JP2007304767A - Control system and control method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To estimate a model of a controlled object and parameters without loading field equipment, and automatically readjust the model and parameters with secular changes in the controlled equipment. <P>SOLUTION: A control system comprises a system decision apparatus 1 for selecting a model of a controlled system including the controlled object 4 and identifying model parameters, a controller 20 for controlling the controlled object 4 such that an observed value and a desired value of the controlled object 4 match according to the model and model parameters identified by the system decision apparatus 1, and a notification means for sending model change observation data to the system decision apparatus 1 when determining a control performance degradation. The system decision apparatus 1 reselects a model of the controlled object 4 and reidentifies model parameters when receiving the model change observation data. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a control system and a control method in process control.

Conventionally, in process control, an operation amount given to a control target and a control amount that is a response of the control target to the operation amount are collected, and the parameters of the model of the control target are estimated from the collected data. A self-tuning controller that determines the parameters of the PID controller based on the results has been proposed (see, for example, Patent Document 1).
Similarly, a process identification method for collecting data and identifying a model to be controlled has been proposed (see, for example, Patent Document 2).

JP-A-3-152601 JP-A-9-120303

  According to the techniques disclosed in Patent Documents 1 and 2, it is possible to identify a model to be controlled and its parameters, but this identification process requires a lot of data and calculation capability, and requires a large-scale memory and a high capacity. Since a performance CPU is required, it is not a good idea to install a model or parameter identification function in a field device such as a controller. In particular, as in the technique disclosed in Patent Document 2, if a model to be controlled is to be identified, a complicated calculation is required and a load is imposed on the site-side equipment. Therefore, the model identification function is installed in the site-side equipment. That was difficult in practice.

  In addition, the techniques disclosed in Patent Documents 1 and 2 have a problem that the secular change of the device to be controlled is not taken into consideration. If the device to be controlled changes over time, the model or parameter to be controlled must be readjusted. Despite the aging of the equipment, leaving the model and parameters unadjusted could significantly impair control performance. However, in the techniques disclosed in Patent Documents 1 and 2, in order to readjust the model and parameters, it is necessary to manually give an instruction for readjustment to the identification apparatus.

  The present invention has been made to solve the above-described problem, and can identify a model and parameters of a control target without applying a load to a field side device. It is another object of the present invention to provide a control system and a control method that can automatically readjust models and parameters.

The control system of the present invention includes a system modeling means for selecting a control system model including a control target, a system identification means for identifying a model parameter of the control system model, the model selected by the system modeling means, and the Based on the model parameter identified by the system identification means, when it is determined that the control calculation means for controlling the control target so that the observed value of the control target and the target set value match, and the control performance has deteriorated, A notification means for notifying the system modeling means and the system identification means of deterioration in control performance, and the system modeling means reselects the model of the control system when notified from the notification means, The system identification unit re-identifies the model parameter when receiving a notification from the notification unit. It is obtained by way.
In one configuration example of the control system of the present invention, the control calculation means determines a control gain based on the model selected by the system modeling means and the model parameter identified by the system identification means. is there.
In addition, one configuration example of the control system according to the present invention further includes a state estimation calculation for estimating an observation value of the control object based on the model selected by the system modeling unit and the model parameter identified by the system identification unit. And the control calculation means uses the estimated value estimated by the state estimation calculation means instead of the observed value of the control target.

The control method of the present invention is connected to a system modeling procedure for selecting a control system model including a controlled object by a computer, a system identification procedure for identifying a model parameter of the control system model by a computer, and the computer. The control calculation means controls the control target based on the model selected in the system modeling procedure and the model parameter identified in the system identification procedure so that the observed value of the control target matches the target set value. A control calculation procedure, and a notification procedure for notifying the computer of the deterioration of the control performance when it is determined that the control performance has deteriorated, and when the computer receives the notification of the deterioration of the control performance, A control system model is selected again, and the model parameters are identified again.
Further, one configuration example of the control method of the present invention further includes a gain determination procedure for determining the gain of the control arithmetic means based on the model selected in the system modeling procedure and the model parameter identified in the system identification procedure. Is provided.
In addition, one configuration example of the control method of the present invention further includes a state estimation calculation for estimating an observation value of the control object based on a model selected in the system modeling procedure and a model parameter identified in the system identification procedure. And a control calculation means that uses the estimated value estimated in the state estimation calculation procedure instead of the observed value of the controlled object.

  According to the present invention, since the model of the control system is selected by the system modeling means and the model parameter is identified by the system identification means, a load is not applied to the controller of the field side equipment. If a high-performance computer is used for the system modeling means and the system identification means, modeling and parameter estimation matching the real environment based on complex analysis can be realized, and control performance can be improved. Further, in the present invention, when the control performance is deteriorated, the notification means notifies the system modeling means and the system identification means of the deterioration of the control performance, and the system modeling means selects the control system model again, and the system identification means Since the model parameters are identified again, there is no need for the user to manually instruct readjustment or the user himself or herself to readjust the model or model parameters. In addition, even when a secular change occurs in the device to be controlled, it is possible to avoid deterioration in control performance.

  In the present invention, the gain of the control calculation unit is determined based on the model selected by the system modeling unit and the model parameter identified by the system identification unit. If the gain is determined by an external computer, the load on the controller can be reduced.

  In the present invention, the observation value of the controlled object is obtained by providing the state estimation calculating means for estimating the observed value of the controlled object based on the model selected by the system modeling means and the model parameter identified by the system identifying means. Even when measurement is impossible or when an observed value is buried in noise, control can be performed by calculating an estimated value of the observed value. If the state estimation calculation is performed by an external computer, the load on the controller can be reduced. On the other hand, if the state estimation calculation is performed by the controller, the controller and the state estimation calculation means are connected. Network load can be reduced.

[First Embodiment]
Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a control system according to the first embodiment of the present invention. The control system of FIG. 1 includes a system determination device 1 that estimates a model and parameters to be controlled and parameters of a controller of a field device, and a field side that controls a control object 4 such as a furnace via an actuator 3 such as a heater. It is comprised from the apparatus 2 and the network 5 which connects the system determination apparatus 1 and the field side apparatus 2. FIG.

The system determination device 1 is composed of a high-performance computer. Note that the system determination device 1 requires a server function to be connected to the field device 2 via the network 5, and the system determination device 1 is connected to a plurality of computers in order to perform analysis by various approaches during modeling. It may be divided.
The field side device 2 includes a controller 20 that controls the control target 4 via the actuator 3, and a user gives an instruction to the controller 20 or an instruction to the system determination device 1 via the network 5. You may have a user registration PC.

  FIG. 2 is a block diagram illustrating a configuration example of the system determination device 1. The system determination device 1 includes a receiving unit 10 that receives data transmitted from the controller 20, a system modeling unit 11 that selects a model of a control system including the control target 4, a system identification unit 12 that identifies model parameters, Transmitting means 13 for transmitting a program and data to the controller 20.

FIG. 3 is a block diagram illustrating a configuration example of the controller 20. The controller 20 includes a control calculation unit 22 that calculates a control output u, a control output unit 23 that outputs the control output u to an actuator 3 such as a heater, and a sensor output that receives an output of a sensor such as a thermocouple or a resistance temperature detector. Recognizing means 24, transmitting means 25 for transmitting data to system determining apparatus 1, and receiving means 26 for receiving a program and data transmitted from system determining apparatus 1. The sensor output recognizing means 24 constitutes a notifying means for notifying the system determining apparatus 1 of the deterioration of the control performance when it is determined that the control performance has deteriorated.
An example of the network 5 is the Internet, for example.

FIG. 4 is a diagram showing data transition in the control system of FIG. The sensor output recognition means 24 of the controller 20 receives an observed value (control amount) y that is a response of the control object 4 from a sensor (not shown) in the form of current or voltage. Then, the sensor output recognition unit 24 passes the observation value y to the transmission unit 25 and also passes the model change observation sensor output to the transmission unit 25 as model change observation data.
The transmission unit 25 transmits the observation value y and the model change observation data to the system determination device 1 via the network 5. Further, when receiving a control output u from the control calculation unit 22 as will be described later, the transmission unit 25 transmits the control output u to the system determination device 1.

The receiving means 10 of the system determining device 1 receives the observation value y, the control output u, and the model change observation data transmitted from the controller 20.
The system modeling unit 11 collects time series data including a set of the control output u and the observation value y when the model is updated. Then, the system modeling means 11 selects a mathematical model of a control system composed of the actuator 3, the controlled object 4, and the controller 20 that meets the specifications required by the user. Typical examples of the control system model include a transfer function model, a state space model, a polynomial model, and the like, and each model has various classes. For example, in the state space model, there are classes such as linear / non-linear, deterministic / probabilistic, and presence / absence of dead time.

  Based on the collected time series data, the system identification unit 12 identifies the parameters of the control system model by a known method. For example, for parameter identification of a state space model, input / output data is converted into Markov parameters or transfer functions, and a minimum realization is obtained (references “BLHo and REKalman, Relelungstechnik, 14-12, p.545-548, 1966 '', literature `` L. Silverman, IEEE T-AC, 16-6, p.554-567, 1971 '', literature `` J. Rissanen, SIAM. J. on Control, 9-3, p. 420-43O, 1971), and a method for obtaining a minimum internal description directly using input / output data (reference “B. Gopinath, Bell System Tech., 48-5, p. 1101-1113, 1969”), reference “MABudin” , IEEE T-SMC, 2-3, p.396-402, 1972), subspace method (see "PV Overschee and BDMoor, Automatica, 30-1, p.75-93, 1994") There is.

The transmission means of the system determination device 1 transmits the model data representing the mathematical expression of the control system selected by the system modeling means 11, the model parameter identified by the system identification means 12, and the control calculation program corresponding to the mathematical expression model via the network 5. Via the controller 20.
The receiving unit 26 of the controller 20 receives model data, model parameters, and a control calculation program transmitted from the system determination device 1 and passes them to the control calculation unit 22.

  The control calculation means 22 determines a control parameter (controller gain) from the model data and model parameters by a known method according to the control calculation program. As a method for deriving a controller gain and a method for deriving a control output, there are an LQ control law and the like in a deterministic system, and an LQG control law and the like in a stochastic system.

  Further, the control calculation means 22 calculates the control output u and passes it to the control output means 23 and the transmission means 25. In PID control, an ideal response is realized by adjusting three parameters of P (proportional band) / I (integration time) / D (differential time) which are gains of the controller. PID control is limited to a feedback (closed loop) system, but like a certain two-degree-of-freedom PID control system, the number of adjustment parameters is increased to increase the degree of freedom of the controller, and the expanded parameters are foreseeable information. It is also possible to add as a feed forward (open loop) element which means Well-known PID-controlled auto-tuning technology and self-tuning technology also perform well with limited models. In this system, the system determination device 1 determines a model, and auto-tuning technology and self-tuning technology are programmed for the model and provided to the controller. Therefore, effective auto-tuning technology and self-tuning for individual users are provided. Technology can be provided.

  In the model-based control, the control output u output from the controller 20 to the control target 4, the control output u transmitted from the controller 20 to the system determination device 1, and the observed value y output from the control target 4 to the controller 20 , Based on the data flow of the observation value y transmitted to the system determination device 1 via the controller 20. System modeling means 11 is necessary to grasp the control system in detail. The system modeling means 11 is intended to select a mathematical model whose input / output relationship matches that of the control system. The controller 20 automatically determines an optimal controller gain for a predetermined index of the control system determined by the system determination device 1.

  In the case of a control system in which a control system identification method using blind identification typified by independent component analysis can be used, the observation value y is not required without requiring the control output u transmitted from the controller 20 to the system identification means 12. Model parameters can be derived only by

Next, the process flow of the control system of the present embodiment will be described.
First, the system modeling means 11 of the system determination device 1 selects a mathematical model of a control system based on time series data collected from the controller 20 in advance. This model selection is performed when the operation is started after starting the control system and when model change observation data described later is output from the controller 20. 5 and 6 are flowcharts showing the operation of the system modeling means 11.

  The system modeling unit 11 determines whether there is a modeling request (step S100). The system modeling means 11 determines that there is a modeling request when starting the control system or receiving model change observation data from the controller 20. Subsequently, the system modeling unit 11 creates a control system model (step S101), determines an identification rule (step S102), and determines whether there is consistency with time-series data (step S103). When the created model and the identification rule are consistent with the time series data, the system modeling unit 11 determines the estimation rule (step S105) and determines whether the estimation request is satisfied (step S106). If it is determined in step S103 that the model and the identification rule are not consistent with the time-series data, the system modeling unit 11 determines whether another identification rule exists (step S104), and the other identification is performed. If there is a rule, the process returns to step S102, and if there is no other identification rule, the process returns to step S101 to create a model again.

When it is determined in step S106 that the estimated measurement satisfies the estimation request, the system modeling unit 11 determines a control law (step S108) and determines whether the control request is satisfied (step S109). If it is determined in step S106 that the estimated measurement does not satisfy the estimation request, the system modeling unit 11 determines whether another estimation rule exists (step S107), and if another estimation rule exists. Returns to step S105, and if there is no other estimation rule, returns to step S101.
If it is determined in step S109 that the control request is satisfied, the selection of the mathematical model for the control system is completed and model data is created (step S111). If it is determined in step S109 that the control request is not satisfied, the system modeling unit 11 determines whether there is another control law (step S110). If another control law exists, step S108 is performed. If there is no other control law, the process returns to step S101.
Next, the system modeling means 11 creates a controller identification program (steps S113 and S114) when the controller 20 performs a model parameter identification calculation (YES in step S112). Further, when the controller 20 performs a state estimation calculation described later (YES in step S115), the system modeling unit 11 creates a controller estimation program (steps S116 and S117).

  Furthermore, when the controller 20 performs a control calculation (YES in step S118), the system modeling unit 11 creates a control calculation program for the controller (steps S119 and S120). Thus, the operation of the system modeling unit 11 is completed.

  Next, the system identification unit 12 of the system determination device 1 identifies the model parameter based on the collected time series data. FIG. 7 is a flowchart showing the operation of the system identification unit 12. The system identification unit 12 determines whether there is a model parameter identification factor (step S200). If there is an identification factor, the system identification unit 12 determines whether the controller 20 identifies the model parameter (step S201).

  When the controller 20 does not identify the model parameter, the system identification unit 12 determines whether there is a model change factor (step S202). The system identification unit 12 determines that there is a model change factor when the control system is activated or when model change observation data is received from the controller 20. And the system identification means 12 identifies a model parameter (step S203, S204). Thus, the operation of the system identification unit 12 is completed.

  The transmission means 13 of the system determination device 1 transmits model data, model parameters, a controller identification program, a controller estimation program, and a controller control calculation program to the controller 20 via the network 5.

  Next, the operation of the control calculation means 22 of the controller 20 will be described. FIG. 8 is a flowchart showing the operation of the control calculation means 22. The control calculation means 22 determines whether or not to derive the control parameter (step S300). When the control parameter is derived, the control parameter is determined according to the model data, the model parameter, and the controller control calculation program transmitted from the system determination device 1. Is obtained by a known method (steps S301 and S302). The control calculation means 22 sets the derived control parameter for itself.

Subsequently, the control calculation means 22 determines whether or not it is necessary to perform control calculation (step S303), and when it is necessary to perform control calculation, the observed value y of the control object 4 matches the target set value. Thus, the control output u is calculated (step S304), and the calculated control output u is passed to the control output means 23 and the transmission means 25 (step S305).
The control output means 23 outputs the control output u calculated by the control calculation means 22 to the actuator 3 in the form of voltage or current. The transmission means 25 transmits the control output u to the system determination device 1 via the network 5.

A flowchart summarizing the entire operation is shown in FIG. 9 is the operation of the system modeling unit 11, step S401 is the operation of the system identification unit 12, and steps S402 and S404 are the operation of the control calculation unit 22.
The sensor output recognition unit 24 of the controller 20 receives the observation value y of the control target 4 from a sensor (not shown) and passes it to the control calculation unit 22 and the transmission unit 25 (step S403 in FIG. 9). The transmission unit 25 transmits the observation value y to the system determination device 1 via the network 5.
As described above, the control calculation unit 22 calculates the control output u and passes the calculated control output u to the control output unit 23 and the transmission unit 25 (step S404).

  Next, the sensor output recognition unit 24 of the controller 20 determines whether or not the control performance has deteriorated (step S405). If it is determined that the control performance has deteriorated, the model change observation data is transmitted. Pass to 25. The transmission means 25 transmits model change observation data to the system determination apparatus 1 via the network 5 (step S406). Whether or not the control performance has deteriorated can be determined using the observation value of the model change observation sensor that is not directly required for the identification of the model or the control of the control object 4. This model change observation sensor will be described later.

If there is no deterioration in the control performance, the controller 20 repeats the processing of steps S403 to S405 for each sampling until, for example, the user instructs to end the control (YES in step S407).
When the model determination observation data is received from the controller 20, the system determination device 1 returns to step S400 (S100 in FIG. 5).

  As described above, in this embodiment, the system determination device 1 identifies the mathematical model and the model parameter of the control system, so that no load is applied to the site-side equipment 2. Further, in the present embodiment, when a change occurs in the controlled object 4, model change observation data is transmitted from the controller 20 to the system determination device 1, and the system determination device 1 automatically readjusts the model and model parameters. This eliminates the need for the user to manually instruct readjustment. In addition, even when a secular change occurs in the device to be controlled, it is possible to avoid deterioration in control performance.

[Second Embodiment]
When the error between the identified model and the actual control system falls within the allowable range, the state can be estimated using the model, that is, the observation value y of the controlled object 4 can be estimated. By performing state estimation, the observed value y of the control object 4 that cannot be measured can be generated in a pseudo manner. In addition, when the actual observation value y is buried in noise, an uncertain phenomenon that cannot be covered by a deterministic system in modeling is treated as noise having statistical characteristics of a stochastic system by using a Kalman filter or the like. The observation value y can be generated in a pseudo manner. In the case where the Kalman filter is used, error covariance is obtained in the estimation calculation process. Since this value can be used as an index of estimation error, when this value exceeds a certain threshold during system operation, an identification factor can be generated to re-identify model parameters. For example, when a temperature sensor cannot be attached to a heating target heated in a furnace, the temperature of the heating target can be estimated by performing state estimation.

  Also in the present embodiment, the configuration of the entire control system is the same as that of the first embodiment, and therefore, description will be made using the reference numerals in FIG. FIG. 10 is a block diagram illustrating a configuration example of the system determination device 1 according to the present embodiment. The system determination apparatus 1 according to the present embodiment is obtained by adding a state estimation calculation unit 14 to the first configuration.

  FIG. 11 is a diagram showing data transition in the control system of the present embodiment. The state estimation is performed based on the control system model selected by the system modeling unit 11, the model parameter identified by the system identification unit 12, the control output u and the observation value y transmitted from the controller 20. However, since the observation value y of the control object 4 can be measured only at the time of system identification, another observation value related to the observation value y is used instead of the observation value y during the actual operation. As another observation value, for example, when the object to be heated is heated in a furnace, the temperature of the heater or the temperature in the furnace can be used.

  The state estimation calculation unit 14 passes the estimated value y ′ of the observed value y to the transmission unit 13. The transmission unit 13 transmits the estimated value y ′ to the controller 20 via the network 5. As the state estimation calculation means 14, when a probability model is selected in determining a control system model, a state estimator represented by a known Kalman filter can be used. When a deterministic model is selected in determining a control system model, a state observer such as a Luenberger observer can be used. There are many state estimators and state observers for nonlinear systems.

  The control calculation means 22 of the controller 20 determines the control parameter as in the first embodiment, calculates the control output u according to the estimated value y ′ of the observation value y transmitted from the system determination device 1, The calculated control output u is transferred to the control output means 23 and the transmission means 25. Since the estimation and control can be executed separately, the control method and the estimation method can be determined independently.

Next, the process flow of the control system of the present embodiment will be described. FIG. 12 is a flowchart showing the overall operation of the control system of the present embodiment, and the same processes as those in FIG. 9 are denoted by the same reference numerals.
First, the system modeling means 11 of the system determination device 1 selects a mathematical model of a control system as in the first embodiment (step S400 in FIG. 12, steps S100 to S120 in FIG. 5 and FIG. 6). However, in the present embodiment, it is assumed that the observed value y of the controlled object 4 cannot be measured, so it is necessary to be able to acquire time series data when selecting a model. For example, when a heating target is heated in a furnace, a temperature sensor is attached to the test heating target at the time of modeling and model parameter identification, and the heat treatment is experimentally performed in the furnace to obtain an observation value y (temperature of the heating target). It will be necessary.

  The process of the model parameter identification (step S401) by the system identification unit 12 of the system determination apparatus 1 and the control parameter determination (step S402) by the control calculation unit 22 of the controller 20 are the same as those in the first embodiment. .

  Next, the state estimation calculation unit 14 determines an estimation rule parameter (estimated gain) of the state estimator (step S408). The measuring means (not shown) measures the temperature of the heater 33 and the temperature of the heat sink 34 as an estimated observation value that can be measured instead of the observation value y (step S409). Then, the state estimation calculation unit 14 calculates an estimated value y ′ of the observed value y of the control object 4 (step S410). FIG. 13 is a flowchart showing the operation of the state estimation calculation means 14. First, the state estimation calculation means 14 determines whether or not the controller 20 derives an estimation rule parameter (step S500), and if the controller 20 does not derive, the time series data of the control output u and the observed value y Based on the model data and model parameters, estimation rule parameters are derived (steps S501 and S502). When deriving the estimation rule parameter, it is necessary to be able to obtain the observation value y temporarily as in the case of model selection and model parameter identification.

  Subsequently, the state estimation calculation means 14 determines whether or not the controller 20 performs the state estimation calculation (step S503). If the controller 20 does not perform the state estimation calculation, the state estimation calculation unit 14 can perform measurement instead of the observation value y. Based on the observed value and the control output u, an estimated value y ′ of the observed value y of the controlled object 4 is calculated (step S504), and the calculated estimated value y ′ is passed to the transmission means 13 (step S505). The transmission unit 13 transmits the observation value y ′ to the controller 20 via the network 5.

  The control calculation means 22 of the controller 20 calculates the control output u so that the estimated value y ′ of the observed value y of the controlled object 4 matches the target set value, and transmits the calculated control output u to the control output means 23. The data is transferred to the means 25 (step S411 in FIG. 12). The operation of the controller 20 is as described in the first embodiment except that the estimated value y ′ is used instead of the observed value y. The control output unit 23 outputs the control output u calculated by the control calculation unit 22 to the actuator 3 in the form of voltage or current, and the transmission unit 25 transmits the control output u to the system determination device 1 via the network 5. . In the initial sampling during the actual operation, the state estimation calculation unit 14 cannot calculate the estimated value y ′, and therefore, a preset initial value such as an initial estimated value is used.

The processing in steps S405 to S407 is the same as that in the first embodiment. If there is no change in the control object 4, the control calculation means 22 and the state estimation calculation means 14 perform the processing of steps S409, S410, S411, and S405 until the user gives an instruction to end control (YES in step S407). Repeat for each sampling.
When the model determination observation data is received from the controller 20, the system determination device 1 returns to step S400 (S100 in FIG. 5).

  In the present embodiment, the same effect as in the first embodiment can be obtained, and even if the observation value y of the controlled object 4 cannot be measured by providing the state estimation calculation means 14, Control can be performed by calculating an estimated value y ′ of the observed value y.

[Third Embodiment]
Next, a third embodiment of the present invention will be described. In this embodiment, a specific example of the second embodiment will be described. Here, let us consider a case where it is desired to heat a heating target at a set temperature for a certain period of time by controlling the heater power. It is assumed that the temperature of the heating target cannot be measured, and there is a relationship that can be described by a mathematical formula between the temperature of the heater and the temperature of the heating target. In such a case, it is possible to accurately estimate the temperature of the heating target and control the heater power so that the estimated temperature becomes equal to the set temperature. In the present embodiment, the model of the control target is a linear time-invariant system, and the heater power is controlled so that the estimated temperature of the heating target becomes equal to the set temperature.

  FIG. 14 is a block diagram showing the configuration of the control system according to the present embodiment, and the same components as those in FIG. 1 are denoted by the same reference numerals. In FIG. 14, 30 is a heating target, 31 is a furnace, 32 is a hot plate on which the heating target 30 is placed, 33 is a heater (actuator 3) attached to the hot plate 32, 34 is a heat sink (exhaust heat fan), and 35 is a heater. A thermocouple which is a temperature sensor attached to 32, and 36 is a thermocouple attached to the heat sink 34.

The data flow in the control system of FIG. 14 is the same as that of FIG. However, since the observation value y of the control object 4 (temperature of the heating object 30) can be measured only at the time of system identification, during the actual operation, instead of the observation value y, another estimation is related to the observation value y. As observation values, the temperature of the heater 33 and the temperature of the heat sink 34 are used.
First, the mathematical model of the control system will be described. Here, a linear first-order differential type is considered as a system model. The state equation is expressed as equation (1).

  x (•) is a state vector of 3 rows × 1 column, u (•) is a control output of scalar quantity, A is a system matrix (heat transfer coefficient) of 3 rows × 3 columns, and B is a drive of 3 rows × 1 column. It is a matrix (heater power conversion efficiency coefficient). The observation equation is assumed to be a temperature value obtained by converting the electromotive force of the thermocouple by the sensor output recognition means 24 as shown in the equation (2).

y (•) is the observation vector of 3 rows × 1 column, T _H (•) is the heater temperature of the scalar amount, T _O (•) is the heating target temperature of the scalar amount, and T _S (•) is the heat sink temperature of the scalar amount. It is. However, in Formula (2), Formula (3) shall be materialized.

  Next, model parameter identification performed by the system identification unit 12 of the system determination device 1 will be described. For the model parameter identification of the state space model shown in Equation (1) and Equation (2), a method for converting the input / output data into a Markov parameter or transfer function to obtain a minimum realization (references “BLHo and REKalman, Relelungstechnik”). , 14-12, p.545-548, 1966 '', literature `` L. Silverman, IEEE T-AC, 16-6, p.554-567, 1971 '', literature `` J. Rissanen, SIAM. J. on Control 9-3, p. 420-43O, 1971) and a method for obtaining a small internal description directly using input / output data (see “B. Gopinath, Bell System Tech., 48-5, p”). .1101-1113,1969 ”, literature“ MABudin, IEEE T-SMC, 2-3, p.396-402,1972 ”), subspace method (literature“ PVOverschee and BDMoor, Automatica, 30-1 ”) , p.75-93, 1994)). Here, we explain how to obtain the minimum realization of the internal description directly using the input / output data proposed by Gopinath. First, equation (1) can be expressed as equation (4).

  When the matrix of equation (5) is applied to both sides of equation (4) and the sum from k = 1 to N is taken, equation (6) is obtained.

  If the matrix related to [A B] is regular, an inverse matrix exists, a solution of [A B] is obtained, and a model parameter can be obtained. In this example, if the control output u (k) is sufficient to excite the mode of the system, the matrix over [A B] can be regularized with N = 3, but the control output used for normal operation is used. Therefore, it is necessary to select N> 3. Therefore, the model parameter [A B] to be obtained can be obtained as in Expression (7).

In order to identify the model parameter described above, it is necessary to acquire the temperature T _O of the heating object 30 as an observation value. However, in this embodiment, the temperature T _O of the heating object 30 cannot be observed. The temperature T _O of the heating target 30 at the time of parameter identification is obtained by attaching a temperature sensor to the test heating target and experimentally heating the heating target in the furnace 31.

  Next, state estimation performed by the state estimation calculation unit 14 of the system determination device 1 will be described. In the present embodiment, it is assumed that the temperature of the heating object 30 cannot be directly acquired during the actual operation except during the identification of model parameters or during maintenance. In the control, the target temperature transition (target temperature rise orbit) is determined for the heating object 30. Therefore, since the temperature of the heating target 30 is required in the control, it is necessary to derive an estimated value of the temperature of the heating target 30. Although the state vector can be obtained from the equation (1) based on the model parameter obtained by the system identification means 12 and the data of the control output u (k), this method cannot correct the estimated value using the available sensor observation value. Therefore, accurate estimation results cannot be expected. Therefore, in the present embodiment, a case will be described in which a one-dimensional observer is used as an example using an observer that considers correcting an estimated value with another observed value related to an observed value to be estimated.

  The equation of state is equation (1), and the observation equation is assumed to be the temperature conversion direct value of the electromotive force of the thermocouples 35 and 36 installed in the heater 33 and the heat sink 34 as shown in equation (8).

  C is an observation matrix of 2 rows × 3 columns, and is expressed as in Equation (9).

  As an estimation model, consider a system such as equation (10) that has the same dimensions as the system shown in equation (1).

  Hat x (•) is a state estimation vector of 3 rows × 1 column (hereinafter, “∧” appended on the character is also called a hat), and K is an estimation gain of 3 rows × 2 columns. When the estimation error e (k + 1) is obtained by subtracting the state equation of Expression (1) from the expression of the estimation model shown in Expression (10), Expression (11) is obtained.

  Therefore, if the eigenvalues of the matrix {A-KC} can be arbitrarily set, Expression (12) can be achieved.

  That is, a state observer with an estimation error of 0 can be configured. This is possible when (C, A) is observable in the present embodiment. It can be assumed that (C, A) is observable in most cases where the above-described model parameter identification method is used. Therefore, in the present embodiment, since it is a third-order model, the matrix K when three poles of the determinant | zI−A + KC | are selected in the unit circle of the complex plane is applied to the expression (10). Thus, the same dimensional observer can be configured. In the present embodiment, a deterministic system is targeted as a control system model. However, in a situation where the state vector of Expression (1) and the observation vector of Expression (8) are disturbed by Gaussian noise, the Kalman filter (reference “ REKalman, Trans. ASME. Basic Eng, 82D-1, p. 35-45, 1960 ") may be used.

  Next, control performed by the control calculation unit 22 of the controller 20 will be described. Using the system matrix A and the drive matrix B obtained by the system identification means 12 and the observation matrix C introduced in the equation (9), the state equation of the equation (1) and the observation equation of the equation (8) are configured. Here, a control problem in which an arbitrary time function is set as a target value and the estimated temperature value of the heating target 30 follows the target value will be described. The purpose here is to provide a feedback control law that minimizes the quadratic evaluation function J of equation (13).

r (·) is a known target trajectory vector of 3 rows × 1 column, Q _F is a known state termination weighting matrix of 3 rows × 3 columns, Q is a known state weighting matrix of 3 rows × 3 columns, R is 1 It is a known control output weighting factor of row × 1 column. Hereinafter, a method for deriving such a feedback control law will be described.
By using the relationship of Formula (14), the relationship of Formula (15) is obtained.

  Moreover, the relationship of Formula (17) is obtained by using the relationship of Formula (16) in arbitrary time functions q (k).

  If a formula obtained by doubling the formula (15) and the formula (17) is added to the formula (13), the formulas (18) and (19) can be obtained.

  Therefore, the control output u (k) that minimizes the equation (19) is expressed by the equation (20).

  Here, it can be seen that P (k) and P (N + 1) and time functions q (k) and q (k + 1) that minimize Equation (19) are as follows.

  As described above, the optimum control output u (k) at the time point k can be obtained by backward calculation from the end point N in the equations (20) to (24).

Since the processing flow of the control system of this embodiment is the same as that of the second embodiment, the operation of the control system will be described with reference to FIG.
First, the controller user operates the user registration PC 21 to input requests such as the apparatus operating environment, control performance, and estimated performance. This input may be a method in which a Web screen is prepared in advance as a Web server function of the system determination device 1 and a user inputs from a browser, or may be input from something like a personal computer tool. In the present embodiment, as the control performance desired to be requested, the temperature rise time, the overshoot allowable amount, the transient state allowable deviation range, the steady state allowable deviation range, and the control output allowable amount are input. It is assumed that an allowable error range of the estimated state temperature and an allowable error range of the estimated steady state temperature are input.

  Subsequently, the controller user transmits the input request content from the user registration PC 21 to the system determination device 1. At this time, it goes without saying that the user registration PC 21 is already connected to the network 5 and can send data to the system determination device 1 via the network 5. The system determination device 1 stores the received request content.

  Next, the controller user attaches a temperature sensor to the test heating target and places the heating target in the furnace 31. Then, the user gives an instruction to the controller 20 from the user registration PC 21 to start a test for acquiring time-series data. Upon receiving an instruction from the user registration PC 21, the controller 20 receives the control output u from the system determination device 1 via the network 5 and outputs this control output u to the heater 33. At this time, the system determination device 1 generates a control output u within the allowable control output amount requested by the user for testing and transmits it to the controller 20.

  The controller 20 transmits time series data including a set of the control output u, the temperature of the test heating target, the temperature of the heater 33, and the temperature of the heat sink 34 to the system determination device 1 via the network 5. Data may be sent sequentially, or once stored in the controller 20 and sent for a fixed time.

Next, before entering the actual operation, the system modeling unit 11 of the system determination device 1 selects a control system model as in the first embodiment (step S400 in FIG. 12), and the system identification unit 12 further The model parameters are identified (step S401).
The state estimation calculation unit 14 of the system determination device 1 determines an estimation rule parameter (step S408).

Next, when the controller user puts the heating target 30 into the furnace 31 and the main operation is started, the state estimation calculation unit 14 of the system determination device 1 estimates the estimated value y ′ (heating) of the observation value y of the control target 4. (Estimated temperature of the object 30) is calculated, and this estimated value y ′ is transmitted to the controller 20 via the network 5 (step S410).
The control calculation means 22 of the controller 20 calculates the control output u and outputs it to the controlled object 4 and transmits it to the system determination device 1 via the network 5 (step S411). The processing in steps S405 to S407 is the same as that in the first embodiment.

  Here, a model change observation sensor for determining whether or not the control performance has deteriorated will be described. In the control system of the present embodiment, for example, a pressure sensor that measures the pressure inside the heater 33 is a model change observation sensor. If the pressure inside the heater 33 when the shutter of the furnace 31 is closed and sealed changes, the heat transfer from the heater 33 to the heat sink 34 naturally changes, so the model or model parameters must be updated. Good control and state estimation will not be possible. Therefore, when the change in the output of the model change observation sensor is equal to or greater than a predetermined threshold, the sensor output recognition unit 24 determines that the control performance has deteriorated and transmits the model change observation data.

[Fourth Embodiment]
In the first to third embodiments, the system identification unit is provided in the system determination device 1, but the system identification unit may be provided in the controller 20. FIG. 15 is a block diagram illustrating a configuration example of the controller 20 according to the present embodiment. Also in the present embodiment, the configuration of the entire control system is the same as that of the first embodiment, and therefore, description will be made using the reference numerals in FIG. FIG. 16 is a diagram showing data transition in the control system of the present embodiment.

  The system identification unit 27 identifies a model parameter from the control output u and the observed value y, similarly to the system identification unit 12 of the first embodiment, according to the controller identification program transmitted from the system determination device 1. FIG. 17 is a flowchart showing the operation of the system identification unit 27.

  The system identification unit 27 determines whether or not the controller 20 identifies a model parameter (step S600), and when performing model parameter identification, determines whether or not a model change factor exists (step S601). The system identification unit 27 determines that there is a model change factor when the control system is activated or when model change observation data is generated. Then, the system identification unit 27 identifies the model parameter (step S602), and sets the identified model parameter in the control calculation unit 22 (step S603). Thus, the operation of the system identification unit 27 is completed. When the system identification unit 27 identifies the model parameter, the determination is YES in step S201 in FIG. 7, and it goes without saying that the system identification unit 12 does not identify the model parameter.

  Other operations are the same as those in the first embodiment. The system identification unit 27 can be easily applied to the second and third embodiments.

[Fifth Embodiment]
In the first to third embodiments, the control calculation means is provided in the controller 20, but the control calculation means may be provided in the system determination device 1. FIG. 18 is a block diagram illustrating a configuration example of the system determination device 1 according to the present embodiment. Also in the present embodiment, the configuration of the entire control system is the same as that of the first embodiment, and therefore, description will be made using the reference numerals in FIG. FIG. 19 is a diagram showing data transition in the control system of the present embodiment.

  The control calculation means 15 determines the control parameter from the model data and the model parameter according to the control calculation program, and calculates the control output u. FIG. 20 is a flowchart showing the operation of the control calculation means 15. The control calculation means 15 determines whether or not the controller 20 derives the control parameter (step S700). When the controller 20 does not derive the control parameter, the control parameter is determined according to the model data, the model parameter, and the control calculation program. Obtained (steps S701 and S702). The control calculation unit 15 sets the derived control parameter for itself.

  Subsequently, the control calculation means 15 determines whether or not the controller 20 performs the control calculation (step S703). If the controller 20 does not perform the control calculation, the observation value y transmitted from the controller 20 is the target set value. The control output u is calculated so as to match (step S704), and the calculated control output u is passed to the transmission means 13 (step S705). The transmission unit 13 transmits the control output u to the controller 20 via the network 5. The receiving unit 26 of the controller 20 passes the control output u to the control output unit 23.

When the control calculation means 15 performs control parameter derivation and control calculation, it goes without saying that the determination N0 is made in steps S300 and S303 in FIG. 8, and the control calculation means 22 does not perform control parameter derivation and control calculation.
Other operations are the same as those in the first embodiment. The control calculation means 15 can be easily applied to the second and third embodiments.

[Sixth Embodiment]
In the second and third embodiments, the state estimation calculation means is provided in the system determination device 1, but the state estimation calculation means may be provided in the controller 20. FIG. 21 is a block diagram illustrating a configuration example of the controller 20 according to the present embodiment. Also in the present embodiment, the configuration of the entire control system is the same as that of the first embodiment, and therefore, description will be made using the reference numerals in FIG. FIG. 22 is a diagram showing data transition in the control system of the present embodiment.

  The state estimation calculation unit 28 determines an estimation rule parameter (estimated gain) of the state estimator, and calculates an estimated value y ′ of the observation value y of the control target 4. FIG. 23 is a flowchart showing the operation of the state estimation calculation means 28. First, the state estimation calculation unit 28 determines whether or not the controller 20 derives the estimation rule parameter (step S800). When the controller 20 performs the derivation, the time series data of the control output u and the observed value y, Based on the model data and model parameters transmitted from the system determination device 1, estimation rule parameters are derived (steps S801 and S802). When deriving the estimation rule parameter, it is necessary to be able to obtain the observation value y temporarily as in the case of model selection and model parameter identification.

  Subsequently, the state estimation calculation means 28 determines whether or not the controller 20 performs the state estimation calculation (step S803), and when the controller 20 performs the state estimation calculation, the estimation observation that can be measured instead of the observation value y. Based on the value and the control output u, an estimated value y ′ of the observed value y of the controlled object 4 is calculated (step S804), and the calculated estimated value y ′ is passed to the control calculating means 22 and the transmitting means 25. The transmission unit 25 transmits the observation value y ′ to the system determination device 1 via the network 5.

  When the state estimation calculation unit 28 performs the estimation rule parameter derivation and the state estimation calculation, the determination is YES in steps S500 and S503 in FIG. 13, and the state estimation calculation unit 14 does not perform the estimation rule parameter derivation and the state estimation calculation. Needless to say. Other operations are the same as those in the second and third embodiments.

  In the first to sixth embodiments, it is determined whether the control performance has deteriorated based on the output of the model change observation sensor. However, the second, third, In the sixth embodiment, a temperature sensor is attached to a test object to be heated, and a heat treatment is experimentally performed in a furnace to obtain an observed value y (temperature of the object to be heated). It is possible to determine whether or not the control performance has deteriorated by looking at the difference from the estimated value y ′ by the calculation means.

  In other words, it is assumed that the test heating target is heated in an environment where state estimation is performed. When the difference between the observed value y obtained by the temperature sensor attached to the test heating object at that time and the estimated value y ′ calculated by the state estimation calculating means 28 and 54 is equal to or greater than a predetermined threshold value, sensor output recognition is performed. The means 24 determines that the control performance has deteriorated and transmits model change observation data. Therefore, in terms of apparatus maintenance, the system can be calibrated by periodically performing a heating test using a test heating target with a temperature sensor.

If the estimated temperature is the state as in this example and the temperature can be measured directly (the estimated state is the same as the temperature data from the temperature sensor attached to the test heating object), the comparison can be made as it is. However, when the state to be estimated and the physical quantity measured by the test sensor are different, it is necessary to convert them into quantities that can be compared based on the system model. Such a function can also be realized by the controller 20 when the controller 20 has the state estimation calculation means 28, and the system determining device 1 transmits data necessary for state estimation from the controller 20 side. This can also be realized by the system determination device 1.
Further, the user may recognize a change in the device environment or the like, and instruct readjustment of the model or model parameter.

  In the first to sixth embodiments, the system modeling unit 11 selects a model of the control system. The designer selects such a model in advance and fixes the model at the time of selection. (That is, the model is not re-selected).

  As described above, the system determination device 1, the controller 20, and the user registration PC 21 in the first to sixth embodiments are each based on a computer having a CPU, a storage device, and an interface, and a program for controlling these hardware resources. Can be realized. The CPU executes processing as described in the first to sixth embodiments according to the program stored in the storage device.

  The present invention can be applied to process control.

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 structural example of the system determination apparatus in the control system of the 1st Embodiment of this invention. It is a block diagram which shows the structural example of the controller in the control system of the 1st Embodiment of this invention. It is a figure which shows transition of the data in the control system of the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the system modeling means of the system determination apparatus in the control system of the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the system modeling means of the system determination apparatus in the control system of the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the system identification means of the system determination apparatus in the control system of the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the control calculating means of the controller in the control system of the 1st Embodiment of this invention. It is a flowchart which shows the operation | movement of the whole control system of the 1st Embodiment of this invention. It is a block diagram which shows the structural example of the system determination apparatus in the control system of the 2nd Embodiment of this invention. It is a figure which shows transition of the data in the control system of the 2nd Embodiment of this invention. It is a flowchart which shows the operation | movement of the whole control system of the 2nd Embodiment of this invention. It is a flowchart which shows operation | movement of the state estimation calculating means of the system determination apparatus in the control system of the 2nd Embodiment of this invention. It is a block diagram which shows the structure of the control system which concerns on the 3rd Embodiment of this invention. It is a block diagram which shows the structural example of the controller in the control system of the 4th Embodiment of this invention. It is a figure which shows transition of the data in the control system of the 4th Embodiment of this invention. It is a flowchart which shows operation | movement of the system identification means of the controller in the control system of the 4th Embodiment of this invention. It is a block diagram which shows the structural example of the system determination apparatus in the control system of the 5th Embodiment of this invention. It is a figure which shows transition of the data in the control system of the 5th Embodiment of this invention. It is a flowchart which shows operation | movement of the control calculating means of the system determination apparatus in the control system of the 5th Embodiment of this invention. It is a block diagram which shows the structural example of the controller in the control system of the 6th Embodiment of this invention. It is a figure which shows transition of the data in the control system of the 6th Embodiment of this invention. It is a flowchart which shows operation | movement of the state estimation calculating means of the controller in the control system of the 6th Embodiment of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... System determination apparatus, 2 ... Site side apparatus, 3 ... Actuator, 4 ... Control object, 5 ... Network, 10 ... Receiving means, 11 ... System modeling means, 12 ... System identification means, 13 ... Transmission means, 14 ... State Estimation calculation means, 15 ... control calculation means, 20 ... controller, 21 ... user registration PC, 22 ... control calculation means, 23 ... control output means, 24 ... sensor output recognition means, 25 ... transmission means, 26 ... reception means, 27 ... System identification means, 28 ... State estimation calculation means, 30 ... Heating object, 31 ... Furnace, 32 ... Heat plate, 33 ... Heater, 34 ... Heat sink, 35, 36 ... Thermocouple.

Claims (6)

System modeling means for selecting a model of a control system including a controlled object;
System identification means for identifying model parameters of the model of the control system;
Based on the model selected by the system modeling means and the model parameters identified by the system identification means, control arithmetic means for controlling the control object so that the observed value of the control object and the target set value match,
A notification means for notifying the system modeling means and the system identification means of the deterioration of the control performance when it is determined that the control performance has deteriorated;
The system modeling means, when receiving a notification from the notification means, reselects the model of the control system,
The system identification unit re-identifies the model parameter when receiving a notification from the notification unit. The control system according to claim 1,
The control calculation means determines a control gain based on the model selected by the system modeling means and the model parameter identified by the system identification means. The control system according to claim 1,
And a state estimation calculating means for estimating an observation value of the controlled object based on the model selected by the system modeling means and the model parameter identified by the system identifying means,
The control calculation means uses the estimated value estimated by the state estimation calculation means in place of the observed value of the control target. A system modeling procedure for selecting a control system model including a controlled object by a computer;
A system identification procedure for identifying the model parameters of the model of the control system with a computer;
Based on the model selected in the system modeling procedure and the model parameter identified in the system identification procedure, the control calculation means connected to the computer is such that the observed value of the controlled object and the target set value match. A control calculation procedure for controlling the control object;
A notification procedure for notifying the computer of the deterioration of the control performance when it is determined that the control performance has deteriorated;
When the computer receives a notification of deterioration of the control performance, the computer again selects a model of the control system and re-identifies the model parameter. The control method according to claim 4, wherein
The control method further comprises a gain determination procedure for determining a gain of the control calculation means based on the model selected in the system modeling procedure and the model parameter identified in the system identification procedure. The control method according to claim 4, wherein
Furthermore, a state estimation calculation procedure for estimating an observation value of the control target based on the model selected in the system modeling procedure and the model parameter identified in the system identification procedure,
The control method, wherein the control calculation means uses an estimated value estimated by the state estimation calculation procedure instead of the observed value of the control target.

Images