JP2008123354A - Temperature controller, temperature control method, and temperature control program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To optimize a design parameter for model prediction control after applying model prediction control to multipoint temperature regulation control on a control target having a plurality of temperature control areas. <P>SOLUTION: In a temperature control system, a multipoint temperature regulation means 21 performing temperature regulation control of respective zones Z<SB>1</SB>, Z<SB>2</SB>, ..., Z<SB>n</SB>of a cylinder 31 cooperatively based on the model prediction control is installed, while a personal computer 11 is installed in the upper level of a panel computer 22. The personal computer 11 is provided with a model prediction control support means 12 optimizing the design parameter in the model prediction control. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a temperature control device, a temperature control method, and a temperature control program, and is particularly suitable when applied to a temperature control method for a cylinder of an injection molding machine.

  There is a method using PID control in the temperature control of a cylinder of a conventional injection molding machine. In this PID control, a PID control parameter is determined while performing auto-tuning using a method such as a step response method or a limit cycle method, and the PID control parameter is determined by repeatedly performing auto-tuning for system fluctuations. Adaptive control is performed while changing to a system corresponding to system fluctuations.

Moreover, there are internal model control and model predictive control as a control method considering interference occurring between a plurality of control objects.
Internal model control is a method for designing a controller based on an inverse function of a model transfer function. In model predictive control, the predicted value of the future controlled variable is obtained using a model corresponding to the control target, and the manipulated variable in the control input section is determined so that the predicted value approaches the set value (or reference trajectory) in the predicted section. In this method, only the input value corresponding to the current time among the obtained manipulated variables is input.

Further, for example, in Patent Document 1, by using a model corresponding to a control target, a control weight coefficient that is one of design parameters of model predictive control is determined based on a robust stability condition, and A method for calculating a PID control parameter by using a control weighting factor according to a calculation formula based on the relationship is disclosed.
JP 2003-195905 A

However, in the conventional PID control, a PID control parameter is determined by approximating a low-order model even for a complicated control target, so that a desired control characteristic cannot be obtained depending on the control target. there were. Further, in the conventional PID control, there is a problem that control that optimizes the entire system cannot be performed because control in consideration of interference occurring between a plurality of control targets cannot be performed.
In addition, in adaptive control, it is necessary to repeatedly perform auto-tuning according to system fluctuations, and there is a problem that control automation and control stability according to system fluctuations are not realized.

Further, the internal model control has a problem that the internal model control cannot be applied when the inverse matrix of the model transfer function cannot be obtained.
In model predictive control, it is necessary to optimize design parameters such as predictive horizon, control horizon, input weight, and output weight in order to achieve both control performance and robustness. There was a problem.

Further, in the method disclosed in Patent Document 1, only the control weight coefficient is parameter tuned, and other design parameters such as prediction horizon and control horizon are not parameter tuned. Therefore, an optimal solution as a PID control parameter is not necessarily obtained. There was a problem that I could not.
Accordingly, an object of the present invention is to provide a temperature control capable of optimizing the design parameters of the model predictive control after applying the model predictive control to the multi-point temperature control of the control target having a plurality of temperature control regions. An apparatus, a temperature control method, and a temperature control program are provided.

In order to solve the above-described problem, according to the temperature control device of claim 1, the multipoint temperature control is performed while linking the multipoint temperature control of the control target having a plurality of temperature control regions based on the model predictive control. It is characterized by comprising adjusting means and model predictive control support means for optimizing design parameters in the model predictive control above the model predictive control.
Further, according to the temperature control apparatus of claim 2, the multipoint temperature adjusting means performs the model predictive control according to the design parameter optimized by the model predictive control support means, so that the multipoint temperature control means. A model predictive control unit that calculates a set value in the control; and a multi-loop temperature controller that performs multipoint temperature control of the cylinder based on the set value calculated by the model predictive control unit. To do.

  According to the temperature control apparatus of claim 3, the model predictive control support means is identified by the model identifying means for identifying a model that is a calculation target of design parameters for model predictive control, and the model identifying means. Auto-tuning means for calculating an optimized design parameter in the model predictive control for the model, and a simulation means for simulating the operation of the multi-point temperature adjusting means when the design parameter in the model predictive control is given And data logging means for storing data in the multipoint temperature control.

According to the temperature control apparatus of the fourth aspect, the model prediction control means calculates a set value so that the output in the prediction horizon coincides with the reference trajectory.
According to the temperature control apparatus of the fifth aspect, the model predictive control support means optimizes design parameters in the model predictive control by using particle swarm optimization.

According to the temperature control apparatus of claim 6, the model identification unit applies the least square method to the set value and the control amount for the step input of each zone of the cylinder, thereby obtaining the transfer function of the model. It is characterized by seeking.
According to the temperature control apparatus of claim 7, the model identification means applies the least square method to the set value and the control amount for the step input of each zone of the cylinder, thereby obtaining the transfer function of the model. The transfer function of the interference term is characterized by modeling only the thermal interference from zone 2 to zone 1.

According to the temperature control apparatus of claim 8, the simulation unit simulates model prediction control using the design parameter calculated by the auto tuning unit and is calculated by the model prediction control. It is characterized by simulating multi-point temperature control with a set value as input and displaying the execution result on a screen.
According to the temperature control apparatus of the ninth aspect, the data logging means displays a set value, a control amount, and an operation amount in the multipoint temperature control on a screen.

  According to the temperature control method of claim 10, the step of optimizing the design parameter in the model predictive control, and the model predictive control according to the optimized design parameter, the multipoint temperature control control The method includes a step of calculating a set value, and a step of performing multi-point temperature control of a control target having a plurality of temperature control regions based on the set value calculated in the model predictive control.

  According to the temperature control program of the eleventh aspect of the present invention, the step of optimizing the design parameter in the model predictive control and the model predictive control according to the optimized design parameter are performed, thereby performing multipoint temperature control. Causing the computer to execute a step of calculating a set value in the control and a step of performing multi-point temperature control of a control target having a plurality of temperature control regions based on the set value calculated in the model predictive control It is characterized by that.

  As described above, according to the present invention, it is possible to optimize the design parameters of the model predictive control after applying the model predictive control to the multi-point temperature control of the control target having a plurality of temperature control regions. This makes it possible to control the temperatures of a plurality of temperature control regions while coordinating multipoint temperature control based on model predictive control while achieving both control performance and robustness. For this reason, even when interference or system fluctuation occurs in a multi-input multi-output system, it is possible to realize control automation and control stability according to the system fluctuation and to occur between a plurality of temperature control regions. It is possible to perform control in consideration of interference, and it is possible to improve the accuracy of multipoint temperature control.

Hereinafter, a temperature control device according to an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a schematic configuration of a temperature control device according to an embodiment of the present invention, and FIG. 2 is a block diagram showing a data flow of the temperature control system of FIG.
In Figure 1, the cylinder 31 for injecting resin is provided in the injection molding machine, the cylinder 31 is n (n is an integer of 2 or more) number of zones _{Z 1,} Z 2, · · ·, is divided into _{Z n} Yes. As shown in FIG. 2, the cylinder 31 is provided with heaters 32 a to 32 _n that heat the cylinder 31 for each of the zones Z ₁ , Z ₂ ,. _{Z 1,} Z 2, ···, a temperature measuring means 33a~33n for measuring respectively for each _{Z n} is provided. For example, thermocouples can be used as the temperature measuring means 33a to 33n.

The temperature control system is provided with multi-point temperature adjusting means 21 that performs temperature control of each zone Z ₁ , Z ₂ ,..., Z _n of the cylinder 31 in cooperation with each other based on model predictive control. Yes. Here, the multi-point temperature control unit 21, the zone Z _1, Z 2 of the cylinder _31, ..., each independently multi-loop temperature controller meter 24 and model predictive control performed by the temperature control of Z _n Based on this, a panel computer 22 is provided to coordinate the temperature control of the zones Z ₁ , Z ₂ ,..., Z _n of the cylinder 31. The multi-loop temperature controller 24 and the panel computer 22 are connected via a communication line 18.

For example, RS-485 can be used as the communication line 18. This RS-485 is compatible with bus-type multipoint connection, and can realize a multiple-to-multiple connection of up to 32 units.
Here, in the multi-loop temperature controller 24, as shown in FIG. 2, the temperature control of the zones Z ₁ , Z ₂ ,..., Z _n of the cylinder 31 is performed by PID (proportional integral derivative) control. PID control units 24a to 24n to perform are provided, and the panel computer 22 calculates the current set values SV1 ′, SV2 ′,..., SVn ′ of the PID control units 24a to 24n based on the model predictive control. Model prediction control means 23 is provided.

In addition, as a method for performing the temperature control of the zones Z ₁ , Z ₂ ,..., Z _n of the cylinder 31, other control methods such as proportional control, integral control, proportional integral control are used in addition to PID control. You may do it.
Further, a personal computer 11 is provided above the panel computer 22, and the personal computer 11 is provided with model prediction control support means 12 for optimizing design parameters in model prediction control, and a display for displaying various data. A device 17 is connected. The panel computer 22 and the personal computer 11 are connected via a communication line 19. For example, Ethernet (registered trademark) can be used as the communication line 19.

Here, the model prediction control support means 12 is provided with a model identification means 13, an auto tuning means 14, a simulation means 15, and a data logging means 16. And the model identification means 13 can identify the model in model predictive control which identifies the model used as the calculation object of the design parameter of model predictive control. The auto-tuning unit 14 can calculate optimized design parameters in model predictive control for the model identified by the model identifying unit 13. The simulation means 15 can simulate the operation of the multipoint temperature adjustment means 21 when design parameters in model predictive control are given. The data logging means 16 can accumulate data relating to temperature control of each zone Z ₁ , Z ₂ ,..., Z _n of the cylinder 31.

The model identification unit 13, the auto tuning unit 14, and the simulation unit 15 can operate offline, and the data logging unit 16 can operate both offline and online.
Then, the cylinder 31, the zone _{Z 1,} Z 2 at the heaters 32a through 32n, · · ·, is heated every _{Z n,} each zone _{Z 1,} Z 2, · · ·, temperature of _{Z n} is temperature measurement It measures by means 33a-33n, respectively. The temperatures measured by the temperature measuring means 33a to 33n are sent to the PID control units 24a to 24n as control amounts PV1, PV2,.

When the PID control units 24a to 24n receive the control amounts PV1, PV2,..., PVn from the temperature measuring means 33a to 33n, the control amounts PV1, PV2,. The operation amounts MV1, MV2,..., MVn are calculated so as to coincide with ′, SV2 ′,. Then, the PID control units 24a to 24n operate the operation amounts MV1, MV2,..., SVn ′ so that the control amounts PV1, PV2,..., PVn respectively match the next set values SV1 ′, SV2 ′,. ..., calculating the MVn, the operation amount MV1, MV2, ..., by controlling the output of the heater 32a through 32n based on MVn, the zone _Z 1 of the heater 32a through _{32n, Z} 2, · · · to control the heating for each _{Z n.} Incidentally, PID control by the PID controller 24a~24n Each zone _{Z 1, Z 2, ···,} can be performed independently for each _{Z n.} Moreover, the period of PID control can be set to 50 ms, for example.

Here, the model prediction control means 23 for calculating the next set values SV1 ′, SV2 ′,..., SVn ′ of the PID control units 24a to 24n is provided above the multi-loop temperature controller 24. . Each zone _Z 1 at PID controller _24a-24n, Z 2, · · ·, the PID control of _{Z n} is performed, the current operation amount MV1, MV2, · · ·, MVn, the controlled variable PV1, PV2,..., PVn and set values SV1, SV2,..., SVn are sent to the model prediction control means 23.

  Then, the model prediction control means 23 uses the current manipulated variables MV1, MV2,..., MVn, the controlled variables PV1, PV2,..., PVn and the set values SV1, SV2,. When received from the meter 24, the optimal control problem is solved at a certain time while predicting the future control amounts PV1, PV2,..., PVn for the model corresponding to the cylinder 31, and the future control amounts PV1, PV2 are solved. ,..., PVn can be determined to be as close as possible to the target value, and set values SV1 ′, SV2 ′,..., SVn ′ at that time can be determined and output to the multi-loop temperature controller 24.

For example, the model prediction control means 23 can calculate the next set values SV1 ′, SV2 ′,..., SVn ′ so that the output in the prediction horizon matches the reference trajectory.
Here, the plant output at the current time k is represented by y (k), and the set value trajectory at an arbitrary time t that the plant output y (k) should follow is represented by s (t). If the error between the plant output y (k + i) after i steps and the set value trajectory is s (t + i) is ε (k + i), the reference trajectory r (k + i | k) can be expressed by the following equation.
r (k + i | k) = s (k + i) −ε (k + i)
= S (k + i) -e- ^{iεTs / Tref} (k)

Ts is a sampling period, and Tref is a time constant of the reference trajectory. The notation r (k + i | k) means that the reference trajectory depends on the condition at time k.
When the section to predict the plant output y (k) is called a prediction horizon _{H P,} the input trajectory u between the prediction horizon _{H P (k + i | k} ) (i = 0,1, .., H P -1 ) is assumed to be applied, the plant output y (k) is the prediction horizon H reference trajectory between _P r (k + i | can select k) | k) to as close as possible to the input trajectory u (k + i . When the future input trajectory u (k + i | k) is selected, only the first element of the trajectory can be applied to the plant as an input signal.

  Note that examples of the model used in the model prediction control means 23 include an impulse response model and a step response model. Further, in order to be able to handle an integration process and an unstable process, a transfer function model or an ARX (Auto-Regressive eXogenous input) model may be used. Alternatively, a state space model may be used in order to facilitate extension to a multivariable system.

FIG. 3 is a block diagram showing an example of a schematic configuration of the model prediction control means of FIG.
3, model predictive control means 41, the plant output y (k + i | k) reference trajectories between the prediction horizon _{H P} is r (k + i | k) input orbit as close as possible u (k + i | k) (i = 0, 1, ..., H _U -1), and the first of the sequences of input trajectories u (k + i | k) (i = 0, 1, ..., H _u -1) Only the element is extracted as the input signal u (k) and applied to the PID control units 24a to 24n. The delay means 43 delays the input signal u (k) by one step and inputs it to the model prediction control means 41. A section for controlling the input trajectory u (k + i | k) is called a control horizon H _u .

Here, the model predictive control means 41, each zone _{Z 1,} Z 2, · · ·, control the amount of time that is measured by _{Z n} PV1, PV2, · · ·, with PVn are input, the model prediction The design parameter optimized by the control support means 12 is input. The design parameter optimized by the model predictive control support means 12 is, for example, the coincidence point Pset1 between the plant output y (k) and the reference trajectory r (k + i | k) at the current time k between predicted horizons. , Pset2,..., Psetn, and time constants Tref1, Tref2,..., Trefn of the reference trajectory r (k + i | k) can be selected.

Then, the model predictive control means 41 solves the optimal control problem for each fixed time while predicting the future control amounts PV1, PV2,..., PVn for the model corresponding to the cylinder 31, and the future control amount PV1. , PV2,..., PVn are calculated so that the sequence of input trajectories u (k + i | k) (i = 0, 1,..., H _U −1) is as close as possible to the target value. Then, the model prediction control means 41 uses only the first element in the sequence of the input trajectory u (k + i | k) (i = 0, 1,..., H _u −1) as the input signal u (k). Extract and output to the PID controllers 24a to 24n. The cycle of model prediction control can be set to 1000 ms, for example.
Specifically, when the set of coincidence points i is P, the model prediction control means 41 has an error between the reference trajectory r (k + i | k) and the plant output y (k + i | k) as shown in the following equation. The least squares solution can be calculated so that the square sum of is minimized.

For example, when the unit step signal is applied and the response of the system after the _HP step is S (H _P ), the plant output y (k + i | k) is the reference trajectory by calculating the least squares solution of the following equation. A sequence of input trajectories u (k + i | k) (i = 0, 1,..., H _U −1) can be calculated so as to be as close as possible to r (k + i | k).
ΔU = θ \ [T−Y _f ]
Note that \ (backslash operator) means the calculation of the least squares solution. _{Further, ΔU, θ, T, Y} f and S can be expressed by the following equation.

However, y _f is the free response of the plant. Here, the free response is a response obtained at a coincidence point when the future input trajectory remains at the latest value u (k−1). If the response of the system after the _HP step is S (H _P ), the predicted output at time k + H _P can be expressed by the following equation.
_{y (k + H P | k} ) = y f (k + H P | k) + S (H P) Δu (k | k)
However, Δu (k | k) is a change from the current input u (k−1) to the predicted input u (k | k), and can be expressed by the following equation.
Δu (k | k) = u (k | k) −u (k−1)

When the model prediction control means 41 calculates the vector ΔU, it extracts the first element Δu (k | k) of the vector ΔU. Then, the model prediction control means 41 constitutes an input signal u (k) to be input to the PID control units 24a to 24n from the first element Δu (k | k) of the vector ΔU as shown in the following equation. .
u (k) = Δu (k | k) + u (k−1)
When the model prediction control means 41 configures the input signal u (k) input to the PID control units 24a to 24n, the model prediction control means 41 outputs the input signal u (k) to the PID control units 24a to 24n, respectively, and delays. The input signal u (k−1) can be output to the system response prediction means 41 via the means 43.
In the next step, when the plant output y (k + 1) is measured, these series of calculations can be repeated.

Further, in FIG. 2, a model prediction control support means 12 for optimizing design parameters in model prediction control is provided above the model prediction control means 23.
When model predictive control is performed by the model predictive control means 23, the current manipulated variables MV1, MV2,..., MVn, controlled variables PV1, PV2,..., PVn and set values SV1, SV2,. SVn is sent to the model predictive control support means 12.

  Then, the model prediction control support means 12 performs model prediction on the current operation amounts MV1, MV2,..., MVn, control amounts PV1, PV2,..., PVn and set values SV1, SV2,. When received from the control means 23, these values are set as initial values. Then, an evaluation function including an adjustment parameter and a constraint condition are prepared, and repeated calculation is performed so as to minimize the evaluation function under the constraint condition. Then, the coincidence points Pset1, Pset2,..., Psetn and time constants Tref1, Tref2,..., Trefn obtained as a result of the repeated calculation are output to the model prediction control means 23 as optimized design parameters. To do. Note that optimization of design parameters in model predictive control can be executed on the personal computer 11 in an offline environment.

Here, as a design parameter optimization method in model predictive control, for example, particle swarm optimization can be used.
This particle swarm optimization is a modern heuristic (MH) method developed through simulation of a simplified social model, and represents the movement of a flock of birds in a two-dimensional space of continuous variables. Developed through that.

  For particle swarm optimization, see J.A. Kennedy and R.K. “Particle Swarm Optimization” by Eberhart (Proc. Of IEEE International Conference on Neural Networks, Vol. IV, pp. 1942–1948, Perth, Australia, et al. "Study of Voltage Reactive Power Control Method by Swarm Optimization" (The Institute of Electrical Engineers of Japan B, Vol.119, No.12, December 1999), Japanese Patent Application Laid-Open No. 2000-11603 “Voltage Reactive Power Control Method”, Japanese Patent Application Laid-Open No. 2002-51464 It is described in the “state estimation method”.

In particle swarm optimization, the position (state quantity) of each agent (in the particle swarm optimization, each group is called an agent) is represented by x and y coordinates, and the velocity of each agent is represented by vx. (Velocity in the x direction) and vy (velocity in the y direction).
Then, the position of each agent at the next time point can be updated from the position and speed information of each agent.

Based on this concept, the following optimization can be considered if the whole flock of birds takes action that optimizes some objective function.
That is, each agent remembers the best value pbest of the objective function that has been most evaluated so far in each search, and the x and y coordinates indicating the position (state quantity). In addition, each agent shares the highest value among the group of best values pbest held by each agent, that is, the best value gbest of the objective function so far of the group.

Each agent has a current position (x, y coordinates) and a current speed (vx, vy).
Each agent moves the direction to the position where the best values pbest and gbest exist according to the distance between the current position (x, y coordinates) and velocity (vx, vy) and the best values pbest and gbest. Try to change.
The action to be changed is expressed by speed. Then, the speed of each agent can be corrected as shown in equation (1) using the current speed and the best values pbest and gbest.
V _i ^{k + 1} = w × V _i ^k + c ₁ × rand () × (pbest _i −s _i ^k )
+ C ₂ × rand () × (gbest-s _i ^k ) (1)

Where V _i ^k is the speed of agent i, rand () is a uniform random number from 0 to 1, s _i ^k is the position of k-th search (search point) of agent i, and pbest _i is agent i The best values pbest and w are weighting functions for the agent speed, and c ₁ and c ₂ are weighting coefficients for each term.
By using this equation (1), it is possible to obtain a speed V _i ^{k + 1} that probabilistically approaches the best value pbest of each agent and the best value gbest of the group, and thus, the current value of each agent is determined. The position (search point) s _i ^k can be corrected as shown in equation (2).
s _i ^{k + 1} = s _i ^k + V _i ^{k + 1} (2)
Here, particle swarm optimization is a multi-point search having a plurality of search points as in the genetic algorithm, and each search point is a probability using the best value pbest of each search point and the best value gbest of the group. Thus, a global optimum solution (best solution) can be obtained.

  Specifically, the state variable of the particle swarm optimization agent is a design parameter in model predictive control. For example, as a state variable of an agent of particle swarm optimization, coincidence points Pset1, Pset2,... Between a plant output y (k) and a reference trajectory r (k + i | k) at a current time k between predicted horizons. , Psetn, and time constants Tref1, Tref2,..., Trefn of the reference trajectory r (k + i | k) can be selected.

Next, the agent position s _i and the velocity v _i are set as initial values of the agent. Note that the number i of agents is set in advance.
Next, the optimal solution is searched by repeating the calculations of the equations (1) and (2). The number of repetitions k is set in advance. When searching for an optimal solution, the evaluation of each agent and the update of the best value pbest of each search point and the best value gbest of the group are performed for each step.

Here, each agent can be evaluated by the following procedure.
First, an agent state variable s _i is set as a design parameter in model predictive control.
Next, the model state quantity data of the model to be controlled and the state quantity data for each zone Z ₁ , Z ₂ ,..., Z _{n of the} cylinder 31 are calculated by N points for a certain range of control output.
Next, the model state quantity of the control target model and the state quantity for each zone Z ₁ , Z ₂ ,..., Z _{n of the} cylinder 31 are evaluated by the following equations.

However, Xm (n) is the state quantity data for each zone Z ₁ , Z ₂ ,..., Z _n of the cylinder 31, Xe (n) is the model state quantity data of the controlled object model, and W (n) Is a weight function for data points, and F is an evaluation result.
Based on the evaluation results F of all agents, the best value pbest of each search point and the best value gbest of the group are determined.
Then, when the designated number of times is finished, the state quantity of the best value gbest of the group is set as a design parameter and stored as an optimized design parameter.
This makes it possible to optimize the design parameters for model predictive control by using a simple algorithm. Even when interference or system fluctuation occurs in a multi-input / multi-output system, control according to the system fluctuation is possible. It is possible to realize automation and control stability, and it is possible to perform control in consideration of interference occurring between a plurality of temperature control regions.

FIG. 4 is a flowchart showing the operation of the temperature control system of FIG.
In FIG. 4, the model prediction control support means 12 reads the data file F when optimizing the design parameters in the model prediction control (step S20). Here, data for constructing a model corresponding to the cylinder 31 is described in the data file F. For example, settings for step inputs of the zones Z ₁ , Z ₂ ,..., Z _n of each cylinder 31 are described. Values SV1, SV2,..., SVn and control amounts PV1, PV2,..., PVn can be registered. Also, the model can be changed for each injection material and temperature zone, and the data file F can be constructed for each model.

When the data file F corresponding to the model to be identified is read, the identification function is executed by the model identification means 13 and the transfer function identification parameter is calculated (step S21). Incidentally, the identification function may be a function of each zone _{Z 1,} Z 2 of the cylinder 31, ..., the set value SV1 to a step input of _{Z n,} SV2, ..., SVn and controlled variable PV1, PV2 ,..., PVn can be applied to obtain a model transfer function corresponding to the cylinder 31 by applying the least square method.

Each zone of the cylinder 31 is affected by the heat of the adjacent zone. In the injection molding machine, the zone Z ₁ is the tip of the cylinder 31 is a nozzle portion, since heat capacity is small compared to other zones, strongly it receives a thermal influence of the adjacent zone Z _2.
In general, the influence of the adjacent zone is called interference. Moreover, interference is expressed by a transfer function, and the one introduced as an element of the transfer function matrix is called an interference term. In the transfer function matrix, elements other than diagonal elements are called interference terms. For example, in the case of a transfer function matrix of 4 rows and 4 columns, 12 elements other than diagonal elements correspond to interference terms.
In the transfer function matrix of the cylinder described by n rows and n columns, the interference term introduces the influence on the zone Z ₁ with respect to the zone Z ₂ , that is, only the first row and second column elements.

By identifying the transfer function of the interference to the zone Z ₁ with respect to the zone Z _2, the model predictive control can be performed using the transfer function matrix in consideration of the important interference. Since the interference of the zone Z _{2 with} the zone Z ₁ is reflected in the model, the model prediction control means 23 can calculate the input trajectory u with a small model error.
Further, in order to identify only the interference to the zone Z ₁ for the zone Z _2, it is possible to reduce the number of steps to identify other interference term.
Then, when the model is identified by the model identification means 13, it is determined whether or not the design parameter in the model predictive control is optimized by the auto tuning function (step S22), and the design parameter in the model predictive control is determined by the auto tuning function. If not optimized, the design parameters in the model predictive control are manually input, and the design parameters are output to the multi-loop temperature controller 24 (step S23).

  And the simulation means 15 will perform a simulation function, if the design parameter in model prediction control is input manually (step S24). The simulation function can be converted into a function. That is, the simulation means 15 simulates the model prediction control for the multi-loop temperature controller 24 while using the design parameters input manually, and inputs the set values calculated by the model prediction control. By simulating the temperature control, the time series data of the set values SV1, SV2,..., SVn and the control amounts PV1, PV2,. It is displayed on the display means 17 as a simulation result.

Then, it is determined whether or not to end the operation of the model predictive control support means 12 (step S25). When the operation of the model predictive control support means 12 is continued, the process returns to step S23 and the above operations are repeated.
On the other hand, when the design parameters in the model predictive control are optimized by the auto-tuning function, the auto-tuning means 14 uses the particle swarm optimization for the model identified by the model identifying means 13 to The optimized design parameter in the predictive control is calculated, and the design parameter is output to the multi-loop temperature controller 24 (step S26).

  Then, when the design parameter in the model predictive control is optimized by the auto tuning function, the simulation unit 15 executes the simulation function (step S27). That is, the simulation unit 15 simulates the model predictive control for the multi-loop temperature controller 24 while using the design parameters optimized by the auto-tuning function, and inputs the set value calculated by the model predictive control. By simulating the multipoint temperature control, the time series data of the set values SV1, SV2,..., SVn and the controlled variables PV1, PV2,. The time series data is displayed on the display means 17 as a simulation result.

Then, it is determined whether or not to end the operation of the model prediction control support means 12 (step S28), and when the operation of the model prediction control support means 12 is continued, the process returns to step S22 and the above operations are repeated.
On the other hand, the multipoint temperature adjusting means 21 is provided with a function for initial setting of the temperature adjustment function and a function for the temperature adjustment function by model predictive control.
Then, the multipoint temperature adjusting means 21 acquires the identification parameter identified by the identification function (step S11) and predicts the design parameter manually input or the design parameter optimized by the auto tuning function by model prediction. Obtained from the control support means 12 (step S12).
Then, by inputting the identification parameter and the design parameter into the function for initial setting of the temperature adjustment function, the temperature adjustment function is initialized by model predictive control, and the value of the normal process check result is output (step S13).

2 includes control amounts PV1, PV2,..., PVn and current set values SV1, SV2,..., SVn actually measured by the temperature measuring means 33a to 33n. Are input via the multi-loop temperature controller 24, and the next set values SV1 ′, SV2 ′,..., SVn ′ are calculated by model predictive control (step S14).
Note that an algorithm based on model predictive control is described in the function of the temperature control function based on model predictive control, and the same algorithm as the function of the simulation function of the model predictive control support means 12 can be used. However, the function of the temperature control function by model predictive control uses measured values actually measured by the temperature measuring means 33a to 33n as the controlled variables PV1, PV2,. As the function of the simulation function of the control support means 12, calculated values calculated using models as control amounts PV1, PV2,..., PVn are used.

Then, when the next set values SV1 ′, SV2 ′,..., SVn ′ are calculated in the model predictive control, the multi-loop temperature controller 24 sets the control amounts PV1, PV2,. , MVn are calculated so as to coincide with the set values SV1 ′, SV2 ′,..., SVn ′, and the operation amounts MV1, MV2 are calculated via the PID control units 24a to 24n. , ..., by outputs of MVn to the heater 32a through 32n, the zone _{Z 1,} Z 2 of the cylinder 31, ..., performs temperature control for each _{Z n} (step S15).

Then, the zone _{Z 1,} Z 2 of the cylinder 31, ..., the temperature control is performed for each _{Z n,} the zone _Z 1 temperature of the cylinder 31 is at a temperature measuring means _{33a~33n, Z 2,} ·· · are respectively measured for each _{Z n,} an injection molding machine, the amount of control that is actually measured respectively at a temperature measuring means 33a~33n PV1, PV2, ···, PVn and current set value SV1, SV2, · · · SVn is output to the model prediction control means 23 via the multi-loop temperature controller 24 and is also output to the model prediction control support means 12 (step S16).

  The model predictive control support means 12 performs injection molding on the control amounts PV1, PV2,..., PVn and the current set values SV1, SV2,..., SVn actually measured by the temperature measuring means 33a to 33n. When the data is input from the machine, the data logging means 16 executes the logging function, thereby controlling the control amounts PV1, PV2,..., PVn and the current set value SV1, respectively measured by the temperature measuring means 33a to 33n. SV2,..., SVn are accumulated (step S29).

Then, the data logging means 16 determines whether or not to end the logging function (step S30), and when continuing the logging function, returns to step S29 and repeats the above operation.
Further, the multipoint temperature adjusting means 21 determines whether or not to end the temperature adjustment function by the model prediction control (step S17), and when continuing the temperature adjustment function by the model prediction control, returns to step S14 and performs the above operation. repeat.

Thereby, after applying the model predictive control to the multipoint temperature control of the cylinder 31 divided for each of the zones Z ₁ , Z ₂ ,..., Z _n , the design parameters of the model predictive control are optimized. Cylinders divided into zones Z ₁ , Z ₂ ,..., Z _n while coordinating multi-point temperature control based on model predictive control while achieving both control performance and robustness. The temperature of 31 can be controlled. For this reason, even when interference between the zones or system fluctuation occurs in the cylinder 31 divided for each of the zones Z ₁ , Z ₂ ,..., Z _n , control automation and control stability according to the system fluctuation In addition, it is possible to perform control in consideration of interference occurring between a plurality of temperature control regions, and it is possible to improve the accuracy of multipoint temperature control.

It is a block diagram which shows schematic structure of the temperature control system which concerns on one Embodiment of this invention. It is a block diagram which shows the data flow of the temperature control system of FIG. It is a block diagram which shows an example of schematic structure of the model prediction control means of FIG. It is a flowchart which shows operation | movement of the temperature control system of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Personal computer 12 Model prediction control support means 13 Model identification means 14 Auto tuning means 15 Simulation means 16 Data logging means 17 Display means 18, 19 Communication line 21 Multipoint temperature control means 22 Panel computer 23, 41 Model prediction control means 24 Many Loop temperature controller 31 Cylinder 24a-24n PID controller 32a-32n Heater 33a-33n Temperature measuring means 43 Delay means

Claims (11)

Multi-point temperature adjusting means for performing multi-point temperature control of a control target having a plurality of temperature control regions while linking based on model predictive control;
A temperature control apparatus comprising: model predictive control support means for performing optimization of design parameters in the model predictive control at a higher level than the model predictive control. The multipoint temperature adjusting means includes:
Model predictive control means for calculating a set value in the multi-point temperature control by performing model predictive control according to the design parameter optimized by the model predictive control support means;
The temperature control apparatus according to claim 1, further comprising: a multi-loop temperature controller that performs multi-point temperature control of the cylinder based on the set value calculated by the model predictive control unit. The model predictive control support means includes
Model identification means for identifying a model that is a calculation target of design parameters for model predictive control;
Auto-tuning means for calculating a design parameter optimized in the model predictive control for the model identified by the model identifying means;
Simulation means for simulating the operation of the multi-point temperature adjustment means when given design parameters in the model predictive control;
The temperature control apparatus according to claim 1, further comprising a data logging unit that accumulates data in the multipoint temperature control.   4. The temperature control apparatus according to claim 2, wherein the model predictive control means calculates a set value such that an output in the predictive horizon coincides with a reference trajectory.   5. The temperature control apparatus according to claim 1, wherein the model predictive control support unit optimizes design parameters in the model predictive control by using particle swarm optimization. 6. .   6. The model identification unit according to claim 3, wherein the model identification unit obtains a transfer function of the model by applying a least square method to a set value and a control amount with respect to a step input of each zone of the cylinder. The temperature control apparatus according to 1.   The model identification means obtains the transfer function of the model by applying a least square method to the set value and control amount for the step input of each zone of the cylinder, and the transfer function of the interference term is changed from zone 2 to zone 1. The temperature control apparatus according to any one of claims 3 to 5, wherein only the thermal interference is modeled.   The simulation means simulates model predictive control using the design parameter calculated by the auto-tuning means and also simulates multi-point temperature control using the set value calculated by the model predictive control as an input. The temperature control apparatus according to claim 3, wherein the execution result is displayed on a screen.   The temperature control apparatus according to claim 3, wherein the data logging unit displays a set value, a control amount, and an operation amount in the multipoint temperature control on a screen. A step of optimizing design parameters in model predictive control;
Performing model predictive control according to the optimized design parameters to calculate a set value in multi-point temperature control;
A temperature control method comprising: performing multi-point temperature control of a control target having a plurality of temperature control regions based on the set value calculated in the model predictive control. A step of optimizing design parameters in model predictive control;
Calculating a set value in multi-point temperature control by causing model predictive control to be performed according to the optimized design parameters;
A temperature control program that causes a computer to execute multi-point temperature control of a control target having a plurality of temperature control regions based on a set value calculated by the model predictive control.

Images