阅读说明：本技术 机器学习装置、校正参数调整装置及机器学习方法 (Machine learning device, correction parameter adjustment device, and machine learning method ) 是由 藤田智哉 木全敏章 井藤达也 井内幸弘 佐藤刚 于 2018-06-14 设计创作，主要内容包括：机器学习装置(100)对在具有多种校正功能的电动机驱动装置中在用于控制电动机的指令值的校正时所使用的校正参数进行学习，该机器学习装置(100)的特征在于，具备：状态观测部(101)，其将校正参数、基于电动机驱动装置的驱动数据及校正功能的种类计算出的特征量作为状态变量而进行观测；以及学习部(102)，其按照基于状态变量创建的训练数据组，针对各个校正功能对校正参数进行学习。(A machine learning device (100) learns a correction parameter used for correcting a command value for controlling a motor in a motor drive device having a plurality of correction functions, the machine learning device (100) comprising: a state observation unit (101) that observes, as state variables, the correction parameters and feature quantities calculated based on the drive data of the motor drive device and the type of correction function; and a learning unit (102) that learns the correction parameters for each correction function according to a training data set created based on the state variables.)

1. A machine learning device which learns a correction parameter used in correction of a command value for controlling a motor in a motor drive device having a plurality of correction functions,

the machine learning device is characterized by comprising:

a state observation unit that observes, as state variables, the correction parameters and feature quantities calculated based on drive data of the motor drive device and the type of the correction function; and

and a learning unit that learns the correction parameters for each of the correction functions in accordance with a training data set created based on the state variables.

2. The machine learning apparatus of claim 1,

the drive data includes at least 1 of feedback data from a motor that drives the motor drive device and drive data of a machine device that is a control target of the motor.

3. The machine learning apparatus of claim 1 or 2,

the feature amount is a multidimensional vector, and is data representing a characteristic of the drive data for each of the correction functions.

4. The machine learning apparatus of any one of claims 1 to 3,

the learning unit includes:

a reward calculation unit that calculates a reward based on the state variable; and

and a function updating unit that updates a function for determining the correction parameter based on the report.

5. The machine learning apparatus of claim 4,

the function updating unit updates the function for each of the correction functions.

6. A calibration parameter adjustment device is characterized by comprising:

the machine learning device of any one of claims 1 to 5;

a correction function selection unit that selects 1 of the plurality of correction functions;

a drive data acquisition unit that acquires the drive data;

a feature value extracting unit that extracts the feature value based on the selected correction function and the acquired drive data;

an adjustment condition input unit that inputs an adjustment condition of the correction parameter; and

and a correction parameter determination unit that determines the correction parameter that satisfies the adjustment condition based on a learning result of the learning unit.

7. The correction parameter adjustment device according to claim 6,

the drive data includes at least 1 of feedback data from a motor that drives the motor drive device and drive data of a machine device that is a control target of the motor.

8. The correction parameter adjustment device according to claim 7,

the drive data of the mechanical device is detection data obtained from a sensor provided in the mechanical device.

9. The correction parameter adjustment device according to any one of claims 6 to 8,

the correction parameter determining section includes:

a correction parameter calculation unit that calculates a correction parameter based on the learning result and transmits the calculated correction parameter to the motor drive device;

a correction parameter evaluation unit that evaluates whether or not a drive result obtained based on the calculated correction parameter satisfies the adjustment condition; and

and an end determination unit that determines whether or not to end adjustment of the correction parameter based on the evaluation result of the correction parameter evaluation unit.

10. The correction parameter adjustment device according to claim 9,

the correction parameter evaluation unit evaluates the correction parameter using at least 1 of drive data obtained when the motor drive device is actually driven and drive data obtained when the operation of the motor drive device is simulated.

11. A machine learning method for learning a correction parameter used in correction of a command value for controlling a motor in a motor drive device having a plurality of correction functions,

the machine learning method is characterized by comprising the following steps:

observing the correction parameter, a feature amount calculated based on drive data of the motor drive device and a type of the correction function as a state variable; and

the correction parameters are learned for each of the correction functions in accordance with a training data set created based on the state variables.

Technical Field

The present invention relates to a machine learning device, a correction parameter adjustment device, and a machine learning method for learning correction parameters used for a correction function of a motor drive device having a plurality of correction functions.

Background

The motor driving device driven by a motor is, for example, a machine tool, an industrial machine, a robot, or a conveyor. The motor drive device controls a control target such as a tool, a processing target, or a robot so as to follow a given command value. However, the mechanical structure and mechanical elements constituting the motor drive device have various inherent causes of errors and disturbances, and therefore the accuracy of following the command value may be lowered. The error causes include friction of the sliding surface, insufficient rigidity of the structural member, assembly error, delay in control, heat generation of the motor, and the like. Therefore, it is conceivable to correct the command value in consideration of the error and use the corrected command value. The optimum value of the correction condition such as the correction amount of the command value and the correction timing (timing) differs depending on the difference in the structure of the motor drive device, the individual difference, the difference in the installation environment, and the like.

Since the operator needs to perform adjustment of the correction conditions for each motor drive device, it takes time and effort to perform the adjustment, and it takes time until the correction conditions become appropriate. Therefore, a device for automatically adjusting the correction conditions has been studied. Patent document 1 discloses a technique for automatically determining a correction condition using a correction parameter for determining a correction condition for a motion error caused by the influence of friction. In this technique, a correction parameter for determining a correction condition for a motion error caused by the influence of friction is automatically determined by changing a correction parameter for determining a correction condition for a torque command and repeating correction of a torque command and update of a correction torque until a response error caused by circular motion is equal to or smaller than a threshold value.

Patent document 1: japanese patent laid-open publication No. 11-24754

Disclosure of Invention

However, in the technique described in patent document 1, since there is only an algorithm for adjusting only the correction parameters of a single correction function, only the correction parameters of a single correction function can be adjusted. The creation of the algorithm for adjusting the correction parameters of the correction function is a very time-consuming and labor-intensive task, and when the motor drive device has a plurality of correction functions, it is necessary to create the algorithm for adjusting the correction parameters individually in accordance with each correction function, which causes a problem that the time and labor required for creating the algorithm are enormous.

The present invention has been made in view of the above problems, and an object of the present invention is to provide a machine learning device and a correction parameter adjusting device that can easily adjust a correction parameter for determining a correction condition of a command value of a motor driving device even in a motor driving device having a plurality of correction functions.

In order to solve the above problems and achieve the object, the present invention provides a machine learning device for learning a correction parameter used for correcting a command value for controlling a motor in a motor drive device having a plurality of correction functions, the machine learning device including: a state observation unit that observes, as state variables, the correction parameters and the feature quantities calculated based on the drive data of the motor drive device and the type of the correction function; and a learning unit that learns the correction parameters for each correction function in accordance with a training data set created based on the state variables.

ADVANTAGEOUS EFFECTS OF INVENTION

The machine learning device according to the present invention has an effect that even in a motor drive device having a plurality of correction functions, it is possible to easily adjust a correction parameter for determining a correction condition of a command value of the motor drive device.

Drawings

Fig. 1 is a diagram showing a functional configuration of a correction parameter adjustment device according to embodiment 1 of the present invention.

Fig. 2 is a diagram showing a configuration of the feature extraction unit shown in fig. 1.

Fig. 3 is a diagram showing a configuration of the machine learning device shown in fig. 1.

Fig. 4 is a diagram showing a configuration of the correction parameter determining unit shown in fig. 1.

Fig. 5 is a diagram showing a structure of the motor drive device shown in fig. 1.

Fig. 6 is a diagram showing the structure of the mechanical apparatus shown in fig. 5.

Fig. 7 is a diagram showing a configuration of the servo control unit shown in fig. 5.

Fig. 8 is a diagram showing an example of the hardware configuration of the correction parameter adjustment apparatus shown in fig. 1.

Fig. 9 is a flowchart showing an operation of the correction parameter adjusting apparatus shown in fig. 1.

Fig. 10 is a diagram showing a configuration of the learning unit shown in fig. 3 when the machine learning device uses reinforcement learning.

Fig. 11 is a diagram showing an operation of a learning unit in a case where the machine learning device shown in fig. 2 uses reinforcement learning.

Fig. 12 is a diagram showing a configuration of a correction parameter adjustment device according to embodiment 2 of the present invention.

Fig. 13 is a diagram showing a configuration of a correction parameter adjustment device according to embodiment 3 of the present invention.

Fig. 14 is a diagram showing a configuration of the correction parameter determining unit shown in fig. 13.

Detailed Description

Hereinafter, a machine learning device, a correction parameter adjustment device, and a machine learning method according to embodiments of the present invention will be described in detail with reference to the drawings. The present invention is not limited to this embodiment.