JP2014081758A - Control device and control method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce a risk which may be generated when the arithmetic function or communicating function of a device on which advanced control is mounted stop.SOLUTION: An upper side control arithmetic part 10 calculates manipulated variables MV2 by an advanced control arithmetic operation. An input/output data storage part 12 stores the input variables of a polynomial model as input result data and the manipulated variables MV2 as output result data. An execution instruction part 13 instructs the execution of inductive analysis in a preliminarily specified timing. An inductive analysis part 14 executes inductive analysis by using the nearest data set stored in the input/output data storage part 12, and calculates the coefficient of the polynomial model. A lower side control arithmetic part 20 calculates manipulated variables MV1 by the polynomial model. A manipulated variable output part 22 adopts the manipulated variables MV2 as manipulated variables MV in a control cycle in which the manipulated variables MV2 are set, and adopts the manipulated variables MV1 as the manipulated variables MV in a control cycle in which the manipulated variables MV2 are not set.

Description

  The present invention relates to a control device and a control method for realizing a backup function that compensates for stoppage of higher-order control such as advanced control by lower-order control.

  Conventionally, advanced control (for example, Advanced Process Control used in a large-scale plant) based on a technique such as multivariable model predictive control is known. Multivariable model predictive control is control for optimal operation and economic operation by modeling the relationship between multivariables such as control variables and manipulated variables and giving mutual relationships such as constraint conditions and economic efficiency. . The upper side of such multivariable model predictive control (see Patent Document 1) or multipoint temperature equalization control (see Patent Document 2), the upper side of target value shaping control (see Patent Document 3), and the opening of a simulation base As a technique for realizing hybrid control of altitude control such as loop control (see Patent Document 4) and PID control, there is a technique of setting the output of the higher-level altitude control as the set value SP of the lower-level PID control. Conceivable.

  FIG. 6 is a block diagram showing the configuration of a hybrid control system combining altitude control and PID control. In many cases, the advanced control arithmetic device 100 employs a multi-input algorithm that takes in a plurality of state observation values, and through simulations and optimization calculations, values to be maintained for state quantities such as temperature and pressure (targets) Value). Such an advanced control arithmetic device 100 is realized by high-spec hardware such as a PC. Then, in order to maintain a single variable state quantity such as temperature and pressure at the target value, the temperature controller 101 uses the target value calculated by the altitude control arithmetic device 100 as the set value SP (see FIG. 6). In the example, an operation amount MV, which is an instruction value to a power regulator or a valve of a semiconductor manufacturing apparatus) is calculated by PID control calculation.

  Therefore, even if the hardware of the host-side advanced control arithmetic device 100 fails, the PID control is continued while the set value SP is maintained, so that the state quantities such as temperature and pressure do not run away. That is, the control method is a backup structure (distributed structure) that compensates for abnormalities in advanced control with PID control.

  On the other hand, there is a case where the advanced control arithmetic device 100 directly calculates the operation amount MV (FIG. 7). In this case, the control system has a structure (centralized structure) that should be called a centralized system using high-spec hardware.

Japanese Patent No. 3758862 Japanese Patent No. 4358673 Japanese Patent No. 4639821 JP-A-10-285804

  In many cases, the advanced control employs a complex algorithm with multiple inputs that takes in a plurality of state observation values, and therefore the amount of computation is also large. Further, since the advanced control with a large amount of calculation (calculation and communication becomes intensive), the risk that the calculation function and the communication function are stopped cannot be ignored. Therefore, the apparatus has a centralized structure as shown in FIG. Implementing is an instrument with high risk.

  When the calculation function and communication function of the advanced control arithmetic device 100 are stopped in the centralized structure as shown in FIG. 7, for example, state quantity control (temperature control or pressure) in the manufacturing process of semiconductor, petrochemical, steel, etc. In the case of control, etc., a big problem such as production stoppage occurs. In addition, in the air conditioning control of buildings including energy management such as heat source optimization, problems such as excess power demand arise.

  If the distributed structure as shown in FIG. 6 is used, the abnormality of the advanced control arithmetic device 100 can be compensated by the lower temperature controller 101. However, as shown in FIG. In the case of directly calculating the MV, it is difficult to use the commercially available temperature controller 101 as it is.

  The present invention has been made in order to solve the above-described problem, and a production stoppage that can occur when an arithmetic function or a communication function of an advanced control arithmetic device in which advanced control for directly calculating an operation amount MV is implemented, It is an object of the present invention to provide a control device and a control method that can reduce risks such as excess power demand and control performance degradation.

  The control device of the present invention includes a high-order control arithmetic device that calculates an operation amount by advanced control arithmetic, and a low-order control arithmetic device that calculates an operation amount by a polynomial model that can perform advanced control. The arithmetic device is a high-order side control arithmetic means for calculating an operation amount MV2 from an input variable by advanced control arithmetic at every predetermined control period, and the low-order side control arithmetic value MV2 calculated by the high-order side control arithmetic means. The input result data and the output results are obtained by using the upper operation amount setting means to be set in the arithmetic device, the input variables of the polynomial model as input result data, and the operation amount MV2 calculated by the upper control operation means as output result data. Input / output data storage means for storing data as a data set, and execution instruction means for instructing to perform inductive analysis at a predetermined timing In response to an instruction from the execution instruction unit, an inductive analysis is performed using a data set collected in the most recent collection period among the data sets stored in the input / output data storage unit, and the polynomial model Recursive analysis means for calculating the coefficient of the above, and coefficient output means for setting the coefficient of the polynomial model calculated by the recursive analysis means in the low-order control arithmetic device, the low-order control arithmetic device is The lower-order control calculation means for calculating the operation amount MV1 from the input variable by the polynomial model for each predetermined control cycle, and the operation amount MV to be output to the control target in the control cycle in which the operation amount MV2 is set. In the control cycle in which the operation amount MV2 is adopted and the operation amount MV2 is not set, the operation amount MV1 is adopted as the operation amount MV to be output to the control target, and the operation amount MV It is characterized in that it has a manipulated variable output means for outputting the control target.

Further, in one configuration example of the control device of the present invention, the higher-order control arithmetic device is a computer that can execute more complex arithmetic than the lower-order control arithmetic device, and the lower-order control arithmetic device is the upper control arithmetic device. It is a control device designed to operate more stably than the side control arithmetic unit, and the advanced control arithmetic is an operation at a level that cannot be processed by the lower side control arithmetic unit.
Further, in one configuration example of the control device of the present invention, the control target has a property that a series of control steps are repeated, and the predetermined timing is a timing at which the series of control steps are completed. .
In one configuration example of the control device of the present invention, the control target is a state quantity in a manufacturing process, and the predetermined timing is a timing at which a series of control steps of the manufacturing process is completed.
Moreover, in one structural example of the control apparatus of this invention, the said control object is a state quantity in air-conditioning control, and the said pre-determined timing is a timing when the series of control process used as the execution unit of energy management was completed. is there.
Further, in one configuration example of the control device of the present invention, the formula of the polynomial model is a formula that realizes PID control calculation, and the lower-level control calculation means functions as PID control calculation means.

  Further, the control method of the present invention includes a high-order control calculation step in which the high-order control calculation means of the high-order control calculation device calculates the manipulated variable MV2 from the input variable by the advanced control calculation for each predetermined control cycle; An upper operation amount setting means for setting the operation amount MV2 calculated in the upper control calculation step in the lower control operation device; The input / output data storage means sets the input variable of the polynomial model as input result data, the operation amount MV2 calculated in the higher-level control calculation step as output result data, and sets the input result data and output result data as a data set. As the input / output data storage step and the execution instruction means of the higher-level control calculation means perform inductive analysis at a predetermined timing. And an inductive analysis means of the higher-level control calculation means, in response to an instruction by the execution instruction step, the latest collection of the data sets stored in the input / output data storage means An inductive analysis step of calculating a coefficient of the polynomial model by performing an inductive analysis using a data set collected in a period, and a coefficient output means of the higher-order control arithmetic means in the inductive analysis step A coefficient output step for setting the calculated coefficient of the polynomial model in the lower-order control arithmetic device, and lower-order control arithmetic means of the lower-order control arithmetic device predetermine the manipulated variable MV1 from the input variable by the polynomial model. The operation amount MV2 is set in the lower-order side control operation step for calculating every control cycle and the operation amount output means of the lower-order control operation device. In the control cycle, the operation amount MV2 is employed as the operation amount MV to be output to the control target, and in the control cycle in which the operation amount MV2 is not set, the operation amount MV1 is employed as the operation amount MV to be output to the control target. And an operation amount output step for outputting the operation amount MV to the control target.

  According to the present invention, a host control arithmetic unit that calculates an operation amount by advanced control computation includes a host control calculation unit, a host operation amount setting unit, an input / output data storage unit, an inductive analysis unit, and a coefficient output unit. Advanced control algorithm for directly calculating the manipulated variable MV by providing lower-order control computation means and manipulated variable output means in the lower-order control computation device that computes the manipulated variable using a polynomial model that can perform advanced control Can be reduced in risks such as production stoppage, excess power demand, and control performance degradation that can occur when the computation function and communication function of the host-side control computation device in which is installed. Further, in the present invention, polynomial approximation can be easily performed by limiting the variation range of the input variable using the latest result data. Further, according to the present invention, it is possible to cope with not only a hardware failure of the host control arithmetic device but also a planned maintenance for correcting the parameters of the advanced control.

  Further, in the present invention, the control target is of a nature in which a series of control steps are repeated, and the recursive analysis is executed at this timing with the timing at which the series of control steps are completed as a predetermined timing. Since an instruction is issued, polynomial approximation can be easily performed using the latest performance data.

  Further, in the present invention, the lower-level control calculation means of the lower-level control calculation device can be caused to function as the PID control calculation means by using a mathematical expression that realizes the PID control calculation as the mathematical expression of the polynomial model.

It is a block diagram which shows the structure of the control system which concerns on the 1st Embodiment of this invention. It is a block diagram which shows the detailed structure of the control apparatus which concerns on the 1st Embodiment of this invention. It is a flowchart explaining operation | movement of the advanced control arithmetic unit which concerns on the 1st Embodiment of this invention. It is a flowchart explaining operation | movement of the exclusive controller which concerns on the 1st Embodiment of this invention. It is a figure which shows an example of input performance data and output performance data. It is a block diagram which shows the structure of the hybrid control system which combined altitude control and PID control. It is a block diagram which shows the structure of the control system of single advanced control.

[Principle of Invention 1]
As hardware that can perform advanced control by distributed arrangement, there is a programmable logic controller (PLC) that can be used as a dedicated controller. When such a dedicated controller is used, a complex algorithm with multiple inputs cannot be implemented. Therefore, it is assumed that a compact polynomial model obtained by an inductive analysis method or the like is mounted on a dedicated controller and arranged on the lower side (backup side) of the advanced control arithmetic device.

When using a dedicated controller, it is reasonable to think that it is difficult to create a polynomial that covers the entire variable range of an input variable that has multiple inputs. The reason is as follows (A) and (B).
(A) It is difficult and inefficient to generate data to be used in an inductive analysis method by a simulation that covers all input variable ranges in advance.
(B) If the input / output relationship of the strongly nonlinear system is complicated, it is difficult to obtain a polynomial that can replace (approximate) the entire variable range with a practical level of accuracy.

  Therefore, the inventor creates a simple polynomial model using the latest data (preliminary high-level control data) specified in advance, and uses a dedicated controller for distributed arrangement as an emergency backup of the high-level control algorithm. I came up with periodically downloading the coefficients of the polynomial model. Such a configuration is more rational than creating a polynomial model that covers the entire variable range of the input variable. The reason is as follows (C) and (D).

(C) Not all of the many input variables fluctuate at the same pace, and it is reasonable to think that some input variables include those that fluctuate slower than the pace of regular download of polynomial coefficients. is there. Therefore, such an input variable has a substantially constant value in the latest specified range and is not substantially treated as a variable, so that polynomial approximation is easily performed.

(D) Many of the many input variables are not observed and measured over the entire range from the minimum value to the maximum value, and the range of variation is limited depending on the process of the manufacturing process. Therefore, since the nonlinearity between input and output can be alleviated by limiting the fluctuation range using the latest performance data, it is easy to perform polynomial approximation.

[Principle of Invention 2]
In the case of a manufacturing process for semiconductors, petrochemicals, steel, etc., it is preferable to define the latest performance data based on the repetitive steps of the manufacturing process. For example, in the case of temperature control, a single temperature increasing / decreasing operation is set as a series of control steps, or three times of temperature increasing / decreasing operations are set as a series of control steps. Is the timing to update. And external environmental factors (cooling water temperature, air pressure) such as utilities for supplying cooling water and compressed air to the manufacturing process, and internal environmental factors (maximum heat generation, maximum flow rate) due to heater deterioration and piping contamination Corresponds to “slow fluctuation”.

  In the case of air-conditioning control including energy management, it is preferable to define the latest performance data based on the energy management process corresponding to the outside air fluctuation cycle. For example, one day is set as a series of control steps, and three days are set as a series of control steps, and the timing at which these series of control steps are completed is set as the timing for updating the polynomial model. The seasonal variation element with small fluctuations in days corresponds to “slow fluctuations”.

  If the latest result data is defined as described above, the result data can be collected in such a way that the necessary input variable is observed and measured only in the required fluctuation range. That is, it is possible to collect performance data that can create a polynomial model for reproducing the most recent series of control processes. According to such a method for defining actual data, it is not necessary to store data other than the latest actual data (a large amount of past actual data).

[First Embodiment]
Hereinafter, embodiments of the present invention will be described with reference to the drawings. This embodiment shows an example of a framework with a broad concept. FIG. 1 is a block diagram showing a configuration of a control system according to the present embodiment. The control system of the present embodiment includes an advanced control arithmetic device 1 that is a higher-level control arithmetic device, a dedicated controller 2 that is a lower-level control arithmetic device, and a control object 3. In FIG. 1, a semiconductor manufacturing apparatus used in a semiconductor manufacturing process is taken as an example of the control target 3.

FIG. 2 is a block diagram showing a detailed configuration of the control device of the present embodiment. The control device according to the present embodiment includes the advanced control arithmetic device 1 and the dedicated controller 2 shown in FIG.
The advanced control calculation device 1 includes an upper control calculation unit 10 that calculates an operation amount MV2 from input variables by an advanced control calculation for each predetermined control cycle, and an operation amount output to be described later in the dedicated controller 2 in the dedicated controller 2. The input actual data and the output actual data are set as the upper operation amount setting unit 11 set in the unit, the input variable of the polynomial model as the input actual data, and the operation amount MV2 calculated by the upper control calculation unit 10 as the output actual data. In accordance with an instruction from the execution instruction unit 13 and an input / output data storage unit 12 that stores the data as a data set, an execution instruction unit 13 that instructs execution of inductive analysis at a predetermined timing, Perform inductive analysis using the data set collected in the most recent collection period among the data sets stored in the output data storage unit 12, and determine the relationship between input and output Having a recursive analysis unit 14 for calculating the coefficients of the to polynomial model, the coefficients of the calculated polynomial model, and a coefficient output section 15 for outputting to the coefficient setting unit to be described later in the dedicated controller 2.

  The dedicated controller 2 uses a polynomial model to calculate the operation amount MV1 from input variables for each predetermined control cycle, and a lower-side operation amount for setting the operation amount MV1 in an operation amount output unit described later. In the control cycle in which the setting unit 21 and the operation amount MV2 are set, the operation amount MV2 is adopted as the operation amount MV to be output to the control target 3, and in the control cycle in which the operation amount MV2 is not set, the control target 3 is set. The operation amount MV1 is adopted as the operation amount MV to be output, and the coefficient of the polynomial model is input from the operation amount output unit 22 that outputs the operation amount MV to the control target 3 and the coefficient output unit 15 in the advanced control arithmetic device 1. A coefficient setting unit 23 that sets the coefficient as a coefficient of a polynomial model of the lower-level control calculation unit 20.

  Each of the advanced control arithmetic device 1 and the dedicated controller 2 can be realized by a computer having a CPU, a storage device, and an interface, and a program for controlling these hardware resources. The CPU of each device executes the following processing according to a program stored in the storage device.

  As an operation of the control device of the present embodiment, first, an operation at a stage of creating an initial model of a polynomial model will be described with reference to FIGS. FIG. 3 is a flowchart for explaining the operation of the advanced control arithmetic device 1, and FIG. 4 is a flowchart for explaining the operation of the dedicated controller 2.

The higher-level control calculation unit 10 of the advanced control calculation device 1 calculates an operation amount MV2 that is assumed to be output to the controlled object 3 by the following control calculation formula (step S100 in FIG. 3).
MV2 = f (X1, X2, X3, X4) (1)

  In Expression (1), f is a function defined in advance by an advanced and complex simulation-based operation amount calculation algorithm (advanced control algorithm), and X1, X2, X3, and X4 are input variables (temperatures measured by the temperature sensor). Such as a state quantity measurement value). In this embodiment, it is assumed that there are four input variables, but the number is not limited to four.

  Among the input variables X1, X2, X3, and X4, those that change over a particularly long time such as input variables that change according to the season are included. Such an input variable that changes over a long period of time is assumed to have a substantially constant value in the data set when the recursive analysis is automatically executed. The input variables X1, X2, X3, and X4 include those in which only a part of the entire range from the minimum value to the maximum value that can be assumed is measured. Such an input variable is assumed to have a substantially linear effect on the manipulated variable MV2 in the data set when the inductive analysis is automatically executed.

  Subsequently, the upper operation amount setting unit 11 of the advanced control arithmetic device 1 sets the operation amount MV2 calculated by the upper control arithmetic operation unit 10 in the operation amount output unit 22 in the dedicated controller 2 (step S101 in FIG. 3). ).

The input / output data storage unit 12 of the advanced control arithmetic unit 1 uses the input variables X1, X2, X3, and X4 of the polynomial model observed in the current control cycle as input result data, and performs higher-level control computation in the current control cycle. The operation amount MV2 calculated by the unit 10 is stored as output record data, and the input record data and output record data are sequentially stored as a data set Ai (step S102 in FIG. 3).
Ai = (X1i, X2i, X3i, X4i, MV2i) (2)

  X1i, X2i, X3i, and X4i are the input variables X1, X2, X3, X4, and MV2i in the current control cycle i, respectively, and the manipulated variable MV2 calculated in the current control cycle i. FIG. 5 is a diagram showing an example of input record data and output record data. In FIG. 5, X1 and X4 are variables that are constant values or substantially constant values, and X3 corresponds to a variable in which only a part of the entire range from the minimum value to the maximum value is measured. FIG. 5 shows a fictitious data example in which the data image actually obtained in the manufacturing process is simplified for easy understanding.

Here, since the polynomial model for calculating the manipulated variable MV1 is in an undetermined stage (NO in step S200 in FIG. 4), the lower control operation unit 20 and the lower manipulated variable setting unit 21 of the dedicated controller 2 execute processing. do not do.
The operation amount output unit 22 of the dedicated controller 2 sets the operation amount MV2 from the higher-order operation amount setting unit 11 of the advanced control arithmetic device 1 in the current control cycle (YES in step S203 in FIG. 4), so The operation amount MV2 is adopted as the operation amount MV to be output, and this operation amount MV is output to the control object 3 (the actual output destination is a power regulator, a valve, etc.) (step S204 in FIG. 4).

  In the advanced control arithmetic unit 1, the processes in steps S <b> 100 to S <b> 102 are repeatedly executed for each control cycle until the series of control steps is completed (YES in step S <b> 106 in FIG. 3). The processes of S203 and S204 are repeatedly executed for each control cycle until the polynomial model is confirmed (YES in step S200 in FIG. 4) or until a series of control steps is completed (YES in step S208 in FIG. 4). The control cycle of the advanced control arithmetic device 1 and the control cycle of the dedicated controller 2 may be the same.

  The timing at which a series of control steps are completed is defined in advance, so that the timing at which the data set Ai (i = 1 to n) of the series of control steps is accumulated in the input / output data storage unit 12 of the advanced control arithmetic device 1 Then, an instruction is sent from the execution instructing unit 13 to execute the inductive analysis (YES in step S103 in FIG. 3).

  For example, one temperature raising / lowering operation of the semiconductor manufacturing process is a series of control steps for collecting the data set Ai, and when the manipulated variable MV2 rises and returns to a low value below a predetermined threshold as shown in FIG. However, when it is defined that it is the timing when the control process is completed, the execution instructing unit 13 instructs to perform the inductive analysis when the operation amount MV2 returns to a low value that is equal to or lower than the threshold value after increasing. Put out. In addition, in the case of a scheduled manufacturing process or air conditioning control, the timing for completing a series of control steps for collecting the data set Ai may be designated in advance by time. Thereby, the execution instruction | indication part 13 can issue an instruction | indication to perform inductive analysis, when the designated time comes.

In response to an instruction from the execution instruction unit 13, the inductive analysis unit 14 of the advanced control arithmetic device 1 stores the most recent data set Ai (i = 1 to n) (that is, the most recent data set 12 stored in the input / output data storage unit 12). The coefficients R1_a, R2_a, R3_a, R4_a, R5_a of the following polynomial models are calculated by inductive analysis (for example, SVR (Support Vector Regression) or multiple regression analysis) using the input / output data) (FIG. 3 step S104).
MV2 = R1_a.X1 + R2_a.X2 + R3_a.X3 + R4_a.X4 + R5_a
... (3)

  In the present embodiment, the equation of the polynomial model is a first order polynomial, but it may be a second or higher order polynomial. In this embodiment, the input variable of the polynomial model and the input variable of the advanced control algorithm are the same, but the input variable of the polynomial model and the advanced control algorithm may be different. The input / output data storage unit 12 may store the input variables of the polynomial model.

  The coefficient output unit 15 of the advanced control arithmetic device 1 outputs the calculated coefficients R1_a, R2_a, R3_a, R4_a, R5_a to the coefficient setting unit 23 in the dedicated controller 2 as coefficients of the initial model of the polynomial model (FIG. 3 step S105).

When the coefficient setting unit 23 of the dedicated controller 2 outputs the coefficients R1_a, R2_a, R3_a, R4_a, and R5_a of the polynomial model from the coefficient output unit 15 of the advanced control arithmetic device 1 (YES in step S206 in FIG. 4). These coefficients R1_a, R2_a, R3_a, R4_a, and R5_a are set as coefficients of the polynomial model of the lower-level control calculation unit 20 (step S207 in FIG. 4).
MV1 = R1_a.X1 + R2_a.X2 + R3_a.X3 + R4_a.X4 + R5_a
... (4)

  In addition, the polynomial itself of Expression (4) is mounted in advance in the lower-level control calculation unit 20. A polynomial model for calculating the manipulated variable MV1 is determined by setting the coefficients R1_a, R2_a, R3_a, R4_a, and R5_a.

Next, as the operation of the control device of the present embodiment, a backup operation for compensating for the stop of the advanced control arithmetic device 1 by the control by the dedicated controller 2 will be described.
The operations (steps S100 and S101 in FIG. 3) of the upper control arithmetic unit 10 and the upper operation amount setting unit 11 of the advanced control arithmetic device 1 are as described above. The input / output data storage unit 12 of the advanced control arithmetic unit 1 uses the input variables X1, X2, X3, and X4 of the polynomial model observed in the current control cycle as input result data, and performs higher-level control computation in the current control cycle. The operation amount MV2 calculated by the unit 10 is stored as output record data, and the input record data and output record data are sequentially stored as a data set Bi (step S102 in FIG. 3).
Bi = (X1i, X2i, X3i, X4i, MV2i) (5)

Here, since the polynomial model is determined by the above processing (YES in step S200 in FIG. 4), the lower-level control calculation unit 20 of the dedicated controller 2 assumes the operation amount MV1 that is output to the control target 3. Is calculated by the polynomial model represented by the equation (4) (step S201 in FIG. 4).
The lower operation amount setting unit 21 of the dedicated controller 2 sets the operation amount MV1 calculated by the lower control calculation unit 20 in the operation amount output unit 22 (step S202 in FIG. 4).

  The operation amount output unit 22 of the dedicated controller 2 sets the operation amount MV2 from the higher-order operation amount setting unit 11 of the advanced control arithmetic device 1 in the current control cycle (YES in step S203 in FIG. 4), so The operation amount MV2 is adopted as the operation amount MV to be output, and this operation amount MV is output to the control object 3 (the actual output destination is a power regulator, a valve, etc.) (step S204 in FIG. 4).

  In the advanced control arithmetic unit 1, the processes in steps S100 to S102 described above are repeatedly executed at every control cycle until the series of control steps is completed (YES in step S106 in FIG. 3). In the dedicated controller 2, steps S200 to S200 are performed. The process of S204 is repeatedly executed for each control cycle until a series of control steps is completed (YES in step S208 in FIG. 4). However, when the advanced control arithmetic unit 1 stops for some reason on the way, it switches to the following processing.

  If the altitude control arithmetic unit 1 stops for some reason as shown in FIG. 5, the upper control arithmetic unit 10, the upper operation amount setting unit 11, and the input / output data storage unit 12 cannot execute processing. .

  The operations (steps S201 and S202 in FIG. 4) of the lower control operation unit 20 and the lower operation amount setting unit 21 of the dedicated controller 2 are as described above. The operation amount output unit 22 of the dedicated controller 2 has not been set with the operation amount MV2 from the higher-order operation amount setting unit 11 of the advanced control arithmetic device 1 in the current control cycle (NO in step S203 in FIG. 4), so that the control target 3 The operation amount MV1 set from the lower operation amount setting unit 21 is adopted as the operation amount MV to be output to the control object 3, and this operation amount MV is output to the control object 3 (step S205 in FIG. 4).

  When advanced control arithmetic unit 1 is stopped, the processes in steps S200 to S203 and S205 described above are repeatedly executed for each control cycle until a series of control steps is completed (YES in step S208 in FIG. 4). Even at the timing when a series of control steps are completed, the advanced control arithmetic unit 1 is stopped, and the data set Bi (i = 1 to m) is not accumulated for the series of control steps. The execution instructing unit 13 of the device 1 does not issue an inductive analysis execution instruction.

Next, as an operation of the control device of the present embodiment, an operation in which the dedicated controller 2 backs up the advanced control by the advanced control arithmetic device 1 and updates the polynomial model will be described.
The operations (steps S100 and S101 in FIG. 3) of the upper control arithmetic unit 10 and the upper operation amount setting unit 11 of the advanced control arithmetic device 1 are as described above. The input / output data storage unit 12 of the advanced control arithmetic unit 1 uses the input variables X1, X2, X3, and X4 of the polynomial model observed in the current control cycle as input result data, and performs higher-level control computation in the current control cycle. The operation amount MV2 calculated by the unit 10 is stored as output record data, and the input record data and output record data are sequentially stored as a data set Ci (step S102 in FIG. 3).
Ci = (X1i, X2i, X3i, X4i, MV2i) (6)

The operations (steps S201 and S202 in FIG. 4) of the lower control operation unit 20 and the lower operation amount setting unit 21 of the dedicated controller 2 are as described above.
The operation amount output unit 22 of the dedicated controller 2 sets the operation amount MV2 from the higher-order operation amount setting unit 11 of the advanced control arithmetic device 1 in the current control cycle (YES in step S203 in FIG. 4), so The operation amount MV2 is adopted as the operation amount MV to be output, and this operation amount MV is output to the control object 3 (step S204 in FIG. 4).

  In the advanced control arithmetic unit 1, the processes in steps S100 to S102 are repeatedly executed for each control period until the series of control steps is completed (YES in step S106 in FIG. 3). In the dedicated controller 2, the processes in steps S200 to S204 are performed. The process is repeated for each control period until a series of control steps is completed (YES in step S208 in FIG. 4).

  The timing at which a series of control steps are completed is defined in advance, so that the data set Ci (i = 1 to n) of the series of control steps is accumulated in the input / output data storage unit 12 of the advanced control arithmetic unit 1 Then, an instruction is sent from the execution instructing unit 13 to execute the inductive analysis (YES in step S103 in FIG. 3).

  In response to an instruction from the execution instruction unit 13, the inductive analysis unit 14 of the advanced control arithmetic device 1 stores the most recent data set Ci (i = 1 to n) (that is, the most recent data set 12 stored in the input / output data storage unit 12). The coefficients R1_a, R2_a, R3_a, R4_a, R5_a of the polynomial model as shown in the equation (3) are calculated by inductive analysis using (input / output data) (step S104 in FIG. 3). The coefficient output unit 15 of the advanced control arithmetic device 1 outputs the calculated coefficients R1_a, R2_a, R3_a, R4_a, and R5_a to the coefficient setting unit 23 in the dedicated controller 2 as coefficients after updating the polynomial model (FIG. 3 step S105).

  When the coefficient setting unit 23 of the dedicated controller 2 outputs the coefficients R1_a, R2_a, R3_a, R4_a, and R5_a of the polynomial model from the coefficient output unit 15 of the advanced control arithmetic device 1 (YES in step S206 in FIG. 4). These coefficients R1_a, R2_a, R3_a, R4_a, and R5_a are set as coefficients of the polynomial model of the lower-level control calculation unit 20 (step S207 in FIG. 4). The polynomial model for calculating the manipulated variable MV1 is updated by setting the coefficients R1_a, R2_a, R3_a, R4_a, and R5_a.

  In this embodiment, when data sets Ai, Bi, Ci,... Of a series of control processes are obtained, each data set Ai, Bi, Ci is treated as the latest input / output data. I made it. However, the present invention is not limited to such a configuration. In updating the polynomial model, the data set Bi and the data set Ai collected before the data set Bi are collected at the timing when the data set Bi is accumulated, and used as the latest input / output data. At the timing when the data set Ci is accumulated, the data set Ci and the data set Bi collected before the data set Ci may be integrated into the latest input / output data. That is, in a period in which the control characteristics do not change significantly, the period that can be handled as the latest input / output data can be freely extended.

  For temperature control of manufacturing processes such as semiconductors, petrochemicals, and steels, it is not necessary to perform a single temperature raising / lowering operation as a series of control steps for collecting data sets and updating a polynomial model. The temperature operation may be a series of control steps for collecting a data set and updating a polynomial model. In the case of air conditioning control including energy management, it is not necessary to perform the control for each day as a series of control steps for collecting the data set and updating the polynomial model, and for the three-day control to collect the data set. Thus, a series of control steps for updating the polynomial model may be performed.

  In the present embodiment, the calculation function and communication function of the advanced control arithmetic device 1 are stopped due to a hardware failure or the like, and the operation amount MV2 is not set. Even when the advanced control arithmetic device 1 is stopped, the backup function by the dedicated controller 2 can be used. In this case, the operation amount output unit 22 may be provided with a switching function so that it can be switched to the backup mode in advance during maintenance.

  Moreover, each component mounted in the advanced control arithmetic device 1 shown in FIG. 2 is not necessarily mounted on one computer. For example, since the input / output data can be used for purposes other than the present invention, the input / output data storage unit 12 may be provided as a database system in a computer separate from the inductive analysis unit 14 and the like.

[Second Embodiment]
Next, a second embodiment of the present invention will be described. This embodiment shows a special form of the present invention. Also in the present embodiment, since the configuration of the control device is the same as that of the first embodiment, description will be made using the reference numerals in FIGS.

In the present embodiment, the polynomial model formula set in advance in the lower-level control calculation unit 20 of the dedicated controller 2 realizes the PID control calculation. That is, the lower-level control calculation unit 20 functions as a PID control calculation unit and calculates the operation amount MV1 (step S201 in FIG. 4). The polynomial model at this time is shown below.
MV1 = Kp · Er + (Kp / Ti) · ΣEr + (Kp · Td) · ΔEr
= Kp · SP-Kp · PV + (Kp / Ti) · ΣEr + (Kp · Td) · ΔEr
... (7)

  In the equation (7), SP is a set value set by the operator of the control device (for example, in the case of temperature control in the semiconductor manufacturing process, the temperature setting value in the furnace of the semiconductor manufacturing device), and PV is measured by a sensor (not shown). Control amount (for example, furnace temperature). Kp is a proportional gain, Ti is an integration time, and Td is a differentiation time. Er is a control deviation (difference SP-PV between the set value SP and the control amount PV), ΣEr is an integral value of the control deviation, and ΔEr is a differential value of the control deviation. Each of Kp, −Kp, (Kp / Ti), and (Kp · Td) is a coefficient of a polynomial model. Each of SP, PV, ΣEr, and ΔEr becomes an input variable of the polynomial model.

  In a series of control steps for collecting a data set and updating a polynomial model, the set value SP becomes a constant value, and the control amount PV, the integral value ΣEr, and the differential value ΔEr change only within a limited range. To do. In that case, even if the higher-level advanced control algorithm is an algorithm corresponding to a strongly nonlinear system, the lower-level side can be backed up appropriately as a PID control that is a linear control method.

  The input / output data storage unit 12 of the advanced control arithmetic unit 1 inputs the input variables SP, PV, ΣEr, and ΔEr of the polynomial model observed in the current control cycle as input result data, and the higher-level control arithmetic unit in the current control cycle. The operation amount MV2 calculated by 10 may be stored as output record data, and the input record data and output record data may be stored as a data set (step S102 in FIG. 3).

The inductive analysis unit 14 of the advanced control arithmetic unit 1 uses a data set (that is, the latest input / output data) stored in the input / output data storage unit 12 to perform a polynomial model such as the following equation by inductive analysis. The coefficients Kp, −Kp, (Kp / Ti), and (Kp · Td) may be calculated (step S104 in FIG. 3).
MV2 = Kp · Er + (Kp / Ti) · ΣEr + (Kp · Td) · ΔEr
= Kp · SP-Kp · PV + (Kp / Ti) · ΣEr + (Kp · Td) · ΔEr
... (8)

  As an inductive analysis method at this time, for example, the technique disclosed in Japanese Patent No. 4223894 can be applied. In this way, the coefficients Kp, −Kp, (Kp / Ti), and (Kp · Td) can be calculated and the polynomial model can be determined so that the control operation based on PID becomes an operation that can perform advanced control.

  The other configuration of the control device and the flow of processing are as described in the first embodiment. Note that the coefficient of the initial model of the polynomial model may be determined by a generally known PID adjustment method without being calculated by the inductive analysis unit 14.

  The present invention can be applied to a technique for supplementing stoppage of higher-order control such as advanced control with lower-order control.

  DESCRIPTION OF SYMBOLS 1 ... Advanced control arithmetic device, 2 ... Dedicated controller, 3 ... Control object, 10 ... Upper side control arithmetic part, 11 ... Upper side operation amount setting part, 12 ... Input / output data storage part, 13 ... Execution instruction part, 14 ... Inductive analysis unit, 15 ... coefficient output unit, 20 ... lower control operation unit, 21 ... lower operation amount setting unit, 22 ... operation amount output unit, 23 ... coefficient setting unit.

Claims (12)

A high-order control arithmetic device that calculates an operation amount by advanced control arithmetic, and a low-order control arithmetic device that calculates an operation amount by a polynomial model that can perform advanced control,
The upper control arithmetic unit is
High-order side control calculation means for calculating the operation amount MV2 from input variables by an advanced control calculation for each predetermined control cycle;
Upper operation amount setting means for setting the operation amount MV2 calculated by the upper control calculation device in the lower control operation device;
Input / output data storage means for storing the input variables of the polynomial model as input result data, the operation amount MV2 calculated by the higher-level control arithmetic means as output result data, and storing the input result data and output result data as a data set. When,
Execution instruction means for instructing to perform an inductive analysis at a predetermined timing;
In response to an instruction from the execution instruction unit, an inductive analysis is performed using a data set collected in the most recent collection period among the data sets stored in the input / output data storage unit, and the polynomial model An inductive analysis means for calculating a coefficient;
Coefficient output means for setting the coefficient of the polynomial model calculated by the inductive analysis means in the lower-level control arithmetic unit;
The lower-level control arithmetic unit is
Low-order side control calculation means for calculating the manipulated variable MV1 from the input variable by the polynomial model for each predetermined control cycle;
In the control cycle in which the operation amount MV2 is set, the operation amount MV2 is adopted as the operation amount MV to be output to the control target, and in the control cycle in which the operation amount MV2 is not set, the operation amount MV to be output to the control target. A control device characterized by comprising an operation amount MV1 and an operation amount output means for outputting the operation amount MV to a controlled object. The control device according to claim 1,
The upper control arithmetic device is a computer capable of performing more complex arithmetic than the lower control arithmetic device,
The lower control arithmetic unit is a control device designed to operate more stably than the upper control arithmetic unit,
The control apparatus according to claim 1, wherein the advanced control calculation is a calculation at a level that cannot be processed by the lower-level control calculation apparatus. The control device according to claim 1 or 2,
The control object is of a nature in which a series of control steps are repeated,
The control device characterized in that the predetermined timing is a timing at which the series of control steps are completed. The control device according to claim 3,
The control object is a state quantity in the manufacturing process,
The control device characterized in that the predetermined timing is a timing at which a series of control steps of a manufacturing process is completed. The control device according to claim 3,
The control object is a state quantity in air conditioning control,
The control apparatus according to claim 1, wherein the predetermined timing is a timing at which a series of control steps as an energy management unit is completed. The control device according to any one of claims 1 to 5,
2. The control apparatus according to claim 1, wherein the formula of the polynomial model is a formula for realizing a PID control calculation, and the lower-level control calculation means functions as a PID control calculation means. A high-order side control calculation means for calculating a manipulated variable MV2 from an input variable by an advanced control calculation for each predetermined control cycle;
An upper operation amount setting means in which the upper operation amount setting means of the upper control calculation means sets the operation amount MV2 calculated in the upper control calculation step in the lower control arithmetic device;
The input / output data storage means of the upper control calculation means uses the input variables of the polynomial model as input record data, and the operation amount MV2 calculated in the higher control calculation step as output record data. An input / output data storage step for storing data as a data set;
An execution instruction step for instructing the execution instruction means of the higher-order side control calculation means to perform an inductive analysis at a predetermined timing;
The recursive analysis means of the higher-level control calculation means uses a data set collected in the most recent collection period among the data sets stored in the input / output data storage means in response to an instruction by the execution instruction step. Performing an inductive analysis and calculating a coefficient of the polynomial model;
A coefficient output step of setting the coefficient of the polynomial model calculated in the recursive analysis step in the lower-level control calculation device;
A low-order side control arithmetic step in which the low-order side control arithmetic means of the low-order side control arithmetic device calculates the manipulated variable MV1 from the input variable by the polynomial model for each predetermined control cycle;
Control in which the operation amount output means of the lower-level control arithmetic unit adopts the operation amount MV2 as the operation amount MV to be output to the control target in the control cycle in which the operation amount MV2 is set, and the operation amount MV2 is not set. A control method comprising: an operation amount MV1 as an operation amount MV to be output to a control target in a cycle; and an operation amount output step of outputting the operation amount MV to the control target. The control method according to claim 7,
The upper control arithmetic device is a computer capable of performing more complex arithmetic than the lower control arithmetic device,
The lower control arithmetic unit is a control device designed to operate more stably than the upper control arithmetic unit,
The control method according to claim 1, wherein the advanced control computation is a computation at a level that cannot be processed by the lower-level control computation device. The control method according to claim 7 or 8,
The control object is of a nature in which a series of control steps are repeated,
The control method according to claim 1, wherein the predetermined timing is a timing at which the series of control steps are completed. The control method according to claim 9, wherein
The control object is a state quantity in the manufacturing process,
The control method according to claim 1, wherein the predetermined timing is a timing at which a series of control steps of a manufacturing process is completed. The control method according to claim 9, wherein
The control object is a state quantity in air conditioning control,
The control method according to claim 1, wherein the predetermined timing is a timing at which a series of control steps as an energy management execution unit is completed. The control method according to any one of claims 7 to 11,
The formula of the polynomial model is a formula that realizes PID control calculation, and the lower-level control calculation step functions as a PID control calculation step.

Images