Information processing method, information processing device, system, manufacturing method for article, program, and recording medium

Abstract

To provide a method for properly extracting learning data for machine learning of features during operation in a normal state with respect to machine facilities changing in way of operation according to states of succeeding processes although repeatedly performing the same operation like a robot arranged in a production line.SOLUTION: An information processing method comprises: acquiring measured values related to machine facilities in a normal state in a first period; using the measured values and profile information on measured value when the machine facilities are in an operation state to specify a period in the machine facilities are in the operation state in the first period; extracting, as machine learning data, feature values based upon the measured values in the period in which the machine facilities are in the operation state in the first period; generating a learning-completed model through machine learning using the machine learning data; and using the learning-completed model to determine a state of the machine facilities at the time of evaluation.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an information processing method, an information processing device, a machine equipped with an information processing device, a control program, and a computer-readable recording medium, which are used when detecting a sign of a failure of a machine based on data from a sensor, etc. In particular, the present invention relates to the creation of learning data used when generating a failure prediction model of the machine by machine learning. [Background technology]

[0002] The operating state of machinery and equipment can change from moment to moment due to changes in the state of its components. When the operating state of the machinery and equipment is within the allowable range in light of the intended use of the machinery and equipment, it is called a normal state, and when it is outside the allowable range, it is called a faulty state (abnormal state). For example, if a production machine goes into a faulty state (abnormal state), it will produce defective products or cause the production line to stop, resulting in other problems.

[0003] In production machinery, etc., maintenance work is generally performed on a regular or irregular basis in order to prevent breakdowns from occurring as much as possible, even when the same work is performed repeatedly and continuously. In order to increase preventive safety, it is effective to shorten the intervals between maintenance work, but since production machinery, etc. are stopped during maintenance work, if the frequency of maintenance work is increased too much, the operating rate of production machinery, etc. will decrease. Therefore, it is desirable to be able to detect when a breakdown is imminent, even if the machinery, etc. is still in a normal state. If it is possible to detect that a breakdown is imminent (to predict the occurrence of a breakdown), maintenance work for the machinery, etc. can be performed at that point, preventing the operating rate from decreasing more than necessary.

[0004] As a method for predicting the occurrence of a failure, a trained model is created in advance by machine learning the state of the mechanical equipment, and the state of the mechanical equipment at the time of evaluation is evaluated using the trained model. For example, a method is known in which a trained model is generated by machine learning the characteristics of the normal state of the mechanical equipment, the degree of deviation between the state at the time of evaluation and the machine-learned normal state is calculated using the trained model, and the occurrence of a failure is predicted based on the calculated degree of deviation. In order to improve the prediction accuracy, it is important to build a trained model suitable for predicting failures, but in order to do so, the suitability of the learning data used in machine learning becomes an issue.

[0005] For example, Patent Document 1 describes extracting a vector based on a sensor signal indicating the state of a machine equipment, and selecting a feature to be used based on a data check of the feature vector. It also describes creating a normal model of the machine equipment based on the selected learning data.

[0006] Furthermore, Patent Document 2 describes that, with regard to the selection of multiple learning data prepared for each season according to seasonal fluctuations, etc., a sensor signal to be focused on and selected according to anomalies is selected based on the anomaly measure, which is the result of multivariate analysis, and the evaluation results of the influence of each sensor signal. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] JP 2011-70635 A [Patent Document 2] JP 2011-59790 A Summary of the Invention [Problem to be solved by the invention]

[0008] In mechanical equipment, measurement data is acquired for various parameters to manage the operating state of the equipment. When creating machine learning data to model the normal state of the equipment, it is important to appropriately extract the measurement data during operation when the equipment is in a normal state. However, in the case of machinery and equipment such as industrial robots installed on production lines, it has been difficult to appropriately extract data during operation when the machinery and equipment is in a normal state.

[0009] Robots placed on a production line generally perform the same operation repeatedly to repeatedly produce the same product, but even if the robot itself is in a normal state, its operation may be affected by the status of the machines in charge of the processes before and after the robot. For example, consider a case where a robot receives a workpiece from a machine in charge of the previous process, performs the work of the process it is in charge of (e.g., assembling parts), and passes the workpiece to a machine in charge of the next process. Even if the robot is in a normal state, if the cycle time of the robot is shorter than the cycle time of the machine in charge of the previous process, the robot will have waiting times and will not always be in operation. Similarly, if the cycle time of the robot is shorter than the cycle time of the machine in charge of the next process, the robot will have waiting times and will not always be in operation.

[0010] Even if the cycle time of the preceding and following processes and the cycle time of the robot are set to be equal, if an abnormality occurs in the preceding or following process, it may cause a problem in the transfer of workpieces, and the robot may have to stop even if it is in a normal state. For example, the robot may have to wait until the workpiece is sent from the machine in charge of the preceding process, or it may have to stop operating until the machine in charge of the following process is ready to receive it.

[0011] In this way, even if the robot is in a normal state, the continuous measurement data includes measurement data from when the robot is on standby or stopped, and the learning data for machine learning of the characteristics of the robot during normal operation contains noise.

[0012] In addition to noise during standby or stoppage, various measurement data when performing repeated work operations also includes measurement data that does not represent the characteristics of normal operation. For example, when a robot operates with six-axis control, some axes operate frequently, while others operate infrequently or do not operate at all, depending on the contents of the programmed work operations. For this reason, data measuring the drive status of each axis includes measurement data that does not reflect the characteristics of normal operation. Therefore, if all of the measurement data were to be used, the learning data for machine learning of the characteristics of normal operation would be redundant or contain noise.

[0013] Although Patent Document 1 and Patent Document 2 recognize the selection of learning data, they have not fully considered specific methods for selecting and acquiring learning data for machinery and equipment that performs repetitive operations such as production machinery. For this reason, it has been difficult to create a trained model with high prediction accuracy using conventional methods.

[0014] Therefore, there was a need for a method to appropriately extract learning data to use machine learning to learn the characteristics of normal operation of mechanical equipment, such as robots installed on a production line, which perform the same operation repeatedly but whose behavior changes depending on the situation in the previous and following processes. [Means for solving the problem]

[0015] The present invention is an information processing method comprising the steps of acquiring measurement values ​​related to mechanical equipment, detecting that the number of times per unit time that the measurement values ​​reach a predetermined value is a predetermined number of times, thereby identifying an operating period in which the mechanical equipment was in an operating state, acquiring features from the measurement values ​​corresponding to the operating period as machine learning data, and acquiring a trained model for acquiring the state of the mechanical equipment by machine learning using the machine learning data.

[0016] The present invention is also an information processing device characterized by comprising a control unit that acquires measurement values ​​related to mechanical equipment, identifies an operating period in which the mechanical equipment was in operation by detecting that the number of times per unit time that the measurement values ​​reach a predetermined value is a predetermined number of times, acquires features from the measurement values ​​corresponding to the operating period as machine learning data, and acquires a trained model for acquiring the state of the mechanical equipment by machine learning using the machine learning data. Effect of the Invention

[0017] According to the present invention, for example, for a robot installed on a production line, which repeats the same operation but changes the way it operates depending on the situation of the previous and next processes, it is possible to appropriately extract learning data for machine learning of the characteristics of normal operation. As a result, it is possible to create a failure prediction model with high prediction accuracy. [Brief description of the drawings]

[0018] [Figure 1] 1 is a schematic functional block diagram for explaining functional blocks included in a failure prediction system according to an embodiment; [Diagram 2] FIG. 2 is a schematic diagram for explaining a hardware configuration of the embodiment. [Diagram 3] 5A to 5C are schematic diagrams for explaining a feature extraction method according to the embodiment. [Figure 4] (a) Example of measurement data for one cycle of operation. (b) Example of unit record data (unit record) when operating continuously. (c) Example of unit record data (unit record) when a period of non-operation is included. [Diagram 5] 1A is a diagram for explaining an operation profile, FIG. 1B is a diagram showing a case where it is determined that the device is in operation, and FIG. 1C is an example showing a case where it is determined that the device is not in operation. [Figure 6] 1 is a schematic diagram for explaining a machine learning method according to an embodiment. [Figure 7] 5A and 5B are schematic diagrams for explaining a method of determining a determination threshold value in the embodiment. [Figure 8] 1 is a schematic diagram for explaining a failure prediction method in an embodiment. [Figure 9] 4 is a flowchart showing a processing procedure for generating a failure prediction model in the embodiment. [Figure 10] 4 is a flowchart showing a procedure for failure prediction according to the embodiment. [Figure 11] FIG. 1 is a perspective view showing a six-axis robot as an example of mechanical equipment. [Figure 12] A diagram showing the movement of each rotary joint of a six-axis robot. [Figure 13] 5A and 5B are schematic diagrams for explaining a method of determining an operation profile in the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] With reference to the drawings, embodiments of the present invention will be described, including a failure prediction system, a control method, a control device, a mechanical equipment equipped with a control device, a control program, a computer-readable recording medium, etc., used to predict failures in mechanical equipment.

[0020] [Function block configuration] FIG. 1 is a schematic functional block diagram for explaining the configuration of the functional blocks included in the failure prediction system of the embodiment. In FIG. 1, functional elements necessary for explaining the features of the embodiment are shown as functional blocks, but general functional elements that are not directly related to the problem-solving principle of the present invention are omitted. In addition, each functional element shown in FIG. 1 is a functional concept, and does not necessarily have to be physically configured as shown in the figure. For example, the specific form of distribution and integration of each functional block is not limited to the illustrated example, and all or part of them can be functionally or physically distributed and integrated in any unit depending on the usage situation, etc. As shown in FIG. 1, the failure prediction system of the embodiment includes a mechanical equipment 10 to be diagnosed, and a failure prediction device 100.

[0021] The mechanical equipment 10 is various industrial equipment such as an articulated robot or a production device installed in a production line. Various sensors 11 for measuring the state of the mechanical equipment 10 are installed in the mechanical equipment 10. For example, when the mechanical equipment 10 is an articulated robot, a sensor for measuring the current value of a motor that drives a joint, a joint angle sensor, a sensor for measuring speed, vibration, and sound, etc. may be installed. However, this is merely an example, and an appropriate type and number of sensors may be installed as the sensor 11 at an appropriate position depending on the type of the mechanical equipment 10 and the work purpose. Various sensors such as a force sensor, a torque sensor, a vibration sensor, a sound sensor, an imaging sensor, a distance sensor, a temperature sensor, a humidity sensor, a flow rate sensor, a pH sensor, a pressure sensor, a viscosity sensor, and a gas sensor may be used as the sensor 11. Note that in FIG. 1, the sensor 11 is shown as a single number for convenience of illustration, but usually, multiple sensors are installed. The mechanical equipment 10 is connected to the failure prediction device 100 via wired or wireless communication so as to be able to communicate with the failure prediction device 100, and the failure prediction device 100 can acquire data measured by the sensor 11 via communication.

[0022] In the stage of creating a failure prediction model, the failure prediction device 100 uses data collected from the sensor 11 to select features highly correlated with the occurrence of a failure in mechanical equipment, and performs machine learning using the selected features to generate and store a trained model (failure prediction model). In addition, in the evaluation stage (failure prediction stage), the data collected from the sensor 11 at the time of evaluation is input to the trained model, and the degree of deviation is calculated using the input and output of the trained model to determine whether or not a failure is approaching. Below, the functional blocks of the failure prediction device 100 will be explained in order. The failure prediction device 100 includes a control unit 110, a memory unit 120, a display unit 130, and an input unit 140.

[0023] The control unit 110 includes a plurality of functional blocks, which are configured, for example, by a control program stored in a storage device, read and executed by the CPU of the failure prediction device 100. Alternatively, some or all of the functional blocks may be configured by hardware such as an ASIC provided in the failure prediction device 100.

[0024] The storage unit 120 includes a sensor data storage means 121, a feature amount storage means 122, an operation profile storage means 123, an operating state determination storage means 124, a failure prediction model condition storage means 125, and a failure prediction model storage means 126. These means included in the storage unit 120 are appropriately allocated to storage areas of storage devices such as a hard disk drive, RAM, and ROM. The storage unit 120 is a data acquisition unit that acquires and stores various data necessary for processing to execute failure prediction.

[0025] The display unit 130 and the input unit 140 are user interfaces provided in the failure prediction device 100. The display unit 130 uses a display device such as a liquid crystal display or an organic EL display, and the input unit 140 uses an input device such as a keyboard, a jog dial, a mouse, a pointing device, or a voice input device.

[0026] The sensor data collection means 111 of the control unit 110 acquires measurement data from the sensors 11 of the mechanical equipment 10 and stores it in the sensor data storage means 121. That is, for example, measurement data related to the state of the mechanical equipment, such as current, speed, pressure, vibration, sound, temperature of each part, etc. measured in the mechanical equipment 10, is collected and stored in the sensor data storage means 121.

[0027] The feature extraction means 112 extracts a feature indicating a feature of the state of the mechanical equipment 10 based on the measurement data stored in the sensor data storage means 121, and stores the feature in the feature storage means 122. For example, the maximum and / or minimum values ​​of the sensor measurement values ​​collected in one operation cycle of the mechanical equipment may be extracted as the feature data, or an average value may be calculated. Alternatively, for example, the sensor measurement values ​​for a predetermined period may be integrally converted into a time-series frequency domain. Alternatively, the sensor measurement values ​​may be differential values ​​or second derivative values ​​with respect to time of the measurement values ​​arranged in a time series. Furthermore, if the sensor measurement values ​​(raw data) themselves are useful as a judging material for detecting whether or not a failure is approaching, the measurement values ​​themselves may be treated as feature data. In this embodiment, the feature extraction means 112 extracts or calculates a feature based on the sensor measurement data, creates time-series feature data, and stores the feature in the feature storage means 122. The extraction of the feature will be specifically described later with reference to FIG. 3.

[0028] The state determination means 113 acquires sensor data from the sensor data storage means 121, feature amounts from the feature amount storage means 122, and a definition of an operation profile for determining an operation state from the operation profile storage means 123. The definition of the operation profile is information related to criteria for determining whether the mechanical equipment 10 is operating (operating) or stopped (not operating). The state determination means 113 identifies a period during which the mechanical equipment 10 was operating (operating) based on the acquired information. Then, information relating to the period during which it was determined to be operating (operating) is stored in the operation state determination storage means 124. Note that the state determination means 113 does not necessarily have to acquire both the sensor data and the feature amounts, and may acquire only the sensor data if, for example, the operation state of the mechanical equipment 10 can be determined based on the sensor data alone.

[0029] When generating a trained model, the data extraction means 114 extracts features from the feature storage means 122 based on information related to the operating period of the mechanical equipment stored in the operating state determination storage means 124 and information stored in the failure prediction model condition storage means 125. The extracted features are output to the failure prediction model generation means 115 as data for machine learning. The failure prediction model condition storage means 125 stores information in advance that identifies, among various features, features that represent characteristics of the mechanical equipment in a normal state. For this reason, features that represent characteristics of the mechanical equipment in a normal state are selected as data for machine learning, and only the portion corresponding to the period during which the mechanical equipment is in operation is extracted.

[0030] Furthermore, at the time of evaluation, the data extraction means 114 extracts features from the feature storage means 122 based on information relating to the operation period of the mechanical equipment 10 stored in the operation state determination storage means 124 and information stored in the failure prediction model condition storage means 125. The extracted features are output as evaluation feature data to the failure determination means 116. For this reason, the same types of features used during machine learning are selected as the evaluation feature data, and only the portions corresponding to the operation period of the mechanical equipment are extracted.

[0031] The failure prediction model generation means 115 generates a trained model (failure prediction model) using the machine learning data input from the data extraction means 114 when generating the trained model, and stores it in the failure prediction model storage means 126.

[0032] The fault determination means 116 inputs the feature data for evaluation input from the data extraction means 114 during evaluation into a trained model (fault prediction model) stored in the fault prediction model storage means 126, and calculates the degree of deviation between the input and the output. Then, the means 116 compares the degree of deviation with a determination threshold value to determine whether or not there is a sign of a fault. The fault notification means 117 notifies an external device of the determination result of the fault determination means 116 and displays it on the display unit 130 .

[0033] [Hardware configuration] FIG. 2 shows a schematic diagram of an example of a hardware configuration of the failure prediction system according to the embodiment. As shown in FIG. 2, the failure prediction system can include PC hardware including a CPU 1601 as a main control means, a ROM 1602 as a storage device, and a RAM 1603. The ROM 1602 can store information such as a processing program and an inference algorithm for implementing the failure prediction method described below. The RAM 1603 is used as a work area for the CPU 1601 when executing the control procedure. An external storage device 1606 is connected to the control system. The external storage device 1606 is composed of an HDD, an SSD, an external storage device of another system mounted on a network, or the like.

[0034] The processing program of the CPU 1601 for realizing the failure prediction method of this embodiment described later can be stored in a storage unit such as an external storage device 1606 consisting of an HDD or SSD, or a ROM 1602 (for example, an EEPROM area). In this case, the processing program of the CPU 1601 for realizing the failure prediction method can be supplied to each of the above-mentioned storage units via a network interface (NIF) 1607, and can be updated to a new (different) program. Alternatively, the processing program of the CPU 1601 for realizing the failure prediction method can be supplied to each of the above-mentioned storage units via various storage means such as magnetic disks, optical disks, and flash memories, and a drive device therefor, and can be updated. Various storage means, storage units, and storage devices in a state in which a program capable of executing the processing of the CPU 1601 for realizing the failure prediction method is stored constitute a computer-readable recording medium storing the failure prediction procedure of the present invention.

[0035] The sensor 11 shown in Fig. 1 is connected to the CPU 1601. In Fig. 2, for the sake of simplicity, the sensor 11 is shown as being directly connected to the CPU 1601, but it may be connected via, for example, IEEE488 (so-called GPIB) or the like. The sensor 11 may also be configured to be connected to the CPU 1601 via a network interface 1607 and a network 1608.

[0036] The network interface 1607 can be configured using a communication standard for wired communication such as IEEE 802.3, or wireless communication such as IEEE 802.11 and 802.15. The CPU 1601 can communicate with other devices 1104 and 1121 via the network interface 1607. For example, if the mechanical equipment to be the target of failure prediction is a robot, the devices 1104 and 1121 may be a supervisory control device such as a PLC or sequencer arranged for controlling and managing the robot, or a management server.

[0037] In the example shown in Fig. 2, an operation unit 1604 and a display device 1605 related to the input unit 140 and the display unit 130 shown in Fig. 1 are connected as a UI device (user interface device). The operation unit 1604 can be configured by a terminal such as a handy terminal, or devices such as a keyboard, a jog dial, a mouse, a pointing device, a voice input device, etc. (or a control terminal equipped with them). The display device 1605 may be any device that can display information related to the processing executed by the state determination means 113, the failure prediction model generation means 115, the failure determination means 116, etc. on a display screen, and for example, a liquid crystal display device can be used.

[0038] [How to predict failures] In this embodiment, the failure prediction model generating means 115 of the failure prediction device 100 constructs a learned model (failure prediction model) by so-called unsupervised learning. To learn the characteristics of mechanical equipment failures by so-called unsupervised learning, machine learning is performed using only operation data in a state without failures, i.e., when the mechanical equipment is operating normally. Unsupervised learning is a method in which a learning device is provided with a large amount of only input data, so that the device learns how the input data is distributed. In other words, it is a method in which a device that performs processes such as compression, classification, and shaping of input data learns the process without providing it with teacher output data corresponding to the input data.

[0039] A failure prediction method using an unsupervised learning technique will be specifically described. A case where the operating state of the mechanical equipment is within an allowable range in light of the intended use of the mechanical equipment is regarded as a normal state, and machine learning is performed using only the operation data of the mechanical equipment in a normal state. In this embodiment, an autoencoder is used as the unsupervised learning model.

[0040] This embodiment is characterized by the method of extracting data used for machine learning. That is, features that represent the behavior of the machinery when it is in a normal state are selected, and only data during the period when the machinery is in operation is extracted from the continuous data of the selected features and used as learning data.

[0041] First, the extraction of the feature amount will be described with reference to FIG. 3. The sensor 11 of the mechanical equipment 10 to be subjected to the failure prediction includes a sensor 1 which is a current sensor, a sensor 2 which is a speed sensor, and a sensor 3 which is a pressure sensor, as illustrated in FIG. 3. In the failure prediction method according to the present embodiment, a feature amount indicating the operating state of the mechanical equipment 10 is extracted based on the measurement data of each sensor included in the sensor 11. The feature amount is extracted by performing processes such as integral conversion to the frequency domain, calculation of the first and second derivatives with respect to time, filtering, and extraction of maximum and minimum values ​​of periodic operation, on the time-series measurement data of the sensor. Note that the above-mentioned sensors and measurement data processing are merely examples, and any sensor and measurement data processing may be used as long as data suitable for grasping the state of the mechanical equipment can be obtained. Furthermore, if the state of the mechanical equipment can be easily analyzed by the measurement data of the sensor itself, the measurement data itself may be used as the feature amount without performing any special processing. FIG. 3 shows a schematic diagram of the state in which three types of processing are performed on the measurement data from Sensor 1, Sensor 2, and Sensor 3, and time-series data for a total of nine types of features, as shown on the right, are extracted.

[0042] Next, a method for determining whether the mechanical equipment 10 is operating (in operation) or stopped (not in operation) and for identifying the operating (in operation) period will be described. Here, an articulated robot that is installed in a production line and is responsible for one process in the manufacturing process is given as an example of the mechanical equipment 10. The articulated robot receives a workpiece from a machine responsible for the preceding process, performs the task of that process (e.g., assembling parts), and passes the workpiece to a machine responsible for the following process, but it repeats the same operation in order to repeatedly produce the same product.

[0043] FIG. 11 shows an external view of a six-axis articulated robot, which is an example of the mechanical equipment 10. Links 200 to 206 are connected in series by six rotary joints J1 to J6. Each rotary joint is provided with a sensor that measures the rotation speed of the motor, a sensor that measures the rotation angle of the joint, a torque sensor, and the like. A robot hand 210 can be attached to the link at the tip. A teaching pendant 102 is connected to a control device 101 that controls the operation of the robot, allowing an operator to teach the robot operations.

[0044] Robots installed on production lines perform cycle operations, i.e., repeat fixed operations, but in the case of a six-axis articulated robot, some axes may operate frequently and some may not operate at all depending on the contents of the cycle operation. For example, in a process in which a workpiece is rotated at a fixed position, only J6, the rotary joint at the tip, needs to operate, and the other rotary joints J1 to J5 do not operate. On the other hand, in a process in which a workpiece is moved horizontally, mainly rotary joint J1 operates, and J5 and J6 near the tip may not operate.

[0045] 12 shows the direction of movement of each rotary joint of a 6-axis articulated robot, and whether or not each rotary joint is driven can be known by, for example, measuring the rotation speed of the motor that drives the rotary joint with a sensor. However, as described above, the operation of each rotary axis differs depending on the contents of the cyclic operation, so in order to determine whether or not the robot is in operation, it is better to measure the rotation speed of the rotary joint that is driven most frequently in the cyclic operation performed by the robot.

[0046] Here, a method for selecting the rotary joint that is most frequently driven while the robot is performing a cyclic operation will be described. FIG. 13 shows the measurement results of the rotational speed sensors of the rotary joints J1 to J6 when the robot is stably continuing a certain cyclic operation (for example, assembly work). One period of the cyclic operation is 10 seconds, and the sensor output waveforms for six periods are shown. As shown in the figure, it can be seen that in this cyclic operation (for example, assembly work), the rotary joint J5 is driven most frequently throughout the entire period, and conversely, the rotary joints J1 and J6 are not driven. Therefore, it can be said that the most suitable way to determine whether the robot performing this cyclic operation is in operation is to monitor the rotational speed of the rotary joint J5.

[0047] To automatically select the rotating joint that is operating most frequently, the failure prediction device 100 acquires rotation speed data of each rotating joint that is in stable operation from the sensor 11, and measures the number of times the rotation speed becomes zero (the number of times the graph crosses zero). The control unit 110 compares the number of times it becomes zero for each rotating joint, and determines the rotating joint that has the most number of times (J5 in the example of FIG. 13) as the rotating joint to be used for the operation profile. This is because the rotating joint with the most number of times has the most speed changes, and is therefore suitable for determining operation. In other words, profile information, which will be described later, is set based on the measurement values, among those measured by multiple sensors, that have the greatest degree of change when the mechanical equipment performs a predetermined repetitive operation.

[0048] Figure 4(a) shows an example of measurement data of the rotation speed of the motor that drives the rotary joint J5 determined in this way, and shows the waveform for the period during which one process is performed (one cycle of operation). Here, one cycle of operation is performed for 10 seconds as an example.

[0049] Meanwhile, the sensor data collection means 111 collects measurement data outputted from the sensor 11 in time series and stores it in the sensor data storage means 121. At that time, for convenience of handling, the measurement data that is continuous in time is divided into predetermined time intervals (for example, 60 seconds), and the measurement data for the predetermined time is treated as one unit of record data (unit record). The period corresponding to each unit of record data (unit record) is called a unit record period.

[0050] Fig. 4(b) is a graph showing an example of unit record data (unit record) when a robot in a normal state is continuously operating (working). On the other hand, as already explained, a robot installed in a production line may temporarily stop operating (become in a non-operating state) due to the status of the machines in charge of the preceding and following processes. In that case, as shown in Fig. 4(c), the unit record data (unit record) will include measurement data from the non-operating state.

[0051] In the unit recording period shown in Fig. 4(b), the robot is continuously operating, so there is no problem in using the feature values ​​corresponding to this unit recording period as learning data. However, the unit recording period shown in Fig. 4(c) includes a period in which the robot was not operating, even if it was in a normal state, so the feature values ​​corresponding to this unit recording period contain noise as learning data.

[0052] Therefore, in this embodiment, the state determination means 113 checks whether or not the mechanical equipment (robot) has become in a non-operating state for a unit recording period corresponding to each unit recording data (unit record) stored in the sensor data storage means 121.

[0053] Specifically, profile information relating to the characteristics of one cycle of repeated operation is stored in advance in the operation profile storage means 123. In this example, as shown in Fig. 5(a), the speed of the motor, which repeatedly accelerates and decelerates, has the characteristic of becoming zero eight times within one cycle (10 seconds), so the number of zero crossings of the speed is 0.8 times / second is stored in advance as profile information indicating the operating state. In other words, the profile information is set based on the number of times the measurement value of the speed sensor becomes zero within a unit time.

[0054] The state determination means 113 acquires profile information indicating an operating state (the number of zero crossings of the speed is 0.8 times / second) from the operation profile storage means 123. Then, for each unit record data (unit record) of the motor speed stored in the sensor data storage means 121, it determines whether it is in an operating state (whether a non-operating state is included).

[0055] For example, as shown in Figure 5(b), if the robot is operating continuously without stopping, 48 zero crossings are counted in a 60-second unit record of data, which is calculated as 0.8 crossings per second. This matches the profile, so the robot is determined to be in operation.

[0056] On the other hand, if the unit recording data (unit record) includes a period during which the robot was stopped, as shown in Figure 5(c), the number of zero crossings is counted only 24 times, which is calculated as 0.4 times / second, and since this does not match the profile, it is determined that the period includes non-operating periods.

[0057] The profile information for determining "operating" does not necessarily have to be set to a fixed value (0.8 times / second). Taking into consideration fluctuations in the robot operation speed on the production line (fluctuations in takt time), it may be set with a predetermined width, for example, within a range of ±20% centered on 0.8 times / second. Alternatively, for example, 0.7 times / second may be set as a threshold value, and anything above that may be determined as "operating" and anything below that may be determined as "not operating, etc."

[0058] In this manner, it is determined whether the robot was "in operation" or "including non-operation" in the unit recording period corresponding to each unit recording data (unit record). In this manner, the state determination means 113 identifies the unit recording period in which the robot was in operation, and stores the information in the operation state determination storage means 124. For example, among the information on the time axis (calendar information), a label of "in operation" is assigned to a unit recording period that is determined to be in operation, and a label of "noise" is assigned to a unit recording period that included non-operation, and these are stored in the operation state determination storage means 124.

[0059] When performing machine learning, the data extraction means reads information on a unit recording period labeled with "operating" from the operation state determination storage means 124, and extracts features corresponding to the unit recording period from the feature storage means 122. The data extraction means outputs the extracted features to the failure prediction model generation means 115 as learning data.

[0060] In this way, according to this embodiment, by selectively extracting only the feature data corresponding to the unit recording period during which the robot was in operation (in motion) from the time-series feature data, it is possible to create data for machine learning with less noise.

[0061] Next, with reference to the schematic diagram of Figure 6, a method of having an autoencoder perform machine learning using the learning data extracted as described above will be described. An autoencoder is a type of neural network that compresses (encodes) input learning data into a small amount of information and then restores (decodes) it. Through machine learning, it learns "parameters for effectively compressing and restoring input data," i.e., the characteristics of the input data.

[0062] An autoencoder encodes the input value x and compresses it to the intermediate layer z. It then decodes the intermediate layer z and restores it as the output value y. At that time, machine learning is performed to reduce the restoration error J between the input value and the output value.

[0063] That is, the autoencoder determines W and b in Equation 1 and W' and b' in Equation 2 so as to reduce the recovery error J in Equation 3 shown below, where s is an activation function.

[0064]

number

[0065]

number

[0066]

number

[0067] If data with similar characteristics to the training data is input to a trained autoencoder (sometimes called a trained model or a failure prediction model), the output will have a small restoration error due to encoding and decoding using the parameters acquired during training. On the other hand, if data with different characteristics from the training data is input to a trained model, the parameters acquired during training cannot compress and decode the data well, resulting in a large restoration error.

[0068] In order to utilize this characteristic for predicting a fault state (abnormal state), in this embodiment, a feature value corresponding to a period during which the robot was in a normal state and in operation (in motion) is given as an input value x to an autoencoder for machine learning.

[0069] Furthermore, when predicting failures, from the features extracted during evaluation, features of the same type as those selected during learning and corresponding to the period during which the machine was in operation (in operation) are extracted, and input to the trained model as input value x, and output value y is output. Then, the restoration error of output value y for input value x is calculated, and the restoration error, i.e., the degree of deviation between input and output, is treated as an index showing the degree to which the machine equipment deviates from its normal state.

[0070] In this embodiment, a judgment threshold used when judging whether or not a breakdown of the mechanical equipment is imminent using the deviation is set in advance. To set the judgment threshold, first, feature amounts based on sensor data of the actual mechanical equipment from a normal state until a breakdown occurs are extracted and input to the trained model, and the change over time in the deviation until the breakdown occurs is examined.

[0071] In this embodiment, the feature data used in this case is the same type of feature as that selected during learning (feature obtained by performing the same processing on measurement data from the same sensor) and is extracted as feature corresponding to the period during which the machine was in operation (in operation). The period during which the machine was in operation (in operation) is identified by a judgment using an operation profile, as described in the creation of learning data. A judgment threshold is set based on the change in deviation over time to judge whether a failure is approaching. If the deviation is equal to or greater than the judgment threshold, it is judged that a failure of the mechanical equipment is imminent, i.e., there is a sign of failure.

[0072] Fig. 7 is a diagram for specifically explaining a method for determining a judgment threshold. The horizontal axis of the graph in Fig. 7 is time (hours), and the vertical axis is an index value (deviation between input and output of a trained model) indicating the degree of approaching the occurrence of a fault, and shows the change over time of the index value from the beginning of a normal state until the occurrence of a fault. For convenience of illustration, the deviation calculated based on the feature amount corresponding to the period when the device was in operation (in operation) is shown as a graph that is continuous over time.

[0073] Assume that you want to secure a certain period of time t as operating time from when the failure prediction device predicts and notifies you that a failure is approaching until the failure actually occurs, that is, when you want the failure prediction device to predict a certain period of time t before the failure occurs. In this case, as shown in the figure, the numerical value of the index value (the degree of deviation between the input and output of the trained model) at a point in time going back a certain period of time t from the occurrence of the failure is set as the judgment threshold T for failure prediction (judgment threshold setting process).

[0074] Next, a failure prediction method using the above-mentioned trained model and the judgment threshold value will be described. Fig. 8 is a schematic diagram for explaining a failure prediction method using an autoencoder.

[0075] Evaluation data showing the operating state of the machinery at the time of evaluation is input to the trained model, and the input and output values ​​are used to calculate the degree of deviation showing how far the machinery is from the trained normal state. The evaluation data used is data from the time of evaluation for the same type of features selected during training (features obtained by performing the same processing on measurement data from the same sensor), and features corresponding to the period during which the machinery was in operation (in operation) are extracted. The period during which the machinery was in operation (in operation) is identified by a judgment using the operation profile, as described in the creation of training data.

[0076] Specifically, as shown in Fig. 8, evaluation data is input to the failure prediction model, and the recovery error J of the resulting output value y and input value x of the failure prediction model is calculated and treated as the deviation from the normal state. In this embodiment, this deviation is treated as an index value indicating the degree of approach to the occurrence of a failure. If the deviation (recovery error J) is equal to or greater than the judgment threshold value T, it is determined that the period until the occurrence of a failure is equal to or less than the predetermined period t, that is, there is a sign of a failure. Conversely, if the deviation (recovery error J) is less than the judgment threshold value T, it is determined that the period until the occurrence of a failure is longer than the predetermined period t, that is, there is no sign of a failure.

[0077] [Processing Procedure] Next, the procedure of the process executed by the failure prediction device 100 will be described with reference to the flowcharts of FIGS. [Generate Model] FIG. 9 is a flowchart showing a processing procedure for generating a failure prediction model.

[0078] First, in step S101, the sensor data collection means 111 of the failure prediction device 100 acquires measurement data from the sensor 11 for measuring the state of the mechanical equipment 10, and stores the data in the sensor data storage means 121. That is, the measurement values ​​related to the mechanical equipment in a normal state are acquired during a first period (measurement data acquisition process).

[0079] Next, in step S102, the feature extraction means 112 extracts features indicating characteristics of the operating state of the mechanical equipment 10 based on the sensor data stored in the sensor data storage means 121, and stores them in the feature storage means 122 (feature extraction process). Next, in step S103, the state determination means 113 reads out unit record data (unit record) from the sensor data storage means 121.

[0080] Next, in step S104, the state determination means 113 compares the read unit recording data (unit record) with the profile stored in the operation profile storage means 123, and determines whether the robot was operating during that unit recording period. If it is determined that the robot was operating (step S104: yes), the process proceeds to step S105, where the state determination means 113 labels that unit recording period of the calendar information as "operating" and stores it in the operation state determination storage means 124. If it is determined that the unit recording period includes non-operating periods (step S104: no), the process proceeds to step S106, where the state determination means 113 labels that unit recording period of the calendar information as "noise" and stores it in the operation state determination storage means 124.

[0081] Next, in step S107, it is determined whether the number of unit recording periods labeled with "in operation" has reached a predetermined number. Here, the "predetermined number" is a number that is set in advance to ensure a sufficient amount of machine learning data to generate a highly accurate failure prediction model. If the number of unit recording periods labeled with "in operation" has not reached the predetermined number (step S107: no), steps S103 and onwards are repeated until the predetermined number is reached. If the number of unit recording periods labeled with "in operation" has reached the predetermined number (step S107: yes), proceed to step S108.

[0082] In step S108, the data extraction means 114 extracts features from the feature storage means 122 based on information about the period during which the mechanical equipment is in operation stored in the operating state determination storage means 124 and information stored in the failure prediction model condition storage means 125. The failure prediction model condition storage means 125 prestores information that identifies the feature that represents the normal state of the mechanical equipment among various feature values ​​(for example, a feature selected from the nine feature values ​​shown on the right side of FIG. 3). In addition, the operating state determination storage means 124 stores information that identifies the period during which the mechanical equipment is in operation. For this reason, the feature that represents the characteristic of the normal state of the mechanical equipment is selected as data for machine learning, and only the portion corresponding to the period during which the mechanical equipment is in operation is extracted (learning data extraction process). The extracted feature values ​​are output to the failure prediction model generation means 115 as data for machine learning.

[0083] Next, in step S109, the failure prediction model generation means 115 generates a trained model (failure prediction model) using the machine learning data input from the data extraction means 114, and stores it in the failure prediction model storage means 126 (trained model generation process). By executing the above series of processes, a trained model (fault prediction model) can be generated.

[0084] [Failure prediction] Next, a procedure of processing executed by the failure prediction device 100 using the generated trained model (failure prediction model) when determining whether or not the mechanical equipment 10 is approaching the occurrence of a failure will be described.

[0085] 10 is a flowchart showing the processing procedure. The processing for determining whether the mechanical equipment 10 is approaching a breakdown is started, for example, by a user using the input unit 140 of the failure prediction device 100 to instruct the start of the processing. Alternatively, the control program of the failure prediction device 100 may be configured so that the processing is automatically started according to the operating time of the mechanical equipment 10.

[0086] When the process starts, in step S201, the sensor data collection means 111 of the failure prediction device 100 acquires measurement data from the sensor 11 for measuring the state of the mechanical equipment 10, and stores the data in the sensor data storage means 121. That is, measurement values ​​related to the mechanical equipment during the evaluation period are acquired. Next, in step S202, the feature extraction means 112 extracts features indicating characteristics of the operating state of the mechanical equipment 10 based on the sensor data stored in the sensor data storage means 121, and stores them in the feature storage means 122.

[0087] Next, in step S203, the state determination means 113 reads out unit record data (unit record) from the sensor data storage means 121 and compares it with the profile stored in the operation profile storage means 123. Then, it is determined whether the robot was in operation during that unit recording period. If it is determined that the robot was in operation, the state determination means 113 attaches a label of "in operation" to that unit recording period of the calendar information and stores it in the operation state determination storage means 124. If it is determined that the unit recording period includes non-operation, the state determination means 113 attaches a label of "noise" to that unit recording period of the calendar information and stores it in the operation state determination storage means 124. It is desirable to repeat steps S201 to S203 until a sufficient number of samples of "in operation" labels are secured to perform highly accurate failure prediction. When a sufficient number of "in operation" labels are labeled in the calendar information to perform highly accurate evaluation, the process proceeds to step S204.

[0088] In step S204, the data extraction means 114 extracts features from the feature storage means 122 based on information about the period during which the mechanical equipment is in operation stored in the operating state determination storage means 124 and information stored in the failure prediction model condition storage means 125. The failure prediction model condition storage means 125 stores in advance information that specifies the feature that represents the normal state of the mechanical equipment among various features (for example, a feature selected from the nine types of feature shown in the right of FIG. 3). In addition, the operating state determination storage means 124 stores information that specifies the period during which the mechanical equipment is in operation. For this reason, the same type of feature used when the learning data was created is selected as the evaluation data, and only the portion corresponding to the period during which the mechanical equipment is in operation is extracted (evaluation data extraction process). The extracted feature is output to the failure determination means 116 as the evaluation data.

[0089] Next, in step S205, the fault judgment means 116 inputs the evaluation feature data input from the data extraction means 114 into the trained model (fault prediction model) stored in the fault prediction model storage means 126, and calculates the deviation between the input and the output.

[0090] Next, in step S206, the failure determination means 116 compares the calculated deviation with a determination threshold value to determine whether or not the mechanical equipment 10 is approaching the occurrence of a failure, that is, whether or not there is a sign of a failure. If the deviation is equal to or greater than the judgment threshold (step S206: yes), it is judged that a breakdown of the mechanical equipment 10 is imminent, and the process proceeds to step S207.

[0091] In step S207, the fault determination means 116 issues a notification command to the fault notification means 117. The fault notification means 117, which has received the notification command, notifies the user of the determination result of the fault determination means 116. When making the notification, the user is notified via a user interface, and information related to the determination may be stored in the storage unit 120 or provided to an external device via an external interface. To notify the user, processing such as displaying on the display unit 130 of the fault prediction device 100, issuing a voice message, or printing and outputting on a medium such as paper may be performed. When notification to the user is completed, the processing ends (END).

[0092] If the deviation is less than the judgment threshold (step S206: no), it is judged that the occurrence of a failure in the mechanical equipment 10 is far away, i.e., there is no sign of a failure, and the process ends (END). Note that even if it is judged that there is no sign of a failure, the result may be notified to the user, information related to the judgment may be stored in a storage device, or the information may be provided to an external device via an external interface.

[0093] As described above, in this embodiment, for example, for a robot installed on a production line, which repeats the same operation but changes the way it operates depending on the status of the previous and next processes, various feature amounts are extracted based on the measurement data of the sensor. Among these feature amounts, feature amounts suitable for machine learning the normal operation of the robot are selected.

[0094] In addition, from the sensor measurement data, measurement data that changes significantly when the robot performs its operation is selected, and profile information is set to identify that the robot is in an operating state (operating). Then, by comparing the measurement data for a given period with the profile information, it is determined whether the robot was in an operating state during that period. This makes it possible to create machine learning data by extracting only the features for the period when the robot was in an operating state from the time-series feature data. By creating machine learning data with less noise, it is possible to create a trained model (failure prediction model) with higher prediction accuracy than conventional methods. In addition, when setting a judgment threshold and when creating evaluation data, it is possible to extract only the features for the period when the robot was in an operating state, thereby improving the accuracy of failure prediction using the trained model (failure prediction model).

[0095] [Other embodiments] The present invention is not limited to the above-described embodiment, and many modifications are possible within the scope of the technical concept of the present invention. For example, in the above-described embodiment, a failure prediction model was created using an autoencoder by a so-called unsupervised learning method, but the present invention can also be implemented by creating a failure prediction model using a so-called supervised learning method. Supervised learning is a method of providing a large amount of data sets of certain inputs and results (labels) to a learning device, learning the features in those data sets, and constructing a model that estimates results from inputs, that is, a learned model that inductively acquires the relationship between inputs and outputs.

[0096] In the above embodiment, a method using a neural network has been described as an example of machine learning, but the machine learning method is not limited to this, and for example, genetic programming, functional logic programming, support vector machines, etc. may be used. A general-purpose computer or processor may be used as a device for performing machine learning, but high-speed processing is possible when a graphics processing unit with a GPGPU function or a large-scale PC cluster is used.

[0097] Furthermore, machine learning is not limited to one time, and additional learning may be performed. In that case, additional learning is performed by extracting only the feature amounts during the period when the machine equipment was in operation.

[0098] In addition, for the operating profile, it is advisable to select the parameters that change most frequently when the machinery or equipment performs its operation; however, the rotational speed of a rotary joint is just one example, and parameters for determining the operating state can be selected as appropriate depending on the type of machinery or equipment and its operation.

[0099] In addition, in the embodiment, the number of zero crossings per unit time is used as an index to determine whether the system is in operation or has a non-operational period, but the index is not limited to this. For example, the number of inflection points or the number of extreme values ​​per unit time may be used as an index.

[0100] In the above embodiment, the failure prediction device extracts features from the acquired sensor data, identifies the period during which the device is in operation, and extracts features for that period from all of the features, but the processing method is not limited to this example. For example, the period during which the device is in operation may be identified based on the sensor data first, and features may be extracted only from the sensor data corresponding to that period.

[0101] The failure prediction device of the present invention can be applied to failure prediction of various machines and equipment, such as industrial robots, service robots, processing machines operated by computer numerical control, etc. The failure prediction device may be integrated with the machine equipment to form a failure prediction system, or the failure prediction device may be provided as a part of the machine equipment.

[0102] The present invention can also be realized by a process in which a program for implementing one or more functions of the embodiments is supplied to a system or device via a network or a storage medium, and one or more processors in a computer of the system or device read and execute the program. The present invention can also be realized by a circuit (e.g., ASIC) for implementing one or more functions. [Explanation of symbols]

[0103] 10···Machine equipment / 11···Sensor / 100···Failure prediction device / 110···Control unit / 111···Sensor data collection means / 112···Feature extraction means / 113···State determination means / 114···Data extraction means / 115···Failure prediction model generation means / 116···Failure determination means / 117···Failure notification means / 120···Memory unit / 121···Sensor data storage means / 122···Feature storage means / 123···Operation profile storage means / 124···Operation state determination storage means / 125···Failure prediction model condition storage means / 126···Failure prediction model storage means / 130···Display unit / 140···Input unit

Claims

1. Acquire measurements related to machinery and equipment, By detecting the number of times that the measurement value reaches a predetermined value per unit time, an operating period during which the mechanical equipment was in an operating state is identified; acquiring feature values ​​from the measurement values ​​corresponding to the operating period as machine learning data; Obtaining a trained model for acquiring the state of the mechanical equipment by machine learning using the machine learning data.

23. An information processing method comprising:

2. The predetermined number of times has a predetermined upper limit value and a predetermined lower limit value.

2. The information processing method according to claim 1,

3. The unit time is a time having a predetermined temporal length.

2. The information processing method according to claim 1,

4. Acquiring the state of the mechanical equipment using the trained model.

4. The information processing method according to claim 1, wherein the first and second inputs are input to the first and second inputs.

5. In determining the condition of the machinery and equipment, Obtaining measurement values ​​related to the machinery and equipment during an evaluation period; Identifying the operating period during which the mechanical equipment was in operation within the evaluation period based on the measurement value of the mechanical equipment during the evaluation period and the number of times the measurement value per unit time became the predetermined value; acquiring, as an evaluation feature, a feature based on a measurement value corresponding to the operation period in the evaluation period; acquiring an index value indicating the degree to which the mechanical equipment has deviated from a normal state based on the evaluation feature amount and the trained model, and determining the state of the mechanical equipment during the evaluation period; 5. The information processing method according to claim 4.

6. data on the mechanical equipment from the normal state to the abnormal state, for features of the same type as the features acquired as the machine learning data, is input to the trained model, a deviation between the input data input to the trained model and output data output from the trained model is calculated, and a judgment threshold is obtained based on a change in the deviation over time during the period in which the mechanical equipment from the normal state to the abnormal state is calculated; determining a state of the mechanical equipment during the evaluation period using the index value and the judgment threshold value; 6. The information processing method according to claim 5,

7. In acquiring the judgment threshold value, data from the operation period of the mechanical equipment from the normal state to the abnormal state is acquired and input into the trained model.

7. The information processing method according to claim 6,

8. In obtaining the trained model, the trained model is obtained by machine learning using an autoencoder.

8. The information processing method according to claim 1,

9. The control unit notifies the result determined in determining the state of the mechanical equipment.

9. The information processing method according to claim 1,

10. The operating state is a state in which the mechanical equipment repeatedly executes a predetermined operation.

10. The information processing method according to claim 1,

11. the measurement value includes a plurality of measurement values ​​measured by a plurality of sensors; and identifying the operating period based on a measurement value that has a large degree of change when the mechanical equipment repeatedly executes the predetermined operation among the plurality of measurement values.

11. The information processing method according to claim 10.

12. The predetermined value is zero.

12. The information processing method according to claim 1,

13. The unit time is 10 seconds, and a period during which the number of occurrences is 8 is specified as the operation period.

13. The information processing method according to claim 12.

14. The predetermined value includes at least one of an inflection point of a measurement value and an extreme value of a measurement value.

14. The information processing method according to claim 1,

15. Labeling the active period as active; 15. The information processing method according to claim 1,

16. Identifying a non-operating period during which the machine equipment was in a non-operating state based on the number of times the measurement value became the predetermined value per unit time.

16. The information processing method according to claim 1,

17. labeling at least a portion of the non-operating period as noise based on the number of times the measured value reaches the predetermined value per unit time; 17. The information processing method according to claim 16.

18. The predetermined value or the number of times can be set by a user.

18. The information processing method according to claim 1,

19. A reference value of the number of times is set.

19. The information processing method according to claim 1,

20. A range is set for the reference value, 20. The information processing method according to claim 19.

21. The range is a range of ±20% around the reference value.

21. The information processing method according to claim 20.

22. The measurements are information about speed, 14. The information processing method according to claim 1,

23. The mechanical equipment is a robot having at least two joints; identifying the operating period based on the measurement value of the joint that has the highest number of times that the measurement value of the joint has reached the predetermined value; 23. The information processing method according to claim 1,

24. The mechanical equipment is a robot, The robot repeats an operation of receiving a workpiece, an operation of assembling the workpiece, and an operation of handing over the workpiece.

23. The information processing method according to claim 1,

25. The machine learning is unsupervised learning or supervised learning.

25. The information processing method according to claim 1,

26. Acquire measurements related to machinery and equipment, By detecting the number of times that the measurement value reaches a predetermined value per unit time, an operating period during which the mechanical equipment was in an operating state is identified; acquiring feature values ​​from the measurement values ​​corresponding to the operating period as machine learning data; Obtaining a trained model for acquiring the state of the mechanical equipment by machine learning using the machine learning data. A control unit is provided.

23. An information processing apparatus comprising:

27. A machine equipped with the information processing device according to claim 26.

28. A method for manufacturing an article, comprising the steps of: manufacturing an article by using the machinery and equipment according to claim 27;

29. 26. A program enabling a computer to execute the information processing method according to any one of claims 1 to 25.

30. 30. A computer-readable recording medium storing the program according to claim 29.

Images