JP2021002398A - Failure prediction device, failure prediction system, and failure prediction method - Google Patents

Abstract

To provide a failure prediction system which enables accurate failure prediction in accordance with circumstances.SOLUTION: A failure prediction system 1 includes a machine learning device 5 which learns condition associated with failure of an industrial machine 2. The machine learning device 5 includes: a status observation part 52 which observes a status variable composed of, for example, output data of a sensor 11, internal data of a control software, or calculation data obtained based on the aforementioned two types of data while the industrial machine 2 is in operation or stationary; a determination data acquisition part 51 which acquires determination data indicating the presence or absence of failure of the industrial machine 2 or the degree of failure; and a learning part 53 which learns condition associated with the failure of the industrial machine 2 through supervised learning in accordance with a training data set created based on a combination of a status variable and the determination data.SELECTED DRAWING: Figure 1

Description

The present invention relates to a machine learning method and a machine learning device for learning failure conditions, and a failure prediction device and a failure prediction system provided with the machine learning device.

In industrial machinery, it may be required to detect abnormalities in components in advance in order to improve the yield or prevent the occurrence of serious accidents. For example, a method of comparing the output value of a sensor with a predetermined threshold value and detecting an abnormality based on the result is known. Here, the term "industrial machine" means not only a machine controlled by an industrial robot or a computer numerical control (CNC) device, but also a machine including a service robot and various mechanical devices. And.

Patent Document 1 discloses a failure prediction diagnosis method for predicting a failure of a robot by comparing a reference motion pattern of a robot in a normal state with a motion pattern of a robot in operation.

Patent Document 2 describes a robot by comparing the difference between the work rate on the load side based on the actual operating state of the drive shaft and the work rate on the drive side based on the operation command to the drive shaft with the determination value. A failure prediction method for evaluating the presence or absence of deterioration and the deterioration level of the mechanical part is disclosed.

JP-A-63-123105 Japanese Unexamined Patent Publication No. 10-039908

However, as industrial machines become more complicated or sophisticated, the factors that lead to failures are also becoming more complicated. Therefore, conventional failure prediction methods performed according to certain criteria may not be applicable or inaccurate in actual situations. Therefore, there is a demand for a failure prediction device that enables accurate failure prediction according to the situation.

According to the first invention of the present application, it is a machine learning device that learns conditions associated with a failure of an industrial machine, and controls output data of a sensor that detects the state of the industrial machine or the surrounding environment, and the industrial machine. A state observation unit that observes the internal data of the control software and a state variable including at least one of the output data or the calculation data obtained based on the internal data while the industrial machine is operating or stationary, and the industry. It is associated with the failure of the industrial machine according to the judgment data acquisition unit that acquires the judgment data indicating the presence or absence of the failure of the machine or the degree of the failure, and the training data set created based on the combination of the state variable and the judgment data. A machine learning device including a learning unit that learns conditions by supervised learning is provided.
According to the second invention of the present application, it is a machine learning device that learns conditions associated with a failure of an industrial machine, and controls output data of a sensor that detects the state of the industrial machine or the surrounding environment, and the industrial machine. A state observation unit that observes the internal data of the control software and a state variable including at least one of the output data or the calculation data obtained based on the internal data while the industrial machine is operating or stationary, and the industry. It is associated with the failure of the industrial machine according to the judgment data acquisition unit that acquires the judgment data indicating the presence or absence of the failure of the machine or the degree of the failure, and the training data set created based on the combination of the state variable and the judgment data. A machine learning device including a learning unit for learning conditions by unsupervised learning is provided.
According to the third invention of the present application, in the machine learning device according to the first or second invention, the learning unit learns the conditions according to the training data set created for a plurality of industrial machines. It is configured as follows.
According to the fourth invention of the present application, in the machine learning device according to any one of the first to third inventions, the learning unit learns the normal state only for a certain period of time, and thereafter, the determination data acquisition. It is configured to detect the occurrence of a failure by a unit.
According to the fifth invention of the present application, in the machine learning device according to any one of the first to fourth inventions, the learning unit acquires the determination data indicating the failure of the industrial machine by the determination data acquisition unit. At that time, the determination data included in the training data set is weighted according to the length of time that goes back from the time of failure occurrence to the time of acquisition of the determination data, and the condition is updated.
According to the sixth invention of the present application, it is a failure prediction device for predicting a failure of the industrial machine provided with the machine learning device according to any one of the first to fifth inventions, and the learning unit performs the training. Failure prediction is further provided with a failure information output unit that outputs failure information indicating the presence or absence of failure or the degree of failure of the industrial machine in response to the current input of the state variable based on the result learned according to the data set. Equipment is provided.
According to the seventh invention of the present application, in the failure prediction device according to the sixth invention, the learning unit follows an additional training data set created based on the combination of the current state variable and the determination data. It is configured to relearn the above conditions.
According to the eighth invention of the present application, in the failure prediction device according to the sixth or seventh invention, the machine learning device is connected to the industrial machine via a network, and the state observation unit is connected to the industrial machine via the network. It is configured to acquire the current state variable.
According to the ninth invention of the present application, in the failure prediction device according to the eighth invention, the machine learning device exists on a cloud server.
According to the tenth invention of the present application, in the failure prediction device according to any one of the sixth to eighth inventions, the machine learning device is built in the control device for controlling the industrial machine.
According to the eleventh invention of the present application, in the failure prediction device according to any one of the sixth to tenth inventions, the learning result by the machine learning device is shared by a plurality of the industrial machines.
According to the twelfth invention of the present application, a failure prediction device according to any one of the sixth to eleventh inventions, a sensor that outputs the output data, a failure information notification unit that notifies the operator of the failure information, and the like. A failure prediction system is provided.
According to the thirteenth invention of the present application, in the failure prediction system according to the twelfth invention, the time when the failure information is notified to the operator by the failure information notification unit is the first time before the time when the failure occurs. It is before the time specified in the prescribed period.
According to the 14th invention of the present application, in the failure prediction system according to the 13th invention, the time when the failure information is notified to the operator by the failure information notification unit is the first time before the time when the failure occurs. It is before the time specified in the predetermined period and after the time specified in the second predetermined period, which is longer than the first predetermined period, retroactively from the time when the failure occurs.
According to the fifteenth invention of the present application, it is a machine learning method for learning conditions associated with a failure of an industrial machine, and controls output data of a sensor for detecting the state of the industrial machine or the surrounding environment, and the industrial machine. A state variable containing at least one of the internal data of the control software and the output data or the calculation data obtained based on the internal data is observed while the industrial machine is operating or stationary, and the failure of the industrial machine is observed. A machine that acquires judgment data indicating the presence or absence or the degree of failure, and learns the conditions associated with the failure of the industrial machine by supervised learning according to the training data set created based on the combination of the state variable and the judgment data. A learning method is provided.
According to the 16th invention of the present application, it is a machine learning method for learning conditions associated with a failure of an industrial machine, and controls output data of a sensor for detecting the state of the industrial machine or the surrounding environment, and the industrial machine. A state variable containing at least one of the internal data of the control software and the output data or the calculation data obtained based on the internal data is observed while the industrial machine is operating or stationary, and the failure of the industrial machine is observed. A machine that acquires judgment data indicating the presence or absence or the degree of failure, and learns the conditions associated with the failure of the industrial machine by unsupervised learning according to a training data set created based on the combination of the state variable and the judgment data. A learning method is provided.

These and other objectives, features and advantages of the present invention will become more apparent with reference to the detailed description of exemplary embodiments of the invention shown in the accompanying drawings.

The machine learning device and the machine learning method according to the present invention learn the conditions associated with the failure of an industrial machine according to a training data set created based on a combination of state variables and determination data. Since the failure conditions are learned while actually operating the industrial machine, accurate failure conditions according to the actual usage conditions are learned. Further, according to the failure prediction device and the failure prediction system according to the present invention, since the machine learning device capable of machine learning the failure conditions is provided, accurate failure prediction according to the actual usage situation becomes possible.

FIG. 1 is a block diagram showing an example of a failure prediction system according to an embodiment. FIG. 2 is a flowchart showing an example of the flow of the learning process in the machine learning device. FIG. 3 is a diagram showing a configuration example of a neural network. FIG. 4 is a diagram for explaining an example of a learning period in an unsupervised learning method. FIG. 5 is a diagram for explaining an example of a recurrent neural network. FIG. 6 is a block diagram showing an example of a failure prediction system according to another embodiment. FIG. 7 is a diagram (No. 1) for explaining an example of an index value indicating the degree of failure in the failure prediction system according to the embodiment. FIG. 8 is a diagram (No. 2) for explaining an example of an index value indicating the degree of failure in the failure prediction system according to the embodiment. FIG. 9 is a flowchart showing an example of the flow of failure prediction using the learning result.

Hereinafter, with reference to the accompanying drawings, a machine learning method and a machine learning device according to the present invention, and an embodiment of a failure prediction device and a failure prediction system provided with the machine learning device will be described. The components of the illustrated embodiment have been scaled appropriately to aid in the understanding of the present invention. Also, the same reference numerals are used for the same or corresponding components.

FIG. 1 is a block diagram showing an example of a failure prediction system according to an embodiment. The failure prediction system 1 can learn conditions (hereinafter, may be referred to as “failure conditions”) associated with a failure of an industrial machine by using a machine learning device 5 having a machine learning function. Further, the failure prediction system 1 can create failure information according to the state of the industrial machine and its surrounding environment based on the result learned by the machine learning device 5.

As used herein, "industrial machine" shall mean a variety of machines, including industrial robots, service robots and machines controlled by computer numerical control (CNC) devices. Further, in the present specification, "failure of industrial machine" includes failure of component parts of industrial machine. That is, the "failure of an industrial machine" is not limited to a state in which the intended function of the industrial machine cannot be performed, and includes, for example, a state in which the normal operation cannot be reproduced temporarily or permanently.

The "failure information" created by the failure prediction system 1 includes information indicating the presence or absence of failure of the industrial machine or information indicating the "degree of failure". The "fault information" may include information indicating that the industrial machine is in a normal state. "Degree of failure" means the seriousness of failure. The "degree of failure" may be limited to either the maximum value or the minimum value. The "degree of failure" may be a continuous quantity or a discrete quantity. The operator can determine whether the target component should be replaced or repaired immediately or at the next maintenance work, depending on the "degree of failure".

In the following description, the failure prediction system 1 used for predicting the failure of the robot 2 will be described. However, one of ordinary skill in the art will recognize that the present invention can be applied to any other industrial machine as well.

The robot 2 illustrated in FIG. 1 is a 6-axis vertical articulated robot in which each joint is driven by a motor. The robot 2 is connected to the robot control device 3 by a known communication means. The robot control device 3 creates a command for the robot 2 according to the control program.

The robot control device 3 is a digital computer including a CPU, a ROM, a RAM, a non-volatile memory, and an interface connected to an external device. As shown in FIG. 1, the robot control device 3 includes a failure determination unit 31.

The failure determination unit 31 determines the failure of the robot 2 by using a known failure diagnosis method. The failure determination unit 31 determines the presence or absence of a failure or the degree of failure of the robot 2 independently of the failure information created by the failure prediction system 1. For example, when the disturbance torque detected by the torque sensor or the vibration amplitude of the output data of the sensor exceeds a predetermined threshold value, the failure determination unit 31 determines that a failure has occurred. Alternatively, the failure determination unit 31 may determine that a failure of the robot 2 has occurred based on the internal data of the control software stored in the robot control device 3. In this way, the failure determination unit 31 determines a failure based on various factors. The determination result by the failure determination unit 31 is input to the determination data acquisition unit 51 of the machine learning device 5 described later.

In another embodiment, the machine learning device 5 is configured so that failure information is input to the determination data acquisition unit 51 in response to an operator's input operation that discovers or knows the failure of the robot 2. You may.

The failure prediction system 1 further includes a robot 2 or a sensor 11 that detects the state of the surrounding environment. The sensor 11 is at least one of a force sensor, a torque sensor, a vibration sensor, a sound collecting sensor, an imaging sensor, a distance sensor, a temperature sensor, a humidity sensor, a flow rate sensor, a light amount sensor, a pH sensor, a pressure sensor, a viscosity sensor, and an odor sensor. One may be included. The data output from the sensor 11 (hereinafter, may be simply referred to as “output data”) is input to the state observation unit 52 of the machine learning device 5.

The machine learning device 5 learns the failure conditions of the robot 2. In one embodiment, the machine learning device 5 may be a digital computer connected to the robot 2 via a network and separate from the robot control device 3.

In another embodiment, the machine learning device 5 may be built in the robot control device 3. In that case, the machine learning device 5 executes machine learning using the processor of the robot control device 3. In yet another embodiment, the machine learning device 5 may exist on the cloud server.

As shown in FIG. 1, the machine learning device 5 includes a determination data acquisition unit 51, a state observation unit 52, and a learning unit 53.

The determination data acquisition unit 51 acquires determination data from the failure determination unit 31. The determination data is input from the determination data acquisition unit 51 to the learning unit 53, and is used when the machine learning device 5 learns the failure condition. The determination data is data for determining the presence or absence of a failure or the degree of failure. The determination data does not have to include data indicating that there is a failure, that is, the robot 2 is in an abnormal state.

The state observation unit 52 observes a state variable as an input value for machine learning while the robot 2 is operating or stationary. In the embodiment in which the machine learning device 5 is connected to the robot 2 and the sensor 11 via a network, the state observation unit 52 acquires a state variable via the network.

The state variable may include the output data of the sensor 11. The state variable may include internal data of the control software that controls the robot 2. Internal data may include at least one of torque, position, velocity, acceleration, jerk, current, voltage and estimated disturbance value. The estimated disturbance value is, for example, a disturbance value estimated by the observer based on the torque command and the velocity feedback.

State variables may include calculated data obtained based on output data or internal data. The calculated data may be acquired using at least one of frequency analysis, time frequency analysis and autocorrelation analysis. Of course, the calculated data may be obtained using a simpler calculation, such as coefficient multiplication or calculus.

The learning unit 53 learns the failure condition according to the training data set created based on the combination of the state variable output from the state observation unit 52 and the determination data output from the determination data acquisition unit 51. The training data set is data in which state variables and judgment data are associated with each other.

An example of the learning process in the machine learning device 5 will be described with reference to FIG. When the learning is started, in step S201, the state observation unit 52 acquires a state variable including output data, internal data, calculation data, and the like. In step S202, the determination data acquisition unit 51 acquires determination data based on the determination result by the failure determination unit 31.

In step S203, the learning unit 53 learns the failure condition according to the training data set created based on the combination of the state variable acquired in step S201 and the determination data acquired in step S202. The processes of steps S201 to S203 are repeatedly executed until the machine learning device 5 sufficiently learns the failure conditions.

In one embodiment, the learning unit 53 of the machine learning device 5 may learn the failure conditions according to the neural network model. FIG. 3 shows an example of a neural network model. Neural network comprises l neurons _{x 1, x 2, x 3} , ···, an input layer comprising a x _l, m neurons _{y 1, y 2, y 3} , ···, a y _m It is composed of an intermediate layer (hidden layer) and an output layer containing _n neurons z ₁ , z ₂ , z ₃ , ..., Z _n . Although only one intermediate layer is shown in FIG. 3, two or more intermediate layers may be provided. The machine learning device 5 (neural network) may use a general-purpose computer or processor, but if GPGPU (General-Purpose computing on Graphics Processing Units), a large-scale PC cluster, or the like is applied, the processing speed is higher. It is possible.

The neural network learns the failure conditions associated with the failure of the robot 2. The neural network is subjected to so-called supervised learning according to the training data set created based on the combination of the state variable observed by the state observation unit 52 and the judgment data acquired by the judgment data acquisition unit 51. Learn the relationship with failure occurrence, that is, failure conditions. Supervised learning is a model in which a large number of sets of data of a certain input and a result (label) are given to a learning device to learn the features in those data sets and estimate the result from the input, that is, the relationship thereof. Can be obtained inductively.

Alternatively, the neural network can accumulate only the state variables in the state without failure, that is, when the robot 2 is operating normally, and can learn the failure condition by so-called unsupervised learning. For example, if the frequency of robot 2 failures is extremely low, an unsupervised learning method may be effective. Unsupervised learning is to learn how the input data is distributed by giving a large amount of input data to the learning device, and compress the input data without giving the corresponding teacher output data.・ It is a method to learn a device that performs classification and shaping. The features in those datasets can be clustered among similar people. Using this result, it is possible to predict the output by setting some criteria and allocating the output to optimize it. In addition, as an intermediate problem setting between unsupervised learning and supervised learning, there is also what is called semi-supervised learning, in which only a part of the input and output data sets exist, and the others are input only. This is the case when it is data.

FIG. 4 is a diagram for explaining an example of a learning period in an unsupervised learning method. Here, the horizontal axis indicates time (elapsed time), and the vertical axis indicates the degree of failure. As shown in FIG. 4, the above-mentioned unsupervised learning method has a learning period of a certain period, for example, several weeks, starting from immediately after the robot 2 is shipped or immediately after maintenance. Update the state variable only when and define it as normal. After that, the state variables are not updated, and the "degree of failure" is output from the output result output from the neural network based on the distance from the normal model, and only the abnormality judgment is performed to detect the abnormality. You can achieve what you do.

Further, in the present embodiment, for example, it is effective to use a neural network called a recurrent type in order to model time series data having a time correlation. A recurrent neural network (RNN) does not form a learning model using only the state of the current time, but also uses the internal state of the time so far. A recurrent neural network can be treated in the same way as a general neural network by expanding and considering a network on the time axis. Here, there are various types of recurrent neural networks, but as an example, a simple recursive network (Elman network) will be described.

FIG. 5 is a diagram for explaining an example of a recurrent neural network, FIG. 5 (a) shows the time-axis expansion of the Elman network, and FIG. 5 (b) shows the backpropagation method (backpropagation). : Backpropagation) backpropagation time through time (BPTT: Back Propagation Through Time) is shown. Here, backpropagation can be applied if the structure of the Elman network is as shown in FIG. 5A.

However, in the Elman network, unlike a normal neural network, as shown in Fig. 5 (b), the error propagates back in time, and such backpropagation is referred to as backpropagation through time (BPTT). Call. By applying such a neural network structure, it is possible to estimate an output model based on the transition of the input so far. For example, it fails to determine whether the estimated output value is an abnormal value. It can be used in relation to the outbreak.

When performing failure prediction described later, the output layer outputs information indicating the presence or absence of a failure corresponding to the above-mentioned failure information or "degree of failure" in response to a state variable input to the input layer of the neural network. The possible value of the "degree of failure" may be a value in which either the maximum value or the minimum value is limited, a continuous amount, or a discrete amount.

According to the machine learning device and the machine learning method according to the above-described embodiment, it is possible to learn more accurate failure conditions according to the actual usage situation than the failure conditions based on the determination data output from the determination data acquisition unit 51. As a result, even when the factors leading to the failure are complicated and it is difficult to set the failure conditions in advance, it is possible to predict the failure with high accuracy.

In one embodiment, when the determination data acquisition unit 51 acquires the determination data representing the failure of the robot 2, the learning unit 53 retraces the determination data from the time when the failure occurs to the time when each determination data is acquired. The failure condition may be updated by weighting each according to the length. Here, it is presumed that the shorter the time from the acquisition of the determination data to the actual occurrence of the failure, the closer to the state directly linked to the occurrence of the failure. Therefore, if the determination data is weighted according to the elapsed time from the acquisition of the training data set, the failure condition can be effectively learned.

In one embodiment, the learning unit 53 may learn the failure conditions according to the training data sets created for the plurality of robots 2. The learning unit 53 may acquire training data sets from a plurality of robots 2 used at the same site, or training data collected from a plurality of robots 2 operating independently at different sites. The failure condition may be learned by using the set. Further, the robot 2 that collects the training data set can be added to the target on the way, or conversely, can be removed from the target.

Next, as a method of sharing (sharing) training data sets of a plurality of robots 2, three examples are given below, but it goes without saying that other methods can be applied. First, as a first example, there is a method of sharing the model of the neural network so as to be the same. For example, for each weight coefficient of the network, the difference between the robots 2 is transmitted by using a communication means. It reflects. Further, as a second example, the weight of the learning device 5 can be shared by sharing the input and output data sets of the neural network. Further, as a third example, a certain database is prepared, and the state is shared (similar model) by accessing the database and loading a more appropriate model of the neural network.

FIG. 6 is a block diagram showing an example of a failure prediction system according to another embodiment. The failure prediction system 1 includes a failure prediction device 4 that creates failure information of the robot 2 by using the result learned by the machine learning device 5.

The failure prediction device 4 includes a state observation unit 41 and a failure information output unit 42. The state observation unit 41 functions in the same manner as the state observation unit 52 described with reference to FIG. 1, and acquires state variables that reflect the states of the robot 2 and the surrounding environment. The failure information output unit 42 responds to the input of the state variable via the state observation unit 41 based on the result of learning by the learning unit 53 of the machine learning device 5 according to the training data set, and causes the failure information of the robot 2. Is output.

As shown in FIG. 6, the robot control device 3 can include a notification unit (fault information notification unit) 32. The notification unit 32 notifies the operator of the failure information output by the failure information output unit 42. The mode in which the failure information is notified is not particularly limited as long as the operator can know it. For example, the presence or absence of a predicted failure or the degree of failure may be displayed on a display device (not shown), or a warning sound may be generated according to the content of the failure information.

7 and 8 are diagrams for explaining examples of index values (1st to 4th examples) indicating the degree of failure in the failure prediction system according to the embodiment. Here, in FIGS. 7 (a), 7 (b), 7 (c) and 8, the horizontal axis represents time and the vertical axis represents the degree of failure. First, as shown in FIG. 7A, for example, in the first example, the index value indicating the "degree of failure" is set to increase as the failure approaches, and the index value obtained by learning is used as it is. It can be configured so that the failure information output unit 42 outputs the failure information. Further, as shown in FIG. 7B, for example, in the second example, a threshold value is set for the above-mentioned index value, and if it is above the threshold value, it is abnormal, and if it is below the threshold value, it is normal. It is also possible to configure the failure information output unit 42 to output the represented information as failure information. Further, as shown in FIG. 7 (c), for example, in the third example, a plurality of threshold values (threshold values 1 to 3) are provided for the above-mentioned index values, and the levels separated by each threshold value (failure level 1 to failure). It is also possible to configure the failure information output unit 42 to output the level 4) as the failure information.

As shown in FIG. 8, for example, in the fourth example, the relationship between the above-mentioned index value and the time until the failure is obtained based on the data (teacher data) leading to a plurality of failures, and based on the relationship. , The first threshold value for satisfying that it is before the time specified in the first predetermined period retroactively from the time when the failure occurs is obtained. In addition, a second threshold value is set to satisfy that the time is later than the time specified in the second predetermined period retroactively from the time when the failure occurs. Then, when at least one of the index value being less than the first threshold value and the index value being equal to or more than the second threshold value is satisfied, the index value itself or the level obtained by dividing the index value by the threshold value is determined. The failure information output unit 42 can also output as failure information. The method of determining the threshold value in this case is, for example, a threshold value can be set so that the past teacher data satisfies all the conditions, a margin can be set as necessary, and the threshold value can be set. In addition, the threshold value can be set so as to allow a judgment error within a certain probability.

Next, with reference to FIG. 9, an example of failure prediction executed by utilizing the result learned by the machine learning device will be described. In step S501, the state observation unit 41 acquires the current state variable including the output data from, for example, the sensor 11. In step S502, the failure information output unit 42 outputs failure information according to the state variable acquired in step S501 based on the learning result of the machine learning device 5 described above. When the failure prediction system 1 includes the notification unit 32, the step of notifying the operator of the failure information may be executed after step S502.

The failure prediction by the failure prediction device 4 described with reference to FIG. 9 may be performed when the robot 2 executes a predetermined specific operation. Alternatively, the processes of steps S501 to S502 may be continuously executed in parallel while the robot 2 is operating or stationary. Alternatively, failure prediction may be performed periodically at a predetermined time.

In one embodiment, machine learning by the machine learning device 5 may be executed in parallel with executing the failure prediction by the failure prediction device 4. In that case, at the same time that the failure prediction device 4 creates the failure information, the machine learning device 5 learns based on the determination data acquired through the operation of the failure determination unit 31 or the operator and the state variable at that time. Part 53 relearns the failure condition.

Although the embodiment of machine learning using a neural network has been described, machine learning may be performed according to other known methods such as genetic programming, functional logic programming, and support vector machines. Also, again, in the present specification, the term "industrial machine" means various machines including industrial robots, service robots and machines controlled by computer numerical control (CNC) devices. , As mentioned above.

Although various embodiments of the present invention have been described above, those skilled in the art will recognize that other embodiments can also realize the intended effects of the present invention. In particular, the components of the above-described embodiments can be deleted or replaced without departing from the scope of the present invention, or known means can be further added. It will also be apparent to those skilled in the art that the present invention can also be practiced by optionally combining the features of a plurality of embodiments explicitly or implicitly disclosed herein.

1 Failure prediction system 2 Robot 3 Robot control device 4 Failure prediction device 5 Machine learning device 11 Sensor 31 Failure judgment unit 32 Notification unit 41 State observation unit 42 Failure information output unit 51 Judgment data acquisition unit 52 State observation unit 53 Learning unit

Claims (16)

A machine learning device that learns the conditions associated with industrial machine failures.
Includes at least one of the output data of a sensor that detects the state of the industrial machine or the ambient environment, the internal data of the control software that controls the industrial machine, and the output data or the calculated data obtained based on the internal data. A state observation unit that observes state variables while the industrial machine is operating or stationary, and
A judgment data acquisition unit that acquires judgment data indicating the presence or absence of failure or the degree of failure of the industrial machine, and
A learning unit that learns conditions associated with a failure of the industrial machine by supervised learning according to a training data set created based on a combination of the state variables and the determination data.
A machine learning device characterized by that. A machine learning device that learns the conditions associated with industrial machine failures.
Includes at least one of the output data of a sensor that detects the state of the industrial machine or the ambient environment, the internal data of the control software that controls the industrial machine, and the output data or the calculated data obtained based on the internal data. A state observation unit that observes state variables while the industrial machine is operating or stationary, and
A judgment data acquisition unit that acquires judgment data indicating the presence or absence of failure or the degree of failure of the industrial machine, and
A learning unit that learns conditions associated with a failure of the industrial machine by unsupervised learning according to a training data set created based on a combination of the state variables and the determination data.
A machine learning device characterized by that. The learning unit is configured to learn the conditions according to the training data set created for a plurality of industrial machines.
The machine learning device according to claim 1 or 2, wherein the machine learning device is characterized in that. The learning unit learns the normal state only for a certain period of time, and then detects the occurrence of a failure by the determination data acquisition unit.
The machine learning device according to any one of claims 1 to 3, wherein the machine learning device is characterized by the above. When the determination data acquisition unit acquires the determination data representing the failure of the industrial machine, the learning unit obtains the determination data included in the training data set from the time of failure occurrence to the time of acquisition of the determination data. The above condition is updated by weighting according to the length of the retroactive time.
The machine learning device according to any one of claims 1 to 4, wherein the machine learning device is characterized by the above. A failure prediction device for predicting a failure of the industrial machine, comprising the machine learning device according to any one of claims 1 to 5.
A failure information output unit that outputs failure information indicating the presence or absence of failure or the degree of failure of the industrial machine in response to the current input of the state variable based on the result learned by the learning unit according to the training data set. Further prepare,
A failure prediction device characterized by this. The learning unit relearns the conditions according to an additional training data set created based on the combination of the current state variables and the determination data.
The failure prediction device according to claim 6, wherein the failure prediction device is characterized. The machine learning device is connected to the industrial machine via a network.
The state observation unit acquires the current state variable via the network.
The failure prediction device according to claim 6 or 7. The machine learning device exists on a cloud server.
8. The failure prediction device according to claim 8. The machine learning device is built in a control device that controls the industrial machine.
The failure prediction device according to any one of claims 6 to 8, wherein the failure prediction device is characterized by the above. The learning result by the machine learning device is shared by a plurality of the industrial machines.
The failure prediction device according to any one of claims 6 to 10, characterized in that. The failure prediction device according to any one of claims 6 to 11.
A sensor that outputs the output data and
A failure information notification unit for notifying the operator of the failure information is provided.
A failure prediction system characterized by this. The time when the failure information is notified to the operator by the failure information notification unit is earlier than the time determined in the first predetermined period retroactively from the time when the failure occurs.
The failure prediction system according to claim 12. The time when the failure information is notified to the operator by the failure information notification unit is before the time specified in the first predetermined period retroactively from the time when the failure occurs, and also before the time when the failure occurs. , After the time specified in the second predetermined period, which is longer than the first predetermined period.
13. The failure prediction system according to claim 13. A machine learning method that learns the conditions associated with industrial machine failures.
Includes at least one of the output data of the sensor that detects the state of the industrial machine or the ambient environment, the internal data of the control software that controls the industrial machine, and the output data or the calculated data obtained based on the internal data. Observing the state variable while the industrial machine is operating or stationary,
Obtaining judgment data indicating the presence or absence of failure or the degree of failure of the industrial machine,
According to the training data set created based on the combination of the state variable and the determination data, the conditions associated with the failure of the industrial machine are learned by supervised learning.
A machine learning method characterized by that. A machine learning method that learns the conditions associated with industrial machine failures.
Includes at least one of the output data of the sensor that detects the state of the industrial machine or the ambient environment, the internal data of the control software that controls the industrial machine, and the output data or the calculated data obtained based on the internal data. Observing the state variable while the industrial machine is operating or stationary,
Obtaining judgment data indicating the presence or absence of failure or the degree of failure of the industrial machine,
According to the training data set created based on the combination of the state variable and the determination data, the conditions associated with the failure of the industrial machine are learned by unsupervised learning.
A machine learning method characterized by that.

Images