KR20230094088A - Method of Training a Machine Learning Model for Identification of a Faulty Motor in an Industrial Robot - Google Patents

Abstract

Disclosed is a method of training a machine learning model to identify an error motor in a robot including a plurality of motors, which identifies faulty motor-related parts without adding new sensors to a robot. According to the present invention, the method comprises the steps of: setting a pattern in which a robot operates and an intermediate arrival point on a motion trajectory according to the pattern and providing angle values to be provided to each of the plurality of motors when the robot reaches the intermediate arrival point; acquiring a plurality of first trajectory datasets while controlling the robot to move along the motion trajectory in a steady state; acquiring a plurality of second trajectory datasets associated with each of a plurality of motor replacement states while controlling the robot to move along the motion trajectory in the plurality of motor replacement states; performing preprocessing on the plurality of first trajectory datasets and the plurality of second trajectory datasets associated with each of the plurality of motor replacement states to form learning data; and at least partially using the learning data to train a machine learning model.

Description

Method of Training a Machine Learning Model for Identification of a Faulty Motor in an Industrial Robot}

The disclosure below relates to a method for identifying a faulty motor in an industrial robot.

There are many ways to find a motor error, but there is currently no way to find out which motor or motor-related driver board has an error in an industrial robot. In general, an industrial robot is composed of two or more motors and related driver boards. If an error occurs in these motors or driver boards, it is necessary to check which motor part has an error and fix it. However, it is possible to confirm that the robot has an error, but it takes a lot of time to find which motor part has an error. Even if an error actually occurs, it cannot be determined that it is an error unless you look closely. For example, if there is slight wear on the bearing or slight damage to the electric/electronic parts of the motor or driver board, there is a slight difference from normal operation, but it may be judged to be normal operation from the viewpoint of the overall operation of the robot. there is. When these situations are accumulated, a failure actually occurs, which may cause a problem in the production of the product and cause great damage.

The background art described above is possessed or acquired by the inventor in the process of deriving the disclosure of the present application, and cannot necessarily be said to be known art disclosed to the general public prior to the present application.

An object to be solved by the present disclosure is to provide a machine learning method capable of identifying faulty motor-related parts without adding new sensors to the robot.

The problem to be solved by the present disclosure is not limited to the problems mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below.

According to one aspect of the present disclosure, a method for training a machine learning model to identify an erroneous motor in a robot including a plurality of motors is provided. The method sets an intermediate reaching point on an operating pattern of the robot and an operating trajectory according to the pattern, and provides angle values each of the plurality of motors has when the robot reaches the intermediate reaching point, the robot obtaining a plurality of first trajectory data sets while controlling the robot to move along the motion trajectory in the steady state; controlling the robot to move along the motion trajectory in the multiple motor replacement state; Acquiring a plurality of second trajectory data sets associated with each, performing preprocessing on the plurality of first trajectory data sets and the plurality of second trajectory data sets associated with each of the plurality of motor replacement states to obtain learning data constructing, and training the machine learning model using at least in part the training data.

In one embodiment, the steady state is a state in which a plurality of error-free motors are mounted on the robot.

In one embodiment, each of the plurality of first trajectory data sets includes data on angular values and velocities or angular velocity values of each of the plurality of motors at a point in time when the corresponding first trajectory data set is obtained.

In one embodiment, the step of constructing learning data by performing pre-processing on the plurality of first sign data sets and the plurality of second sign data sets associated with each of the plurality of motor replacement states, the plurality of first sign data sets Performing pre-processing on each of the first trajectory data sets, wherein the step of performing pre-processing on each of the plurality of first trajectory data sets includes a number indicating the intermediate arrival point and the robot reaching the intermediate arrival point. and adding data about angular values that the plurality of motors have when reaching the corresponding first trajectory data set.

In one embodiment, the step of performing preprocessing on each of the plurality of first sign data sets may include setting a sequence number (seqNum) indicating an order in which the corresponding first sign data set was acquired to the corresponding first sign data set. It further includes the step of adding to

In one embodiment, the step of performing preprocessing on each of the plurality of first trajectory data sets may include the robot reaching the intermediate arrival point at angle values of the plurality of motors included in the corresponding first trajectory data set The method further includes adding data about difference values obtained by subtracting angle values of the plurality of motors to the corresponding first trajectory data set.

In one embodiment, the step of performing preprocessing on each of the plurality of first sign data sets may include data indicating that the corresponding first sign data set is obtained in the steady state, the corresponding first sign data set It further includes the step of adding to

In one embodiment, each of the plurality of motor replacement states is a state in which one of the plurality of error-free motors in the robot is replaced with a faulty motor.

In one embodiment, each of the plurality of second trajectory data sets associated with each of the plurality of motor replacement states is a plurality of motors including the replaced faulty motor at the time when the corresponding second trajectory data set was obtained. It includes data about each angular value and velocity or angular velocity value of .

In one embodiment, the step of constructing learning data by performing preprocessing on the plurality of first trajectory data sets and the plurality of second trajectory data sets associated with each of the plurality of motor replacement states, the plurality of motor replacement Performing pre-processing on each of a plurality of second trajectory data sets associated with each of the states, wherein performing pre-processing on each of the plurality of second trajectory data sets indicates the intermediate arrival point and adding data about numbers and angular values of the plurality of motors when the robot reaches the intermediate reaching point to the corresponding second trajectory data set.

In one embodiment, the step of performing preprocessing on each of the plurality of second sign data sets includes adding a sequence number indicating an order in which the corresponding second sign data set was obtained to the corresponding second sign data set. Include more steps.

In one embodiment, the step of performing preprocessing on each of the plurality of second trajectory data sets may include the robot reaching the intermediate arrival point at angle values of the plurality of motors included in the corresponding second trajectory data set The method further includes adding data about difference values obtained by subtracting angle values of the plurality of motors to the corresponding second trajectory data set.

In one embodiment, the step of performing preprocessing on each of the plurality of second trajectory data sets may indicate in which motor replacement state the corresponding second trajectory data set is obtained from among the plurality of motor replacement states. and adding data to the corresponding second trajectory data set.

In one embodiment, the machine learning model is a decision tree learning model, a support vector machine (SVM) learning model, or a k-Nearest Neighbor (kNN) learning model.

According to another feature of the present disclosure, a method for identifying a faulty motor in a robot including a plurality of motors using a machine learning model is provided. The method includes setting a pattern in which the robot operates, acquiring a plurality of trajectory data sets while controlling the robot to move according to the set pattern, and performing preprocessing on each of the plurality of trajectory data sets The method may include configuring input data, and inputting the input data to the machine learning model and controlling output data from the machine learning model to be provided.

In one embodiment, each of the plurality of trajectory data sets includes data regarding angle values and velocities or angular velocity values of each of the plurality of motors at a time point when the corresponding trajectory data set is obtained.

In one embodiment, the step of configuring the input data by performing preprocessing on each of the plurality of trajectory data sets includes a number indicating an intermediate arrival point on the motion trajectory according to the set pattern and the robot reaching the intermediate arrival point and adding data about angular values each of the plurality of motors has to the corresponding trajectory data set.

In one embodiment, the step of configuring the input data by performing preprocessing on each of the plurality of trajectory data sets is to assign a sequence number (seqNum) indicating an order in which the corresponding trajectory data set was acquired to the corresponding trajectory data set. It further includes the step of adding

In one embodiment, the step of configuring the input data by performing preprocessing on each of the plurality of trajectory data sets, the robot at the intermediate arrival point at the angle values of the plurality of motors included in the corresponding trajectory data set The method may further include adding data about difference values obtained by subtracting angle values of the plurality of motors to the corresponding trajectory data set when reaching the corresponding trajectory data set.

In one embodiment, the output data includes a number of a faulty motor among the plurality of motors.

According to the disclosed embodiments, there is a technical effect of being able to identify faulty motor-related parts through a machine learning model without adding new sensors to the robot.

1 is a conceptual diagram for explaining a process of identifying an error motor in an industrial robot using a machine learning model learned according to the present disclosure.
2 is a flowchart illustrating an embodiment of a method for learning a machine learning model implemented in a robot controller to identify an error motor in an industrial robot.
3 is a flowchart illustrating an embodiment of a method of identifying a faulty motor in a robot including a plurality of motors by using a machine learning model learned according to embodiments according to the present disclosure.

Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only, and may be changed and implemented in various forms. Therefore, the form actually implemented is not limited only to the specific embodiments disclosed, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical idea described in the embodiments.

Although terms such as first or second may be used to describe various components, such terms should only be construed for the purpose of distinguishing one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element.

It should be understood that when an element is referred to as being “connected” to another element, it may be directly connected or connected to the other element, but other elements may exist in the middle.

Singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, terms such as "comprise" or "have" are intended to designate that the described feature, number, step, operation, component, part, or combination thereof exists, but one or more other features or numbers, It should be understood that the presence or addition of steps, operations, components, parts, or combinations thereof is not precluded.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and unless explicitly defined in this specification, it should not be interpreted in an ideal or excessively formal meaning. don't

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. In the description with reference to the accompanying drawings, the same reference numerals are given to the same components regardless of reference numerals, and overlapping descriptions thereof will be omitted.

1 is a conceptual diagram for explaining a process of identifying an error motor in an industrial robot using a machine learning model learned according to the present disclosure.

The industrial robot 110 is composed of a mechanical frame and a mechanical element 112 consisting of a plurality of motors coupled to the mechanical frame and a robot controller 114. In the following description, the motor is also referred to as a 'joint' or 'axis', but it should be understood to include parts of a motor and a servo driver. Also, in the following description, 'robot' may refer to only the mechanical element 112 and may mean the robot 110 including the mechanical element 112 and the robot controller 114 depending on the context. The robot 110 may operate according to a pattern set by a user. In one embodiment, the pattern may be a pattern such as a straight line, a triangle, a quadrangle, a pentagon, a hexagon, etc., but it should be recognized that the type of pattern is not limited thereto. When a pattern is set, a motion trajectory is specified. The motion trajectory refers to a path that an end effector of the robot 110 takes from a starting position to a target position. When the robot 110 operates along a specified motion trajectory and the length of the motion trajectory is long, several intermediate arrival points may be designated and moved between the start position and the target position. In order for the robot 110 to move between the intermediate reaching points, the robot controller 114 measures the angular values of the motors at the intermediate reaching point, which is the starting point, and based on this measures the angular values that the corresponding motors should have at the intermediate reaching point, which is the ending point of the corresponding motor. deliver to That is, the robot controller 114 controls the end effector to go from an intermediate arrival point to the next intermediate arrival point by providing angles at which the motors should move to the respective motors in advance. However, in this case, if an error occurs in one motor, the error-occurring motor does not operate normally, and problems such as going a little slower or faster occur, and the operation of the error motor affects the moving distance of the other motors. Therefore, if one motor has an error, other motors also generate different types of motions from the normal ones, and accordingly, the robot generally operates in a different type from the normal ones. The robot controller 114 acquires the measured values of the speed/angular velocity and angle of the motors reflecting different types of motions from these normal motions, configures them as input data, and converts them into the machine learning model 116 learned according to the present disclosure. It can be configured to input The user can then identify the faulty motor by referring to the output of the machine learning model 116 . In one embodiment, the output of the machine learning model 116 includes the number of the faulty motor (axis) (e.g., 1-axis, 6-axis, etc. in the case of a 6-axis robot). In one embodiment, the machine learning model 116 provides the number 0 or the character 'No Error' if there are no errors. In one embodiment, machine learning model 116 is a decision tree learning model. In one embodiment, machine learning model 116 is a support vector machine (SVM) learning model. In one embodiment, the machine learning model 116 is a k-Nearest Neighbor (kNN) learning model.

2 is a flowchart illustrating an embodiment of a method for learning a machine learning model implemented in a robot controller to identify an error motor in an industrial robot.

This machine learning model learning method sets an operating pattern of the robot 110 and an intermediate reaching point on a motion trajectory according to the pattern, and when the robot 110 reaches the intermediate reaching point, a plurality of motors provide angle values each have. It starts from step S205. The pattern set here may be a pattern that can well represent the errors of the axes (motors). The motion pattern to be set may be a reciprocating straight line pattern with a long distance capable of causing motion of all axes, a triangle pattern, a square pattern, and the like. The intermediate arrival point may be any one of the intermediate arrival points described above with reference to FIG. 1 or any one of detailed intermediate arrival points thereof. When the robot 110 reaches an intermediate point, angle values each of the plurality of motors have may be calculated based on inverse kinematics or provided by measurement. In step S210, a plurality of first trajectory data sets are acquired while controlling the robot 110 to move along the motion trajectory in a steady state. Here, the normal state may be a state in which a plurality of error-free motors are mounted on the robot 110 . The plurality of first trajectory data sets may be data sets obtained when the robot 110 moves detailed path sections along the motion trajectory. In one embodiment, the number of the plurality of first locus data sets is 1,000 or more. Referring to Table 1 below, which shows an embodiment of the first trajectory data set, the first trajectory data set is each angular value and speed of a plurality of motors (joints) at the time when the first trajectory data set is acquired, or It may contain data about angular velocity values.

[Table 1]

Table 1 illustrates an embodiment of the first trajectory data set obtained in the case of a 6-axis robot with six motors. data (Joint1 to Joint6) and current velocity/angular velocity values (Joint1Vel to Joint6Vel) of each of the current angle values. It should be noted that the first trajectory data set of Table 1 is a data set obtained when the robot 110 is in a steady state. Table 1 shows the first trajectory data set in the case of a 6-axis robot, but those skilled in the art can similarly configure the first trajectory data set for a 4-axis robot or robots with other structures. you will understand that you can In addition, it should be recognized that Table 1 only conceptually shows a method of storing data in association with each other for the purpose of illustration, and the items shown in Table 1 do not illustrate or imply the structure of stored data.

In step S215, a plurality of second trajectory data sets associated with each of a plurality of motor replacement states are acquired while controlling the robot 110 to move along motion trajectories in a plurality of motor replacement states. Here, each of the plurality of motor replacement states may be a state in which one of the plurality of error-free motors in the robot 110 is replaced with a faulty motor. Taking the case of a 4-axis robot as an example, the plurality of motor replacement states include the first motor replacement state in which the first motor is replaced with an error motor and the remaining motors are normal motors without errors, the second motor is replaced with an error motor, and the remaining motors are replaced with an error motor. 2nd motor replacement state where the motors are normal motors without errors, 3rd motor replacement state where the third motor is replaced with a faulty motor and the remaining motors are normal motors without errors, and the 4th motor is replaced with a faulty motor and the remaining motors are replaced without errors A fourth motor replacement state of normal motors may be included. In order to create a motor replacement state, a motor with an error may actually be mounted on the robot 110, but a motor or servo driver error may be generated through simulation, resulting in a motor failure state. Referring to the process of acquiring the second trajectory data set in detail using the case of the above example, first, a plurality of second trajectory data sets are acquired while controlling the robot 100 to move along the motion trajectory in the first motor replacement state. In the present disclosure, the plurality of second locus data sets thus obtained are referred to as a plurality of second locus data sets associated with the first motor replacement state. A plurality of second trajectory data sets are obtained while the robot 100 is controlled to move along the motion trajectory in the second motor replacement state. In the present disclosure, the plurality of second locus data sets obtained in this way are referred to as a plurality of second locus data sets associated with a second motor replacement state. While the robot 100 is controlled to move along the motion trajectory in the third motor replacement state, a plurality of second trajectory data sets are acquired. In the present disclosure, the plurality of second locus data sets obtained in this way are referred to as a plurality of second locus data sets associated with a third motor replacement state. Finally, while the robot 100 is controlled to move along the motion trajectory in the fourth motor replacement state, a plurality of second trajectory data sets are acquired. In the present disclosure, the plurality of second locus data sets obtained in this way are referred to as a plurality of second locus data sets associated with a fourth motor replacement state. In one embodiment, the number of the plurality of second locus data sets obtained in each motor replacement state may be 600 or more. The second locus data set is similar to the first locus data set shown in Table 1, but it should be noted that the second locus data set is a data set obtained in a motor replacement state in which one of the plurality of motors is an error motor.

In step S220, preprocessing is performed on the plurality of first sign data sets and the plurality of second sign data sets associated with each of the plurality of motor replacement states to form learning data. By performing these preprocessing steps, the learning of the machine learning model 116 can be performed more accurately, and a faulty motor can be found more quickly. First, a process of performing preprocessing on each of a plurality of first trajectory data sets will be described with reference to Table 2 below.

[Table 2]

In the pre-processing process, data (targetJoint1 to targetJoint6) about the number (Path) indicating the intermediate reaching point set in step S205 and the angle values each of the plurality of motors have when the robot 110 reaches the intermediate reaching point are first processed. A process of adding to the trajectory data set may be included. The pre-processing may further include adding a sequence number (seqNum) indicating an order in which the first sign data set was obtained to the corresponding first sign data set. Optionally, the pre-processing process determines the angle values (targetJoint1 to targetJoint6) of the plurality of motors when the robot 100 reaches the intermediate point among the angle values (Joint1 to Joint6) of the plurality of motors included in the first trajectory data set. A process of adding data (diffJoint1 to diffJoint6) of difference values obtained by subtraction, respectively, to the corresponding first locus data set may be further included. The pre-processing may further include adding data (errorJoint) indicating that the first sign data set is obtained in a normal state to the corresponding first sign data set. In one embodiment, all errorJoint values included in the first trajectory data set are zero.

Now, a process of performing preprocessing on each of a plurality of second trajectory data sets associated with each of a plurality of motor replacement states will be described with reference to Table 2 above. In the pre-processing process, the data (targetJoint1 to targetJoint6) about the number (Path) indicating the intermediate reaching point set in step S205 and the angle values each of the plurality of motors have when the robot 110 reaches the intermediate reaching point is second A process of adding to the trajectory data set may be included. The pre-processing may further include a process of adding a sequence number (seqNum) indicating an order in which the second sign data set was obtained to the corresponding second sign data set. Optionally, the pre-processing process determines the angle values (targetJoint1 to targetJoint6) of the plurality of motors when the robot 100 reaches the midpoint among the angle values (Joint1 to Joint6) of the plurality of motors included in the second trajectory data set. A process of adding data (diffJoint1 to diffJoint6) on difference values obtained by subtraction, respectively, to the corresponding second locus data set may be further included. The pre-processing may further include adding data (errorJoint) indicating which motor replacement state the corresponding second trajectory data set is obtained from among a plurality of motor replacement states to the corresponding second trajectory data set. In one embodiment, the errorJoint value included in the second trajectory data set may be a number indicating an error motor. For example, an errorJoint value included in the second trajectory data set obtained in a state where the third motor has an error and the third motor is replaced may be 3.

In step S225, the machine learning model 116 is trained at least partially using the training data configured in step S220. The machine learning model may be a decision tree learning model, a support vector machine (SVM) learning model, or a k-nearest neighbor (kNN) learning model.

3 is a flowchart illustrating an embodiment of a method of identifying a faulty motor in a robot including a plurality of motors by using a machine learning model learned according to embodiments according to the present disclosure.

An embodiment of the error motor identification method starts from setting a pattern in which the robot 100 operates (S305). The pattern set as described above may be a pattern that can well represent the errors of the axes (motors), and may be a reciprocating straight line pattern with a long distance, a triangle pattern, a square pattern, etc. that can cause motion of all axes. In step S310, a plurality of trajectory data sets are acquired while controlling the robot 100 to move according to a set pattern. As illustrated in Table 1, each of the obtained plurality of trajectory data sets may include data on angular values and velocities or angular velocity values of each of the plurality of motors at the time when the corresponding trajectory data set is obtained. In step S315, preprocessing is performed on each of a plurality of locus data sets to form input data. In the process of performing the preprocessing, as shown in Table 2, the numbers indicating the intermediate reaching points on the motion trajectories according to the set patterns and the angle values that the plurality of motors each have when the robot 100 reaches the intermediate reaching points It may include adding data to the trajectory data set. The process of performing the preprocessing may further include a process of adding a sequence number (seqNum) indicating the order in which the locus data set was acquired to the corresponding locus data set, as illustrated in Table 2. As illustrated in Table 2, the process of performing the preprocessing is performed by subtracting the angle values of the plurality of motors when the robot 100 reaches the intermediate point from the angle values of the plurality of motors included in the trajectory data set, respectively. A process of adding data about the obtained difference values to a corresponding trajectory data set may be further included. In step S320, the input data is input to the machine learning model 114, and the machine learning model 114 controls to provide output data. Here, the output data may include the number of a motor having an error among a plurality of motors. The user can identify which motor among the plurality of motors is an error motor by referring to the output number.

In a situation where industrial robots are constantly operating 24 hours a day, 365 days a year in almost all factories, if the cause of failure of a robot can be discovered in advance and measures can be taken, the loss of the factory can be minimized. If a machine learning model is trained according to the disclosed embodiments and an error motor is identified in a robot including a plurality of motors using the learned machine learning model, not only can the robot repair time be reduced, but the cost of repair is also reduced. It also has the advantage of being able to minimize.

The embodiments described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, the devices, methods and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate (FPGA). array), programmable logic units (PLUs), microprocessors, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. You can command the device. Software and/or data may be any tangible machine, component, physical device, virtual equipment, computer storage medium or device, intended to be interpreted by or provide instructions or data to a processing device. , or may be permanently or temporarily embodied in a transmitted signal wave. Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on computer readable media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. A computer readable medium may store program instructions, data files, data structures, etc. alone or in combination, and program instructions recorded on the medium may be specially designed and configured for the embodiment or may be known and usable to those skilled in the art of computer software. there is. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler.

The hardware device described above may be configured to operate as one or a plurality of software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with limited drawings, those skilled in the art can apply various technical modifications and variations based on this. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents of the claims are within the scope of the following claims.

110: robot
112: mechanical element
114: robot controller
116: machine learning model

Claims (21)

A method of learning a machine learning model to identify an error motor in a robot including a plurality of motors, comprising:
Setting an intermediate reaching point on a pattern in which the robot operates and an operating trajectory according to the pattern, and providing angle values each of the plurality of motors has when the robot reaches the intermediate reaching point;
Acquiring a plurality of first trajectory data sets while controlling the robot to move along the motion trajectory in a steady state;
Obtaining a plurality of second trajectory data sets associated with each of the plurality of motor replacement states while controlling the robot to move along the motion trajectory in the plurality of motor replacement states;
Constructing learning data by performing pre-processing on the plurality of first sign data sets and the plurality of second sign data sets associated with each of the plurality of motor replacement states, and
and training the machine learning model using at least in part the training data.
According to claim 1,
The normal state is a state in which a plurality of error-free motors are mounted on the robot, the machine learning model learning method.
According to claim 2,
Each of the plurality of first trajectory data sets includes data about each angular value and speed or angular velocity value of the plurality of motors at the time when the corresponding first trajectory data set is acquired, machine learning model learning method .
According to claim 3,
The step of constructing learning data by performing preprocessing on the plurality of first trajectory data sets and the plurality of second trajectory data sets associated with each of the plurality of motor replacement states, each of the plurality of first trajectory data sets Including the step of performing preprocessing on,
In the step of performing preprocessing on each of the plurality of first trajectory data sets, the data indicating the number indicating the intermediate arrival point and the angle values of the plurality of motors when the robot reaches the intermediate arrival point And adding to the corresponding first trajectory data set.
According to claim 4,
The step of performing preprocessing on each of the plurality of first sign data sets includes adding a sequence number (seqNum) indicating an order in which the corresponding first sign data set was obtained to the corresponding first sign data set. Further comprising, a method for learning a machine learning model.
According to claim 5,
In the step of performing preprocessing on each of the plurality of first trajectory data sets, when the robot reaches the intermediate reaching point at the angular values of the plurality of motors included in the corresponding first trajectory data set, the plurality of The machine learning model learning method further comprising adding data about difference values obtained by subtracting angle values of the motor to the corresponding first trajectory data set.
According to claim 6,
The step of performing preprocessing on each of the plurality of first sign data sets includes adding data indicating that the corresponding first sign data set is obtained in the steady state to the corresponding first sign data set. Further comprising, a method for learning a machine learning model.
According to claim 1,
Wherein each of the plurality of motor replacement states is a state in which one of the plurality of error-free motors in the robot is replaced with a faulty motor.
According to claim 8,
Each of the plurality of second trajectory data sets associated with each of the plurality of motor replacement states is an angular value of each of the plurality of motors including the replaced faulty motor at the time when the corresponding second trajectory data set is acquired and data regarding velocity or angular velocity values.
According to claim 9,
The step of constructing learning data by performing preprocessing on the plurality of first trajectory data sets and the plurality of second trajectory data sets associated with each of the plurality of motor replacement states, associated with each of the plurality of motor replacement states Performing preprocessing on each of the plurality of second trajectory data sets;
The step of performing pre-processing on each of the plurality of second trajectory data sets may include a number indicating the intermediate arrival point and data about angle values that the plurality of motors have when the robot reaches the intermediate arrival point And adding to the corresponding second trajectory data set.
According to claim 10,
The step of performing preprocessing on each of the plurality of second sign data sets further comprises adding a sequence number indicating an order in which the corresponding second sign data set was obtained to the corresponding second sign data set. , how to train a machine learning model.
According to claim 11,
In the step of performing preprocessing on each of the plurality of second trajectory data sets, when the robot reaches the intermediate arrival point at the angular values of the plurality of motors included in the corresponding second trajectory data set, the plurality of The machine learning model learning method further comprising adding data about difference values obtained by subtracting angle values of the motor to the corresponding second trajectory data set.
According to claim 12,
The step of performing preprocessing on each of the plurality of second trajectory data sets may include data indicating in which motor replacement state the corresponding second trajectory data set is obtained from among the plurality of motor replacement states, the corresponding first 2 A method for learning a machine learning model, further comprising the step of adding to the trajectory data set.
According to claim 1,
The machine learning model is a decision tree learning model, a support vector machine (SVM) learning model, or a k-Nearest Neighbor (kNN) learning model.
A method of identifying an error motor in a robot including a plurality of motors using a machine learning model learned by the machine learning model learning method of any one of claims 1 to 14,
Setting a pattern in which the robot operates;
Acquiring a plurality of trajectory data sets while controlling the robot to move according to the set pattern;
configuring input data by performing preprocessing on each of the plurality of locus data sets; and
and controlling the machine learning model to provide output data by inputting the input data to the machine learning model.
According to claim 15,
Wherein each of the plurality of trajectory data sets includes data on an angular value and a speed or angular velocity value of each of the plurality of motors at a time point when the corresponding trajectory data set is obtained.
According to claim 16,
The step of constructing input data by performing preprocessing on each of the plurality of trajectory data sets includes a number indicating an intermediate reaching point on an operation trajectory according to the set pattern and the plurality of when the robot reaches the intermediate reaching point. and adding data about angle values each motor has to the corresponding trajectory data set.
According to claim 17,
The step of configuring the input data by performing preprocessing on each of the plurality of trajectory data sets may further include adding a sequence number (seqNum) indicating an order in which the corresponding trajectory data set was obtained to the corresponding trajectory data set Including, a faulty motor identification method.
According to claim 18,
The step of configuring input data by performing pre-processing on each of the plurality of trajectory data sets, when the robot reaches the intermediate arrival point at the angle values of the plurality of motors included in the corresponding trajectory data set, the plurality of Further comprising the step of adding data on difference values obtained by subtracting each of the angle values of the motor of the corresponding trajectory data set, the faulty motor identification method.
According to claim 15,
Wherein the output data includes a number of a faulty motor among the plurality of motors.
A computer program stored in a computer readable recording medium for executing any one of the methods of claims 1 to 20.

Images