JPH03149601A - Method for controlling process - Google Patents

Abstract

PURPOSE:To automate and stabilize process control by executing continuous control based upon a mathematical model capable of improving an operation level in normal time, and at the time of generating abnormality in an internal state or abnormality in the deviation or the like of a prerequisite, executing AI control. CONSTITUTION:An artificial intelligence AI control part 6 for controlling a process by interpreting and inferring knowledge obtained by regulating a control method and a model control part 5 for mathematically modeling the process based upon a prescribed prerequisite and controlling the process are used. Usually, the model control part 5 controls the process, and when the abnormality of the process is detected based upon a parameter and its variable and the detected result satisfies a prescribed condition, the control part 6 controls the process. Consequently, the process control can be automated, the judgement of process abnormality is clarified, and either one of the two control methods is selected by the clarified judgement to stabilize the process.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is based on AI (Artificial Intel
Regarding the hybrid process control method that controls the process to be controlled by combining artificial intelligence control using an expert system using the lligence method and model control using a mathematical model, we will especially clarify the distinction between the scope of application of the two types of control. Concerning what has become.

[Prior Art and Problems to be Solved by the Invention] Process control such as the operation of a Juto blast furnace is performed based on so-called modern control theory, which formulates a process into a mathematical model based on predetermined preconditions. This enables continuous, high-speed, and stable control of the process. In blast furnace model control based on modern control theory using a mathematical model, operation is possible when the inside of the blast furnace is in a steady state and the preconditions of the mathematical model are satisfied. Therefore, in other cases, such as when the blast furnace is in an unstable state such as during start-up or during an abnormality, or when the preconditions are deviated, the operator's experience can be used to bring the mathematical model to a state where it can be applied. The operation was carried out manually and depended on the experience of the photographer, resulting in an unstable situation. For example, when the temperature of the stave in the furnace wall cooling system of a blast furnace suddenly rises, this is thought to be due to deposits on the furnace wall falling off and the hot air flowing inside the furnace hitting the stave directly.

When the deposits fall, heat is taken away from the hot metal, which may lower the temperature of the hot metal and deteriorate the quality of the hot metal. To avoid such a situation, it is necessary to immediately replenish the heat in the furnace when a rise in the temperature of the tape is observed, but conventional mathematical models do not accurately represent this behavior theoretically. In this case, the operator carried out replenishment based on experience.

That is, as mentioned above, model 14 based on the mathematical model is
Since it is not possible to deal with cases where preconditions are not satisfied or abnormalities in the internal state of the process, etc., all control cannot be performed by model control, and manual operation based on the operator's experience is required, making it difficult to completely control process operation. It was difficult to systemize the process, and it was difficult to stabilize the process.

By the way, computer technology has developed in recent years, and artificial intelligence using AI methods such as fuzzy control and control using expert systems! 1111 (hereinafter referred to as ^I Iq control) has made it possible to systematize control based on information containing ambiguity such as empirical rules, which was difficult to systemize in the past, as described above.

However, in the case of control using ^■ methods, especially ^■ control using expert systems, empirical rules are converted into rules and stored in a knowledge base, and control is performed by inference based on the rules of experience, resulting in discontinuous control. , and the accuracy of ms depends on empirical rules.

There is also the problem that it is difficult to formulate accurate empirical rules and it is difficult to improve the accuracy of w control.

Further, when fuzzy control is used, there are problems in that experience is required to set membership functions and the number of control rules increases.

Furthermore, when manual control and automatic model control are combined with ilIWs, the process state is unstable and there is no clear standard for determining whether or not it is abnormal. Therefore, it was not possible to uniformly switch between automatic and manual control.

The present invention has been made in view of the above circumstances, and when an abnormality in the internal state of the process is detected based on operational data and when an abnormality in the internal state of the process is detected based on operational data, model control based on modern control theory using a mathematical model is normally performed in the process When an abnormality is detected, such as when the prerequisites of It is an object of the present invention to provide a process control method that can stabilize the process by selecting two 11m.

(Means for Solving the Problems) The process control method according to the present invention is a process control method in which a process to be controlled is controlled by changing the manipulated variable according to parameter fluctuations, and the process control method interprets knowledge that regularizes the control method. an artificial intelligence control for inferring and controlling the process;
controlling the process using a control, detecting an abnormality in the process based on the parameter and the amount of variation thereof, and controlling the process using the artificial intelligence control when the detection result satisfies a predetermined condition. In the present invention, the process is normally controlled continuously by a model control system using a mathematical model based on predetermined preconditions, abnormalities in the process are detected based on parameters, and the detection results are determined based on the predetermined conditions. Discrete control by artificial intelligence control is performed when the following conditions are met.

(Examples) Hereinafter, the present invention will be described in detail based on drawings showing examples thereof. FIG. 1 is a schematic block diagram showing the configuration of a process control system used for implementing the process control method according to the present invention (hereinafter referred to as the method of the present invention) together with a blast furnace. In the figure, l is a stave-cooled blast furnace, and around the blast furnace 1 there are operating data OD, which are parameters such as temperature sensors, pressure sensors, and gas sensors.
A group of various sensors 2 are provided to collect the information.

Operation data OD from sensor group 2 is input to operation data input section 3.
The signal is then A/D converted, a variation amount to be described later is calculated, and is provided to the determination unit 4. The determination unit 4 uses, for example, 11-rule as shown in Table 1 based on the operation data 00,
Conventionally, this criterion for determining the selection rule for selecting the control form of blast furnace process control was ambiguous and did not have a probability, so it was changed from model control ^I all m
It was difficult to switch to. Therefore, there is a need for an index to distinguish between the two, which is shown in Table 1.

(Left below) In Table 1, the indicators for detecting abnormalities based on operational data OD are classified into serious abnormalities, which are abnormalities in furnace conditions, minor abnormalities, which are unstable furnace conditions, and when applying a mathematical model. Models are classified into three categories: unsuitable models that do not meet the prerequisites. Further, the target hot metal temperature is set according to the classification, and the greater the degree of abnormality or unsuitability, the higher the target hot metal temperature is, so that Mrn- is performed on the safe side.

A feature of the index shown in Table 1 is that it takes into account fluctuations in specific operational data 00 over a long period of time, which is not considered in the model s1m based on the mathematical model.

The amount of variation is calculated using the following formula.

However, D: Amount of variation N: Number of sampling times within the monitoring time n: Current time It can be captured.

For example, when sampling the operation data OD at 10 minute intervals and monitoring for 4 hours, N=24. The monitoring time is set longer for more serious phenomena. The longest period of 8 hours corresponds to the time from when coke, iron ore, and other blast furnace inserts are inserted until they are discharged as hot metal, and it is necessary to confirm whether the effects of the abnormal phenomenon have disappeared. Compatible.

The definition of each symbol in Table 1 is as follows.

However, Kr: Ventilation resistance ΔP: Total pressure loss L in the blast furnace Length μ from the varnish dock level to the tuyere: Gas viscosity coefficient pm = Average gas density inside the blast furnace U: Average gas flow rate 8: Constant Co/CO□: Volume ratio 1a/R of CO gas and COx gas at the top of the furnace: Ratio of reaction rate constants R: C+CO, -42CO reaction rate constant failure: C+
%0. →CO reaction rate constant N2%: N2 concentration of furnace top gas Also, in the judgment section 4, the judgment of the above-mentioned index was simply expressed in the form of the If-Then rule based on the knowledge obtained from experience.

For example, 1P AI temperature fluctuation ≧40° or A4
Temperature fluctuation ≧40゜or Atsu Temperature fluctuation
≧40°or S? Temperature fluctuation ≧40°or S@
Temperature fluctuation ≧40゜1st EN Stave temperature fluctuation large = true However, AI, At = S■ Temperature fluctuation of any one of the temperature sensors A+, At'' and S1 of the varnish tape is 40℃ When the above is reached, the most recent stave temperature fluctuation is used as an index.In this way, not only control but also abnormality or precondition deviation judgment can be established using expert system techniques to establish selection rules. The criterion becomes constant and control becomes stable.

The determination unit 4 generates a determination signal based on the determination shown in Table 1,
It is applied to a model control unit 5 which performs control using a mathematical model, a control unit 6 which performs control using an expert system, and a switch means 7. Then, the switch means 7 is switched in accordance with the determination signal, and the target hot metal temperature Tt is determined accordingly, and control data is sent from the selected control unit 5 or 6 via the control data output unit 8 in accordance with the target hot metal temperature T. The CD is sent to a control device such as a hot blast furnace that controls the blast furnace l.

Here, the control in the model control unit 5 is, for example,
The photographing method disclosed in Japanese Patent No. 5561 is used.

Figure 2 is a diagram showing an example of control by the AI control unit.
The control in the control unit 6 is explained by taking the determination of the amount of tar action as an example.The hot metal temperature level is divided into five level indexes, the temperature gradient of the hot metal temperature is divided into five grade gradient indexes, and the level index and gradient index are divided into five levels. The amount of tar action is determined empirically according to the combination with the amount of tar action (II/hr), and the obtained amount of tar action (II/hr) is increased or decreased to the current value.

This is described in the IF TIIEN rule, and the AIII control of the process is thereby performed.

For example, when the hot metal temperature level is 1515℃ or higher and the temperature gradient is 1θ℃ or higher, IF (Level index = 2 & Gradient index = 2)T anal N-
1000, and the amount of tar action is reduced by 1 from the current value.
0001/hr decrease. In this way, we convert empirical rules into rules and perform step-like control using an expert system.
Stabilize the operation of a blast furnace that is in an abnormal state.

By shifting the threshold value used in FIG. 2 in parallel according to the target hot metal temperature, it can be applied even if the rolling conditions change.

Next, the control procedure of the method of the present invention will be explained.

FIG. 3 is a flowchart showing the control procedure. When the operation data OD is sent from each sensor group 2 to the operation data input section 3, it is taken in every predetermined sampling time (for example, 10 minutes) (step 11). The amount of variation is calculated (step I2), and the obtained amount of variation and operation data 00 are given to the judgment unit 4, where the judgments from steps 13 to I6 are performed, and the model control unit 5 or A114
Control section 6 is selected. In Step I3, it is determined whether there is one or more of the indicators shown in Table 1 that indicate an abnormality in the furnace condition, that is, whether or not an abnormality in the furnace condition has occurred. , the AI control unit 6 is immediately selected (step I8), and the target hot metal temperature rt=t520
AI by expert system for ℃
111 is performed. If no furnace condition abnormality occurs,
In step I4, it is determined whether the index indicating furnace condition instability is 3 or more, and when it is 3 or more, the I control unit 6 is selected (step I8), and the target hot metal temperature T, = 1510°C is set. ■Control is performed. When the index is 2 or less, it is determined in step I5 whether the index indicating model unsuitability is 1 or more, and when it is 1 or more, that is, when the preconditions are deviated, control by the mathematical model is impossible, so AI$ Control unit 6
is selected (step I8), and the target hot metal temperature r, =t
^■ control for 500°C is performed.

When the model unsuitability is less than 1, that is, when the preconditions based on the mathematical model are not deviated from, it is determined in step I6 whether the index indicating furnace condition instability is 2, and when it is 2, the model IlrB part 5 is selected (step-17), model control is performed for the target hot metal temperature r, = 1510°C, and when the index indicating furnace condition instability is not 2, that is, l or O
In the same case, the model control section 5 is selected (step I
7) Model control is performed for the target hot metal temperature Tt = 1500°C.

FIG. 4 is a graph showing the relationship between stave temperature, hot metal temperature, and tar action amount when a blast furnace is operated according to the method of the present invention. The vertical axis shows stave temperature, hot metal temperature, and tar action amount from top to bottom, and the horizontal axis shows time. In addition, the broken line for the hot metal temperature shows the temperature change according to the mathematical model, and the solid line shows the actual measured value.Furthermore, the single-point 1i line for the amount of tar action is when it is based on AI control, the dotted line is when it is based on model control, and the solid line is the amount of tar action in actual operation. are shown respectively.

As is clear from this graph, by performing ^1@ control when the stave temperature suddenly changes, the amount of tar action can be significantly increased compared to the case of model ii] w, and the drop in hot metal temperature can be suppressed. became.

FIG. 5 is a diagram showing a comparison between conventional process control using a conventional mathematical model and control according to the method of the present invention. It represents the automation rate, which shows the percentage of automation by computers. As is clear from this figure, in the case of control using a mathematical model, the accuracy rate for the target range of Si content was 66.8%, and the automation rate was 70-80%, but the present invention When the method combines model control using a mathematical model and Al@@ using an expert system, the hit rate in the target range is 85.5%, the automation rate is improved to 85-95%, and the stability of control is improved. This shows that automation, or systemization, has progressed.

In this example, the method of the present invention is used to operate a blast furnace. However, the present invention is not limited to this.
It goes without saying that the present invention can be used for various process controls that have conventionally been controlled using mathematical models.

〔effect〕

As detailed above, in the present invention, abnormalities in the process and deviations from predetermined preconditions are detected based on parameters, and control is normally performed using a mathematical model that can be continuously controlled and improve the operating level. In the event of an abnormality such as an abnormality in the internal state or deviation from preconditions, the expert system can secure the internal situation of the furnace and stabilize the process.
By performing I control, the judgment of control switching is made clear,
It has excellent effects such as promoting automation and stabilizing process control.

[Brief explanation of the drawing]

FIG. 1 shows the process control method according to the present invention during blast furnace operation.
A schematic block diagram showing the configuration of a process control system used for implementation when used, FIG. 2 is a diagram showing an example of control of the AI control unit, FIG. 3 is a flow chart showing the control procedure of the method of the present invention, FIG. 4 is a graph showing the relationship between stave temperature, hot metal temperature and tar action amount, and FIG. 5 is a diagram showing the effect of the method of the present invention. l...Blast furnace 3...Operation data human power department 4...Judgment part 5...Model control part 6...Drinking control part 7...
・Switch means patent Applicant Sumitomo Metal Industries Co., Ltd. Agent Patent attorney Noboru Kono (initiator) Figure 3 Figure 4

Claims (1)

[Scope of Claims] 1. In a process control method for controlling a process to be controlled by changing a manipulated variable in accordance with a change in a parameter, the process is controlled by interpreting knowledge of regularized control methods and making inferences. Controlling the process using artificial intelligence control and model control that mathematically models the process based on predetermined preconditions and controlling the process accordingly, and detecting an abnormality in the process based on the parameter and the amount of variation thereof. A process control method characterized in that the process is controlled by the artificial intelligence control when the detection result satisfies a predetermined condition.