Controller, control method, and computer-readable recording medium for controlling controlled object with trained model

1 . A controller comprising

a processor configured to execute a process that includes:

acquiring process data for an actually used controlled object;

converting the acquired process data based on at least one of a target set value and parameters for the actually used controlled object; and

outputting a manipulated variable for the actually used controlled object by inputting the converted process data into a trained model; and

a Proportional-Integral-Differential controller configured to calculate a manipulated variable for the actually used controlled object, wherein

upon receiving input of process data for a specific controlled object, the trained model outputs a manipulated variable for approximating process data for the specific controlled object to a specific target set value, and

the parameters include parameters for specifying a relation between manipulated variables for the actually used controlled object and process data obtained by use of the manipulated variables, the trained model being trained using a target set value that may be the same or different from the specific target set value, wherein

the process further includes: adjusting a mapping function and an inverse mapping function based on a ratio between at least one of the target set value and the parameters for the actually used controlled object and at least one of the specific target set value for the specific controlled object and the parameters for the specific controlled object, the inverse mapping function being an inverse function of the mapping function, and

switching a control mode between a first control mode under which the manipulated variable calculated by the processor is fed to the actually used controlled object, and a second control mode under which the manipulated variable calculated by the Proportional-Integral-Differential controller is fed to the actually used controlled object, wherein

the outputting the manipulated variable includes conversion of the acquired process data by use of the adjusted mapping function and inverse conversion of data that is calculated using the converted process data and the trained model, the inverse conversion being performed using the inverse mapping function of the mapping function.

2 . A non-transitory computer-readable recording medium having stored therein a control program configured to cause a computer to execute:

acquiring process data for an actually used controlled object;

converting the acquired process data based on at least one of a target set value and parameters for the actually used controlled object; and

calculating a manipulated variable for the actually used controlled object by use of the converted process data and a trained model, wherein

outputting a manipulated variable for the actually used controlled object by inputting the converted process data into a trained model, wherein

upon receiving input of process data for a specific controlled object, the trained model outputs a manipulated variable for approximating process data for the specific controlled object to a specific target set value, the trained model being trained using a target set value that may be the same or different from the specific target set value, and

the parameters include parameters for specifying a relation between manipulated variables for the actually used controlled object and process data obtained by use of the manipulated variables, wherein

the control program is further configured to execute: adjusting a mapping function and an inverse mapping function based on a ratio between at least one of the target set value and the parameters for the actually used controlled object and at least one of the specific target set value for the specific controlled object and the parameters for the specific controlled object, the inverse mapping function being an inverse function of the mapping function, and

switching a control mode between a first control mode under which the manipulated variable calculated by a processor is fed to the actually used controlled object, and a second control mode under which a manipulated variable calculated by a Proportional-Integral-Differential controller is fed to the actually used controlled object, wherein

the outputting the manipulated variable includes conversion of the acquired process data by use of the adjusted mapping function and inverse conversion of data that is calculated using the converted process data and the trained model, the inverse conversion being performed using the inverse mapping function of the mapping function.

3 . A control method comprising:

acquiring process data for an actually used controlled object;

converting the acquired process data based on at least one of a target set value and parameters for the actually used controlled object; and

outputting a manipulated variable for the actually used controlled object by inputting the converted process data into a trained model, wherein the trained model being trained using feature data representing a state of a process of a specific controlled object, wherein

upon receiving input of process data for the specific controlled object, the trained model outputs a manipulated variable for approximating process data for the specific controlled object to a specific target set value, the trained model being trained using a target set value that may be the same or different from the specific target set value, and

the parameters include parameters for specifying a relation between manipulated variables for the actually used controlled object and process data obtained by use of the manipulated variable, wherein

the control method further comprises: adjusting a mapping function that maps the feature data with the process data, and an inverse mapping function based on a ratio between at least one of the target set value and the parameters for the actually used controlled object and at least one of the specific target set value for the specific controlled object and the parameters for the specific controlled object, the inverse mapping function being an inverse function of the mapping function, and wherein

the outputting the manipulated variable includes conversion of the acquired process data by use of the adjusted mapping function and inverse conversion of data that is calculated using the converted process data and the trained model, wherein the manipulated variable is provided to control the actually used controlled object, the inverse conversion being performed using the inverse mapping function of the mapping function,

wherein the control method further comprises controlling the actually used controlled object by using the provided manipulated variable.

4 . The control method according to claim 3 , wherein the manipulated variable is determined correspond to a highest evaluation value associated with the feature data.

5 . The control method according to claim 3 , wherein a control cycle, in which an acquisition unit acquires process data, is adjusted in proportion to value of a time constant included in the parameters.

6 . The control method according to claim 3 , wherein the feature data is associated with a coefficient, wherein the coefficient is adjusted based on at least one of the target set value and the parameters for the actually used controlled object and at least one of the specific target set value for the specific controlled object and the parameters for the specific controlled object.

7 . The control method according to claim 3 , wherein the feature data is associated with a coefficient, wherein the coefficient is adjusted based on a ratio between at least one of the target set value and the parameters for the actually used controlled object and at least one of the specific target set value for the specific controlled object and the parameters for the specific controlled object.

8 . The control method according to claim 3 , wherein the actually used controlled object comprises a water tank, the process data comprises a liquid surface level in the water tank, the manipulated variable comprises a degree of opening a valve.

9 . The control method according to claim 3 , wherein the parameters include at least one of input and output ranges and a process gain.

10 . The control method according to claim 3 , wherein the parameters include a time constant,

the process further includes adjusting a control cycle based on the time constant included among the parameters, and

the acquiring includes acquiring process data for the actually used controlled object in the adjusted control cycle.

11 . The control method according to claim 3 , wherein the process further includes performing learning for the trained model.

12 . The control method according to claim 3 , wherein the switching

includes switching a control mode to a first control mode at a startup, and switching the control mode to a second control mode when the difference of process data and the target set value is lower than a threshold value.