

[0063] The search processing unit 62 performs pattern matching processing to search for a reference pattern from the position time series data. A time interval in which the reference pattern matches the position time series data is identified as an evaluation interval.

[0064] The evaluation value calculation unit 63 calculates an evaluation value for the evaluation section identified for the position time-series data. The obtained evaluation value is output from the time-series data evaluation unit 53 to the life estimation unit 54 together with the identification information of the reference pattern.

[0065] The evaluation value of an evaluation target such as a reference pattern can be identified by, for example, a program number, a program name, a reference pattern number, a reference pattern name, etc. Identification is performed based on one or a combination of these.

[0066] The movement change monitoring unit 64 monitors whether a program including a registered movement pattern of the robot 1 has been changed and is no longer in use, based on the results of the search processing unit 62 searching for a reference pattern from the position time series data.

[0067] Next, the extraction of a reference pattern performed by the reference pattern creating unit 61 will be described with reference to the flowchart of Fig. 5. Fig. 5 is a flowchart illustrating the process of acquiring a reference pattern from position time-series data.

[0068] The reference pattern creation unit 61 first determines whether the robot 1 is in an active state or an inactive state at each time point of the given position time series data (step S101). An example of the position time series data is shown in Fig. 6. In each of the six graphs in Fig. 6, the vertical axis represents rotational position and the horizontal axis represents time.

[0069] The operating state of a robot refers to a state in which one or more of the six joint axes (in other words, servo motors) are moving, and the inactive state refers to a state in which none of the six axes are moving. The determination of whether the robot is in an operating state or an inactive state can be made based on the change in the rotational position of each axis. The bottom of the graph in Figure 6 shows the results of the operating / inactive state determination at each point in time.

[0070] Normally, the robot 1 is in a non-operating state from the time the program starts to the time the electromagnetic brake is released and the servo motor starts to rotate. The robot 1 is also in a non-operating state when interlock control is performed between the robot 1 and other devices. Such non-operating portions of the time-series data are not suitable for evaluating the state of the robot 1.

[0071] The operating state of the robot 1 may be defined as a state in which all six joint axes (in other words, servo motors) are moving. In this case, the non-operating state means a state in which at least one of the six axes is not moving.

[0072] Although it depends on the complexity of the motion to be taught, in the position time-series data corresponding to one motion pattern, the motion state and the non-motion state usually appear alternately. In step S101, by identifying the portions corresponding to the motion state and the non-motion state of the robot 1 from the time-series data of the status signal, it is possible to determine a reference pattern that does not include the non-motion state portions.

[0073] When the robot 1 performs spot welding, for example, all six axes of the robot 1 are stopped for, for example, one second during the welding operation, and during this time, electricity is applied from the electrodes of the gun disposed at the tip of the robot 1. To prevent such a situation from being determined as an inactive state, in step S101, it is preferable to determine that the robot 1 is not in an active state only if a situation in which all six axes of the robot 1 are not moving continues for a predetermined time.

[0074] Until a predetermined time has elapsed since the program started, the reference pattern creation unit 61 may consider the robot 1 to be in a non-moving state, regardless of whether the axes of the six joints are moving or not. This prevents the above-mentioned alignment operation from being included in the reference pattern.

[0075] Next, the reference pattern creation unit 61 identifies the section with the longest continuous motion section in the position time-series data as a candidate section for extracting a reference pattern (step S102). A continuous motion section is a section in which the above-mentioned motion state continues over time. Figure 6 shows the identified candidate section.

[0076] Next, the reference pattern creation unit 61 determines whether the time length of the candidate section is equal to or greater than the minimum length (step S103). If the time length of the candidate section is less than the minimum length, the process is terminated because a reference pattern of an appropriate length cannot be extracted from the candidate section.

[0077] If the determination in step S103 is that the temporal length of the candidate section is equal to or greater than the lower limit length, the reference pattern creation unit 61 divides the temporal length of the candidate section by the upper limit length to obtain a ratio (step S104). Next, the reference pattern creation unit 61 rounds up the ratio to obtain an integer, and determines the obtained integer as the number of reference patterns to be extracted from the candidate section (step S105). Hereinafter, this number may be referred to as the extraction number. In the example of FIG. 6, the ratio is approximately 1.8, so the extraction number is 2.

[0078] Next, the reference pattern creating unit 61 extracts reference patterns from the position time-series data according to the number of extractions (step S106).

[0079] If the extraction number is 1, the reference pattern creation unit 61 extracts all of the candidate sections identified in step S102 from the position time-series data to obtain one reference pattern. The length of the reference pattern is equal to or greater than the lower limit and equal to or less than the upper limit.

[0080] When the number of extractions is two or more, the reference pattern creation unit 61 extracts a continuous portion of a time interval from the candidate interval to obtain a reference pattern. By differentiating the start and end times of the time intervals extracted from the candidate interval, multiple reference patterns can be obtained from one candidate interval. In FIG. 6, two reference patterns obtained from the candidate interval are shown enclosed in rectangles. For each reference pattern, the length of the extracted time interval matches the upper limit length. Multiple reference patterns are extracted from the candidate interval so that every part of the candidate interval is included in one of the reference patterns. When the number of extractions is three or more, it is preferable that the extracted multiple time intervals be evenly spaced within the candidate interval. When the length of the candidate interval is shorter than the upper limit length multiplied by the number of extractions, the multiple reference patterns are extracted so that they partially overlap each other. In the example of FIG. 6, the two acquired reference patterns partially overlap.

[0081] In the example of Fig. 6, for simplicity, the upper limit length of the reference pattern is set to a certain extent. By setting the lower limit length and upper limit length of the reference pattern to be sufficiently small, many reference patterns can be obtained from one piece of position time-series data.

[0082] The reference pattern can be determined by the above series of processes. After obtaining a certain number of reference patterns as candidates, a selection process may be performed to eliminate inappropriate patterns, and the resulting pattern may be adopted as the reference pattern. For example, it is possible to eliminate patterns in which the change in one or more axes among the multiple axes is always zero or less than a predetermined amount throughout the entire time period of the pattern. Reference patterns in which the load applied to each joint of the robot 1 is light or the change in load is less than a predetermined amount may also be eliminated. The load can be evaluated using, for example, the above-mentioned I2 and the frequency analysis integrated value.

[0083] In step S105, the ratio can be rounded down to an integer to determine the number of reference patterns to extract from the candidate section. In this case, multiple reference patterns can be extracted from one candidate section without overlapping. Rounding can be performed to either round up or round down depending on the value after the decimal point.

[0084] Steps S104 and S105 may be omitted, and the extraction number may always be 1. In this case, in step S106, one reference pattern is extracted from the candidate section, with the upper limit length of the reference pattern as the upper limit. If the length of the candidate section is longer than the upper limit length of the reference pattern, the position from which the reference pattern is extracted is arbitrary. For example, one reference pattern may be extracted from the beginning or end of the candidate section.

[0085] In this embodiment, the storage unit 52 can store up to five different reference patterns. Each of the reference patterns created by the reference pattern creation unit 61 is stored in the storage unit 52.

[0086] When a reference pattern is extracted from the initially acquired position time-series data, the time-series data evaluation unit 53 sets the same time interval as the extraction of the reference pattern as an evaluation interval in the time-series data of the current value. The time-series data evaluation unit 53 extracts only the evaluation interval from the time-series data of the current value and calculates the aforementioned evaluation value (i.e., I2 and the frequency analysis integrated value). This evaluation value is treated as the initial evaluation value. If there are multiple reference patterns, an evaluation value is calculated for each reference pattern.

[0087] Next, the pattern matching process performed by the search processing unit 62 will be described with reference to Fig. 7. Fig. 7 is a flowchart mainly explaining the pattern matching process for searching for a reference pattern from position time-series data.

[0088] Even after the reference pattern created by the reference pattern creation unit 61 is stored in the memory unit 52, the robot 1 continues to perform operations according to the program. At appropriate timing, the time-series data acquisition unit 51 acquires the position time-series data and the time-series data of the value of the status signal, and stores them in the memory unit 52.

[0089] The search processing unit 62 included in the time-series data evaluation unit 53 performs pattern matching processing to search for a reference pattern from the obtained position time-series data.

[0090] When the flow shown in Fig. 7 starts, the search processing unit 62 selects one reference pattern as a search target and appropriately determines the number of slides for the search, taking into account the length of the position time-series data and the length of the reference pattern (step S201). As simply shown in Fig. 8, pattern matching is performed by calculating the dissimilarity between the reference pattern to be searched and the position time-series data while sliding the position of the reference pattern along the time axis relative to the position time-series data. If the number of slides increases, the amount of slide along the time axis per slide (slide amount S1 in Fig. 8) decreases, thereby improving matching accuracy but increasing the calculation load. It is preferable to determine the number of slides taking these factors into consideration.

[0091] Next, the search processing unit 62 calculates the dissimilarity between the reference pattern and the position time-series data for each axis of the robot 1 while sliding the reference pattern in the time axis direction relative to the position time-series data (step S202). FIG. 8 illustrates an example of pattern matching for the position time-series data of one axis. The amount of sliding S1 of the reference pattern per time is determined by calculation so that the entire position time-series data can be covered by sliding the reference pattern the number of times determined in step S201. In this embodiment, the sum of the Euclidean distances (in other words, the absolute values ​​of the differences in rotational position) at each sampling timing is used as the dissimilarity. The search processing unit 62 determines the position where the above-mentioned dissimilarity is minimum (step S203).

[0092] In steps S202 and S203, the search processing unit 62 slides the reference pattern independently for each axis of the robot 1, and finds the position on the time axis where the dissimilarity is closest to zero. In reality, the position on the time axis where the reference pattern best matches the position time series data should be the same for all axes of the robot 1 (in other words, the servo motors of the six joints). However, due to aging, noise, and other reasons, some axes may experience a lead / lag in the time axis direction of the position time series data. By performing a matching process independently for each axis and using the median value of the best match position (described later), the reference pattern can be flexibly matched to such position time series data.

[0093] By the processing of step S203, the position on the time axis at which the reference pattern best matches is obtained for each of the multiple axes of the robot 1. Hereinafter, this position may be referred to as the best match position. The search processing unit 62 acquires the median of the best match positions obtained for the multiple axes as the best match position of the reference pattern taking the multiple axes into consideration comprehensively (step S204). Hereinafter, this best match position may be referred to as the representative match position. Any method can be used to determine the representative match position; for example, the average value of the best match positions may be used instead of the median value.

[0094] Next, the search processing unit 62 calculates the average error E between the position time series data and the reference pattern of the representative match position for each axis of the robot 1 (step S205). This average error E is calculated by dividing the sum of the absolute values ​​of the differences in rotational position at each sampling timing for the time interval corresponding to the reference pattern of the representative match position by the number of samplings corresponding to the length of that time interval. The formula for the average error E is shown in the schematic diagram of FIG. 9. Hereinafter, in the position time series data, the time interval corresponding to the reference pattern of the representative match position may be referred to as the reference pattern match interval.

[0095] Next, the search processing unit 62 calculates the maximum error M between the position time-series data and the reference pattern of the representative match position for each axis of the robot 1, for the reference pattern matching section (step S206). This maximum error is the maximum absolute value of the difference in rotational position at each sampling timing for the reference pattern matching section. The formula for the maximum error M is shown in FIG. 9.

[0096] Next, the search processing unit 62 calculates the difference A for each axis of the robot 1 by subtracting the minimum value from the maximum value of the rotational position indicated by the position time series data for the reference pattern matching section, as shown in Figure 9 (step S207).

[0097] The search processing unit 62 calculates the E / A and M / A values ​​for each axis (step S208). These values ​​represent the relative magnitude of the average and maximum error between the position time-series data in the reference pattern matching section and the reference pattern, with respect to the change in rotational position. For example, due to spike noise or other reasons, the difference in rotational position compared to the reference pattern may become extremely large in an extremely short time region of the position time-series data in the reference pattern matching section. If the reference pattern is long, such noise is unlikely to appear in the E / A value, but can be easily detected by calculating the M / A value.

[0098] The search processing unit 62 compares E / A with a predetermined first threshold value and compares M / A with a predetermined second threshold value for each axis of the robot 1 (steps S209 to S210). The first threshold value and the second threshold value can be set arbitrarily, but for example, the first threshold value can be set to 3% and the second threshold value to 5%.

[0099] If the E / A value is equal to or less than the first threshold and the M / A value is equal to or less than the second threshold for all axes in operation, the search processing unit 62 identifies the reference pattern matching section of the position time-series data as the evaluation section (step S211). Thereafter, the evaluation value calculation unit 63 calculates the evaluation value for the evaluation section (step S212).

[0100] Furthermore, the search processing unit 62 stores in the storage unit 52 that the reference pattern has appeared in the position time-series data (step S213). This appearance record is used by the operation change monitoring unit 64 to determine whether or not a change has occurred in the operation pattern of the robot 1.

[0101] If the E / A value exceeds the first threshold or the M / A value exceeds the second threshold for one or more operating axes, an evaluation value should not be calculated from such position time series data, and therefore steps S211 to S213 are skipped.

[0102] The search processing unit 62 and the evaluation value calculation unit 63 repeat the matching process and the evaluation value calculation process of steps S201 to S213 (step S214) for all of the reference patterns stored in the storage unit 52. When the process has been completed for all of the reference patterns, the series of processes shown in FIG.

[0103] The search processing unit 62 and the evaluation value calculation unit 63 perform the processing shown in the flowchart of Fig. 7 on the position time-series data obtained by the operation of the robot 1. As a result, the number of times each reference pattern has appeared in the last 10 days, for example, can be obtained by calculation.

[0104] Next, the change in the movement pattern of the robot 1 and the monitoring by the movement change monitoring unit 64 will be described.

[0105] Consider a case where the motion pattern of robot 1 described with reference to FIG. 3 is no longer used for some reason, or where its frequency of use drops significantly. For example, in the case of robots used in the automobile industry, this situation may arise when the model of the car being manufactured changes, or when the monthly production schedule for the main vehicle model changes in a mixed production line. In this case, robot 1 reproduces a motion pattern different from the previous one. The transition of the current value corresponding to the new motion pattern is shown in FIG. 10.

[0106] A change in the movement pattern of the robot 1 is synonymous with a change in the program of the playback movement. When the robot 1 performs a new movement corresponding to FIG. 10, the playback identification information is stored in the storage unit 52 in association with the time-series data, as in the case of FIG. 3. As the program of the playback movement is changed, the playback identification information is also changed. Therefore, the CSV files for the movement in FIG. 10 and the movement in FIG. 3 are saved in folders with different names.

[0107] When the motion pattern of the robot 1 is changed, the reference pattern created before the change no longer matches the position time-series data after the change. The motion change monitoring unit 64 determines whether the reference pattern stored in the storage unit 52 continues not to appear with sufficient frequency in the position time-series data. There are various methods for this determination, but it is possible to compare the count value of the number of consecutive days during which the reference pattern does not appear in the position time-series data with a threshold, or to compare the count value of appearances over N days with a threshold.

[0108] If it is determined that the situation in which the stored reference pattern does not appear with sufficient frequency in the position time-series data continues, the operation change monitoring unit 64 deletes the reference pattern from the storage unit 52. When the reference pattern creation unit 61 detects this, it automatically performs the process shown in Fig. 5 to create a reference pattern from the position time-series data at that time and stores it in the storage unit 52. This ensures the continuity of collection of evaluation values.

[0109] The operation pattern may be changed manually by an operator when a production plan is changed, or may be automatically switched depending on the product. When the operation pattern is changed by an operator, the previously registered reference pattern may be deleted manually or automatically.

[0110] Next, the handling of the evaluation value before and after changing the robot's motion pattern will be described.

[0111] When the motion pattern of the robot 1 is changed, the load conditions of the servo motors and reducers of the six joints change. Therefore, simply combining the evaluation values ​​obtained before and after changing the motion pattern cannot be used for trend management for failure prediction or for calculating the remaining lifespan.

[0112] When the motion pattern of the robot 1 is changed, the reference pattern is also automatically changed, except in special cases. Figure 11 shows the transition of I2 based on one of the reference patterns before the motion change and the transition of I2 based on one of the reference patterns after the motion change. When the reference pattern on which the evaluation value is based is changed, the evaluation value usually changes discontinuously. The life estimation unit 54 calculates the predicted life timing by taking into account the evaluation values ​​both before and after the change of the reference pattern, using one of the following 12 methods.

[0113] The first method will be explained below with reference to FIG.

[0114] [a] The life threshold coefficient k1 is a predetermined value, and the life threshold TH1 of the first reference pattern is the initial evaluation value p1 of the first reference pattern multiplied by the life threshold coefficient k1 (TH1 = p1 × k1). The life threshold coefficient k1 is determined appropriately and can be, for example, 1.2. [b] The final evaluation value p2 based on the pre-change reference pattern is divided by the initial evaluation value p1 based on the pre-change reference pattern to obtain an increase ratio ir1 (ir1=p2 / p1). [c] The initial evaluation value q1 based on the changed reference pattern is divided by the increase ratio ir1 to obtain the past pre-increase estimated evaluation value s1 (s1=q1 / ir1). [d] The life threshold TH2 based on the changed reference pattern is calculated by multiplying the past pre-increase estimated evaluation value s1 by the life threshold coefficient k1 (TH2=s1·k1). [e] A trend line TL2 is created based on multiple evaluation values ​​based on the changed reference pattern. The timing at which this trend line TL2 reaches the life threshold TH2 applied to the changed reference pattern is defined as the predicted life timing PL2.

[0115] The life threshold coefficient k1 described in [a] above indicates how many times the life threshold TH1 applied to the operation of the first reference pattern is multiplied by the initial evaluation value p1 based on the operation of the first reference pattern. Given a life threshold TH1, the life threshold coefficient k1 can also be calculated by dividing this life threshold TH1 by the initial evaluation value p1. This also applies to the second and subsequent methods described below.

[0116] In the first method, if the reference pattern is changed multiple times, processes [b] to [e] can be performed for each change. Figure 12 shows a case where the robot 1's motion has been changed twice. However, from the second change onward, the increase ratio ir2 of [b] is calculated by dividing the final evaluation value q2 based on the reference pattern before the change by the previous pre-change estimated evaluation value s1 calculated for the reference pattern before the change (ir2 = q2 / s1). In Figure 12, if the initial evaluation value based on the reference pattern after the second change is r1, the previous pre-change estimated evaluation value s2 based on the reference pattern after the second change is calculated using the formula s2 = r1 / ir2. The lifespan threshold TH3 applied to the reference pattern after the second change is calculated by multiplying the previous pre-change estimated evaluation value s2 by the lifespan threshold coefficient k1 (TH3 = s2·k1). The lifespan threshold coefficient k1 can be used in common regardless of the number of motion changes.

[0117] A trend line TL3 created based on multiple evaluation values ​​based on the reference pattern after the second change is shown in Figure 12. The timing at which this trend line TL3 reaches the life threshold TH3 is the predicted life timing. Due to space limitations in the graph, the predicted life timing based on evaluation values ​​after the second change is omitted in Figure 12.

[0118] In the first method, when the reference pattern is changed multiple times, the latest increase ratio and the latest estimated evaluation value before the increase are calculated by cumulatively taking into account the increasing trend of the evaluation value for each past reference pattern.

[0119] Instead of the final evaluation value in [b], the median or average of the evaluation values ​​obtained most recently, including the final evaluation value, may be used. In this case, the influence of noise contained in the evaluation value can be reduced. The final evaluation value, the median, or the average can be considered as a type of final evaluation value based on one reference pattern. The representative value is not limited to the median or the average.

[0120] Instead of the initial evaluation value in [c], the median or average of multiple evaluation values ​​obtained around the initial evaluation value, including the initial evaluation value, may be used. In this case, the influence of noise contained in the evaluation value can be reduced. The initial evaluation value, the median, or the average can be considered as a type of early evaluation value based on one reference pattern. The representative value is not limited to the median or the average.

[0121] As described above, the state monitoring device 5 of this embodiment monitors the state of the robot 1 that can reproduce predetermined actions. The state monitoring device 5 includes a time-series data acquisition unit 51, a storage unit 52, and a lifespan estimation unit 54. The time-series data acquisition unit 51 acquires state time-series data that reflects the state of the robot 1. The storage unit 52 stores the state time-series data acquired by the time-series data acquisition unit 51. The lifespan estimation unit 54 acquires an evaluation value for evaluating the state of the robot 1 based on the state time-series data, creates a trend line TL1 that represents the tendency for the evaluation value to change over time, and determines the timing at which the trend line TL1 reaches a predetermined lifespan threshold TH1 as the predicted lifespan timing PL1.

[0122] When the robot's motion, which is the basis for acquiring state time-series data, is changed as shown in FIG. 11 , the predicted lifespan timing PL2 is calculated as follows. That is, the lifespan estimation unit 54 calculates the past pre-increase estimated evaluation value s1 for the changed motion. Here, a coefficient indicating how many times the lifespan threshold TH1 applied to the motion before the change is multiplied by the initial evaluation value p1 based on the motion before the change is referred to as the lifespan threshold coefficient k1 for the motion before the change. The lifespan estimation unit 54 calculates the lifespan threshold TH2 to be applied to the changed motion by multiplying the past pre-increase estimated evaluation value s1 for the changed motion by the lifespan threshold coefficient k1 for the motion before the change. The lifespan estimation unit 54 creates a trend line TL2 based on multiple evaluation values ​​based on the changed motion, and determines the timing at which the trend line TL2 reaches the lifespan threshold TH2 applied to the changed motion as the predicted lifespan timing PL2.

[0123] As a result, even if the behavior of the robot 1 that is the basis for the evaluation value is changed, the evaluation values ​​before and after the change can be handled consistently, allowing for a consistent lifespan prediction. Therefore, it is possible to obtain the state time-series data and the evaluation value based on the actual behavior of the robot 1, thereby improving the operating efficiency of the robot 1. A calculation using the past estimated evaluation value before increase can obtain the lifespan threshold value TH2 based on the changed behavior.

[0124] In the first method, the lifespan estimation unit 54 obtains an increase ratio ir1 for the pre-change behavior by dividing the final evaluation value p2 based on the pre-change behavior by the early evaluation value p1 based on the pre-change behavior. The lifespan estimation unit 54 obtains a past pre-increase estimated evaluation value s1 for the post-change behavior by dividing the early evaluation value q1 based on the post-change behavior by the increase ratio ir1 for the pre-change behavior.

[0125] As a result, by using the increase ratio and the past pre-increase estimated evaluation value, the life threshold value TH2 based on the changed operation can be obtained by simple calculation.

[0126] In the first method, when the robot's motion, which is the basis for acquiring the time-series data of the status signal, has changed two or more times as shown in Figure 12, the predicted lifespan timing is calculated as follows. That is, the lifespan estimation unit 54 calculates the increase ratio ir2 for the motion before the most recent change by dividing the final evaluation value q2 based on the motion before the most recent change by the past pre-ascension estimated evaluation value s1 for the motion before the most recent change. Except for the above, the calculation of the predicted lifespan timing is performed in the same way as for the first change in the motion of the robot 1. Regardless of the number of changes, the lifespan threshold coefficient k1 for the motion before the first change is commonly used.

[0127] This allows the influence of multiple changes in the robot 1's behavior on the evaluation value to be estimated with high accuracy.

[0128] In the state monitoring device 5 of this embodiment, the final evaluation value based on the operation before the change is the evaluation value p2 of the last time based on the operation before the change.

[0129] This simplifies the calculation.

[0130] In the condition monitoring device 5 of this embodiment, the final evaluation value based on the operation before the change can also be a representative value such as the median or average of the multiple evaluation values ​​obtained most recently, including the final evaluation value p2 based on the operation before the change.

[0131] In this case, the influence of noise contained in the evaluation value can be reduced.

[0132] In the state monitoring device 5 of this embodiment, the evaluation value at the beginning based on the changed operation is the initial evaluation value q1 based on the changed operation.

[0133] This simplifies the calculation.

[0134] In the condition monitoring device 5 of this embodiment, the evaluation value at the beginning based on the operation before the change can also be a representative value such as the median or average value of multiple evaluation values ​​obtained near the initial time, including the initial evaluation value q1 based on the operation after the change.

[0135] In this case, the influence of noise contained in the evaluation value can be reduced.

[0136] The second method will be explained with reference to FIG. 13 as follows.

[0137] [a] The life threshold coefficient k1 is a predetermined value, and the life threshold TH1 of the first reference pattern is the initial evaluation value p1 of the first reference pattern multiplied by the life threshold coefficient k1 (TH1=p1·k1). [b] A trend line TL1 is created based on a plurality of evaluation values ​​based on the pre-change reference pattern, and the timing at which this trend line TL1 reaches the life threshold value TH1 (the predicted life timing PL1 described above) is calculated. [c] The remaining life rate tr1 is calculated by dividing the time from the timing tp2 of the final evaluation value p2 based on the reference pattern before the change to the predicted life time PL1 by the time from the timing tp1 of the first evaluation value p1 based on the reference pattern before the change to the predicted life time PL1 (tr1 = (PL1 - tp2) / (PL1 - tp1)). [d1] As shown in the following formula, the life threshold coefficient k2 of the changed reference pattern is calculated by subtracting 1 from the life threshold coefficient k1 of the reference pattern before the change, multiplying the result by the remaining life rate tr1, and then adding 1. k2=(k1-1)·tr1+1 [e] The initial evaluation value q1 based on the changed reference pattern is multiplied by the life threshold coefficient k2 of the changed reference pattern to obtain the life threshold TH2 to be applied to the changed reference pattern (TH2 = q1 · k2). [f] A trend line TL2 is created based on multiple evaluation values ​​based on the changed reference pattern. The timing at which this trend line TL2 reaches the life threshold TH2 applied to the changed reference pattern is defined as the predicted life timing PL2.

[0138] In the second method, if the reference pattern is changed multiple times, processes [b], [c], [d1], [e], and [f] are performed for each change. However, the above [d1] is extended as shown below in [d1'] so that it can be applied to changes from the second time onwards.

[0139] Consider a case where the [d1'] reference pattern is changed N times and appears in the order of the first reference pattern, the second reference pattern, ..., and the (N+1)th reference pattern. The life threshold coefficient k(N+1) corresponding to the (N+1)th reference pattern can be calculated by subtracting 1 from the life threshold coefficient k1 calculated for the first reference pattern, multiplying the result by the remaining life rates tr1, tr2, ..., trN calculated from the first reference pattern to the Nth reference pattern, and then adding 1 to the result, as shown in the following equation. k(N+1)=(k1-1)·tr1·tr2·····trN+1

[0140] As described above, in the second method, when the robot's motion, which is the basis for acquiring state time-series data, is changed as shown in FIG. 13 , the predicted lifespan timing is calculated as follows. That is, the lifespan estimation unit 54 calculates a lifespan threshold coefficient k2 for the changed motion, which indicates how many times the lifespan threshold TH2 applied to the changed motion is multiplied by the evaluation value q1 at the beginning of the changed motion. The lifespan estimation unit 54 multiplies the evaluation value q1 at the beginning of the changed motion by the lifespan threshold coefficient k2 for the changed motion to calculate the lifespan threshold TH2 applied to the changed motion. The lifespan estimation unit 54 creates a trend line TL2 based on multiple evaluation values ​​based on the changed motion, and determines the timing at which the trend line TL2 reaches the lifespan threshold TH2 applied to the changed motion as the predicted lifespan timing PL2.

[0141] This allows the life threshold TH2 based on the changed operation to be obtained by simple calculation using the life threshold coefficient.

[0142] As described above, in the second method, when the robot's motion, which is the basis for acquiring state time-series data, is changed as shown in FIG. 13 , the predicted lifespan timing is calculated as follows. Specifically, the lifespan estimation unit 54 creates a trend line TL1 based on multiple evaluation values ​​based on the pre-change motion and calculates the predicted lifespan timing PL1 for the pre-change motion, which is the timing at which this trend line TL1 reaches the lifespan threshold TH1 applied to the pre-change motion. The lifespan estimation unit 54 calculates the remaining lifespan rate tr1 for the pre-change motion by dividing the time from the timing tp2 of the final evaluation value p2 based on the pre-change motion to the predicted lifespan timing PL1 by the time from the timing tp1 of the initial evaluation value p1 based on the pre-change motion to the predicted lifespan timing PL1. Here, the coefficient indicating how many times the lifespan threshold TH1 applied to the pre-change motion is relative to the initial evaluation value p1 based on the pre-change motion is referred to as the lifespan threshold coefficient k1 for the pre-change motion. The life estimation unit 54 calculates the life threshold coefficient k2 for the changed operation by subtracting 1 from the life threshold coefficient k1 for the operation before the change, multiplying the result by the remaining life rate tr1 for the operation before the change, and then adding 1 to the result.

[0143] As a result, by using the life threshold coefficient and the remaining life rate, the life threshold TH2 based on the changed operation can be obtained by simple calculation.

[0144] In the second method, when the robot's motion, which is the basis for acquiring the time-series data of the status signal, has changed two or more times (N times), the predicted life timing is calculated as follows. That is, when focusing on one of the first through Nth changes, the life estimation unit 54 calculates the predicted life timing PL for the motion immediately before the focused change. The predicted life timing PL for the motion immediately before the focused change is calculated by creating a trend line TL based on multiple evaluation values ​​based on the motion immediately before the focused change, and determining the timing at which this trend line TL reaches the life threshold TH applied to the motion immediately before the focused change. Next, the life estimation unit 54 calculates the remaining life rate tr for the motion immediately before the focused change. The remaining life rate tr for the motion immediately before the focused change is calculated by dividing the time from the timing of the final evaluation value based on the motion immediately before the focused change to the predicted life timing PL for the motion immediately before the focused change by the time from the timing of the initial evaluation value based on the motion immediately before the focused change to the predicted life timing PL for the motion immediately before the focused change. The life estimation unit 54 focuses on each of the first through Nth changes and calculates the predicted life timing PL1, PL2,...,PLN and remaining life rate tr1, tr2,...,trN for the operation immediately before the focused change. The life estimation unit 54 subtracts 1 from the life threshold coefficient k1 for the operation before the first change, multiplies the result by the sum of the remaining life rates tr1, tr2,...,trN from the first through Nth changes, and adds 1 to the result to calculate the life threshold coefficient k(N+1) for the operation after the Nth change. The life estimation unit 54 multiplies the initial evaluation value based on the operation after the Nth change by the life threshold coefficient k(N+1) for the operation after the Nth change to calculate the life threshold TH(N+1) to be applied to the operation after the change. The life estimation unit 54 creates a trend line TL(N+1) based on multiple evaluation values ​​based on the operation after the Nth change, and sets the timing at which the trend line TL(N+1) reaches the life threshold TH(N+1) applied to the operation after the Nth change as the predicted life timing PL(N+1).

[0145] This allows the influence of multiple changes in the robot 1's behavior on the evaluation value to be estimated with high accuracy.

[0146] As a modification of the second method described above, it is also possible to perform the following [d2] instead of [d1].

[0147] [d2] The life threshold coefficient k1 calculated for the operation before the change is used as the base, and the remaining life rate tr1 for the operation before the change is used as the exponent to calculate the life threshold coefficient k2 for the operation after the change. k2=k1 tr1

[0148] In this modified example, if the reference pattern is changed multiple times, processes [b], [c], [d2], [e], and [f] can be performed for each change. However, the above [d2] is extended to [d2'] shown below so that it can be applied to changes from the second time onwards.

[0149] [d2'] Consider a case where the reference pattern is changed N times and appears in the order of the first reference pattern, the second reference pattern, ..., and the (N+1)th reference pattern. The life threshold coefficient k(N+1) corresponding to the (N+1)th reference pattern is calculated as shown in the following formula, using the life threshold coefficient k1 obtained from the first reference pattern as the base and exponent Π, which is the sum of the remaining life rates tr1, tr2, ..., trN calculated from the first reference pattern to the Nth reference pattern. k(N+1)=k1 Π

[0150] Under the same conditions, the life threshold coefficients k2,...,k(N+1) will be similar whether they are calculated using [d1] and [d1'] or [d2] and [d2'].

[0151] As described above, in the modified second method, when the robot's motion, which is the basis for acquiring the time-series data of the status signal, is changed once, the predicted life timing is calculated as follows: That is, the life estimation unit 54 calculates, as the life threshold coefficient k2 for the changed motion, a power value in which the life threshold coefficient k1 calculated for the motion before the change is used as the base and the remaining life rate tr1 for the motion before the change is used as the exponent.

[0152] This allows the life threshold TH2 based on the changed operation to be obtained by simple calculation.

[0153] In a modified version of the second method, when the robot's motion, which is the basis for acquiring the time-series data of the status signal, has been changed two or more times (N times), the predicted life timing is calculated as follows: That is, the life estimation unit 54 calculates the life threshold coefficient k(N+1) for the motion after the Nth change by using the life threshold coefficient k1 for the motion before the first change as the base and exponent of the sum of the remaining life rates tr1, tr2, . . . , trN from the first change to the Nth change.

[0154] This allows the influence of multiple changes in the robot 1's behavior on the evaluation value to be estimated with high accuracy.

[0155] The third method will be explained below with reference to FIG.

[0156] [a] The life threshold coefficient k1 is a predetermined value, and the life threshold TH1 of the first reference pattern is the initial evaluation value p1 of the first reference pattern multiplied by the life threshold coefficient k1 (TH1=p1·k1). [b1] A life threshold coefficient k2 of the changed reference pattern is calculated by dividing the life threshold TH1 of the pre-change reference pattern by the final evaluation value p2 based on the pre-change reference pattern (k2=TH1 / p2). [c] The initial evaluation value q1 based on the changed reference pattern is multiplied by the life threshold coefficient k2 of the changed reference pattern to obtain the life threshold TH2 to be applied to the changed reference pattern (TH2 = q1 · k2). [d] A trend line TL2 is created based on multiple evaluation values ​​based on the changed reference pattern. The timing at which this trend line TL2 reaches the life threshold value TH2 of the changed reference pattern is defined as the predicted life timing PL2.

[0157] In the third method, for example, if the life threshold coefficient k1 of the initial (before change) reference pattern is 1.20 and the life threshold coefficient k2 of the changed reference pattern is 1.15, it can be considered that the remaining life of robot 1 has been consumed by the difference of 0.05.

[0158] In the third method, if the reference pattern is changed multiple times, the processes [b1], [c], and [d] may be performed for each change.

[0159] As a modification of the third method described above, it is also possible to perform the following [b2] instead of [b1].

[0160] [b2] Create a trend line TL1 based on multiple evaluation values ​​based on the pre-change reference pattern. Calculate the value indicated by this trend line TL1 at timing tp2 of the final evaluation value p2 based on the pre-change reference pattern. Calculate the life threshold coefficient k2 for the post-change reference pattern by dividing the life threshold TH1 for the pre-change reference pattern by the obtained value.

[0161] As described above, in the third method, when the robot's motion, which is the basis for acquiring state time-series data, is changed as shown in Fig. 14, the lifespan threshold coefficient k2 for the changed motion is obtained by dividing the lifespan threshold coefficient k1 for the motion before the change by the final evaluation value p2 based on the motion before the change. In a variation of the third method, the lifespan threshold coefficient k2 for the changed motion is obtained by dividing the lifespan threshold coefficient k1 for the motion before the change by the value indicated by the trend line TL1 based on the motion before the change at timing tp2 of the final evaluation value p2.

[0162] This allows the life threshold TH2 based on the changed operation to be obtained by simple calculation.

[0163] The fourth method will be explained below with reference to FIG.

[0164] [a] The life threshold coefficient k1 is a predetermined value, and the life threshold TH1 of the first reference pattern is the initial evaluation value p1 of the first reference pattern multiplied by the life threshold coefficient k1 (TH1=p1·k1). [b] Estimate an adjustment coefficient that expresses the ratio of the impact of changing the reference pattern on the evaluation score. For example, the adjustment coefficient ka1 can be calculated by dividing the final evaluation score p2 based on the pre-change reference pattern by the initial evaluation score q1 based on the post-change reference pattern (ka1 = p2 / q1). [c] The life threshold value TH1 of the reference pattern before the change is divided by the adjustment coefficient ka1 to obtain the life threshold value TH2 of the reference pattern after the change (TH2=TH1 / ka1). [d] A trend line TL2 is created based on multiple evaluation values ​​based on the changed reference pattern. The timing at which this trend line TL2 reaches the life threshold TH2 applied to the changed reference pattern is defined as the predicted life timing PL2.

[0165] The adjustment coefficient ka1 described in [b] above is the same as the adjustment coefficient ka1 described in the seventh method described below. Instead of this adjustment coefficient ka1, the adjustment coefficient ka2 described in the ninth method described below, or the adjustment coefficient ka3 described in the eleventh method, etc., can also be used.

[0166] As described above, in the fourth method, when the robot's behavior, which is the basis for acquiring state time-series data, is changed as shown in FIG. 15 , the predicted lifespan timing is calculated as follows. That is, the lifespan estimation unit 54 estimates an adjustment coefficient ka1 that expresses, as a ratio, the effect on the evaluation value due to the change in the behavior of the robot 1. The lifespan estimation unit 54 calculates a lifespan threshold TH2 for the changed behavior by dividing the lifespan threshold TH1 applied to the behavior before the change by the adjustment coefficient ka1. The lifespan estimation unit 54 creates a trend line TL2 based on the evaluation value based on the changed behavior. The lifespan estimation unit 54 determines the timing at which the trend line TL2 reaches the lifespan threshold TH2 applied to the changed behavior as the predicted lifespan timing PL2.

[0167] This allows the life threshold TH2 based on the changed operation to be obtained by simple calculation.

[0168] The fifth method will be explained below with reference to FIG.

[0169] [a] The life threshold coefficient k1 is a predetermined value, and the life threshold TH1 of the first reference pattern is the initial evaluation value p1 of the first reference pattern multiplied by the life threshold coefficient k1 (TH1=p1·k1). [b] A trend line TL2 is created based on a plurality of evaluation values ​​based on the changed reference pattern, and the value indicated by this trend line TL2 at the timing when the elapsed time is zero is determined as the past pre-rise estimated evaluation value st1. [c] The life threshold value TH2 based on the changed reference pattern is calculated by multiplying the past pre-increase estimated evaluation value st1 by the life threshold coefficient k1 (TH2=st1·k1). [d] The timing when the trend line TL2 reaches the life threshold value TH2 based on the changed reference pattern is set as the predicted life timing PL2.

[0170] In the example of Fig. 16, the predicted life timing PL1 based on the pre-change reference pattern and the predicted life timing PL2 based on the post-change reference pattern are significantly different. However, this is because the graphs of Figs. 11 to 23 are primarily intended to explain the method in an easy-to-understand manner. In actual operation, it is believed that the predicted life timing PL2 based on the post-change reference pattern can be obtained with good accuracy.

[0171] In the fifth method, if the reference pattern is changed multiple times, steps [b] to [d] can be performed for each change. The life threshold coefficient k1 initially determined in [a] can be used in common.

[0172] In the fifth method, when the robot's motion is changed multiple times, an Nth trend line TLN is created, and the value indicated by the trend line TLN at the time when elapsed time is zero is determined as the past pre-increase estimated evaluation value stN for the changed motion. The lifespan estimation unit 54 multiplies the past pre-increase estimated evaluation value stN for the changed motion by the above-mentioned lifespan threshold coefficient k1 to determine the lifespan threshold THN to be applied to the changed motion. The lifespan estimation unit 54 determines the timing at which the trend line TLN, created based on multiple evaluation values ​​based on the changed motion, reaches the lifespan threshold THN to be applied to the changed motion as the predicted lifespan timing PLN.

[0173] As described above, in the fifth method, when the robot's motion, which is the basis for acquiring state time-series data, is changed as shown in FIG. 16 , the predicted lifespan timing is calculated as follows. That is, the lifespan estimation unit 54 creates a trend line TL2 based on multiple evaluation values ​​based on the changed motion, and calculates the value indicated by the trend line TL2 at zero elapsed time as the past pre-increase estimated evaluation value st1 for the changed motion. Here, a coefficient indicating how many times the lifespan threshold TH1 applied to the pre-change motion is multiplied by the initial evaluation value p1 based on the pre-change motion is referred to as the lifespan threshold coefficient k1 for the pre-change motion. The lifespan estimation unit 54 calculates the lifespan threshold TH2 to be applied to the changed motion by multiplying the past pre-increase estimated evaluation value st1 for the changed motion by the lifespan threshold coefficient k1 for the pre-change motion. The lifespan estimation unit 54 calculates the predicted lifespan timing PL2 as the timing tp2 at which the trend line TL2 created based on multiple evaluation values ​​based on the changed motion reaches the lifespan threshold TH2 applied to the changed motion.

[0174] As a result, by using the life threshold coefficient k1 and the trend line TLN, the life threshold TH2 based on the changed operation can be obtained by simple calculation.

[0175] The sixth method will be explained below with reference to FIG.

[0176] In the fifth method described above, the value indicated by the trend line TL2 at the timing of elapsed time zero is calculated as the past pre-increase estimated evaluation value st1 for the changed behavior, as shown in Fig. 16. Instead, in the sixth method, the past pre-increase estimated evaluation value st1x is calculated as follows.

[0177] First, a trend line TL2 is calculated based on a plurality of evaluation values ​​based on the post-change operation. In Fig. 17, the trend line TL2 is shown as a dotted line. Next, the life estimation unit 54 calculates the slope of an auxiliary line AL1 (described below) based on the load corresponding to the evaluation values ​​used as the basis for calculating the trend line TL2 and the load for the pre-change operation.

[0178] Generally, the deterioration of a machine progresses faster as the load increases. The slope of the auxiliary line AL1 corresponds to the slope of the trend line TL2 corrected in consideration of the difference in load before and after the change in operation. Let the slope of the trend line TL2 be ΔTL2, and the load corresponding to the evaluation value on which the trend line TL2 is based be L TL2 If the load corresponding to the evaluation value of the operation before the change is L0, the slope ΔAL1 of the auxiliary line AL1 can be obtained by ΔAL1 = ΔTL2 / (L TL2 / L0). Hereinafter, the value of L TL2 / L0 may be referred to as the load correction rate.

[0179] Subsequently, the remaining life estimation unit 54 creates an auxiliary line AL1 that has the obtained slope ΔAL1 and passes through the first evaluation value q1 after the change. An example of the auxiliary line AL1 is shown in FIG. 17. The auxiliary line AL1 estimates the transition of the evaluation value assuming that the robot 1 has also performed the operation after the change before the change. L TL2 When L

[0180] < L0, since the load correction rate is less than 1, the slope of the auxiliary line AL1 becomes larger compared to the slope of the trend line TL2.

[0181] Load L TL2 , L0 can be arbitrarily selected as to what values to use. For example, as the load, the root mean square value of the square of the motor current (I2 described above) can be used. As the load, a value obtained by taking the cube root of the cube mean of the absolute value of the current value can be used.

[0182] The load can also be calculated by multiplying the average load torque Tm by the following formula: TL2 , L0.

number

[0183] The trend line TL2 is calculated based on multiple evaluation values, so the load L TL2 It is preferable that L0 is a representative value (for example, a median or an average value) of the corresponding multiple loads. The load L0 of the operation before the change can also be a representative value such as the median or average value of the multiple loads.

[0184] As described above, in the sixth method, when the robot's motion, which is the basis for acquiring the state time-series data, is changed as shown in Fig. 17, the predicted lifespan timing is calculated as follows. That is, the lifespan estimation unit 54 calculates the load L0 based on the motion after the change by multiplying the load L0 based on the motion before the change by the load L0. TL2 The life estimation unit 54 creates a trend line TL2 based on a plurality of evaluation values ​​based on the operation after the change. The life estimation unit 54 creates an auxiliary line AL1 that passes through the evaluation value q1 at the beginning after the change. The slope ΔAL1 of the auxiliary line AL1 is calculated by dividing the slope ΔTL2 of the trend line TL2 by the load correction rate (L TL2 / L0). The value indicated by this auxiliary line AL1 at the time when the elapsed time is zero is obtained as the past pre-rise estimated evaluation value st1x. The timing when the trend line TL2 created based on multiple evaluation values ​​based on the changed operation reaches the life threshold value TH2 applied to the changed operation is defined as the predicted life timing PL2.

[0185] This allows for consideration of changes in load before and after the change in operation, making it possible to more appropriately predict the lifespan based on the operation after the change.

[0186] The seventh method will be explained below with reference to FIG.

[0187] [a] The final evaluation value p2 based on the pre-change reference pattern is divided by the initial evaluation value q1 based on the post-change reference pattern to find the adjustment coefficient ka1 (ka1=p2 / q1). [b] The evaluation value obtained using the changed reference pattern is multiplied by the adjustment coefficient ka1 to obtain an adjusted evaluation value. [c] A trend line TL2 is created based on the multiple adjustment evaluation values ​​obtained for the changed reference pattern. The timing at which this trend line TL2 reaches the life threshold TH1 defined for the original reference pattern is defined as the predicted life timing PL2.

[0188] The adjustment coefficient ka1 can be considered an estimated value that expresses, as a ratio, the effect on the evaluation value due to a change in the robot 1's behavior. Even after the robot 1's behavior pattern is changed, the evaluation value obtained after the change can be made consistent with the evaluation value obtained before the change by a simple calculation of multiplying the adjustment coefficient ka1. With this method, in order to obtain the trend line TL2, it is necessary to obtain evaluation values ​​for at least two times after the robot 1's behavior is changed. To improve the accuracy of the trend line, the trend line TL2 may be obtained on the condition that evaluation values ​​have been obtained a predetermined number of times, for example, five times.

[0189] As described above, in the seventh method, when the robot's behavior, which is the basis for acquiring state time-series data, is changed as shown in Fig. 18, the predicted lifespan timing is determined as follows. That is, the lifespan estimation unit 54 estimates an adjustment coefficient ka1 that expresses, as a ratio, the effect on the evaluation value due to the change in the behavior of the robot 1. The lifespan estimation unit 54 creates a trend line TL2 based on an adjusted evaluation value obtained by multiplying the evaluation value based on the changed behavior by the adjustment coefficient ka1. The lifespan estimation unit 54 determines the timing at which the trend line TL2 reaches the lifespan threshold TH1 applied to the behavior before the change as the predicted lifespan timing PL2.

[0190] In this way, the adjustment coefficient ka1 can be used to match the evaluation value obtained after the operation change with the evaluation value obtained before the change.

[0191] In the seventh method, when the robot's behavior, which is the basis for acquiring the state time series data, is changed as shown in FIG. 18, the lifespan estimation unit 54 calculates the adjustment coefficient ka1 by dividing the final evaluation value p2 based on the behavior before the change by the initial evaluation value q1 based on the changed reference pattern.

[0192] This allows the adjustment coefficient ka1 to be calculated by a simple calculation. Furthermore, the adjustment coefficient ka1 can be calculated even immediately after the robot's operation is changed.

[0193] The eighth method will be described below with reference to FIG.

[0194] [a] The final evaluation value p2 based on the pre-change reference pattern is divided by the initial evaluation value q1 based on the post-change reference pattern to find the adjustment coefficient ka1 (ka1=p2 / q1). [b] The evaluation value of the pre-change reference pattern is used as is. For the evaluation value obtained from the post-change reference pattern, the adjusted evaluation value obtained by multiplying the evaluation value by the adjustment coefficient ka1 is used instead. [c] Based on multiple evaluation values ​​spanning both before and after the change, an overall trend line (trend line) GTL1 is created. The timing at which this overall trend line GTL1 reaches the life threshold TH1 defined for the reference pattern before the change is set as the predicted life timing PL2.

[0195] The predicted life timing PL2 obtained by the eighth method is almost the same as the predicted life timing PL2 obtained by the seventh method described above.

[0196] In the eighth method, as in the seventh method, the adjustment coefficient ka1 can be considered as an estimated value that expresses, in percentage terms, the effect that a change in the behavior of the robot 1 has on the evaluation value. By using this adjustment coefficient ka1, it is possible to appropriately generate a comprehensive trend line GTL1 that spans the evaluation values ​​before and after the change.

[0197] As time passes and evaluation values ​​based on the changed reference pattern are accumulated as appropriate, a trend line TL2 can be generated based only on the changed evaluation values ​​multiplied by the adjustment coefficient ka1. This trend line TL2 is generated according to the seventh method described above. The display of the overall trend line GTL1 may be stopped simultaneously with the display of this trend line TL2 on the display unit 55, or after a predetermined time has elapsed since the display of the trend line TL2.

[0198] As described above, in the eighth method, when the robot's behavior, which is the basis for acquiring state time-series data, is changed as shown in FIG. 19 , the predicted lifespan timing is determined as follows. That is, the lifespan estimation unit 54 estimates an adjustment coefficient ka1 that expresses, as a ratio, the effect on the evaluation value due to the change in the behavior of the robot 1. The lifespan estimation unit 54 creates an overall trend line GTL1 based on the evaluation value based on the behavior before the change and an adjusted evaluation value obtained by multiplying the evaluation value based on the behavior after the change by the adjustment coefficient ka1. The lifespan estimation unit 54 determines the timing at which the overall trend line GTL1 reaches the lifespan threshold TH1 applied to the behavior before the change as the predicted lifespan timing PL2.

[0199] In this way, the adjustment coefficient ka1 can be used to match the evaluation value obtained after the operation change with the evaluation value obtained before the change.

[0200] The ninth method will be described below with reference to FIG.

[0201] [a] A trend line TL1 is created based on a plurality of evaluation values ​​based on the pre-change reference pattern. [b] Calculate the value of the created trend line TL1 at timing tq1 of the initial evaluation value q1 based on the changed reference pattern. Hereinafter, this value will be referred to as the trend line value tv1. [c] The trend line value tv1 is divided by the initial evaluation value q1 based on the changed reference pattern to obtain an adjustment coefficient ka2 (ka2=tv1 / q1). [d] The evaluation value obtained using the changed reference pattern is multiplied by the adjustment coefficient ka2 to obtain an adjusted evaluation value. [e] A trend line TL2 is created based on the multiple adjustment evaluation values ​​obtained for the changed reference pattern. The timing at which this trend line TL2 reaches the life threshold TH1 defined for the original reference pattern is defined as the predicted life timing PL2.

[0202] The tenth method will be described below with reference to FIG. [a] A trend line TL1 is created based on a plurality of evaluation values ​​based on the pre-change reference pattern. [b] Calculate the value of the created trend line TL1 at timing tq1 of the initial evaluation value q1 based on the changed reference pattern. Hereinafter, this value will be referred to as the trend line value tv1. [c] The trend line value tv1 is divided by the initial evaluation value q1 based on the changed reference pattern to obtain an adjustment coefficient ka2 (ka2=tv1 / q1). [d] The evaluation value of the pre-change reference pattern is used as is. For the evaluation value obtained from the post-change reference pattern, the adjusted evaluation value obtained by multiplying the evaluation value by the adjustment coefficient ka2 is used instead. [e] Create an overall trend line GTL1 based on multiple evaluation values ​​spanning both before and after the change. The timing at which this overall trend line GTL1 reaches the life threshold TH1 applied to the reference pattern before the change is set as the predicted life timing PL2.

[0203] Similarly to the seventh method and the like, the ninth and tenth methods also make it possible to match the evaluation value of the robot 1 after the change in the movement pattern with the evaluation value before the change.

[0204] As described above, in the ninth and tenth methods, when the robot's motion, which is the basis for acquiring state time-series data, is changed as shown in Figure 20 or 21, the adjustment coefficient ka2 is estimated as follows. That is, the lifespan estimation unit 54 creates a trend line TL1 based on multiple evaluation values ​​based on the motion before the change, and calculates a trend line value tv1, which is the value indicated by the trend line TL1 at timing tq1 of the initial evaluation value q1 based on the motion after the change. The lifespan estimation unit 54 calculates the adjustment coefficient ka2 by dividing the trend line value tv1 by the initial evaluation value q1 based on the motion after the change.

[0205] This allows the adjustment coefficient ka2 to be calculated by a simple calculation using the trend line TL1. Furthermore, the adjustment coefficient ka1 can be calculated even immediately after the operation of the robot 1 is changed.

[0206] The eleventh method will be described below with reference to FIG.

[0207] [a] A trend line TL1 is created based on multiple evaluation values ​​based on the reference pattern before the change. The value indicated by the created trend line TL1 at the timing of the change of the reference pattern is calculated. Hereinafter, this value will be referred to as the first trend line value tvx1. In the example of FIG. 22, the timing of the change of the reference pattern is the timing when the operator instructs the controller 90 to change the operating program, but this is not limited to this. In this embodiment, normally, when a new reference pattern is registered, an evaluation value based on that reference pattern is also obtained at the same time. Therefore, the timing when the first evaluation value based on the new reference pattern is obtained can also be used as the timing of the change of the reference pattern. [b] Create a trend line TL2 based on multiple evaluation values ​​based on the changed reference pattern. Calculate the value that the created trend line TL2 indicates at the timing of the change in the reference pattern. Hereinafter, this value will be referred to as the second trend line value tvx2. [c] The adjustment coefficient ka3 is calculated by dividing the first trend line value tvx1 by the second trend line value tvx2 (ka3=tvx1 / tvx2). [d] The evaluation value obtained using the changed reference pattern is multiplied by the adjustment coefficient ka3 to obtain an adjusted evaluation value. [e] A trend line TL2x is created based on the multiple adjustment evaluation values ​​obtained for the changed reference pattern. The timing at which this trend line TL2x reaches the life threshold TH1 defined for the original reference pattern is defined as the predicted life timing PL2.

[0208] The twelfth method will be described below with reference to FIG.

[0209] [a] Create a trend line TL1 based on multiple evaluation values ​​based on the reference pattern before the change. Calculate the value (the first trend line value tvx1) indicated by the created trend line TL1 at the timing of the change of the reference pattern. [b] Create a trend line TL2 based on multiple evaluation values ​​based on the changed reference pattern. Calculate the value indicated by the created trend line TL2 at the timing of the change of the reference pattern (the second trend line value tvx2 described above). [c] The adjustment coefficient ka3 is calculated by dividing the first trend line value tvx1 by the second trend line value tvx2 (ka3=tvx1 / tvx2). [d] The evaluation value of the pre-change reference pattern is used as is. For the evaluation value obtained from the post-change reference pattern, the adjusted evaluation value obtained by multiplying the evaluation value by the adjustment coefficient ka3 is used instead. [e] Create an overall trend line GTL1 based on multiple evaluation values ​​spanning both before and after the change. The timing at which this overall trend line GTL1 reaches the life threshold TH1 applied to the reference pattern before the change is set as the predicted life timing PL2.

[0210] Similarly to the seventh method and the like, the eleventh and twelfth methods also make it possible to match the evaluation value of the robot 1 after the change in the movement pattern with the evaluation value before the change.

[0211] As described above, in the eleventh and twelfth methods, when the robot's motion, which is the basis for acquiring the state time-series data, is changed as shown in FIG. 22 or 23 , the adjustment coefficient ka3 is estimated as follows. The lifespan estimation unit 54 creates a trend line TL1 based on a plurality of evaluation values ​​based on the motion before the change, and determines a first trend line value tvx1 indicated by the trend line TL1 at the timing of the change in the motion of the robot 1. The lifespan estimation unit 54 creates a trend line TL2 based on a plurality of evaluation values ​​based on the motion after the change, and determines a second trend line value tvx2 indicated by the trend line TL2 at the timing of the change in the motion of the robot 1. The lifespan estimation unit 54 determines the adjustment coefficient ka3 by dividing the first trend line value tvx1 by the second trend line value tvx2.

[0212] As a result, the adjustment coefficient ka3, which is a ratio representing the effect on the evaluation value due to a change in the behavior of the robot 1, can be accurately estimated using the two trend lines TL1 and TL2.

[0213] In the seventh to twelfth methods, it is possible that the behavior of the robot 1 may be changed multiple times. In this case, "before a change" should be read as "before the most recent change" and "after a change" should be read as "after the most recent change," and the processing described for each method should be performed for each change.

[0214] Next, the estimation of the life span when parts of the robot 1 are replaced will be described.

[0215] For some reason, such as a failure, a part of the robot 1 may be replaced. The robot 1 has many parts, but here we consider a case where a part that is dominant in terms of the lifespan of the robot 1 is replaced. A typical example of such a part is a reducer. The lifespan estimation described here may be applied not only to cases where a reducer is replaced, but also when, for example, an output gear of a reducer, a motor, or the like is replaced.

[0216] When a part is replaced, the worker operates the status monitoring device 5, identifies the axis of the robot 1 whose part has been replaced, and issues a reset command. As a result, an elapsed time reset signal is input to the status monitoring device 5.

[0217] When an operation to reset the elapsed time is performed, the status monitoring device 5 performs a lifespan prediction for the specified axis of the robot 1 without using the data that was used for the lifespan prediction before the reset. However, the basic operation pattern that was set before the reset continues to be used.

[0218] Data removed from the basis of life prediction by resetting the elapsed time includes, for example, time-series data of the status signal, evaluation value, increase value, estimated evaluation value before increase, predicted life timing, trend line, load correction factor, life threshold coefficient other than the predetermined life threshold coefficient, life threshold other than the predetermined life threshold, remaining life rate, adjustment coefficient, and adjusted evaluation value. When the elapsed time is reset, the values ​​before the reset are not carried over. Therefore, for example, if the elapsed time reset operation is performed at the "present" point in Figure 12, all evaluation values, trend lines, etc. that existed in the plot area are erased. After that, evaluation values ​​are plotted at appropriate times as the number of days elapses from day 0, and trend lines, etc. are recreated.

[0219] The data removed from the basis for life prediction by resetting the elapsed time is deleted from the storage unit 52 of the condition monitoring device 5. However, the data may be continuously stored so that the data can be referenced in the future for purposes such as maintenance. The date and time when the elapsed time is reset may be stored in the storage unit 52, and this date and time may be displayed on the display unit 55, for example, simultaneously with the trend line graph.

[0220] As described above, the lifespan estimation unit 54 resets the elapsed time to zero when an elapsed time reset signal indicating that a part of the robot 1 has been replaced is input. The behavior of the robot 1, which is the basis for acquiring the time-series data of the status signal, is the same before and after the elapsed time is reset. The lifespan estimation unit 54 multiplies the initial evaluation value based on the behavior after the elapsed time is reset to zero by a predetermined lifespan threshold coefficient to obtain the lifespan threshold after the reset. A trend line is created based on multiple evaluation values ​​based on the behavior after the elapsed time has been reset to zero, and the timing at which the trend line reaches the lifespan threshold after the reset is determined as the predicted lifespan timing.

[0221] This allows for appropriate lifespan estimation, taking into account resetting of deterioration due to part replacement.

[0222] Although the preferred embodiment of the present disclosure has been described above, the above configuration can be modified, for example, as follows. A single modification may be made, or multiple modifications may be made in any combination.

[0223] The status monitoring device 5 does not need to be directly connected to the robot 1, and may instead 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 the Internet. In this case, the controller 90 acquires and saves the current values ​​in real time during the playback operation, and transmits the time-series data of the current values ​​together with the program number, information identifying the reference pattern, the date and time the current values ​​were acquired, and information identifying the servo motor by batch processing or the like to the status monitoring device 5.

[0224] The status monitoring device 5 may not be provided separately from the controller 90, but may be built into the controller 90. Furthermore, the status monitoring device 5 may be realized using the computer of the controller 90 of the robot 1, without providing a computer that functions as a CPU, ROM, RAM, auxiliary storage device, etc. of the status monitoring device 5. In this case, the display unit 55 may be configured as a part of the teaching pendant of the robot 1, for example.

[0225] The reference pattern may be extracted from the position time series data manually rather than automatically. For example, an operator may specify a program number, program name, etc., display a graph of the position time series data as shown in Figure 6 on the display unit 55, and specify the start and end of a time interval on the screen to extract the reference pattern.

[0226] In the first method, the increase ratio may be less than 1. In this case, the increase ratio can be considered to be 1. In the second method or its variations, for example, the trend line may be horizontal or slope downward to the right, making it impossible to determine the timing at which the lifespan threshold is reached. In this case, the remaining life rate can be considered to be 1. These two situations are examples of potential problems with appropriate robot lifespan prediction, but are not limited to these. Such situations may be caused by an inappropriate reference pattern on which the evaluation value is based. Therefore, the condition monitoring device 5 may discard the currently used reference pattern and obtain a new reference pattern when it becomes clear that a problem is occurring in the calculation for lifespan prediction.

[0227] In the second and subsequent methods, the evaluation value in the early stages can be a representative value such as the median or average of multiple evaluation values ​​obtained around the first time.

[0228] In the second and subsequent methods, the final evaluation value based on the pre-change behavior may be a representative value such as the median or average of the evaluation values ​​most recently obtained.

[0229] In the eleventh method, the predicted life timing is calculated using the first trend line value and the second trend line value. This method of evaluating the effect of a change in the robot's operation on the evaluation value using two trend line values ​​can be applied to other methods as long as there is no contradiction.

[0230] In the above-mentioned 12 methods, the predicted life timing is calculated by consistently handling the evaluation values ​​before and after the change in the robot's operation by changing the life threshold or adjusting the evaluation value, etc. Alternatively, or in addition to this, the time axis of the transition of the evaluation value after the change in operation may be corrected by an appropriate method.

[0231] The evaluation interval in the state time-series data can be set by any other appropriate method instead of pattern matching with the reference pattern.

[0232] The flowcharts shown in the above embodiments are merely examples, and some processes may be omitted, the contents of some processes may be changed, or new processes may be added.

[0233] The functions of the elements disclosed herein can be performed using circuits or processing circuitry, including general-purpose processors, special-purpose processors, integrated circuits, ASICs (Application Specific Integrated Circuits), conventional circuits, and / or combinations thereof, configured or programmed to perform the disclosed functions. A processor is considered a processing circuit or circuitry because it includes transistors and other circuitry. In this disclosure, a circuit, unit, or means is hardware that performs the recited functions or hardware that is programmed to perform the recited functions. The hardware may be hardware disclosed herein or other known hardware that is programmed or configured to perform the recited functions. Where the hardware is a processor, which is considered a type of circuit, the circuit, means, or unit is a combination of hardware and software, and the software is used to configure the hardware and / or processor.

[0234] From the above disclosure, at least the following technical ideas can be grasped.

[0235] [1] A status monitoring device for monitoring the status of a robot capable of reproducing predetermined actions, a motion state time measurement unit that determines a motion state of the robot and measures a motion state time that is a time during which the motion state of the robot continues; a basic data processing unit that calculates the number of basic data acquisitions by dividing the motion state time by a preset basic data upper limit length, acquires the number of basic data acquisitions from position time-series data of each axis of the motion state of the robot corresponding to the motion state time, and stores at least one basic data; a search processing unit that acquires repeatedly reproduced position time series data for each axis and determines whether or not the data matches the stored basic data; a time series data acquisition unit that acquires time series data of a state signal of each axis that reflects the state of the robot in a time interval that matches the basic data; an evaluation value calculation unit that acquires an evaluation value for evaluating the state of the robot based on the time series data acquired by the time series data acquisition unit; a lifespan estimation unit that creates a trend line that indicates the tendency of the evaluation value to change over time and determines the timing at which the trend line reaches a predetermined lifespan threshold as a predicted lifespan timing; A condition monitoring device comprising:

[0236] This makes it possible to appropriately obtain basic data for predicting the remaining life from position time-series data when the robot is actually operating by executing a program for an actual task.

[0237] [2] [1] A condition monitoring device, Further provided is an operation change monitoring unit that monitors the repeatedly reproduced state signals of each axis, the operation change monitoring unit monitors whether or not a portion matching the basic data appears in a state signal of each axis that reflects the state of the robot; When the number of occurrences within a predetermined monitoring period is equal to or less than a certain number, the basic data processing unit acquires new basic data and replaces the basic data whose number of occurrences is equal to or less than the certain number.

[0238] This allows the basic data to be changed accordingly even when the robot's operation for the actual work is changed, ensuring continuity in the collection of evaluation values.

[0239] [3] A condition monitoring device according to [1] or [2], A status monitoring device that determines that the robot is in an operating state when one or more of the multiple axes possessed by the robot are moving.

[0240] This makes it possible to collect basic data and obtain evaluation values ​​in a state suitable for state evaluation for at least one or more axes of the robot.

[0241] [4] A condition monitoring device according to any one of [1] to [3], A status monitoring device that determines that the robot is not in an operating state when all axes of the robot are stopped for a predetermined period of time or longer.

[0242] This allows the robot to be determined to be in an operating state including the short stoppage period even if all axes stop for a short period of time during operation.

[0243] [5] A condition monitoring device according to [1] or [2], A status monitoring device that determines that the robot is in an operating state when all of the multiple axes of the robot are moving.

[0244] This allows basic data to be collected and evaluation values ​​to be obtained in a state where all axes of the robot are suitable for condition evaluation.

[0245] [6] A condition monitoring device according to any one of [1] to [5], A status monitoring device that does not determine that a robot is in an operating state during an initial predetermined time period during which the operating state of the robot is determined.

[0246] This prevents the basic data from including a positioning operation or the like when the robot performs a positioning operation or the like at the beginning of the execution of the program.

[0247] [7] A condition monitoring device according to any one of [1] to [6], The number of basic data acquisitions is a value obtained by dividing the operating state time by the basic data upper limit length, and rounding up the decimal point to an integer.

[0248] In this case, an appropriate number of basic data can be obtained depending on the situation from position time-series data corresponding to one operating state time. Even if the position time-series data is long, basic data can be obtained to cover the entire data.

[0249] [8][7] A condition monitoring device, The basic data processing unit extracts a plurality of the basic data from the position time-series data so that the basic data overlap with each other.

[0250] By allowing partial overlap of extraction, it is possible to obtain as many basic data as necessary from the position time series data.

[0251] [9] A condition monitoring device according to any one of [1] to [6], The number of basic data acquisitions is an integer obtained by rounding down the decimal point of a value obtained by dividing the operating state time by the basic data upper limit length, in the status monitoring device.

[0252] In this case, an appropriate number of basic data can be obtained depending on the situation from position time-series data corresponding to one operating state time. Since multiple basic data can be extracted without overlapping with each other, substantial duplication of processing can be prevented.

[0253]

[10] A condition monitoring device according to any one of [1] to [6], The basic data processing unit acquires one piece of basic data from position time series data of each axis of the motion state of the robot corresponding to the motion state time, and stores the basic data.

[0254] In this case, one basic data is obtained from the position time series data, which simplifies the processing.

[0255]

[11]

[10] A condition monitoring device, The basic data processing unit acquires one basic data of the basic data upper limit length from the end of the position time series data of each axis of the motion state of the robot corresponding to the motion state time.

[0256] This prevents the basic data from including a positioning operation or the like when the robot performs a positioning operation or the like at the beginning of the execution of the program.

[0257]

[12] A status monitoring device for monitoring the status of a robot capable of reproducing a predetermined motion, a time-series data acquisition unit that acquires time-series data of a status signal that reflects the status of the robot; a storage unit that stores the time series data acquired by the time series data acquisition unit; a lifespan estimation unit that acquires an evaluation value for evaluating the state of the robot based on the time-series data, creates a trend line that indicates the tendency of the evaluation value to change over time, and determines the timing at which the trend line reaches a predetermined lifespan threshold as a predicted lifespan timing; and Equipped with when a robot operation that is a basis for acquiring the time-series data of the state signal is changed, the lifespan estimation unit calculates the lifespan threshold value to be applied to the changed operation; A condition monitoring device that creates a trend line based on a plurality of evaluation values ​​based on the changed operation, and determines the predicted life timing as the timing when the trend line reaches the life threshold applied to the changed operation.

[0258]

[13] A status monitoring device for monitoring the status of a robot capable of reproducing a predetermined motion, a time-series data acquisition unit that acquires time-series data of a status signal that reflects the status of the robot; a storage unit that stores the time series data acquired by the time series data acquisition unit; a lifespan estimation unit that acquires an evaluation value for evaluating the state of the robot based on the time-series data, creates a trend line that indicates the tendency of the evaluation value to change over time, and determines the timing at which the trend line reaches a predetermined lifespan threshold as a predicted lifespan timing; and Equipped with When the operation of the robot that is the basis for acquiring the time-series data of the status signal is changed, the lifespan estimation unit (1) determining the lifespan threshold to be applied to the changed operation, creating a trend line based on the plurality of evaluation values ​​based on the changed operation, and determining the timing at which the trend line reaches the lifespan threshold to be applied to the changed operation; or (2) Estimate an adjustment coefficient that expresses the ratio of the effect of a change in the robot's behavior on the evaluation value, multiply the evaluation value based on the changed behavior by the adjustment coefficient to obtain an adjusted evaluation value, create a trend line based on the multiple adjusted evaluation values ​​based on the changed behavior, and calculate the timing at which the trend line reaches the lifespan threshold applied to the behavior before the change. A condition monitoring device in which the predicted lifespan timing is set as the predicted lifespan timing.

Claims

1. A status monitoring device that monitors the status of a robot that can reproduce a predetermined action, a motion state time measurement unit that determines a motion state of the robot and measures a motion state time that is a time during which the motion state of the robot continues; a basic data processing unit that calculates the number of basic data acquisitions by dividing the motion state time by a preset basic data upper limit length, acquires the number of basic data acquisitions from position time-series data of each axis of the motion state of the robot corresponding to the motion state time, and stores at least one basic data; a search processing unit that acquires repeatedly reproduced position time series data for each axis and determines whether or not the data matches the stored basic data; a time series data acquisition unit that acquires time series data of a state signal of each axis that reflects the state of the robot in a time interval that matches the basic data; an evaluation value calculation unit that acquires an evaluation value for evaluating the state of the robot based on the time series data acquired by the time series data acquisition unit; a lifespan estimation unit that creates a trend line that indicates the tendency of the evaluation value to change over time and determines the timing at which the trend line reaches a predetermined lifespan threshold as a predicted lifespan timing; A condition monitoring device comprising:

2. The condition monitoring device according to claim 1, Further provided is an operation change monitoring unit that monitors the repeatedly reproduced state signals of each axis, the operation change monitoring unit monitors whether or not a portion matching the basic data appears in a state signal of each axis that reflects the state of the robot; When the number of occurrences within a predetermined monitoring period is equal to or less than a certain number, the basic data processing unit acquires new basic data and replaces the basic data whose number of occurrences is equal to or less than the certain number.

3. The condition monitoring device according to claim 1, A status monitoring device that determines that the robot is in an operating state when one or more of the multiple axes of the robot are moving.

4. The condition monitoring device according to claim 1, A status monitoring device that determines that the robot is not in an operating state when all axes of the robot are stopped for a predetermined period of time or longer.

5. The condition monitoring device according to claim 1, A status monitoring device that determines that the robot is in an operating state when all of the multiple axes of the robot are moving.

6. The condition monitoring device according to claim 1, A status monitoring device that does not determine that a robot is in an operating state during an initial predetermined time period during which the operating state of the robot is determined.

7. The condition monitoring device according to claim 1, The number of basic data acquisitions is a value obtained by dividing the operating state time by the basic data upper limit length, and rounding up the decimal point to an integer.

8. The condition monitoring device according to claim 7, The basic data processing unit extracts a plurality of the basic data from the position time-series data so that the basic data overlap with each other.

9. The condition monitoring device according to claim 1, The number of basic data acquisitions is an integer obtained by rounding down the decimal point of a value obtained by dividing the operating state time by the basic data upper limit length, in the status monitoring device.

10. The condition monitoring device according to claim 1, The basic data processing unit acquires one piece of basic data from position time series data of each axis of the motion state of the robot corresponding to the motion state time, and stores the basic data.

11. The condition monitoring device according to claim 10, The basic data processing unit acquires one basic data of the basic data upper limit length from the end of the position time series data of each axis of the motion state of the robot corresponding to the motion state time.

Images