WO2022003833A1

WO2022003833A1 - Positioning control device and machine learning device

Info

Publication number: WO2022003833A1
Application number: PCT/JP2020/025698
Authority: WO
Inventors: 翔堀内; 宏武勅使; 直弥田島
Original assignee: 三菱電機株式会社
Priority date: 2020-06-30
Filing date: 2020-06-30
Publication date: 2022-01-06
Also published as: JPWO2022003833A1

Abstract

The present disclosure relates to a positioning control device that is electrically coupled to a motor via an amplifier and that controls the motor, the positioning control device comprising a parameter storage unit, a vibration data acquisition unit, a machine learning unit, and an outputting unit. The parameter storage unit stores a parameter including information on the motor and a tolerance range for motor vibration and takt time, the parameter being information necessary for determining a motor speed control parameter which is a parameter for controlling the motor. The vibration data acquisition unit acquires vibration data which is vibration, as detected by a vibration sensor, at a location where the motor is installed. Using a learned model obtained by learning, from said parameter and said vibration data, a correlation between vibrations at locations where the motor is to be installed and said motor speed control parameter, the machine learning unit determines, from said parameter and said vibration data, said motor speed control parameter that allows the vibration to fit within said tolerance range. On the basis of the determined motor speed control parameter, the outputting unit outputs, to the amplifier, pulses for controlling the amplifier.

Description

Positioning control device and machine learning device

The present disclosure relates to a positioning control device and a machine learning device that control the positioning of a motor by a motor speed control parameter that controls the speed of the motor.

In a system such as a production line using a motor, the motor is electrically connected to a control device such as a programmable logic controller (PLC) via an amplifier. In such a system, the user inputs the motor speed control parameter, which is a parameter necessary for controlling the speed of the motor. Examples of motor speed control parameters are position, operating pulse speed and acceleration rate. The PLC outputs a pulse signal, which is a command signal, to the amplifier based on the motor speed control parameter input by the user, and the amplifier controls the motor based on the pulse signal. That is, the current consumption of the system, the load on the motor and the takt time are determined by the motor speed control parameters input by the user.

Normally, the motor speed control parameters are determined based on the user's experience, etc., according to the motor information including the motor type, equipment, and product. If the system is not performing the intended operation, the user redetermines and inputs the motor speed control parameters. This process is repeated until the system performs the intended operation. In addition, when determining the motor speed control parameters, for example, if the acceleration parameter is set large to shorten the tact time, and as a result, the vibration of the equipment becomes large or the power consumption becomes large, other than the tact time. On the other hand, it may lead to negative results. Therefore, the user needs to consider the tact time, the vibration of the equipment, the power consumption, etc. so as to be within the permissible range. However, there is a trade-off relationship between takt time, vibration, and power consumption, and it is difficult to derive a balanced speed control parameter with short tact time, low vibration, and low power consumption based on experience. Is.

Patent Document 1 describes a machine that processes a workpiece, a higher-level device that is located above one or more control devices that control the machine and adjusts the servo gain used in machining by the machine, and an adjustment of the servo gain of the machine. A control system including a machine learning device for machine learning and a machine learning device is disclosed. In the control system described in Patent Document 1, the machine learning device determines the adjustment behavior of the servo gain of the machine based on the machine learning result and the state data of the adjustment of the servo gain of the machine, and changes the servo gain of the machine. do.

Japanese Unexamined Patent Publication No. 2018-097680

However, in the control system described in Patent Document 1, machine learning is performed by a higher-level device located higher than the control device that controls the machine, and the servo gain of the machine is adjusted. Therefore, there is a problem that the followability to the position command is only optimized and the tact time cannot be shortened. Further, the control system described in Patent Document 1 does not take into consideration the vibration caused by the motor, which greatly affects the life of the system, and is insufficient in extending the life of the system.

The present disclosure has been made in view of the above, and an object thereof is to obtain a positioning control device capable of shortening the tact time as compared with the conventional case and extending the life of a control system including a motor. And.

In order to solve the above-mentioned problems and achieve the object, the present disclosure is a positioning control device that is electrically connected to a motor via an amplifier and controls the motor, and is a parameter storage unit and a vibration data acquisition unit. , A machine learning unit, and an output unit. The parameter storage unit is information necessary for determining the motor speed control parameter, which is a parameter for controlling the motor, and stores the motor information and the parameters including the allowable range of the vibration and the tact time of the motor. The vibration data acquisition unit acquires vibration data, which is the vibration of the motor installation location detected by the vibration sensor. The machine learning unit uses a trained model that learns the correlation between the vibration of the motor installation location and the motor speed control parameter from the parameters and vibration data, and the vibration is within the allowable range from the parameters and vibration data. Determine motor speed control parameters. The output unit outputs to the amplifier as a pulse for controlling the amplifier based on the determined motor speed control parameter.

The positioning control device according to the present disclosure has the effect of shortening the tact time as compared with the conventional case and extending the life of the control system including the motor.

A block diagram showing an example of a configuration of a control system including a positioning control device according to the first embodiment. A block diagram schematically showing an example of the functional configuration of the machine learning unit included in the positioning control device according to the first embodiment. A flowchart showing an example of the learning processing procedure of the machine learning unit included in the positioning control device according to the first embodiment. A flowchart showing an example of a processing procedure of a motor control method in the positioning control device according to the first embodiment. Block diagram showing another example of the configuration of the control system including the positioning control device according to the first embodiment. Block diagram showing another example of the configuration of the control system including the positioning control device according to the first embodiment. A block diagram showing an example of a configuration of a control system including a positioning control device according to the second embodiment. A flowchart showing an example of the learning processing procedure of the machine learning unit included in the positioning control device according to the second embodiment. A flowchart showing an example of a processing procedure of a motor control method in the positioning control device according to the second embodiment. Block diagram showing another example of the configuration of the control system including the positioning control device according to the second embodiment. Block diagram showing another example of the configuration of the control system including the positioning control device according to the second embodiment. A block diagram showing an example of a configuration of a control system including a positioning control device according to the third embodiment. A flowchart showing an example of the learning processing procedure of the machine learning unit included in the positioning control device according to the third embodiment. A flowchart showing an example of a processing procedure of a motor control method in the positioning control device according to the third embodiment. Block diagram showing another example of the configuration of the control system including the positioning control device according to the third embodiment. Block diagram showing another example of the configuration of the control system including the positioning control device according to the third embodiment. Block diagram showing another example of the configuration of the control system including the positioning control device according to the third embodiment. The figure which shows an example of the speed command in the positioning control apparatus according to Embodiment 3. The figure which shows an example of the learning result of the motor speed control parameter in the positioning control apparatus according to Embodiment 3. The figure which shows typically an example of the hardware composition which realizes the positioning control device by

Embodiments

1, 2, and 3.

Hereinafter, the positioning control device and the machine learning device according to the embodiment of the present disclosure will be described in detail with reference to the drawings.

Embodiment 1.
FIG. 1 is a block diagram showing an example of a configuration of a control system including a positioning control device according to the first embodiment. The control system 1 includes a positioning control device 10, an amplifier 30, a motor 50, and a vibration sensor 61.

The positioning control device 10 is within a predetermined range of the motor 50 to be connected from the information of the motor 50 and the vibration data which is the data acquired from the vibration sensor 61 of the device to which the positioning control device 10 is connected. It is a device that obtains the motor speed control parameters so that the vibration becomes optimum. The positioning control device 10 includes a parameter storage unit 11, a sensor value acquisition unit 12, a machine learning unit 13, a motor speed control parameter output unit 14, and a pulse output unit 15. An example of the positioning control device 10 is a PLC or a positioning motion unit.

The parameter storage unit 11 stores parameters that are information about the motor 50 to be controlled. The parameters are information necessary for determining the motor speed control parameters, and include information of the motor 50 input by the user and operating conditions of the motor 50. The information of the motor 50 is information including the type of the motor 50, the power capacity, the rated speed, and the size of the motor 50. The operating condition is a condition to be satisfied by the control system 1 when the motor 50 is controlled by the set motor speed control parameter. Operating conditions include tolerances and priorities. The permissible range is a condition that defines the tact time during operation of the control system 1, the current consumption of the motor 50, and the maximum value of the vibration of the motor 50. The priority item is a condition indicating an item that the user wants to optimize among the tact time, the current consumption of the motor 50, and the vibration of the motor 50. In the first embodiment, it is assumed that the allowable range for the vibration of the motor 50 and the tact time having a trade-off relationship with the vibration of the motor 50 is set, and the vibration of the motor 50 is set as a priority item.

The sensor value acquisition unit 12 holds vibration data, which is data obtained from the vibration sensor 61, and outputs the vibration data to the machine learning unit 13.

The machine learning unit 13 learns the motor speed control parameter that is the optimum vibration of the motor 50 from the value of the parameter storage unit 11 and the vibration data obtained from the sensor value acquisition unit 12. The machine learning unit 13 corresponds to the machine learning device. The vibration and tact time of the motor 50 may be within the allowable range, but if there are a plurality of motor speed control parameters in which the vibration and the tact time are within the allowable range, one motor is according to a predetermined standard. The speed control parameter is selected. Alternatively, as described in the second embodiment, the optimum vibration is such that the current consumption of the motor 50 is also within the allowable range. At this time, the machine learning unit 13 gives priority to the item specified as the priority item and determines the motor speed control parameter so as not to exceed the allowable range. However, for the items specified as priority items, the motor speed control parameters are set so as not to exceed the allowable range, but for items other than the priority items, the allowable range may be exceeded or the allowable range is exceeded. It may not be. Motor speed control parameters include position, start-up pulse count, run pulse speed, run pulse count and acceleration / deceleration rate.

The motor speed control parameter output unit 14 outputs the motor speed control parameter obtained from the machine learning unit 13 to the pulse output unit 15.

The pulse output unit 15 outputs a pulse, which is a command signal for controlling the amplifier 30, to the amplifier 30 based on the motor speed control parameter obtained from the motor speed control parameter output unit 14. The motor speed control parameter output unit 14 and the pulse output unit 15 correspond to the output unit.

The amplifier 30 is a device that is electrically connected to the positioning control device 10 and the motor 50 and controls the motor 50 by a pulse output from the positioning control device 10.

The motor 50 is an electric power device that can be controlled by the amplifier 30. An example of the motor 50 is a servomotor having an encoder for position detection and a stepping motor without an encoder.

The vibration sensor 61 measures the vibration at the location where the motor 50 is installed, and outputs the measured vibration data to the positioning control device 10.

Here, the details of the machine learning unit 13 provided in the positioning control device 10 will be described. FIG. 2 is a block diagram schematically showing an example of the functional configuration of the machine learning unit included in the positioning control device according to the first embodiment. The machine learning unit 13 includes a data acquisition unit 131, a model generation unit 132, a trained model storage unit 133, and an inference unit 134.

The data acquisition unit 131 acquires motor speed control parameters, motor 50 information, allowable ranges, priority items, and vibration data as learning data. The motor speed control parameter is a value set in the motor speed control parameter output unit 14. The information and the allowable range of the motor 50 are values stored in the parameter storage unit 11. The permissible range includes the tact time, the current consumption of the motor 50 and the vibration of the motor 50, but in the first embodiment, the permissible range regarding the vibration of the motor 50 and the tact time is used. The vibration data is vibration data detected by the vibration sensor 61 and acquired by the sensor value acquisition unit 12 when the motor 50 is driven by the set motor speed control parameters.

The model generation unit 132 has a motor speed that provides optimum vibration according to learning data created based on a combination of motor speed control parameters, motor 50 information, allowable range, and vibration data output from the data acquisition unit 131. Learn control parameters. That is, a trained model that infers the motor speed control parameter that provides the optimum vibration from the information of the motor 50, the vibration of the motor 50, and the allowable range of the tact time, that is, the vibration falls within the allowable range of the tact time. Generate. The trained model is a model that learns the correlation between the vibration of the installation location of the motor 50 and the motor speed control parameter from the information of the motor 50, the allowable range of the vibration and tact time of the motor 50, and the vibration data. be. Here, the learning data is data in which the motor speed control parameter, the information of the motor 50, the allowable range of the vibration and the tact time of the motor 50, and the vibration data are associated with each other.

As the learning algorithm used by the model generation unit 132, known algorithms such as supervised learning, unsupervised learning, and reinforcement learning can be used. As an example, the case where reinforcement learning (Reinforcement Learning) is applied will be described. In reinforcement learning, an agent who is the action subject in a certain environment observes the parameters of the environment in the current state and decides the action to be taken. The environment changes dynamically depending on the behavior of the agent, and the agent is rewarded according to the change in the environment. The agent repeats this process and learns the action policy that gives the most reward through a series of actions. Q-learning and TD-Learning are known as typical methods of reinforcement learning. For example, in the case of Q-learning, the general update equation of the action value function Q (s, a) is expressed by the following equation (1).

(1) In the formula, s _t represents the state of the environment at time t, a _t represents the behavior in time t. By the action a _t, the state is changed to s _{t + 1.} r _{t + 1} represents the reward received by the change of the state, γ represents the discount rate, and α represents the learning coefficient. It is assumed that γ is in the range of 0 <γ ≦ 1 and α is in the range of 0 <α ≦ 1. Motor speed control parameter action a _t becomes, the allowable range of vibration information and the motor 50 of the motor 50 to learn the best action a _t in state s _t of the state s _t, and the time t. In one example, the state of the motor is input as the state, the motor speed control parameter is input as the action, and the vibration of the motor 50 is input as a result. In addition, the permissible range is used as the standard of compensation.

In the update formula represented by the equation (1), if the action value Q of the action a having the highest Q value at time t + 1 is larger than the action value Q of the action a executed at time t, the action value Q is increased. However, in the opposite case, the action value Q is reduced. In other words, the action value function Q (s, a) is updated so that the action value Q of the action a at time t approaches the best action value at time t + 1. As a result, the best behavioral value in a certain environment is sequentially propagated to the behavioral value in the previous environment.

As described above, when a trained model is generated by reinforcement learning, the model generation unit 132 includes a reward calculation unit 141 and a function update unit 142.

The reward calculation unit 141 calculates the reward based on the motor speed control parameter, the information of the motor 50, the allowable range of the vibration and tact time of the motor 50, and the vibration data. The reward calculation unit 141 calculates the reward r based on the magnitude of the vibration obtained from the vibration data and the allowable range of the vibration of the motor 50. The threshold value defined by the allowable range of the magnitude of vibration is defined as the first threshold value. For example, if the magnitude of vibration <first threshold value, the reward r is increased (for example, a reward of "1" is given), while if the magnitude of vibration> the first threshold value, the reward r is decreased. (For example, give a reward of "-1".).

The function update unit 142 updates the function for determining the motor speed control parameter that produces the optimum vibration according to the reward calculated by the reward calculation unit 141, and outputs the function to the trained model storage unit 133. For example, in the case of Q-learning, it is used as a function for calculating the motor speed control parameter to be optimized vibrate (1) Action value function formula Q (s _t, a _t). Repeat the above learning.

Learned model storage unit 133, action value is updated by the function updating unit 142 function _{_{Q (s t, a t)}} , i.e., storing the learned model.

The inference unit 134 infers the motor speed control parameter using the learned model stored in the learned model storage unit 133. That is, by inputting the information of the motor 50 acquired by the data acquisition unit 131, the allowable range of the vibration and the tact time of the motor 50, and the priority items into this trained model, the information of the motor 50, the vibration and the tact of the motor 50 are input. Motor speed control parameters inferred from the time tolerance and priority items can be output. Further, the vibration data acquired by analyzing the value of the vibration data detected by the vibration sensor 61, the information of the motor 50, the allowable range of the vibration and the tact time of the motor 50, and the priority items with the trained model. The value of can be reflected in the motor speed control parameter as a feedback value.

In the above description, the inference unit 134 has been described as outputting the motor speed control parameter using the learned model learned by the model generation unit 132 of the positioning control device 10 connected to the motor 50. However, the inference unit 134 acquires a trained model from the outside such as another positioning control device 10 connected to the other motor 50, and outputs the motor speed control parameter based on the acquired trained model. May be good.

Next, the learning process of the machine learning unit 13 will be described. FIG. 3 is a flowchart showing an example of the procedure of the learning process of the machine learning unit included in the positioning control device according to the first embodiment. First, the data acquisition unit 131 acquires motor speed control parameters, motor 50 information, allowable range, and vibration data as learning data (step S11). Here, an allowable range for vibration and tact time of the motor 50 is set. Further, the data acquisition unit 131 acquires priority items from the parameter storage unit 11 (step S12). Here, it is assumed that vibration is set as the priority item.

Then, the model generation unit 132 calculates the reward based on the motor speed control parameter, the information of the motor 50, and the allowable range. Specifically, the reward calculation unit 141 of the model generation unit 132 acquires the motor speed control parameter, the information of the motor 50, and the allowable range, and rewards based on the relationship between the predetermined vibration magnitude and the first threshold value. Judge the increase or decrease of. The first threshold is, in one example, the value of vibration defined in the permissible range. Here, the reward calculation unit 141 determines whether the value of the vibration data when the motor 50 is operated by the motor speed control parameter is less than the first threshold value (step S13). The first threshold value is the value of vibration allowed for the device connected to the motor 50 when the motor 50 is operated.

When the value of the vibration data is less than the first threshold value (Yes in step S13), the reward calculation unit 141 increases the reward (step S14). When the value of the vibration data is larger than the first threshold value (No in step S13), the reward calculation unit 141 reduces the reward (step S15). When the value of the vibration data is equal to the first threshold value, the reward may be increased or decreased.

After step S14 or step S15, the function updater 142 of the model generating unit 132, based on the calculated compensation by compensation calculation unit 141, action value function Q (s _{_t,} a _t) to update (step S16) .. Action value function Q (s _t, a _t) is a function expressed by learned model storage unit 133 stores (1).

Then, the process returns to step S11. That is, the machine learning unit 13 repeatedly executes the processing up to S16 step S11 above, and stores the generated action-value function Q (s _{_t,} a _t) as a learned model. When a feedback value related to current consumption or takt time other than vibration is acquired, the learning data at the time of acquisition of the data is accumulated.

In the machine learning unit 13 according to the first embodiment, the trained model is stored in the trained model storage unit 133 provided inside the machine learning unit 13, but the trained model storage unit 133 is stored in the machine learning unit 13. It may be prepared outside of.

In addition to "reinforcement learning", "supervised learning", "unsupervised learning", "semi-supervised learning" or other known learning algorithms can be used to control motor speed control parameters, motor 50 information, and motor 50. The motor speed control parameter that provides the optimum vibration may be machine-learned from the permissible range of vibration and tact time and the vibration data. Machine learning using these learning algorithms can also reduce the vibration of the entire system.

Next, a method of inferring motor speed control parameters using the trained model stored in the machine learning unit 13 will be described. FIG. 4 is a flowchart showing an example of a processing procedure of a motor control method in the positioning control device according to the first embodiment.

First, the configuration of the initial control system 1 is determined by the user. Then, the parameter storage unit 11 stores the information including the information of the motor 50, the allowable range, and the value including the priority item (step S31). Information, tolerances and priorities for the motor 50 are set by the user via inputs not shown. Here, an allowable range for vibration and tact time of the motor 50 is set. Further, the sensor value acquisition unit 12 acquires vibration data from the vibration sensor 61 (step S32) and holds it. After that, the inference unit 134 of the machine learning unit 13 analyzes the value of the parameter storage unit 11 and the value of the acquired vibration data using the trained model stored in the trained model storage unit 133, and optimal vibration. The motor speed control parameter is set in the motor speed control parameter output unit 14 (step S33).

The motor speed control parameter output unit 14 outputs the set motor speed control parameter to the pulse output unit 15 (step S34). The pulse output unit 15 outputs a pulse to the amplifier 30 based on the motor speed control parameter from the motor speed control parameter output unit 14 (step S35).

This drives the motor 50. When the motor 50 is driven, the vibration of the motor 50 is detected by the vibration sensor 61 provided in the motor 50. After that, as described in step S32, the sensor value acquisition unit 12 acquires vibration data, which is the vibration of the motor 50. Then, as described above, the processes from steps S32 to S35 are repeatedly executed. In one example, when the value of the vibration data acquired by the sensor value acquisition unit 12 changes beyond the permissible range, the inference unit 134 uses the information of the motor 50, the permissible range, and the priority items in step S33. The value of the vibration data and the value of the vibration data are analyzed using the trained model, and the motor speed control parameter that provides the optimum vibration is set in the motor speed control parameter output unit 14.

Although the above description shows the case where the vibration sensor 61 is provided in the motor 50, the embodiment is not limited to this. FIG. 5 is a block diagram showing another example of the configuration of the control system including the positioning control device according to the first embodiment. In the following, the same components as those in FIG. 1 are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those in FIG. 1 will be described. In FIG. 5, the vibration sensor 61 is not provided in the motor 50, but in the product 52, which is a device including the motor 50 and the drive unit 51 driven by the motor 50. As a result, the machine learning unit 13 can optimize the vibration value of the product 52 including the drive unit 51 and the like as well as the motor 50.

FIG. 6 is a block diagram showing another example of the configuration of the control system including the positioning control device according to the first embodiment. In the following, the same components as those in FIGS. 1 and 5 are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those in FIGS. 1 and 5 will be described. In FIG. 6, the vibration sensor 61 is provided in a system 53 including a plurality of products 52. An amplifier 30 is electrically connected to the motor 50 of each product 52 in the system 53. As an example, such a system 53 is a multi-axis control system including a plurality of motors 50 and products 52. Thereby, the machine learning unit 13 can optimize the vibration value of the system 53 including the plurality of products 52. The motor 50 in FIG. 1, the product 52 in FIG. 5, and the system 53 in FIG. 6 where the vibration sensor 61 is installed are the locations where the motor 50 is installed.

In the first embodiment, the machine learning unit 13 of the positioning control device 10 learns the motor speed control parameter in which the vibration data is within the allowable range according to the learning data, and generates a trained model. The learning data is created based on a combination of vibration data from the vibration sensor 61, information on the motor 50, motor speed control parameters, and allowable ranges for vibration and takt time of the motor 50. The vibration sensor 61 is provided in a motor 50, a product 52 including a drive unit 51 connected to the motor 50, or a system 53 including a plurality of products 52. Then, the machine learning unit 13 sets the motor speed control parameter in which the vibration falls within the permissible range by analyzing the information of the motor 50, the permissible range, and the value of the vibration data using the trained model. By driving the motor 50 with the set motor speed control parameters, the tact time can be shortened, and the product 52 or the system 53 can be operated without giving excessive vibration to the motor 50, which imposes a burden on the motor 50. Can be reduced and the life of the motor 50 can be extended.

Further, when the positioning control device 10 is provided with a higher-level device, the positioning control device 10 is provided with a function related to machine learning. Therefore, the load on the host device can be reduced.

Further, unlike the conventional technique, the servo gain which is a control value is determined by the feedback control from the motor 50, but the motor speed control parameter is determined by the output control of the positioning control device 10. Therefore, it can be applied not only to devices such as servo motors and amplifiers having an encoder, which are expensive and capable of feedback control, but also to devices such as stepping motors and amplifiers which do not have a feedback mechanism.

Embodiment 2.
FIG. 7 is a block diagram showing an example of the configuration of the control system including the positioning control device according to the second embodiment. In the following, the same components as those in the first embodiment are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those in the first embodiment will be described. The control system 1 of the second embodiment further includes a current consumption measuring device 62 in the motor 50. The current consumption measuring device 62 is a device that acquires a current consumption value, which is a value of the current consumed at the installation location. The current consumption measuring device 62 outputs the measured current consumption value to the positioning control device 10.

The positioning control device 10 further includes a current consumption acquisition unit 16. The current consumption acquisition unit 16 acquires and holds the current consumption value output from the current consumption measuring device 62.

The machine learning unit 13 has an allowable range of priority items specified according to the learning data created based on the combination of the motor speed control parameter, the information of the motor 50, the allowable range, and the vibration data and the current consumption value of the motor 50. Learn motor speed control parameters that fit in. Here, it is assumed that the allowable range of takt time is set in advance in addition to the allowable range specified by the priority item. That is, the motor speed control parameter in which the tact time is within the allowable range from the information of the motor 50, the allowable range, the priority item, the vibration data of the motor 50, and the current consumption value, and the specified priority item is within the allowable range. Generate a trained model that infers. The priority item specified is vibration or current consumption. That is, the trained model correlates the vibration and current consumption value of the installation location of the motor 50 with the motor speed control parameter from the information of the motor 50, the allowable range, the vibration data, and the current consumption value. It is a learned model. The machine learning unit 13 may learn motor speed control parameters in which the designated priority items fall within the permissible range, and not only the designated priority items but also items other than the priority items fall within the permissible range. You may learn the speed control parameters. Further, the machine learning unit 13 analyzes the information of the motor 50, the allowable range, the value of the vibration data of the motor 50, and the current consumption value by using the trained model, and the priority item is the motor 50 having the information of the motor 50. Outputs motor speed control parameters that fall within the permissible range.

FIG. 8 is a flowchart showing an example of the procedure of the learning process of the machine learning unit included in the positioning control device according to the second embodiment. First, the data acquisition unit 131 acquires the motor speed control parameter, the information of the motor 50, the allowable range, the vibration data of the motor 50, and the current consumption value as learning data (step S51). Further, the data acquisition unit 131 acquires priority items from the parameter storage unit 11 (step S52). The model generation unit 132 determines whether the priority item is vibration or current consumption (step S53).

When the priority item is vibration (vibration in step S53), the model generation unit 132 calculates the reward based on the motor speed control parameter, the information of the motor 50, and the allowable range. Specifically, the reward calculation unit 141 of the model generation unit 132 acquires the motor speed control parameter, the information of the motor 50, the allowable range, the vibration data of the motor 50, and the current consumption value, and determines the magnitude of the vibration. The increase or decrease of the reward is determined based on the relationship with the first threshold value and the relationship between the current consumption value and the second threshold value. The second threshold value is, in one example, the current consumption value defined in the allowable range. Here, the reward calculation unit 141 determines whether the value of the vibration data when the motor 50 is operated by the motor speed control parameter is less than the first threshold value (step S54).

When the value of the vibration data is less than the first threshold value (Yes in step S54), the reward calculation unit 141 increases the reward (step S55). If items other than the priority items are not considered, the reward will be increased if the vibration data, which is the priority item, is less than the first threshold value. However, when considering items other than the priority items so as to be within the permissible range, the reward according to the difference between the first threshold value and the vibration data value and the difference between the second threshold value and the current consumption value are further considered. It is possible to determine the reward according to the above. In one example, the reward calculation unit 141 increases the reward when the difference between the second threshold value and the current consumption value is positive, as compared with the case where the difference between the second threshold value and the current consumption value is negative. At this time, the reward may be determined depending on whether the difference between the second threshold value and the current consumption value is positive or negative, and the reward is determined according to the magnitude of the difference between the second threshold value and the current consumption value. May be good. That is, even within the permissible range, the reward may be increased as the current consumption value becomes smaller, and the reward may decrease as the current consumption value becomes larger. Further, when the current consumption value becomes larger than the second threshold value outside the permissible range, the reward may be reduced. However, in these cases, vibration is a priority item, so the difference between the first threshold and the vibration data value is larger than the difference between the second threshold and the current consumption value. It is desirable to contribute to the increase or decrease of the reward. As a result, when both the vibration and the current consumption are within the permissible range, the reward is higher and the action value can be enhanced as compared with the case where the vibration is within the permissible range but the current consumption is not within the permissible range.

When the value of the vibration data is larger than the first threshold value (No in step S54), the reward calculation unit 141 reduces the reward (step S56). If items other than the priority items are not considered, the reward will be reduced if the vibration data, which is the priority item, is larger than the first threshold value. However, when considering items other than the priority items so as to be within the permissible range, the reward according to the difference between the first threshold value and the vibration data value and the difference between the second threshold value and the current consumption value are further considered. It is possible to determine the reward according to the above. In one example, the reward calculation unit 141 makes the absolute value of the reward to be reduced smaller when the difference between the second threshold value and the current consumption value is positive than when the difference between the second threshold value and the current consumption value is negative. .. At this time, the reward may be determined depending on whether the difference between the second threshold value and the current consumption value is positive or negative, and the reward is determined according to the magnitude of the difference between the second threshold value and the current consumption value. May be good. That is, even within the permissible range, the absolute value of the reward to be reduced may be reduced as the current consumption value becomes smaller, and the absolute value of the reward to be reduced may be increased as the current consumption value becomes larger. Further, when the current consumption value is larger than the second threshold value outside the permissible range, the absolute value of the reward to be reduced may be increased. However, even in these cases, vibration is a priority item, so the difference between the first threshold and the vibration data value is larger than the difference between the second threshold and the current consumption value. It is desirable to contribute to the increase or decrease of the reward.

When the priority item is current consumption (current consumption in step S53), the model generation unit 132 calculates the reward based on the motor speed control parameter, the information of the motor 50, and the allowable range. Specifically, the reward calculation unit 141 of the model generation unit 132 acquires the motor speed control parameter, the information of the motor 50, the allowable range, the vibration data of the motor 50, and the current consumption value, and determines the magnitude of the vibration. The increase or decrease of the reward is determined based on the relationship with the first threshold value and the relationship between the current consumption value and the second threshold value. Here, the reward calculation unit 141 determines whether the current consumption value when the motor 50 is operated by the motor speed control parameter is less than the second threshold value (step S57).

When the current consumption value is less than the second threshold value (Yes in step S57), the reward calculation unit 141 increases the reward (step S58). If items other than the priority items are not considered, the reward will be increased if the current consumption value, which is the priority item, is less than the second threshold value. However, when considering items other than the priority items so as to be within the permissible range, the reward according to the difference between the first threshold value and the vibration data value and the difference between the second threshold value and the current consumption value are further considered. It is possible to determine the reward according to the above. In one example, the reward calculation unit 141 increases the reward when the difference between the first threshold value and the value of the vibration data is positive, as compared with the case where the difference between the first threshold value and the value of the vibration data is negative. At this time, the reward may be determined depending on whether the difference between the first threshold value and the value of the vibration data is positive or negative, or the reward is determined according to the magnitude of the difference between the first threshold value and the value of the vibration data. May be done. That is, even within the permissible range, the reward may be increased as the value of the vibration data becomes smaller, and the reward may be decreased as the value of the vibration data becomes larger. Further, when the value of the vibration data becomes larger than the first threshold value outside the permissible range, the reward may be reduced. However, in these cases, the current consumption is a priority item, so the difference between the second threshold value and the current consumption value is larger than the difference between the first threshold value and the vibration data value. Is desirable to contribute to the increase or decrease in reward. As a result, when both the vibration and the current consumption are within the permissible range, the reward is higher and the action value can be enhanced as compared with the case where the current consumption is within the permissible range but the vibration is out of the permissible range.

When the current consumption value is larger than the second threshold value (No in step S57), the reward calculation unit 141 reduces the reward (step S59). If items other than the priority items are not considered, the reward will be reduced if the current consumption value, which is the priority item, is larger than the second threshold value. However, when considering items other than the priority items so as to be within the permissible range, the reward according to the difference between the first threshold value and the vibration data value and the difference between the second threshold value and the current consumption value are further considered. It is possible to determine the reward according to the above. In one example, the reward calculation unit 141 sets the absolute value of the reward to be reduced when the difference between the first threshold value and the vibration data value is positive, as compared with the case where the difference between the first threshold value and the vibration data value is negative. Make it smaller. At this time, the reward may be determined depending on whether the difference between the first threshold value and the value of the vibration data is positive or negative, or the reward is determined according to the magnitude of the difference between the first threshold value and the value of the vibration data. May be done. That is, even within the permissible range, the absolute value of the reward to be reduced may be reduced as the value of the vibration data becomes smaller, and the absolute value of the reward to be reduced may be increased as the value of the vibration data becomes larger. Further, when the value of the vibration data becomes larger than the first threshold value outside the permissible range, the absolute value of the reward to be reduced may be made larger. However, even in these cases, the current consumption is a priority item, so the difference between the second threshold value and the current consumption value is larger than the difference between the first threshold value and the vibration data value. Is desirable to contribute to the increase or decrease in reward.

After step S55, S56, S58 or S59, the function updater 142 of the model generating unit 132, based on the calculated compensation by compensation calculation unit 141, action value function Q (s _{_t,} a _t) Update ( Step S60). Action value function Q (s _t, a _t) is a function expressed by learned model storage unit 133 stores (1).

Then, the process returns to step S51. That is, the machine learning unit 13 repeatedly executes the processes from S60 from step S51 described above, and stores the generated action-value function Q (s _{_t,} a _t) as a learned model. Further, when the feedback value regarding the tact time other than the vibration and the current consumption is acquired, the learning data at the time of acquiring the data is accumulated.

Next, a method of inferring motor speed control parameters using the trained model stored in the machine learning unit 13 will be described. FIG. 9 is a flowchart showing an example of a processing procedure of a motor control method in the positioning control device according to the second embodiment.

First, the configuration of the initial control system 1 is determined by the user. Then, the parameter storage unit 11 stores the information including the information of the motor 50, the allowable range, and the value including the priority item (step S71). Information, tolerances and priorities for the motor 50 are set by the user via inputs not shown. Further, the sensor value acquisition unit 12 acquires vibration data from the vibration sensor 61 (step S72) and holds it. Further, the current consumption acquisition unit 16 acquires and holds the current consumption value from the current consumption measuring device 62 (step S73).

After that, the inference unit 134 of the machine learning unit 13 analyzes the value of the parameter storage unit 11, the value of the acquired vibration data, and the current consumption value using the trained model, and the set priority item becomes an allowable range. The motor speed control parameter is set in the motor speed control parameter output unit 14 (step S74). If the priority item is vibration, the motor speed control parameters that provide the optimum vibration within the permissible range are determined. At this time, the motor speed control parameter may be determined so that the current consumption, which is an item other than the priority item, is also within the allowable range. If the priority item is current consumption, the motor speed control parameters that provide the optimum current consumption within the permissible range are determined. At this time, the motor speed control parameter may be determined so that the vibration, which is an item other than the priority item, is also within the allowable range.

The motor speed control parameter output unit 14 outputs the set motor speed control parameter to the pulse output unit 15 (step S75). The pulse output unit 15 outputs a pulse to the amplifier 30 based on the motor speed control parameter from the motor speed control parameter output unit 14 (step S76).

This drives the motor 50. When the motor 50 is driven, the vibration of the motor 50 is detected by the vibration sensor 61 provided in the motor 50. After that, as described in step S72, the sensor value acquisition unit 12 acquires vibration data, which is the vibration of the motor 50. Further, when the motor 50 is driven, the current consumption value of the motor 50 is detected by the current consumption measuring device 62, and the current consumption value of the motor 50 is acquired by the current consumption acquisition unit 16 as described in step S73. Will be done. Then, as described above, the processes from steps S72 to S76 are repeatedly executed. In one example, the value of the vibration data acquired by the sensor value acquisition unit 12 changes beyond the permissible range, or the current consumption value acquired by the current consumption acquisition unit 16 changes beyond the permissible range. When it occurs, in step S74, the inference unit 134 analyzes the information of the motor 50, the permissible range and the priority item, the value of the vibration data, and the current consumption value by using the trained model, and the priority item is permissible. The motor speed control parameter within the range is set in the motor speed control parameter output unit 14.

In the above description, the case where the vibration sensor 61 and the current consumption measuring device 62 are provided in the motor 50 is shown, but the embodiment is not limited to this. FIG. 10 is a block diagram showing another example of the configuration of the control system including the positioning control device according to the second embodiment. In the following, the same components as those in FIGS. 1, 5 and 7 are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those in FIGS. 1, 5 and 7 will be described. In FIG. 10, the vibration sensor 61 and the current consumption measuring device 62 are provided not in the motor 50 but in the product 52 which is a device including the motor 50 and the drive unit 51 driven by the motor 50. As a result, the machine learning unit 13 can optimize the vibration value of the product 52 including the drive unit 51 and the like as well as the motor 50.

FIG. 11 is a block diagram showing another example of the configuration of the control system including the positioning control device according to the second embodiment. In the following, the same components as those in FIGS. 1, 6 and 7 are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those in FIGS. 1, 6 and 7 will be described. In FIG. 11, the vibration sensor 61 and the current consumption measuring device 62 are provided in the system 53 including the plurality of products 52. As an example, such a system 53 is a multi-axis control system including a plurality of motors 50 and products 52. Thereby, the machine learning unit 13 can optimize the vibration value of the system 53 including the plurality of products 52.

In the second embodiment, the machine learning unit 13 of the positioning control device 10 learns the motor speed control parameter in which the set priority item is within the permissible range according to the learning data, and generates a trained model. The learning data is created based on a combination of vibration data from the vibration sensor 61, current consumption value from the current consumption measuring device 62, information on the motor 50, motor speed control parameters, and an allowable range. The vibration sensor 61 and the current consumption measuring device 62 are provided in a motor 50, a product 52 including a drive unit 51 connected to the motor 50, or a system 53 including a plurality of products 52. Then, the machine learning unit 13 analyzes the information of the motor 50, the permissible range, the value of the vibration data, and the current consumption value by using the trained model, so that the tact time is within the permissible range and the set priority is set. Set the motor speed control parameters so that the item is within the allowable range. When vibration is set as a priority item, by driving the motor 50 with the set motor speed control parameter, it is possible to shorten the tact time, and the product 52 without giving excessive vibration to the motor 50. Alternatively, the system 53 can be operated, the load on the motor 50 can be reduced, and the life of the motor 50 can be extended. When the current consumption is set as a priority item, the motor 50 is driven by the set motor speed control parameter, so that the tact time can be shortened and the motor 50 does not consume excessive power. The product 52 or the system 53 can be operated to save power in the entire system 53.

Unlike the conventional technique, the servo gain, which is a control value, is determined by the feedback control from the motor 50, but the motor speed control parameter is determined by the output control of the positioning control device 10. Therefore, not only equipment such as a servomotor and an amplifier having an encoder, which is expensive and capable of feedback control, but also equipment such as a stepping motor and an amplifier having no feedback mechanism, the product 52 or the system 53 in which the motor 50 is provided. Life or energy saving can be improved.

In addition, the machine learning unit 13 is designed to learn motor speed control parameters that fall within the permissible range for all items, not just the items specified in the priority items. This makes it possible to set motor speed control parameters that maintain the minimum current consumption value while suppressing vibration within the permissible range when vibration is prioritized. On the contrary, when the current consumption is prioritized, the motor speed control parameter can be set so as to maintain the minimum vibration while suppressing the current consumption within the allowable range.

Embodiment 3.
FIG. 12 is a block diagram showing an example of the configuration of a control system including the positioning control device according to the third embodiment. Hereinafter, the same configurations as those of the first and second embodiments are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those of the first and second embodiments will be described. The control system 1 of the third embodiment further includes a takt time measuring device 63 in the product 52, which is a device including the motor 50 and the drive unit 51 driven by the motor 50, instead of the motor 50. That is, the tact time measuring device 63 is provided in the configuration of FIG. The takt time measuring device 63 is a device that measures takt time using a camera, a sensor, or the like. The tact time measuring device 63 outputs the measured tact time to the positioning control device 10.

The positioning control device 10 further includes a simulator unit 17 and a tact time acquisition unit 18. The simulator unit 17 is built in the positioning control device 10, and simulates the takt time from the motor speed control parameters output by the motor speed control parameter output unit 14. The simulator unit 17 reflects the tact time acquired by the simulation in the output.

The tact time acquisition unit 18 holds the tact time output from the tact time measuring device 63 and the tact time simulated by the simulator unit 17. The value to be held may be only one of the tact time output from the tact time measuring device 63 and the tact time simulated by the simulator unit 17.

The machine learning unit 13 sets the designated priority items in the allowable range according to the learning data created based on the combination of the motor speed control parameter, the information of the motor 50, the allowable range, and the vibration data and the tact time of the motor 50. Learn the motor speed control parameters that fit. That is, a trained model is generated that infers the motor speed control parameter in which the priority item specified from the information of the motor 50, the vibration data of the motor 50, and the takt time falls within the allowable range. The specified priority is vibration or takt time. The trained model is a model that learns the correlation between the vibration and tact time of the installation location of the motor 50 and the motor speed control parameter from the information of the motor 50, the allowable range, the vibration data, and the tact time. be. Further, the machine learning unit 13 analyzes the information of the motor 50, the permissible range, the value of the vibration data of the motor 50, and the tact time using the trained model, and the priority item is permissible for the motor 50 having the information of the motor 50. Output motor speed control parameters that fall within the range.

FIG. 13 is a flowchart showing an example of the procedure of the learning process of the machine learning unit included in the positioning control device according to the third embodiment. First, the data acquisition unit 131 acquires the motor speed control parameter, the information of the motor 50, the allowable range, the vibration data of the motor 50, and the tact time as learning data (step S91). Further, the data acquisition unit 131 acquires priority items from the parameter storage unit 11 (step S92). The model generation unit 132 determines whether the priority item is vibration or takt time (step S93).

When the priority item is vibration (vibration in step S93), the model generation unit 132 calculates the reward based on the motor speed control parameter, the information of the motor 50, and the allowable range. Specifically, the reward calculation unit 141 of the model generation unit 132 acquires the motor speed control parameter, the information of the motor 50, the allowable range, the vibration data of the motor 50, and the takt time, and obtains the predetermined vibration magnitude and the first. 1 The increase or decrease of the reward is determined based on the relationship with the threshold value and the relationship between the takt time and the third threshold value. The third threshold value is, in one example, the value of the takt time defined in the permissible range. Here, the reward calculation unit 141 determines whether the value of the vibration data when the motor 50 is operated by the motor speed control parameter is less than the first threshold value (step S94).

When the value of the vibration data is less than the first threshold value (Yes in step S94), the reward calculation unit 141 increases the reward (step S95). If items other than the priority items are not considered, the reward will be increased if the vibration data, which is the priority item, is less than the first threshold value. However, when considering items other than the priority items so that they are within the permissible range, the reward according to the difference between the first threshold value and the vibration data value and the difference between the third threshold value and the takt time are further increased. You can set the reward to respond to. In one example, the reward calculation unit 141 increases the reward when the difference between the third threshold value and the takt time is positive, as compared with the case where the difference between the third threshold value and the takt time is negative. At this time, the reward may be determined depending on whether the difference between the third threshold value and the tact time is positive or negative, or the reward may be determined according to the magnitude of the difference between the third threshold value and the tact time. .. That is, even within the permissible range, the reward may be increased as the tact time becomes smaller, and the reward may decrease as the tact time becomes larger. Further, when the tact time becomes larger than the third threshold value outside the permissible range, the reward may be reduced. However, in these cases, vibration is a priority item, so the difference between the first threshold and the vibration data value is more rewarding than the difference between the third threshold and the takt time. It is desirable to contribute to the increase or decrease of. As a result, when both the vibration and the takt time are within the permissible range, the reward is higher and the action value can be enhanced as compared with the case where the vibration is within the permissible range but the takt time is not within the permissible range.

When the value of the vibration data is larger than the first threshold value (No in step S94), the reward calculation unit 141 reduces the reward (step S96). If items other than the priority items are not considered, the reward will be reduced if the vibration data, which is the priority item, is larger than the first threshold value. However, when considering items other than the priority items so that they are within the permissible range, the reward according to the difference between the first threshold value and the vibration data value and the difference between the third threshold value and the takt time are further increased. You can set the reward to respond to. In one example, the reward calculation unit 141 makes the absolute value of the reward to be reduced smaller when the difference between the third threshold value and the takt time is positive than when the difference between the third threshold value and the takt time is negative. At this time, the reward may be determined depending on whether the difference between the third threshold value and the tact time is positive or negative, or the reward may be determined according to the magnitude of the difference between the third threshold value and the tact time. .. That is, even within the permissible range, the absolute value of the reward to be reduced may be reduced as the tact time becomes smaller, and the absolute value of the reward to be reduced may be increased as the tact time becomes larger. Further, when the tact time is larger than the third threshold value outside the permissible range, the absolute value of the reward to be reduced may be increased. However, even in these cases, vibration is a priority item, so the difference between the first threshold and the vibration data value is more rewarding than the difference between the third threshold and the takt time. It is desirable to contribute to the increase or decrease of.

When the priority item is takt time (in the case of takt time in step S93), the model generation unit 132 calculates the reward based on the motor speed control parameter, the information of the motor 50, and the allowable range. Specifically, the reward calculation unit 141 of the model generation unit 132 acquires the motor speed control parameter, the information of the motor 50, the allowable range, the vibration data of the motor 50, and the takt time, and obtains the predetermined vibration magnitude and the first. 1 The increase or decrease of the reward is determined based on the relationship with the threshold value and the relationship between the takt time and the third threshold value. Here, the reward calculation unit 141 determines whether the value of the tact time when the motor 50 is operated by the motor speed control parameter is less than the third threshold value (step S97).

When the takt time is less than the third threshold value (Yes in step S97), the reward calculation unit 141 increases the reward (step S98). If items other than the priority items are not considered, the reward will be increased if the priority item, takt time, is less than the third threshold value. However, when considering items other than the priority items so that they are within the permissible range, the reward according to the difference between the first threshold value and the vibration data value and the difference between the third threshold value and the takt time are further increased. You can set the reward to respond to. In one example, the reward calculation unit 141 increases the reward when the difference between the first threshold value and the value of the vibration data is positive, as compared with the case where the difference between the first threshold value and the value of the vibration data is negative. At this time, the reward may be determined depending on whether the difference between the first threshold value and the value of the vibration data is positive or negative, or the reward is determined according to the magnitude of the difference between the first threshold value and the value of the vibration data. May be done. That is, even within the permissible range, the reward may be increased as the value of the vibration data becomes smaller, and the reward may be decreased as the value of the vibration data becomes larger. Further, when the value of the vibration data becomes larger than the first threshold value outside the permissible range, the reward may be reduced. However, in these cases, the tact time is a priority item, so the difference between the third threshold and the tact time is larger than the difference between the first threshold and the vibration data value. It is desirable to contribute to the increase or decrease of the reward. As a result, when both the vibration and the tact time are within the permissible range, the reward is higher and the action value can be enhanced as compared with the case where the tact time is within the permissible range but the vibration is not within the permissible range.

When the value of the takt time is larger than the third threshold value (No in step S97), the reward calculation unit 141 reduces the reward (step S99). If items other than the priority items are not considered, the reward will be reduced if the priority item, the takt time, is larger than the third threshold value. However, when considering items other than the priority items so that they are within the permissible range, the reward according to the difference between the first threshold value and the vibration data value and the difference between the third threshold value and the takt time are further increased. You can set the reward to respond to. In one example, the reward calculation unit 141 sets the absolute value of the reward to be reduced when the difference between the first threshold value and the vibration data value is positive, as compared with the case where the difference between the first threshold value and the vibration data value is negative. Make it smaller. At this time, the reward may be determined depending on whether the difference between the first threshold value and the value of the vibration data is positive or negative, or the reward is determined according to the magnitude of the difference between the first threshold value and the value of the vibration data. May be done. That is, even within the permissible range, the absolute value of the reward to be reduced may be reduced as the value of the vibration data becomes smaller, and the absolute value of the reward to be reduced may be increased as the value of the vibration data becomes larger. Further, when the value of the vibration data becomes larger than the first threshold value outside the permissible range, the absolute value of the reward to be reduced may be made larger. However, even in these cases, the tact time is a priority item, so the difference between the third threshold and the tact time is larger than the difference between the first threshold and the vibration data value. It is desirable to contribute to the increase or decrease of the reward.

After step S95, S96, S98 or S99, the function updater 142 of the model generating unit 132, based on the calculated compensation by compensation calculation unit 141, action value function Q (s _{_t,} a _t) Update ( Step S100). Action value function Q (s _t, a _t) is a function expressed by learned model storage unit 133 stores (1).

Then, the process returns to step S91. That is, the machine learning unit 13 repeatedly executes the processes from S100 from step S91 described above, and stores the generated action-value function Q (s _{_t,} a _t) as a learned model. Further, when the feedback value regarding the current consumption other than the vibration and the tact time is acquired, the learning data at the time of acquiring the data is accumulated.

Next, a method of inferring motor speed control parameters using the trained model stored in the machine learning unit 13 will be described. FIG. 14 is a flowchart showing an example of a processing procedure of a motor control method in the positioning control device according to the third embodiment.

First, the configuration of the initial control system 1 is determined by the user. Then, the parameter storage unit 11 stores the information including the information of the motor 50, the allowable range, and the value including the priority item (step S111). Information, tolerances and priorities for the motor 50 are set by the user via inputs not shown. Further, the sensor value acquisition unit 12 acquires and holds vibration data from the vibration sensor 61 (step S112). Further, the takt time acquisition unit 18 acquires and holds the takt time from the takt time measuring device 63 (step S113).

After that, the inference unit 134 of the machine learning unit 13 analyzes the value of the parameter storage unit 11, the value of the acquired vibration data, and the takt time using the trained model, and the set priority item is within the allowable range. The speed control parameter is set in the motor speed control parameter output unit 14 (step S114). If the priority item is vibration, the motor speed control parameters that provide the optimum vibration within the permissible range are determined. At this time, the motor speed control parameter may be determined so that the takt time, which is an item other than the priority item, is also within the allowable range. If the priority item is takt time, the motor speed control parameter that provides the optimum takt time within the permissible range is determined. At this time, the motor speed control parameter may be determined so that the vibration, which is an item other than the priority item, is also within the allowable range.

The motor speed control parameter output unit 14 outputs the set motor speed control parameter to the pulse output unit 15 (step S115). The pulse output unit 15 outputs a pulse to the amplifier 30 based on the motor speed control parameter from the motor speed control parameter output unit 14 (step S116).

In parallel with steps S115 and S116, the simulator unit 17 simulates the control system 1 from the motor speed control parameters obtained from the motor speed control parameter output unit 14 and calculates the takt time (step S117). The simulator unit 17 outputs the calculated tact time to the tact time acquisition unit 18.

This drives the motor 50. When the motor 50 is driven, the vibration of the motor 50 is detected by the vibration sensor 61 provided in the motor 50. After that, as described in step S112, the sensor value acquisition unit 12 acquires vibration data, which is the vibration of the motor 50. Further, after the pulse is output to the amplifier 30 in step S116 and after the tact time is calculated in step S117, the tact time of the motor 50 is acquired by the tact time acquisition unit 18 as described in step S113. .. Then, as described above, the processes from steps S112 to S116 are repeatedly executed. In one example, the value of the vibration data acquired by the sensor value acquisition unit 12 changes beyond the permissible range, or the tact time acquired by the tact time acquisition unit 18 changes beyond the permissible range. Then, in step S114, the inference unit 134 analyzes the information of the motor 50, the allowable range and the priority item, the value of the vibration data, and the takt time using the trained model, and the priority item is the allowable range. The motor speed control parameter is set in the motor speed control parameter output unit 14.

In the above description, the case where the vibration sensor 61 and the tact time measuring device 63 are provided in the product 52 is shown, but the embodiment is not limited to this. FIG. 15 is a block diagram showing another example of the configuration of the control system including the positioning control device according to the third embodiment. In the following, the same components as those in FIGS. 1, 6 and 12 are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those in FIGS. 1, 6 and 12 will be described. In FIG. 15, the vibration sensor 61 and the takt time measuring device 63 are provided in a system 53 including a plurality of products 52. As an example, such a system 53 is a multi-axis control system including a plurality of motors 50 and products 52. Thereby, the machine learning unit 13 can optimize the vibration value of the system 53 including the plurality of products 52.

Further, in the above description, the case where machine learning can be performed in consideration of the tact time in addition to the vibration of the motor 50 in the configuration of the first embodiment is shown. However, in the configuration of the second embodiment, machine learning may be performed in consideration of the tact time in addition to the vibration and the current consumption of the motor 50.

FIG. 16 is a block diagram showing another example of the configuration of the control system including the positioning control device according to the third embodiment. In the following, the same components as those in FIGS. 1, 10 and 12 are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those in FIGS. 1, 10 and 12 will be described. The control system 1 of FIG. 16 shows a case where the third embodiment is applied to the configuration of FIG. 10 of the second embodiment. The control system 1 further includes a takt time measuring device 63 in the product 52. Further, the positioning control device 10 further includes a simulator unit 17 and a tact time acquisition unit 18.

FIG. 17 is a block diagram showing another example of the configuration of the control system including the positioning control device according to the third embodiment. In the following, the same components as those in FIGS. 1, 6, 11 and 12 are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those in FIGS. 1, 6, 11 and 12 will be described. explain. The control system 1 of FIG. 17 shows a case where the third embodiment is applied to the configuration of FIG. 11 of the second embodiment. The control system 1 further includes a takt time measuring device 63 in a system 53 including a plurality of products 52. Further, the positioning control device 10 further includes a simulator unit 17 and a tact time acquisition unit 18.

The machine learning unit 13 is designated as a priority item according to the learning data created based on the combination of the motor speed control parameter, the information of the motor 50, the allowable range, the vibration data of the motor 50, the current consumption value and the tact time. Learns motor speed control parameters that are within the permissible range. That is, a trained model is generated that infers the motor speed control parameter in which the priority item specified from the information of the motor 50, the vibration data of the motor 50, the current consumption value, and the takt time falls within the allowable range. The priority items specified are vibration, current consumption or takt time. The trained model is based on the information of the motor 50, the allowable range, the vibration data, the current consumption, the tact time, and the vibration, the current consumption, the tact time, and the motor speed control parameter of the installation location of the motor 50. It is a model that learned the correlation of. The machine learning unit 13 may learn motor speed control parameters in which the designated priority items fall within the permissible range, and not only the designated priority items but also items other than the priority items fall within the permissible range. You may learn the speed control parameters. Further, the machine learning unit 13 analyzes the information of the motor 50, the allowable range and priority items, the value of the vibration data of the motor 50, the current consumption value and the tact time by using the trained model, and has the information of the motor 50. The motor 50 outputs motor speed control parameters such that the priority items are within the permissible range.

The machine learning method in the machine learning unit 13 is a combination of the methods shown in the second and third embodiments, and the basic processing is the same, so the description thereof will be omitted. In one example, in step S53 of FIG. 8, the model generation unit 132 determines whether the priority item is vibration, current consumption, or takt time. Then, in each case, the reward calculation unit 141 increases or decreases the reward based on the relationship between the priority item and the threshold value. Further, at this time, the reward may be determined only by paying attention to whether or not the priority item is within the allowable range, or the reward may be determined by paying attention to whether or not the priority item and the items other than the priority item are within the allowable range. May be determined.

Further, the motor control method using the trained model in the positioning control device 10 is a combination of those shown in the second and third embodiments, and the basic processing is the same, so the description thereof will be omitted. .. In one example, in step S114 of FIG. 14, the inference unit 134 analyzes the value of the parameter storage unit 11, the value of the acquired vibration data, the current consumption value, and the takt time using the trained model, and sets the priority. The motor speed control parameter whose item is within the allowable range is set in the motor speed control parameter output unit 14.

FIG. 18 is a diagram showing an example of a speed command in the positioning control device according to the third embodiment. In this figure, the horizontal axis represents time and the vertical axis represents pulse frequency or velocity. In the speed command, the time from the speed 0 to the command speed is called the actual acceleration time, and the time from the command speed to the speed 0 is called the actual deceleration time. The area of the part where the command speed is continued during the speed command represents the moving distance. Further, the time from speed 0 to accelerating to reach the command speed, then decelerating to reach speed 0 is called takt time.

FIG. 19 is a diagram showing an example of learning results of motor speed control parameters in the positioning control device according to the third embodiment. In this figure, the horizontal axis represents time and the vertical axis represents pulse frequency or velocity. The speed command curve F1 shows a speed command curve with motor speed control parameters set so that the tact time is short. In the speed command curve F1, the actual acceleration time and the actual deceleration time are shortened. That is, sudden acceleration and deceleration are performed, and the current consumption increases and the vibration also increases.

The speed command curve F2 shows a speed command curve with motor speed control parameters set so that the tact time is longer than that of the speed command curve F1. In the speed command curve F2, the actual acceleration time and the actual deceleration time are longer than those of the speed command curve F1. Therefore, the vibration can be reduced as compared with the case of the speed command curve F1, but the acceleration and deceleration are too slow. Further, since the command speed, which is the steady operation speed, is lower than the speed command curve F1, the current consumption becomes large and the tact time becomes long.

The speed command curve F3 has a longer tact time and a smaller command speed than the speed command curve F1, and has a shorter tact time and a larger command speed than the speed command curve F2. The actual acceleration time and the actual deceleration time are between the speed command curve F1 and the speed command curve F2, and the vibration is suppressed. Further, since the command speed is larger than the speed command curve F2, the current consumption can be reduced as compared with the speed command curve F2. Further, the tact time can also be shortened as compared with the speed command curve F2. That is, in the speed command curve F3, the vibration, the current consumption, and the tact time are all within the permissible range and are balanced values. The machine learning unit 13 will set motor speed control parameters such as the speed command curve F3.

In the third embodiment, the machine learning unit 13 of the positioning control device 10 learns the motor speed control parameter in which the set priority item is within the permissible range according to the learning data, and generates a trained model. The learning data is created based on a combination of vibration data from the vibration sensor 61, tact time from the tact time measuring device 63, information on the motor 50, motor speed control parameters, and an allowable range. The vibration sensor 61, the current consumption measuring device 62, and the takt time measuring device 63 are provided in the motor 50, the product 52 including the drive unit 51 connected to the motor 50, or the system 53 including the plurality of products 52. Then, the machine learning unit 13 analyzes the information of the motor 50, the permissible range, the value of the vibration data, and the takt time by using the trained model, and the motor speed control in which the set priority items are within the permissible range. Set the parameters. When vibration is set as a priority item, by driving the motor 50 with the set motor speed control parameter, the product 52 or the system 53 is operated without giving excessive vibration to the motor 50, and the motor 50 is operated. The burden on the motor 50 can be reduced, and the life of the motor 50 can be extended. When the tact time is set as a priority item, the tact time can be shortened and the production efficiency of the entire system 53 can be improved by driving the motor 50 with the set motor speed control parameter.

Unlike the conventional technique, the servo gain, which is a control value, is determined by the feedback control from the motor 50, but the motor speed control parameter is determined by the output control of the positioning control device 10. Therefore, not only equipment such as a servomotor and an amplifier having an encoder, which is expensive and capable of feedback control, but also equipment such as a stepping motor and an amplifier having no feedback mechanism, the product 52 or the system 53 in which the motor 50 is provided. Life, energy saving and production efficiency can be improved. Further, as in the conventional technique, in the feedback control using the position command, the gain is only adjusted and the effect of shortening the tact time is small. However, in the third embodiment, the motor speed control parameter is adjusted, so that the motor speed control parameter is adjusted. Depending on the restrictions, it is possible to significantly reduce the tact time compared to the conventional technology.

Also, not only the items specified in the priority items, but also the motor speed control parameters that fall within the allowable range for all items are learned. This makes it possible to set motor speed control parameters that maintain the minimum current consumption and takt time while suppressing vibration within the permissible range when vibration is prioritized. On the contrary, when the takt time is prioritized, the motor speed control parameter can be set so as to maintain the minimum vibration and the current consumption while keeping the takt time within the allowable range.

Here, the hardware configuration of the positioning control device 10 described in the first, second, and third embodiments will be described. FIG. 20 is a diagram schematically showing an example of a hardware configuration that realizes the positioning control device according to the first, second, and third embodiments.

In the positioning control device 10, the processor 101 and the memory 102 are connected via the bus line 103. An example of the processor 101 is a CPU (Central Processing Unit) or a system LSI (Large Scale Integration). An example of the memory 102 is a RAM (Random Access Memory), a ROM (Read Only Memory), which is a main storage device, an HDD (Hard Disk Drive) or an SSD (Solid State Drive), which is an auxiliary storage device.

A part or all of the functions of the sensor value acquisition unit 12, the machine learning unit 13, the motor speed control parameter output unit 14, the pulse output unit 15, the current consumption acquisition unit 16, the simulator unit 17, and the tact time acquisition unit 18 are performed by the processor 101. When realized, some or all of the functions are realized by the processor 101 and software, firmware, or a combination of software and firmware. The software or firmware is written as a program and stored in the memory 102. By reading and executing the program stored in the memory 102, the processor 101 reads and executes the sensor value acquisition unit 12, the machine learning unit 13, the motor speed control parameter output unit 14, the pulse output unit 15, the current consumption acquisition unit 16, and the simulator. A part or all of the functions of the unit 17 and the tact time acquisition unit 18 are realized.

A part or all of the functions of the sensor value acquisition unit 12, the machine learning unit 13, the motor speed control parameter output unit 14, the pulse output unit 15, the current consumption acquisition unit 16, the simulator unit 17, and the tact time acquisition unit 18 are performed by the processor 101. When realized, the positioning control device 10 includes a sensor value acquisition unit 12, a machine learning unit 13, a motor speed control parameter output unit 14, a pulse output unit 15, a current consumption acquisition unit 16, a simulator unit 17, and a tact time acquisition unit 18. A program in which a step executed by a part or all of the above will be executed as a result is stored in the memory 102. The program stored in the memory 102 is one of the sensor value acquisition unit 12, the machine learning unit 13, the motor speed control parameter output unit 14, the pulse output unit 15, the current consumption acquisition unit 16, the simulator unit 17, and the tact time acquisition unit 18. It can also be said to cause a computer to perform a procedure or method performed by a part or all of them.

The configuration shown in the above embodiments is an example, and can be combined with another known technique, can be combined with each other, and does not deviate from the gist. It is also possible to omit or change a part of the configuration.

1 control system, 10 positioning control device, 11 parameter storage unit, 12 sensor value acquisition unit, 13 machine learning unit, 14 motor speed control parameter output unit, 15 pulse output unit, 16 current consumption acquisition unit, 17 simulator unit, 18 tact Time acquisition unit, 30 amplifier, 50 motor, 51 drive unit, 52 product, 53 system, 61 vibration sensor, 62 current consumption measurement device, 63 tact time measurement device, 131 data acquisition unit, 132 model generation unit, 133 trained model Storage unit, 134 inference unit, 141 reward calculation unit, 142 function update unit.

Claims

A positioning control device that is electrically connected to a motor via an amplifier and controls the motor.
Information necessary for determining the motor speed control parameter, which is a parameter for controlling the motor, and is a parameter storage unit that stores parameters including the information of the motor and the allowable range of vibration and tact time of the motor. When,
A vibration data acquisition unit that acquires vibration data that is vibration of the installation location of the motor detected by the vibration sensor, and a vibration data acquisition unit.
Using a trained model that learns the correlation between the vibration of the installation location of the motor and the motor speed control parameter from the parameter and the vibration data, the vibration is permissible from the parameter and the vibration data. A machine learning unit that determines the motor speed control parameters that fall within the range,
Based on the determined motor speed control parameters, an output unit that outputs to the amplifier as a pulse that controls the amplifier, and
A positioning control device comprising.
Further provided with a current consumption acquisition unit that acquires the current consumption value of the installation location of the motor as measured by the current consumption measuring device.
The parameters further include an allowable range of current consumption of the motor.
In the trained model, the correlation between the vibration of the installation location of the motor, the current consumption value, and the motor speed control parameter is learned from the parameters, the vibration data, and the current consumption value. ,
The positioning control device according to claim 1, wherein the machine learning unit determines the motor speed control parameter from the parameter, the vibration data, and the current consumption value using the trained model.
Using the trained model, the machine learning unit prioritizes the priority items selected by the user among the items including the vibration and the current consumption, and the items other than the priority items are also within the permissible range. The positioning control device according to claim 2, wherein the motor speed control parameter is determined.
Further equipped with a tact time acquisition unit for acquiring the tact time of the installation location of the motor measured by the tact time measuring device.
The parameters further include the takt time tolerance of the motor.
In the trained model, the correlation between the vibration of the installation location of the motor, the tact time, and the motor speed control parameter is learned from the parameters, the vibration data, and the tact time.
The positioning control device according to claim 1, wherein the machine learning unit determines the motor speed control parameter from the parameter, the vibration data, and the takt time using the trained model.
Using the trained model, the machine learning unit prioritizes the priority items selected by the user among the items including the vibration and the takt time, and the items other than the priority items are also within the permissible range. The positioning control device according to claim 4, wherein the motor speed control parameter is determined.
Further equipped with a tact time acquisition unit for acquiring the tact time of the installation location of the motor measured by the tact time measuring device.
The parameters further include the takt time tolerance of the motor.
In the trained model, the vibration of the installation location of the motor, the current consumption value, the tact time, and the motor speed control parameter are obtained from the parameters, the vibration data, the current consumption value, and the tact time. It is a learning of the correlation of
The positioning according to claim 2, wherein the machine learning unit determines the motor speed control parameter from the parameter, the vibration data, the current consumption value, and the takt time using the trained model. Control device.
Using the trained model, the machine learning unit prioritizes the priority items selected by the user among the items including the vibration, the current consumption, and the takt time, and allows items other than the priority items. The positioning control device according to claim 6, wherein the motor speed control parameter within the range is determined.
The machine learning unit
A data acquisition unit that acquires a combination of the parameter and the vibration data,
According to the learning data created based on the combination of the parameters and the vibration data, the correlation between the vibration of the installation location of the motor and the motor speed control parameter is learned, and the trained model is generated. Learning department and
A trained model storage unit that stores the trained model,
The positioning control device according to claim 1, further comprising.
A machine learning device that learns the correlation between the vibration of the motor installation location and the motor speed control parameter, which is a parameter that controls the motor.
Information necessary for determining the motor speed control parameter, the parameter including the information of the motor, the allowable range of vibration and tact time of the motor, and the installation location of the motor detected by the vibration sensor. A data acquisition unit that acquires a combination of vibration data, which is the vibration of
According to the learning data created based on the combination of the parameter and the vibration data, the correlation between the vibration of the installation location of the motor and the motor speed control parameter is learned, and a trained model is generated. With the learning department
A trained model storage unit that stores the trained model,
A machine learning device characterized by being equipped with.