Factory system for machine learning of an actuator

1. A control system comprising:

a factory system configured to control an actuator; and

a learning system configured to extract records for machine learning associated with the actuator from the factory system via a network and to generate an estimation model for estimating a condition of the actuator by machine learning using the records,

wherein the factory system includes a plurality of nodes including a control device to control the actuator, and a readable data construction device connected to the learning system via the network,

wherein the readable data construction device is configured to:

acquire data items associated with the actuator from at least one of the nodes other than the readable data construction device; and

construct readable data which is readable from the learning system, the readable data including the data items associated with the actuator, and

wherein the learning system is configured to:

extract the records from the readable data;

store the records for machine learning in a record storage;

generate the estimation model by machine learning using information stored in the record storage,

store a plurality of estimation models including the estimation model;

evaluate an estimation accuracy of each of the estimation models based, at least in part, on the records for machine learning stored in the record storage and estimation results of the condition of the actuator by the plurality of estimation models;

generate display data including an evaluation result of the estimation accuracy;

receive a model designation from at least one of the nodes of the factory system; and

transmit, to the at least one of the nodes, one of the plurality of estimation models according to the model designation.

2. The control system according to according to claim 1 , wherein the readable data construction device is further configured to store the readable data in a data memory that is accessible from the learning system via the network.

3. The control system according to according to claim 1 ,

wherein the estimation model is configured to output an estimation of the condition of the actuator in response to an input of a status of motion of the actuator, and

wherein at least one of the records includes:

at least one data item indicating the status of motion; and

at least one data item indicating the condition of the actuator.

4. The control system according to claim 3 ,

wherein at least one of the plurality of nodes includes a model storage configured to store the transmitted estimation model, and

wherein at least one of the plurality of nodes is configured to estimate the condition of the actuator based, at least in part, on a present value of at least one of the data items indicating a state of motion of the actuator and the transmitted estimation model stored in the model storage.

5. The control system according to claim 4 , wherein at least one of the plurality of nodes is further configured to estimate a deterioration level of the actuator based, at least in part, on the present value and the transmitted estimation model.

6. The control system according to claim 1 ,

wherein the readable data construction device is further configured to acquire the data items read out, at least in part, from the control device by an interface program for accessing data in the control device, and

wherein the interface program is stored in an interface storage provided in at least one of the nodes which is isolated from the network by the readable data construction device.

7. The control system according to claim 6 , wherein the interface program is an application programming interface configured to read out predetermined data items from the storage of the control device.

8. The control system according to claim 6 ,

wherein the plurality of nodes includes an additional node other than the readable data construction device and the control device, and

wherein the additional node includes the interface storage.

9. The control system according to claim 1 ,

wherein the readable data construction device is further configured to:

acquire data items associated with a plurality of the actuators including the actuator; and

construct the readable data including the data items in association with labels, each of the labels including identification information of at least one of the plurality of actuators, and

wherein the learning system is further configured to extract the records for machine learning from the readable data by designating a label and read out a data item corresponding to the label in the readable data.

10. The control system according to claim 1 ,

wherein at least one of the plurality of nodes includes a model storage configured to store the transmitted estimation, and

wherein at least one of the plurality of nodes is configured to estimate the condition of the actuator based, at least in part, on the transmitted estimation model stored in the model storage.

11. The control system according to claim 10 ,

wherein at least one of the plurality of nodes is further configured to adjust a control parameter of the actuator based, at least in part, on an estimation result of the condition of the actuator, and

wherein the control device is configured to control the actuator based, at least in part, on the control parameter.

12. The control system according to claim 1 , wherein the learning system is further configured to:

select a sample record from the records for machine learning stored in the record storage;

input at least one data item of the sample record indicating a status of motion of the actuator to the estimation models to derive the estimation results; and

compare the estimation results which are derived with at least one data item of the sample record indicating the condition of the actuator to evaluate an estimation accuracy of the estimation models.

13. The control system according to claim 1 , wherein at least one of the plurality of nodes is configured to restrict a transmission of data, from the learning system to the factory system, that is different from the estimation model corresponding to the model designation.

14. A factory system connected to a learning system for machine learning via a network, the factory system comprising:

a plurality of nodes including a control device configured to control an actuator, and a readable data construction device connected to the learning system via the network, wherein the readable data construction device is configured to:

acquire data items associated with the actuator from at least one of the nodes other than the readable data construction device; and

construct readable data which is readable from the learning system, the readable data including the data items,

wherein the learning system is configured to:

extract records for machine learning from the readable data;

store the records for machine learning in a record storage;

generate an estimation model for estimating a condition of the actuator based, at least in part, on the records for machine learning;

store a plurality of estimation models including the estimation model;

evaluate an estimation accuracy of each of the estimation models based, at least in part, on the records for machine learning stored in the record storage and estimation results of the condition of the actuator by the plurality of estimation models;

generate display data including an evaluation result of the estimation accuracy;

receive a model designation from at least one of the nodes of the factory system; and

transmit, to the at least one of the nodes, one of the plurality of estimation models according to the model designation, and

wherein at least one of the plurality of nodes is configured to estimate the condition of the actuator based, at least in part, on the transmitted estimation model.

15. The factory system according to claim 14 , wherein the readable data construction device is further configured to put the readable data in a data memory that is accessible from the learning system via the network.

16. A factory system according to claim 14 ,

wherein at least one of the plurality of nodes includes a model storage configured to store the transmitted estimation model, and

wherein the condition is estimated based, at least in part, on the transmitted estimation model stored in the model storage.

17. The factory system according to claim 14 , wherein at least one of the plurality of nodes is configured to restrict a transmission of data, from the learning system to the factory system, that is different from the estimation model corresponding to the model designation.

18. A method comprising:

acquiring data items associated with an actuator of a factory system from at least one of a plurality of nodes of the factory system, the plurality of nodes including a control device to control the actuator;

constructing readable data which is readable from a learning system, the readable data including the data items;

storing the readable data in a data memory which is accessible from a learning system via a network, wherein the learning system extracts records for machine learning from the readable data and generates an estimation model for estimating a condition of the actuator based, at least in part, on the records for machine learning;

storing a plurality of estimation models including the estimation model;

evaluating an estimation accuracy of each of the estimation models based, at least in part, on the records for machine learning stored in a record storage and estimation results of the condition of the actuator by the plurality of estimation models;

generating display data including an evaluation result of the estimation accuracy;

receiving a model designation from at least one of the nodes of the factory system;

transmitting, to the at least one of the nodes, one of the plurality of estimation models according to the model designation; and

estimating the condition of the actuator based, at least in part, on the transmitted estimation model.

19. The method according to claim 18 , further comprising:

acquiring the transmitted estimation model from the learning system via the network; and

storing the transmitted estimation model in a model storage,

wherein estimating the condition includes estimating the condition based, at least in part, on the transmitted estimation model stored in the model storage.

20. The method according to claim 18 , wherein at least one of the plurality of nodes is configured to restrict a transmission of data, from the learning system to the factory system, that is different from the estimation model corresponding to the model designation.