Tact division data collection system and anomaly detection system

Abstract

To provide a tact-divided data collecting system and an abnormality detecting system capable of realizing tact extraction at the same time as logging and reducing the number of devices, thereby collecting the tact-divided data at a low cost.SOLUTION: The tact-divided data collecting system for solving the above problem is provided with a processor. The processor of the tact-divided data collecting system, on the basis of a trigger waveform which is a part of a tact waveform being a repetitive waveform, and which is previously stored in a storage device, performs the processing of determining whether the tact waveform in time series data acquired from a sensor includes the trigger waveform. Then, when the processor of the tact-divided data collecting system determines that the tact waveform includes the trigger waveform, the tact waveform data including the trigger waveform is collected.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present invention relates to a tact division data collection system and an anomaly detection system. [Background technology]

[0002] The frequency of failures of industrial equipment varies greatly depending on the number of years since its introduction. Statistics show that there are three stages: early failure, which occurs 1-2 years after introduction, when failures occur due to design or manufacturing defects; random failure, which occurs 3-6 years after introduction, when failures are unlikely to occur and operation is stable; and wear-out failure, which occurs 7-8 years after introduction, when failures increase due to the lifespan of the machine. The early failure stage is supported by the manufacturer's response to initial defects, but the wear-out failure stage is difficult to predict as it depends on the machine's usage status and the frequency of maintenance by the user. If the user does not grasp the condition, it will lead to a decrease in operation rate and a decrease in productivity. For equipment in this stage, it is often difficult to introduce the latest monitoring solutions from the manufacturer because the models are old, so there is an increasing need for solutions that add on new general-purpose sensors to monitor and detect abnormalities. Such add-on sensors include temperature sensors, vibration sensors, and in the case of equipment driven by a motor, current sensors that measure the motor drive current are used. When the target equipment is a workpiece-based processing machine such as a machining center, or a vertical articulated robot that transports workpieces, the measurement data is a repetition of "takts," which indicate a series of processes, so abnormality detection typically involves a method of detecting differences by comparing repeated takts.

[0003] Patent Document 1 proposes a method for implementing tact division by performing pattern matching between a predetermined tact reference waveform and acquired data. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2020-187502 A Summary of the Invention [Problem to be solved by the invention]

[0005] In the case of anomaly detection using an add-on sensor, it is often difficult to obtain a trigger signal for the start of a tact from the control system due to the desire to minimize the impact on existing equipment. Since there is often variation between tacts due to waiting for work or waiting for peripheral equipment to work, tact division may fail when extracting a fixed cycle at the tact length, and the challenge is how to achieve highly accurate tact division without a trigger signal.

[0006] In order to satisfy the above requirements, the system described in Patent Document 1 provides a learning phase before operation, identifies and stores a tact reference waveform from previously acquired data, and performs pattern matching between the data acquired up to that point and the tact reference waveform at regular intervals during operation. When the degree of matching exceeds a certain amount, it is determined that a tact exists, and the data is extracted as tact data. However, this method has two drawbacks. The first is that the pattern matching is performed on the entire tact reference waveform, which requires a high calculation load and is considered to be difficult to implement simultaneously in a data logging device. The second is that when matching is established, the extracted acquired data is handled collectively, so if the system is to save the data in a file, the data I / O load is large, which is likely to affect data logging, and is considered to be difficult to implement simultaneously in a data logging device. Due to the above two drawbacks, it is necessary to prepare a tact division device in addition to the data logging device, which increases the cost of the device.

[0007] The present invention has been made in consideration of the above points, and has as its object to realize tact extraction simultaneously with logging, reduce the number of devices, and collect tact division data at low cost. [Means for solving the problem]

[0008] According to a first aspect of the present invention, there is provided a tact division data collection system as described below. That is, the tact division data collection system includes a processor. The processor determines whether a trigger waveform is included in a tact waveform in time-series data acquired from a sensor, based on a trigger waveform that is a part of a tact waveform, which is a repetitive waveform, and that is stored in advance in a storage device. When the processor determines that a trigger waveform is included in the tact waveform, it collects data of the tact waveform including the trigger waveform.

[0009] According to a second aspect of the present invention, the following anomaly detection system is provided. That is, the anomaly detection system includes a tact division data collection module and an anomaly detection module. The tact division data collection module includes a sensor data acquisition unit, a tact waveform storage unit, a trigger waveform storage unit, a tact start determination unit, and a tact data storage unit. The sensor data acquisition unit acquires time series data from the sensor. The tact waveform storage unit stores one of the repeated waveforms included in the learning data as a tact reference waveform. The trigger waveform storage unit stores a waveform for a certain time period from the beginning of the tact reference waveform as a trigger waveform that is a tact start trigger. The tact start determination unit includes a process of extracting data for a trigger waveform length from the operation data and calculating the similarity with the trigger waveform. The tact data storage unit stores the operation data as a file for each tact length according to the calculation result. The anomaly detection module performs normal / abnormal determination using the obtained data for each tact. Effect of the Invention

[0010] According to the present invention, a system is provided that can collect tact division data at low cost by realizing tact extraction at the same time as logging and reducing the number of devices. Note that the problems, configurations and effects other than those described above will become clear from the following description of the embodiment of the invention. [Brief description of the drawings]

[0011] [Figure 1]FIG. 1 is a diagram showing the appearance of a data collection system and a target device according to a conventional example. [Diagram 2] FIG. 1 is a diagram showing a power supply circuit of a device targeted by a data collection system according to a conventional example. [Figure 3A] FIG. 13 is a diagram showing current waveforms obtained by sensing a target device of the data collection system according to the conventional example and the present embodiment, and shows an example of a current waveform including multiple tacts. [Figure 3B] FIG. 13 is a diagram showing current waveforms obtained by sensing a target device of the data collection system according to the conventional example and the present embodiment, and is a diagram showing an example of a reference waveform showing one tact. [Figure 4] FIG. 11 is a flowchart showing the process flow of a conventional tact division data collection system. [Figure 5A] FIG. 13 is a block diagram of a tact division data collection system according to a conventional example, showing a block diagram during learning. [Figure 5B] FIG. 1 is a block diagram of a tact division data collection system according to a conventional example, showing a block diagram during operation. [Figure 6] FIG. 2 is a diagram showing the external appearance of an example of a tact division data collection system and a target device according to the embodiment. [Figure 7] FIG. 2 is a diagram showing an example of a power supply circuit of a device targeted by the data collection system according to the embodiment. [Figure 8] FIG. 4 is a diagram showing an example of a trigger waveform indicating the start of acquisition of one tact of a current waveform obtained by sensing a target device in the tact division data collection system according to the first embodiment. [Figure 9] FIG. 2 is a flowchart showing an example of a process flow of the tact division data collection system according to the first embodiment. [Figure 10A] FIG. 11 is a block diagram showing an example of processing of the tact division data collection system according to the embodiment, showing a block diagram during learning. [Figure 10B] FIG. 2 is a block diagram showing an example of processing of the tact division data collection system according to the embodiment, and shows a block diagram during operation. [Figure 11]FIG. 4 is a diagram showing an example of a trigger waveform indicating the start of acquisition of one tact of a current waveform obtained by sensing a target device in the tact division data collection system according to the first embodiment. [Figure 12] FIG. 11 is a flowchart showing an example of a process flow of a tact division data collection system according to the second embodiment. [Figure 13] FIG. 11 is a flowchart showing an example of a process flow of a tact division data collection system according to the third embodiment. [Figure 14] FIG. 13 is a flowchart showing an example of a process flow of a tact division data collection system according to the fourth embodiment. [Figure 15A] FIG. 13 is a diagram showing a current waveform obtained by sensing a target device in the tact division data collection system according to the fifth embodiment, and is a diagram showing an example of a raw current waveform. [Figure 15B] FIG. 13 is a diagram showing a current waveform obtained by sensing a target device in the tact division data collection system according to the fifth embodiment, and is a diagram showing an example of a moving RMS waveform. [Figure 16] FIG. 13 is a flowchart showing an example of a process flow of a tact division data collection system according to the fifth embodiment. [Figure 17] FIG. 13 is a flowchart showing an example of a process flow of a tact division data collection system according to the sixth embodiment. [Figure 18] FIG. 23 is a diagram showing the external appearance of an example of an anomaly detection system and a target device according to the seventh embodiment. [Figure 19] FIG. 23 is a flowchart showing an example of the processing flow of the anomaly detection system according to the seventh embodiment. [Figure 20A] FIG. 23 is a block diagram showing an example of processing of the anomaly detection system according to the seventh embodiment, in a learning state. [Figure 20B] FIG. 23 is a block diagram showing an example of processing of the anomaly detection system according to the seventh embodiment, in operation. [Figure 21] FIG. 2 is a hardware configuration diagram of an example of a system configuration. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] As described above, in the tact division data collection system using the mechanism for pattern matching with the reference waveform described in Patent Document 1, pattern matching is performed with the acquired data for the tact length and the tact reference waveform, and when a match is established, the data can be saved all at once. As described above, this type of method has a high calculation load and I / O load, and cannot be implemented simultaneously on logging equipment, which may lead to increased costs due to the addition of equipment.

[0013] First, an example of devices and sensor data targeted by a conventional tact division data collection system and an example of a system configuration will be described.

[0014] Fig. 1 is a diagram showing an example of a conventional tact division data collection system and a target device, and the target device is a robot. In this system, the robot is configured to include a control box 200 and a robot unit 300. The robot unit 300 has multiple operating axes, and a first operating axis 301, a second operating axis 302, a third operating axis 303, a fourth operating axis 304, a fifth operating axis 305, and a sixth operating axis 306 can each rotate in the direction of the arrow.

[0015] The tact division data collection system 1 is connected to a control box 200 of a target device. The tact division data collection system 1 includes a storage unit 10 and a processing unit 20. The storage unit 10 stores a tact reference waveform storage unit 11 that stores a reference waveform of a tact, and a tact data storage unit 13 that saves data after division. The processing unit 20 includes a sensor data acquisition unit 21 and a tact start determination unit 22.

[0016] 2 is a diagram showing an example of a sensing target circuit of a target device of the tact division data collection system. Power is supplied from the control box 200 to each of the drive motors in the robot section 300. The robot section 300 includes multiple drive motors, and the rotation of each drive motor is transmitted to a reducer, which then operates a movable shaft provided at the end of the reducer. The reducer is a mechanism that reduces the input rotation speed and increases the torque instead.

[0017] 2, the control box 200 includes a power supply circuit 210 and a control board 220, and power supplied from the switchboard 100 is passed to the power supply circuit 210 and then passed through the control board 220 to be supplied to each motor. The control board 220 can supply power to a plurality of power consuming devices. A first motor 321, a second motor 322, a third motor 323, a fourth motor 324, a fifth motor 325, and a sixth motor 326 are mounted on the robot unit 300 as drive motors.

[0018] A first reducer 311, a second reducer 312, a third reducer 313, a fourth reducer 314, a fifth reducer 315, and a sixth reducer 316 are connected to each of the drive motors. A first movable shaft 301, a second movable shaft 302, a third movable shaft 303, a fourth movable shaft 304, a fifth movable shaft 305, and a sixth movable shaft 306 are connected to each of the reducers. In the robot section 300, the motors connected to each movable shaft are driven by the magnitude and frequency of the current output from the control board 220 according to the operation program. Therefore, it is known that these current waveforms are important physical quantities that reflect the deterioration state of the motors of each movable shaft of the robot and the reducers and movable shafts that are mechanical parts connected to the motors. In the conventional example, a current sensor is clamped to a power supply circuit connected to the first motor to perform sensing.

[0019] Fig. 3 is a simplified waveform showing a current waveform acquired by a current sensor, Fig. 3A is an example of a current waveform including multiple tacts, and Fig. 3B is a diagram showing an example of a reference waveform showing one tact. One tact indicates the minimum unit of a series of repeated movements of a robot, and Fig. 3A shows continuous data including four tacts, and the tact division data collection system is a system that acquires the continuous data and outputs the data for one tact in Fig. 3B as a file. Fig. 4 is a diagram showing the process flow of a conventional tact division data collection system, and Fig. 5 is a block diagram of the conventional tact division data collection system.

[0020] The learning process of the conventional tact division data collection system will be described with reference to Figures 1 to 5. As described above, a current sensor is clamped to the wiring between the first motor that operates the first movable axis 301 of the robot part 300 in Figures 1 and 2 and the control board 220, and a continuous current waveform in which multiple tacts are connected in Figure 3A is acquired. This flow is the procedure shown by items 0 and 1 (S1, S2) in Figure 4, and corresponds to the block diagram during learning in Figure 5A. The tact division data collection system 1 observes this continuous current waveform, identifies repeated tact waveforms, and stores them as tact reference waveforms.

[0021] The sensor data acquisition unit 21 collects and saves sensor data for a predetermined period of time from 100 msec to 1 sec as a file. In this example, data for 1 sec is collected and saved in one temp file (temporary file), and the tact length is, for example, 60 sec.

[0022] In the identification process in the tact reference waveform storage unit 11, the saved 1-second files are linked and analyzed, for example, for one hour, i.e., 3,600 files, to determine the existence of a 60-second tact, and the 60-second tact reference waveform is saved. Next, the process during operation will be explained. During operation, as shown in item 2 (S3) of Figure 4, 60 temp files are acquired, joined, and saved, and the process of comparing with the reference waveform in items 3 and 4 (S4, S5) is prepared. In the conventional example, the comparison means uses the statistics shown in the following formula.

[0023]

number

[0024] Here, RMS is the Root Mean Square, n is the amount of data for the tact length, and x iindicates the sensor value, i.e., the current value, and one value is calculated for the entire tact waveform. In item 3, the RMS of the tact reference waveform identified and stored in advance and the RMS of the acquired combined data for 60 seconds obtained during operation are calculated. In item 4, it is determined whether the absolute value of this RMS difference is less than a predetermined value; for example, if the RMS value of the tact reference waveform is 8 [Arms] and the RMS value of the acquired combined data is 8.5 [Arms], the absolute value of the RMS difference will be 0.5 [Arms].

[0025] If the threshold value is 0.3 [Arms], the absolute difference value of the RMS is equal to or greater than the threshold value, so the process moves to step 5 (S6). In step 5, the acquired combined data is deemed not to match the tact reference waveform, so the process deletes one file starting from the oldest in order to update the acquired combined data. By doing this, when returning to step 2, a new file is added, and the process moves to the tact start judgment step again.

[0026] This loop of steps 2 to 5 is repeated, and when the absolute difference value of the RMS becomes equal to or less than the threshold, it is assumed that the tact was present in the acquired combined data, and the process proceeds to step 6 (S7). Step 6 is a process for saving the acquired combined data as a tact file, and after saving it in the tact data storage unit, step 7 (S8) is performed to delete all temp files related to the acquired combined data. Basically, a tact does not overlap with the next tact, so the data is reset in this way and the next tact is awaited. In step 0 (S1), the processing unit 20 executes the sensor data acquisition unit 21, and in steps 1 to 7 (S2 to S8), the processing unit 20 executes the tact start determination unit 22.

[0027] It is possible to build a takt division data collection system in the conventional example using the above flow, but since statistical value calculation processing for the takt time is performed in item 3, there is a possibility that it will affect sensor data acquisition work that is processed in real time. Therefore, as shown by the dotted arrow in Figure 4, there is a disadvantage that the cost of the equipment will increase by dividing the equipment that performs the processing. In addition, since it is necessary to save the data for the takt time in a lump in order to save the data for the takt time in item 6, there is a possibility that it will affect the real-time sensor data acquisition process from the viewpoint of I / O, and there is a disadvantage that the cost of the equipment may increase in this respect as well.

[0028] Next, an embodiment that solves these problems will be described. The embodiment is an example for explaining the present invention, and is omitted and simplified as appropriate for clarity of explanation. The present invention can be implemented in various other forms. Unless otherwise specified, each component may be singular or plural. The position, size, shape, range, etc. of each component shown in the drawings may not represent the actual position, size, shape, range, etc., in order to facilitate understanding of the invention. Therefore, the present invention is not necessarily limited to the position, size, shape, range, etc. disclosed in the drawings.

[0029] In the following embodiments, the calculation load and I / O load can be reduced, and a low-cost tact division data collection system can be realized. From the viewpoint of reducing costs, resource saving can also be achieved.

[0030] In the following embodiments, the same reference numerals denote the same configurations even if the drawing numbers are different, and the same effects are achieved. In addition, the description of the contents that have already been described or the similar contents may be omitted.

[0031] First Embodiment The tact division data collection system in the first embodiment has a configuration shown in FIG. 6, and the power supply configuration and sensor installation locations of the target device can be the same as those of the conventional example described above with reference to FIG. 2, as shown in FIG. 7, as an example. That is, the tact division data collection system 3 includes a storage unit 30 and a processing unit 40. The target device can be a robot, and the robot includes a control box 500 and a robot unit 600. The control box 500 includes a power supply circuit 510 and a control board 520, and the power supplied from the switchboard 400 is transferred to the power supply circuit 510 and passed through the control board 520 to be supplied to each motor. The robot unit 600 includes each motor (621-626), each reducer (611-616), and each movable shaft (601-606). FIG. 8 is a diagram showing an example of a trigger waveform according to the first embodiment, FIG. 9 is a flowchart showing an example of a processing flow of the tact division data collection system according to the first embodiment, and FIG. 10 is a block diagram showing an example of a processing of the tact division data collection system according to the first embodiment.

[0032] A common feature of the embodiments of the present invention is the double frame and diagonal line filling process described in items 2, 4, and 7 (S12, S14, and S17) in the flowchart of Fig. 9. Here, during learning, the trigger waveform is identified after the tact reference waveform is identified.

[0033] This trigger waveform is data that is a part of the tact reference waveform, and its role is to determine the presence or absence of a tact with a data amount less than the tact length. As shown in Fig. 10A, the tact reference waveform is identified in the same procedure as the conventional example, and the tact reference waveform is stored in the tact reference waveform storage unit 31.

[0034] Then, using that data, a unique waveform as shown in FIG. 8 is identified in the trigger waveform storage unit 32, which exists within the tact and does not exist repeatedly within the tact. Here, this unique waveform is assumed to be, for example, the first 10 seconds. After identification, the unique waveform is stored as a trigger waveform in the trigger waveform storage unit 32. This procedure corresponds to steps 0 to 2 (S10 to S12) in FIG. 8.

[0035] Then, during operation, as shown in item 3 (S13) of FIG. 9, only the data for 10 seconds for the trigger is acquired, combined, and stored. In the next item 4 (S14), the same statistical calculation and comparison as in the conventional example is performed on the data for the trigger waveform length. If the statistical difference absolute value is equal to or less than the threshold value in item 5 (S15), the process proceeds to item 7 (S17). In item 7 (S17), the acquired data is added to the data for the trigger waveform, and the remaining data for the tact length is acquired from the file one by one. Then, when the tact length is reached, the data is saved collectively as a tact file as described in item 8 (S18). In addition, all of the combined temp files are deleted in item 9 (S19).

[0036] In this embodiment, as shown in Fig. 10B, the tact start determination calculation can be performed using data for a trigger waveform length that is shorter than the tact length. For example, as described above, when the tact length is 60 seconds and the trigger waveform length is 10 seconds, the tact start determination calculation can be reduced to 1 / 6 compared to the conventional example, and the calculation cost of the tact division data collection system 3 can be reduced.

[0037] Also, as shown in Figure 11, it is possible to take the middle part of the tact reference waveform as the trigger waveform, and in that case, it is necessary to temporarily store data from the beginning to the start of the trigger waveform together with the data for the trigger waveform. Since only the data for the length of the trigger waveform is used in the tact start judgment calculation, the same effect of reducing calculation costs can be obtained, but considering the memory cost of temporarily storing it, it is possible to suppress the increase in memory costs by using a unique waveform from the beginning as the trigger waveform. Also, based on the same viewpoint, the trigger waveform at the end part of the tact reference waveform may be used.

[0038] In the first embodiment, as an example, the processing unit 40 executes the sensor data acquisition unit 41 in item number 0 (S10), and executes the tact start determination unit 42 in items 1 to 9 (S11 to S19).

[0039] <Second embodiment> Fig. 12 is a flow chart showing the flow of processing of the tact division data collection system according to the second embodiment, and the tact division data collection system in the second embodiment has the same configuration as that shown in Fig. 6, and the power supply configuration of the target device and the sensor installation locations are the same as the example explained using Fig. 7. Also, the block diagram of the processing of the tact division data collection system according to the second embodiment is the same as that of Fig. 10.

[0040] Hereinafter, common features of the embodiments of the present invention are the temporary data storage format described in item 0 in the flowchart of FIG. 12 and the use of normalized cross correlation (NCC) for tact start determination described in items 4 and 5.

[0041] In the first embodiment, data is temporarily stored in a file, whereas in the second embodiment, data is temporarily stored in memory. This makes it possible to reduce the I / O load when saving acquired data. However, since extra memory storage area is required, it is necessary to select each configuration according to the device to be used. The main effect of the second embodiment is achieved by using another feature, the NCC. Here, the NCC is calculated by the following formula.

[0042]

number

[0043] Here, n is the amount of data for the trigger waveform length, and x i is the length of the acquired trigger waveform data, y i indicates a trigger waveform identified in advance. NCC is an index used when evaluating the similarity between two waveforms. Compared to using statistics as shown in the first embodiment, by using NCC, the similarity of the entire waveform can be evaluated and the threshold can be determined by a relative value called similarity. For example, NCC can calculate the similarity of two waveforms as a numerical value between 0 and 1, which has the effect of reducing the SI man-hours required to select a threshold for each target.

[0044] According to this configuration, it is possible to obtain the effect of reducing the man-hours required for setting the threshold value for tact start judgment while improving the accuracy of tact start judgment. In addition, although the judgment method using NCC has been described in this embodiment, other representative methods for calculating the similarity between waveforms, such as sum of squared differences (SSD) and sum of absolute differences (SAD), can also be used, and it is possible to improve the accuracy of tact start judgment in the same way as when NCC is used.

[0045] In the second embodiment, as an example, the processing unit 40 executes the sensor data acquisition unit 41 in item number 0 (S20), and executes the tact start determination unit 42 in items 1 to 9 (S21 to S29).

[0046] <Third embodiment> Fig. 13 is a flow chart showing an example of the process flow of the tact division data collection system according to the third embodiment, and the tact division data collection system in the third embodiment has the same configuration as that shown in Fig. 6, and the power supply configuration of the target device and the sensor installation locations are the same as the example explained using Fig. 7. Also, the block diagram of the process of the tact division data collection system according to the third embodiment is the same as that of Fig. 10.

[0047] A common feature of the embodiments of the present invention described below is the double border and diagonal line filling process described in items 7 and 8 in the flowchart of Fig. 13. In the third embodiment, when saving data after it is determined that a tact has started in the tact start determination in item 5 (S35), the data held for the trigger is saved all at once, and the data for the remaining trigger length thereafter is added to the file sequentially each time it is obtained from the file. In this way, by saving most of the data sequentially rather than all at once, it is possible to reduce the I / O processing load, which was a problem in the conventional example.

[0048] According to this configuration, it is possible to obtain an effect of suppressing the I / O processing load when saving tact data, and suppressing the influence on real-time processing of acquiring sensor data. In the third embodiment, as an example, in steps 7 to 8 (S37 to S38), the processing unit 40 executes the tact start determination unit 42.

[0049] <Fourth embodiment> Fig. 14 is a flow chart showing an example of the process flow of the tact division data collection system according to the fourth embodiment, and the tact division data collection system in the fourth embodiment has the same configuration as that shown in Fig. 6, and the power supply configuration of the target device and the sensor installation locations are the same as the example described using Fig. 7. Also, the block diagram of the process of the tact division data collection system according to the fourth embodiment is the same as that of Fig. 10.

[0050] A common feature of the embodiments of the present invention is the double border and diagonal line filling process described in items 1 and 2 in the flowchart of Fig. 14. In the fourth embodiment, the identification of the tact reference waveform and trigger waveform is automated by the system, rather than being performed by the system builder through manual analysis. This makes it possible to reduce the number of steps required for system construction work. Each automation method will be described below.

[0051] In the automation of tact reference waveform identification described in item 1, continuous data including a plurality of tacts is used as learning data, and the autocorrelation function shown in formula (3) is first derived.

[0052]

number

[0053] Here, n is the amount of continuous data, and x i indicates continuous data. Whereas cross-correlation is the similarity between waveforms, ACF(t) calculates the change in similarity when one of two waveforms is shifted slightly. When ACF(t) is calculated for continuous data that contains multiple tacts, the ACF(t) value has a peak at the same period as the tact length. By applying this property, it is possible to identify the tact reference waveform.

[0054] Next, the automation of trigger waveform identification uses the continuous data and the identified tact reference waveform. First, the NCC is calculated for the continuous data and the tact reference waveform, and the number of tact reference waveforms contained in the continuous data is calculated from the similarity. Next, a tentative trigger waveform is created by cutting out a predetermined length from the beginning of the tact reference waveform. The NCC is calculated for the continuous data and the tentative trigger waveform, and the number of tentative trigger waveforms contained in the continuous data is calculated from the similarity. Finally, it is determined whether the tentative trigger waveform can become a formal trigger waveform by whether the number of tact reference waveforms and the number of tentative trigger waveforms match. If the numbers do not match, the tentative trigger waveform is extended by a specified amount from the specified length, and the same calculation is performed, and this is repeated until the numbers match. If this calculation is performed until the tentative trigger waveform reaches the tact reference waveform length, a relationship is obtained in which the number of tentative trigger waveforms gradually approaches the number of tact reference waveforms from an excessive state, and the first tentative trigger waveform length that matches the number of tact reference waveforms becomes the length of the trigger waveform to be identified, and automatic identification of the trigger waveform is achieved.

[0055] According to the present configuration equipped with both of the above-mentioned automation techniques, the tact reference waveform identification and trigger waveform identification, which had to be performed manually, can be automated to reduce the number of manual labor steps. In steps 1 and 2 (S41 and S42), the processing unit 40 executes the tact start determination unit 42.

[0056] <Fifth embodiment> Fig. 15 shows an example of the current waveform and its statistical waveform targeted by the tact division data collection system according to the fifth embodiment, and Fig. 16 is a flow chart showing an example of the process flow of the tact division data collection system according to the fifth embodiment, the tact division data collection system in the fifth embodiment has the same configuration as that shown in Fig. 6, and the power supply configuration of the target device and the sensor installation locations are the same as the example described using Fig. 7. Also, the block diagram of the process of the tact division data collection system according to the fifth embodiment is the same as that of Fig. 10.

[0057] A common feature of the embodiments of the present invention is the double frame and diagonal line filling process described in item 1 in the flowchart of FIG. 16. In the above embodiments, simplified sensor data has been handled, but the actual current value flowing through the motor is an AC waveform as shown in FIG. 15A, and the magnitude of the value changes in a short period. Therefore, if the similarity such as NCC is calculated using the waveform as it is, it may lead to a decrease in accuracy. Therefore, in the fifth embodiment, a process is added in which the AC waveform as shown in FIG. 15A is simplified using statistics as shown in item 1 (S51). In the fifth embodiment, as an example, the AC waveform is simplified using a moving RMS. The moving RMS is a method of calculating the RMS shown in Equation 1 for each predetermined section, and a waveform such as that shown in FIG. 15B is obtained. In this embodiment, the waveform in FIG. 15B is used in calculations such as tact start judgment, and the AC waveform in FIG. 15A is saved when the final tact file is saved.

[0058] According to this configuration, even if the sensor data has a complex waveform such as an AC waveform, the accuracy of the tact division can be maintained. In step 1 (S51), the processing unit 40 executes the tact start determination unit 42.

[0059] Sixth embodiment Fig. 17 is a flow chart showing an example of the process flow of the tact division data collection system according to the sixth embodiment, and the tact division data collection system according to the sixth embodiment has the same configuration as that shown in Fig. 6, and the power supply configuration of the target device is the same as the example described using Fig. 7. Also, the block diagram of the process of the tact division data collection system according to the sixth embodiment is the same as that of Fig. 10. The sixth embodiment is intended for a case where a plurality of sensor data are acquired, for example, the case where all sensor data of the first motor to the sixth motor in Fig. 7 are acquired.

[0060] A common feature of the embodiments of the present invention described below is the double border and diagonal line filling process described in item 1 in the flowchart of Fig. 17. In the embodiments up to this point, it was assumed that the sensor data was from a single channel, but in this embodiment, the case where data from multiple channels is acquired collectively is targeted. In this embodiment, as described in item 1 (S71), instead of performing tact start determination calculations for each channel, calculations such as tact start determination are performed by selecting one of the channel data, and the results are used to divide the data for all channels into tacts.

[0061] According to this configuration, even if the number of channels increases, the amount of calculation of the tact division data collection system can be suppressed to the same level as that of one channel. In step 1 (S71), the processing unit 40 executes the tact start determination unit 42.

[0062] Seventh embodiment Fig. 18 is a diagram showing an example of an anomaly detection system and its target according to the seventh embodiment, and Figs. 19 and 20 are a flowchart and a block diagram showing an example of the processing flow of the anomaly detection system according to the seventh embodiment. The power supply configuration of the target device and the sensor installation locations are the same as the example explained using Fig. 7.

[0063] The seventh embodiment is characterized by the double border and diagonal line filling process described in item 10 in the flowchart of Fig. 19. Other processes can be the same as those described above, and in this embodiment, a function is added to detect an abnormality in the device as shown in item 10 (S100) using the tact division data described in the embodiments so far.

[0064] Specifically, the processing unit 40 executes an anomaly detection algorithm to determine whether the stored data is normal or not. Then, the processing unit 40 outputs the result to the detection result display unit 50. Here, the detection result display unit 50 can be configured using an appropriate display. Note that the processing unit 40 can execute an appropriate program in the display process related to the anomaly detection. Also, the manner of display is not particularly limited as long as an operator or the like can grasp the result of the anomaly detection.

[0065] According to this embodiment, an anomaly detection system is provided that includes a tact division data collection module that collects data for each tact, and an anomaly detection module that uses the obtained data for each tact to perform normal / abnormal judgment. This configuration has the effect of being able to perform anomaly detection at the same time as data collection by utilizing the divided tact data. In addition, in item 10 (S100), the processing unit 40 executes the anomaly detection unit 43.

[0066] As described above, according to the embodiment of the present invention, in the case of anomaly detection using an add-on sensor for discrete processes such as machining on a workpiece basis or robot transport, tact extraction after data logging is essential. However, when it is difficult to obtain a trigger signal, tact extraction is performed by pattern matching using the entire tact reference waveform. Therefore, in terms of processing load, it is necessary to prepare a processing device separately from the logging device, which increases costs. This solves the problem.

[0067] Although the embodiments of the present invention have been described above in detail, the present invention is not limited to the above-described embodiments, and various design changes can be made without departing from the spirit of the present invention described in the claims. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those having all of the configurations described. In addition, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

[0068] The tact division data collection system 3 and the anomaly detection system 5 are, for example, configured using one or more computers, and can be a system equipped with a processor, a storage device, and an interface used for inputting and outputting data, as shown in FIG. 21. The processing unit 40 described in the embodiment can be configured using a processor, and the storage unit 30 can be configured using a storage device. Here, the processor can be, for example, a CPU (Central Processing Unit), but other semiconductor devices may also be used. The storage device can be, for example, an SSD (Solid State Drive). The storage device may include a RAM (Random Access Memory) into which the processor reads a program, and the storage device may be, for example, an HDD (Hard Disk Drive).

[0069] The sensor data acquisition unit 41, the tact start determination unit 42, and the anomaly detection unit 43 described in the embodiment are programs executed by a processor. The tact reference waveform storage unit 31, the trigger waveform storage unit 32, and the tact data storage unit 33 are arranged in a storage device as a database. Note that the storage device may store other data, for example, data related to anomaly detection.

[0070] As an example, the tact division data collection system 3 and the anomaly detection system 5 are placed at the site where the robot performs the work, but they may be placed on a cloud different from the site. This allows data to be managed on the cloud.

[0071] The target device from which the tact data is collected is not limited to the configuration of the robot described in the embodiment, and may be, for example, other different industrial equipment that performs multiple types of repetitive operations.

[0072] The tact division data collection system 3 and the anomaly detection system 5 may acquire and process data from other types of sensors that output similar values ​​in the respective tacts. For example, data based on a force sensor may be acquired and processed in the same manner.

[0073] The abnormality detection system 5 may notify the operator of the abnormality in other ways. For example, the notification may be made by a lamp or a buzzer.

[0074] The tact division data collection system 3 and the anomaly detection system 5 may input and output data using a portable non-volatile storage device connected thereto. [Explanation of symbols]

[0075] 3: Tact division data collection system; 30: Memory section, 31: tact reference waveform storage unit, 32: Trigger waveform memory unit, 33: Takt data storage unit, 40: Processing unit, 41: Sensor data acquisition unit, 42: Tact start determination unit, 400: Switchboard, 500: Control Box, 510: Power supply circuit, 520: control board, 600: Robot Club, 601: First moving axis, 602: second movable axis, 603: Third moving axis, 604: Fourth moving axis, 605: Fifth moving axis, 606: Sixth moving axis.

Claims

1. A processor is provided. The processor, Based on a trigger waveform that is a part of a tact waveform, which is a repetitive waveform, and that is stored in advance in a storage device, it is determined whether a trigger waveform is included in the tact waveform in the time series data acquired from the sensor; When it is determined that the tact waveform includes a trigger waveform, data of the tact waveform including the trigger waveform is collected; The processor, In the judgment, data of the length of the trigger waveform is extracted from the tact waveform, and a similarity with the stored trigger waveform is calculated; if it is determined that the data is similar based on the similarity value, the data of the extracted length is written to a file, and then the data of the remaining length is written to a file. A tact division data collection system.

2. 2. The tact division data collection system according to claim 1, The storage device includes: The waveform for a certain period of time from the beginning of the tact waveform is stored as a trigger waveform. A tact division data collection system.

3. 2. The tact division data collection system according to claim 1, The storage device includes: A certain period of time in the middle of the tact waveform is stored as a trigger waveform. The processor, Collecting data of the tact waveform up to the start of the trigger waveform based on the temporarily stored time series data; A tact division data collection system.

4. 2. The tact division data collection system according to claim 1, The storage device includes: A certain period of time at the end of the tact waveform is stored as a trigger waveform. The processor, Collecting data of the tact waveform up to the start of the trigger waveform based on the temporarily stored time series data; A tact division data collection system.

5. A tact division data collection system as claimed in claim 1, The similarity is calculated based on cross-correlation. A tact division data collection system.

6. 2. The tact division data collection system according to claim 1, The processor, Calculate statistics of the time series data obtained from the sensor and perform processing using the calculated statistics. A tact division data collection system.

7. 2. The tact division data collection system according to claim 1, When acquiring time series data from multiple sensors, The processor, Processing is performed using time series data obtained from any one of the sensors. A tact division data collection system.

8. 2. The tact division data collection system according to claim 1, The tact division data collection system is deployed on a cloud. A tact division data collection system.

9. A processor is provided. The processor, Based on a trigger waveform that is a part of a tact waveform, which is a repetitive waveform, and that is stored in advance in a storage device, it is determined whether a trigger waveform is included in the tact waveform in the time series data acquired from the sensor; When it is determined that the tact waveform includes a trigger waveform, data of the tact waveform including the trigger waveform is collected; The storage device includes: In continuous data that contains multiple tact waveforms, the number of tact waveforms is calculated, The length of the identified trigger waveform is adjusted so that the number of trigger waveforms cut out from the tact waveforms included in the continuous data matches the number of calculated tact waveforms, and the identified trigger waveform is stored. A tact division data collection system.

10. The tact division data collection system according to claim 9, The storage device includes: The waveform for a certain period of time from the beginning of the tact waveform is stored as a trigger waveform. A tact division data collection system.

11. The tact division data collection system according to claim 9, The storage device includes: A certain period of time in the middle of the tact waveform is stored as a trigger waveform. The processor, Collecting data of the tact waveform up to the start of the trigger waveform based on the temporarily stored time series data; A tact division data collection system.

12. The tact division data collection system according to claim 9, The storage device includes: A certain period of time at the end of the tact waveform is stored as a trigger waveform. The processor, Collecting data of the tact waveform up to the start of the trigger waveform based on the temporarily stored time series data; A tact division data collection system.

13. The tact division data collection system according to claim 9, The tact waveform in the continuous data is identified using the autocorrelation function. A tact division data collection system.

14. The tact division data collection system according to claim 9, The processor, Calculate statistics of the time series data obtained from the sensor and perform processing using the calculated statistics. A tact division data collection system.

15. The tact division data collection system according to claim 9, When acquiring time series data from multiple sensors, The processor, Processing is performed using time series data obtained from any one of the sensors. A tact division data collection system.

16. The tact division data collection system according to claim 9, The tact division data collection system is deployed on a cloud. A tact division data collection system.

17. a sensor data acquisition unit that acquires time series data from a sensor; A tact waveform storage unit that stores one of the repetitive waveforms included in the learning data as a tact reference waveform; A trigger waveform storage unit that stores a waveform for a certain period of time from the beginning of the tact reference waveform as a trigger waveform that is a tact start trigger; A tact start determination unit that includes a process of extracting data for a trigger waveform length from the operation data and calculating a similarity with the trigger waveform; a tact data storage unit that stores operation time data as a file for each tact length according to the calculation result of the similarity; A tact division data collection module comprising: and an anomaly detection module that performs normal / abnormal determination using the obtained data for each tact. The anomaly detection system according to the present invention is characterized in that

Images