JPS61275905A - Controller for controlling multi-input multi-output system - Google Patents

Abstract

PURPOSE:To add a process stabilizing control function, an additional state decision function, and a model selecting function by using optimum control parameters based on dynamic model identification due to time series data analysis of process data. CONSTITUTION:In an input part 1 to which process data is inputted, time series analog signals are selected and inputted successively for every variable through isolators 4...by a multiplexer 5. Plural data are read in by the multiplexer in a time very shorter than the sampling period, and the input value signal is converted to digital data by an A/D converter 6 and is taken into an arithmetic processing part 2 from a bus line 7. the arithmetic processing part consists of a central processing unit CPU 8, a memory 9, and a data storage file 10. A magnetic bubble memory or the like is used for the data storage file 10 to register and preserve parameters for execution of the control of process models, an optimum control gain, etc. and log data of the process.

Description

Detailed Description of the Invention (Technical Field of the Invention) The present invention relates to multi-input multi-output system control based on the structure of a target process in a distributed control system applied to various continuous process and batch process plants in industry. The present invention relates to a miniaturized control device that enables stabilization control using a model and detection of process abnormalities.

(Prior art and its problems) In recent years, distributed control systems using miniaturized controllers have become mainstream in measurement and control systems for process control. Compared to a centralized control method that directly controls multiple points using a large computer, a distributed control method has the following advantages.

■Easy to set and change control models for each part of the target process.

(There is little risk that abnormal operation of the φ subsystem will affect the entire system.

The equipment can be introduced in stages and is economical in terms of price.

Small controllers with these advantages have become rapidly popular in recent years due to the higher performance and lower prices of vgropros, and large process computers and general-purpose minicomputers are used as host systems for these controllers to display the status of the entire process. , records, and integrated management are increasingly being taken care of.

By the way, all these small computers have l input l
It is limited to proportional control, integral control, differential control, and their combinations using classical control models for output systems, and its applicability to control of complex processes is limited.

-J,m In many industry processes, there are signal feedback loops within the process, where many variables (e.g. temperature, pressure, etc. and their associated manipulated variables) are interrelated. They are moving together. When a conventional controller is applied to such a process, each partial process is treated as one input and one output, and the entire process is stabilized using multiple controllers. However, in actual processes, the interference between variables cannot be ignored, and it is virtually impossible to stabilize the entire process using conventional controllers.

That is, when a conventional controller is used to control a multi-input multi-output system in which variables interfere with each other, control cannot be achieved with a single loop (1 input, 1 output) combination.

Conventional single-manpower, single-output models, as typified by so-called PID controllers, are limited to grasping the internal structure of the target process as a lumped constant and turning it into a black box response function. Furthermore, the freedom of the conventional model is also useful when integrating partial models to set goals at the level of the entire plant process and finding a balance between the set values of each partial process. The degree is extremely small.

In addition, the interaction between process variables within a process does not necessarily have a constant magnitude with respect to changes in the state of each variable, and there are many cases where the steady state model no longer applies in transient or abnormal conditions. If deviation from the steady state can be detected, countermeasures can be taken, such as switching models or shifting to -F manual operation, but with conventional controllers that can only perform limit checks, there is a tendency to be slow in discovering abnormalities and taking countermeasures.

Due to the performance limitations of these conventional controllers, the application of current distributed control methods, despite having many of the advantages mentioned above, is limited to simple processes with relatively fast responses (multiple inputs and multiple outputs). Also, the interaction can be ignored.
Processes that can be expressed by a combination of one-manpower and one-output system are limited to processes that have a large stationary region and are stable.

Therefore, for multi-input multi-output systems that include complex feedback loops within the process, such as cement kiln process, process, various thermal furnace processes, and various chemical plants, analysis, model estimation, and control system design are required. Despite the fact that a series of proven effective methods have been established, the current situation is that the introduction of effective control is slow due to the cost of expensive computers and minicomputers and the difficulty of introducing centralized control.

(Object of the Invention) The present invention has been made in view of the above-mentioned circumstances, and its purpose is to control a multi-input system for controlling a complex system that has a feedback loop within the process. An object of the present invention is to provide a multi-output system control device.

(Features of the Invention) The features of the present invention are "model registration,""datacollection," and "
The aim is to realize a compact control device with a series of functions such as calculation, operation signal output, status display, and recording, and to achieve optimal control based on dynamic model identification through time-series data analysis of process data. The purpose is to enable stabilization control of the process by using parameters.

It is also characterized by its multifunctional configuration, which allows the addition of functions such as additional state determination and model selection.

The functional requirements of the control device realized by the present invention are listed below.

■As a distributed control device, it has a matrix-type parameter to handle multiple-input, multiple-output system models.

- Parameters representing model characteristics can be optimal control parameters based on dynamic model identification through process time-series data analysis, and additional state determination and model selection functions can be added as necessary.

■Includes a comparison check function between the prediction function based on the registered structural model and the actual measured value, and dynamically determines the state of the process.

■The control device registers model parameters, collects data,
It has a series of functions related to control execution such as calculation, operation signal output, status display, and recording, and can also include communication functions with other controllers and host systems.

(Operation of the Invention) Data such as controlled variables and manipulated variables are read as variables from the target process of the multi-input multi-output system. These are input simultaneously as a plurality of analog data by a multiplexer or the like.

Based on the above input data, the arithmetic processing unit consisting of the central processing unit CPU and various storage devices operates the process model according to preset parameters, predicts the state transition of the process, and outputs control using the optimal control gain. Calculate.

The control outputs calculated by the above calculations are output as set values for the manipulated variables.

Regarding the above description, the relationship between the control model and variables will be explained with reference to FIGS. 4 and 5. As shown in FIG. 4, a control model is constructed based on the mutual relationship between controlled variables and manipulated variables of the target process, and the target process functions based on the output of the manipulated variable Y.

Since the above-mentioned controlled variables and manipulated variable 1, x and i represent vectors, for example, * = (xi, x2゜depth3), ? In the case of = (yl, 2), a multi-input multi-output system control model is constructed as shown in FIG.

(Embodiments of the Invention) Fig. 1 is a functional block diagram showing an embodiment of a control device for controlling a multi-input multi-output system according to the present invention, and Fig. 2 is an explanation showing an embodiment of the display screen of the device. It is a diagram.

In the figure, reference numeral 1 denotes an input section into which process data is input, and to this input section 1, a time-series analog signal is sent to an isolator 40 for each variable. The signals are sequentially selected and input by the multiplexer 5 via the multiplexer 5. The reading of multiple data by the multiplexer is carried out in an extremely short time compared to the sampling period, and in fact can be considered as simultaneous reading. 7 to the arithmetic processing section 2.

The arithmetic processing unit 2 is composed of a central processing unit ((:PU) 8, a memory 9, and a data storage file 10.The data storage file 10 uses a magnetic bubble memory or the like to retain its contents even when the power is turned off. It has a function to register and store process models, parameters for executing control such as optimal control gain, and process log data, and all programs for executing control reside in memory 9. The other part is used for performing calculations.

Reference numeral 11 denotes a keyboard for manually inputting various parameters such as the above-mentioned process model, optimum control gain, and control conditions necessary for performing control. Reference numeral 12 is a display part using a liquid crystal display or the like, and as an example shown in Figure 2, it displays the status of each variable in the target process by adding set values and actual measured values for each variable, and output values for manipulated variables. The variables are compared and displayed together with additional information such as the upper and lower limit ranges set for each variable.

12 also has a guide display for manual operations such as registering parameters for the device, setting and changing control conditions, etc.
Reference numeral 13 denotes a printer for printing process status, recording, registered parameters, etc.

Further, the output section 3 latches the control output for each manipulated variable as digital data, converts it into analog data using the D/A converter 14, and then amplifies and outputs it via the isolation amplifier 15.

The control device based on the above configuration allows matrix-type parameters in order to realize control of a multi-input multi-output system, so basically various models can be registered and used depending on the characteristics of the target process. As a typical example, we will explain the application of a control model based on statistical analysis of a dynamic system.

Before implementing control, a control model is determined in advance.The control model is constructed by collecting data on assumed input and output variables for the controlled variables and manipulated variables of the target process, as shown in Figure 5. start. After organizing the collected data and analyzing the self- and cross-correlation functions, spectra, etc., a model for a multi-input multi-output system can be identified by statistical estimation. If this is illustrated for the example shown in FIG. 5, it can be expressed as a process model as shown in the following equation.

In equation (1), xi(n), x2(n),! 3 (31)
is a controlled variable! 1°! 2. ! 3 at time n, and its value is the controlled variable xl(n-f ), x2(n-1), x3(n-i) and the manipulated variables yl(n-i), y2(n-f), which are obtained by multiplying each actual measured value by a coefficient matrix. Here, L is a lag, which is determined by the data analysis described above.

Further, the second term of equation (1) is random noise whose influence is statistically negligible.

Next, a control system is designed based on the characteristics of the process expressed by equation (1), and by replacing the time series data with state space representation and performing optimization calculations, the control system is designed for time point n (n yl(
n-1), y2(n-1) are the values at time (n-1) of the manipulated variables for controlling the controlled variable (1) values χ1(n), χ2(n), and χ3(n) at time n. shows the value of In the same equation, ξ1,1(n-1) and below are state vectors according to the lag, and the controlled variables χ1゜X2. As described above, the coefficient matrix of the process model, the parameters of the optimal control gain matrix, and the related term constants corresponding to the deviation from the set value of χ3 are determined and registered, thereby making preparations for control.

Next, the outline of the control operation will be explained according to the flowchart shown in FIG.

Actual measured values of process data at a plurality of points, such as controlled variables and manipulated variables, are read into the input unit 1 by periodic interrupts. These are input as a plurality of analog data from the isolator 4 to the multiplexer 5. The actual measured values of the above data are converted into industrial units, etc. according to the program in the memory 9.

The above measured values are based on preset standards, and the CPU
In step 8, the upper and lower limits, including the rate of change, are checked.The display section 12 shows, for example, the measured values PVI, 2, 3, the set values SV+, 2.3, and 2.3, as shown in the display screen of FIG. Output value MV3 is displayed. Also, the process tolerance is
Graphs are displayed in an easy-to-understand manner.

Next, according to a preset control mode, the output value of the manipulated variable is calculated.
) is selected, the arithmetic processing unit 2 consisting of the data storage file JL/10, CPU 8, and various storage devices is selected.
According to the equation (1) taken into the equation (1), when the actual measured value PVI, 2.3 read at any time is input, the process model is operated based on the arithmetic processing by the CPU 8, and the state transition of the process is predicted.

Furthermore, in the CPU 8, the optimum operation amount for the n point in the process is determined by the above equation (2)! l,! 2 is calculated.

Manipulated variable Y1. as a control output calculated as described below.
The upper and lower limits and rate of change of y2 are checked, and if the rate of change exceeds a certain level, it is corrected to within a certain range and output.

This control output is converted to a predetermined signal level and output to the process. Each of the above parameters is registered in the data storage file io, and the process model is updated at any time as described above.

In the above control, by calculating the deviation between the actual measured value PVI, 2.3 and its predicted value, abnormal trends in the process can be detected at an early stage.Although not shown, alarms etc. can be issued based on the detection of refusal. It can be connected to devices.

In the above embodiment, controlled variables and manipulated variables are illustrated as models, but it is of course applicable to various process models that are not limited thereto.

(Effects of the Invention) The present invention enables control of a multi-input multi-output system that could not be achieved with conventional controllers, and achieves process stabilization control, prediction-based abnormality detection, and response to process state changes.

This makes it possible to stabilize complex process targets with multiple inputs and multiple outputs that have signal feedback loops inside the process, which was difficult with conventional controllers, and to ensure efficient operation over long periods of time. .

In addition, early abnormality detection and status determination based on predictions, combined with the above-mentioned stabilization, can reduce the work of complex plant operators, and depending on the target, unmanned operation is also possible. Furthermore, this controller, which includes a structural model that reflects the true mechanism of the target system, works in conjunction with other partial systems to achieve the required level of optimal production balance for the entire plant, contributing to the realization of optimal operation of the entire factory.

[Brief explanation of the drawing]

FIG. 1 is a functional block diagram showing an embodiment of a control device for controlling a multi-input multi-output system according to the present invention, FIG. 2 is an explanatory diagram showing an embodiment of the display screen of the device, and FIG. A schematic flowchart of the control operation, and FIGS. 4 and 5 are explanatory diagrams of the relationship between variables and target processes. ! 10. Input section, 298. Arithmetic processing unit, 309. Output section, 8°0. Central processing circuit (CPU), θ01. External storage device.

Claims (2)

[Claims] (1) An input section that reads variable data from multiple points for the purpose of process control of a multi-input, multi-output system where variables have non-negligible interference with each other, and an input section that operates a process model according to preset parameters to control the process. A control device for controlling a multi-input multi-output system, which includes an arithmetic processing section that predicts state transition and calculates an optimal control output using a preset optimal control gain, and an output section that outputs the control output at a plurality of points. (2) In the apparatus according to claim 1, the process model is operated to predict the state transition of the process, and the deviation between the predicted value and the actual measured value is sequentially calculated as a prediction error, and the deviation is calculated based on a preset standard. A control device for multi-input, multi-output system control that detects abnormal trends in processes by checking.