JPH08155616A - Optimum control type change-over system for molten steel level in mold of continuous caster - Google Patents

Abstract

PURPOSE: To enable the selection of an optimum type at the time of the current operation from operational information in plural optimum control types to each variation of the different kinds of the molten steel level. CONSTITUTION: When the operational data of the operational conditions, etc., collected from a continuous caster through an operational data collecting part 32 are input into the preset level variation restraining decision tree 34 and level variation promoting decision tree 36, a control type for restraining the level variation with the former tree and a control type for promoting the level variation with the latter tree are decided, respectively. These deciding results are inputted to a control type change-over judging part 40 to judge the optimum control type based on this result. Based on this judged result, the optimum type is selected in three control types of an adaptive control, disturbant observer control and H infinitive control with an optimum control type change-over part 42 and an actuator command value needed to this control is outputted to a cylinder 28A for actuator.

Description

Detailed Description of the Invention

[0001]

The present invention makes it possible to apply a plurality of control methods to the molten steel level control in a mold of a continuous casting machine, and to determine the most molten steel level based on the operating conditions at the time of operation and the operating information such as the type of steel to be cast. By selecting a control method that can suppress fluctuations, level fluctuations can be minimized under any steel type and operating conditions, and stable operation can be performed, which allows optimum control of molten steel level in the continuous casting machine mold. Regarding the system.

[0002]

2. Description of the Related Art In continuous casting equipment, a continuous casting machine (hereinafter also referred to as a continuous casting machine) for continuously producing slabs such as slabs and billets from molten metal is constructed, for example, as shown in FIG. . In this continuous casting machine, the molten steel 10 in the ladle 12 is poured into a tundish 14 and then through a dipping nozzle 16 into a mold 18 which is a box-shaped mold made of copper and opened up and down. The mold 18 is usually water-cooled, and the molten steel 1 injected into the mold 18 is
0 solidifies from the outside in contact with the wall surface.

When a certain degree of solidified shell is formed on the outside of the mold 18 in contact with the mold 18, the solidified shell portion is continuously pulled out from the mold 18 by a pinch roll 20,
After being further cooled by a water-cooled spray (not shown) and completely solidified to the inside, it is cut into a predetermined length by a cutter 22 and sent as a cast piece 24 to the rear, for example, a rolling process.

By the way, in this continuous casting process, the molten steel 10 injected into the mold 18 overflows from the mold 18, or the molten steel insufficiently solidified into the mold 1.
It is necessary to prevent leakage from the pull-out port at the bottom of 8. Therefore, the amount of the cast slab 24 pulled out from the mold 18 and the amount of the molten steel 10 poured into the mold 18 are made equal to each other,
The level of molten steel in the mold is controlled so that the level of molten steel in the mold 18 is kept constant.

Further, in this molten steel level control in the mold, inclusions existing in the molten steel portion which has not yet solidified in the mold are re-engaged downward due to the level fluctuation when floating on the molten steel level surface and solidified as it is. It is also important to prevent the deterioration of the quality of the rolled material due to internal defects.

Therefore, generally, in continuous casting, the deviation between the target level position and the measured value (signal) of the molten steel level obtained by measuring the molten steel level in the mold 18 with the level meter 26 is fed back to the control system. Then, the deviation is reduced, for example, by sending a stopper position command value to an actuator such as a stopper 28 or a sliding nozzle for adjusting the flow rate of the molten steel 10 injected from the tundish 14 into the mold 18, so that the mold 18 is supplied to the mold 18. Molten steel level control is performed in which the molten steel flow rate to be injected is controlled so that the molten steel level in the mold matches the target level position.

[0007]

However, in controlling the level of molten steel in the mold, there are various factors that cause level fluctuations.

For example, when molten steel is poured from the tundish 14 into the mold 18, alumina or the like gradually adheres to the inner wall of the immersion nozzle 16, so that the molten steel level for the actuator operation is changed. The level changes due to the change in sensitivity.

Further, due to the exfoliation of alumina and the like adhering to the inner wall of the immersion nozzle 16, a rapid change in the molten steel level occurs.

Further, the cast piece 2 pulled out from the mold 18
4 is swollen at a portion which is not in contact with the pinch roll 20, and the swollen portion is compressed when it comes into contact with the pinch roll 20 and the unsolidified portion inside is pushed upward, which is called unsteady bulging. Periodic level fluctuations occur.

Further, the sensitivity of the actuator operation to the molten steel level is affected by dead time or delay time by the inert gas blown to prevent the immersion nozzle 16 from clogging, so that the molten steel level in the mold fluctuates. Occurs.

Immersion nozzle 16 as described above
For a slow change in the time constant (a large change in the time constant) such as a level change caused by adhesion of alumina or the like to the inner wall of the glass, for example, JP-A-59-30460 and JP-A-63-63.
As disclosed in 192545, the sensitivity of the actuator to the molten steel level is estimated by fitting the level data to the model in the frequency domain, and the optimum proportional integral derivative (P
This can be suitably dealt with by applying the adaptive control for controlling the molten steel level by obtaining the ID) gain and updating the parameters of the PID controller.

For a rapid change in the time constant (disturbance or fluctuation having a relatively small time constant) such as a rapid level change such as separation of alumina or the like adhering to the inner wall of the immersion nozzle 16, for example, Japanese Unexamined Patent Publication (Kokai) Publication No. A corrected output to the actuator is obtained by a disturbance estimator that estimates the disturbance flow rate from the input signal to the actuator and the level signal proposed in 5-23811, and the molten steel level is obtained by adding this to the output of the PID controller. By applying the disturbance observer control (disturbance estimation control) to control, it can respond suitably.

Further, for a periodic level fluctuation such as unsteady bulging, for example, Japanese Patent Application No.
By applying the H infinity control theory proposed by the applicant of the present application to the state equation and the output equation in 234385, the H infinity control capable of frequency shaping of the controller is suitably applied. be able to.

As described above, the three control methods described above are
It is not necessarily applicable to all molten steel level fluctuations, and there is a type of molten steel level fluctuation most effectively applied to each. In addition, the frequency and type of factors that cause level fluctuations vary depending on the amount of inert gas blown into the immersion nozzle, the composition of the molten steel to be cast, the casting speed, the molten steel temperature, the secondary cooling conditions, etc. It is difficult to determine which control scheme is best to choose.

The present invention has been made in order to solve the above-mentioned conventional problems, and is based on operation information at the time of operation from among a plurality of control methods capable of optimally responding to different types of molten steel level fluctuations. An object of the present invention is to provide an optimum control system switching system for the molten steel level in a continuous casting machine mold, which enables selection of an optimum control system at that time.

[0017]

According to the present invention, when controlling the molten steel level in a mold of a continuous casting machine, adaptive control for adapting to large fluctuations of the time constant and updating to an optimum control gain, time constant comparison The disturbance observer control that estimates and compensates for extremely small disturbances and fluctuations and the H infinity control that suppresses disturbances and fluctuations in the frequency band of molten steel level fluctuations in the mold are based on the operation information at the time of operation The above object is achieved by adopting a configuration including control method switching means for selecting a control method that is determined to be capable of suppressing the molten steel level fluctuation in the mold.

According to the present invention, in the optimum control method switching system, the control method switching means has a function of selecting an optimum control method using a control method decision tree created based on past operation information. It is configured to be.

[0019]

In the present invention, a plurality of control methods capable of optimally controlling the molten steel level according to the type of molten steel level fluctuation in the mold are adopted, and based on the operating information at the time of operation,
Since the optimum control method that can suppress the molten steel level fluctuation at that time can be selected, stable molten steel level control can always be performed.

As a method of selecting the control method at this time, for example, a method of preparing a control method decision tree in advance and determining and selecting the optimum control method using the decision tree can be mentioned.

In other words, it is judged whether the type of level fluctuation is one with a large time constant, one with a small time constant, or whether the fluctuation is periodic, and control is performed in accordance with each level fluctuation. While selecting and switching the method,
The control method that suppresses the level fluctuation by learning the conditions when the level fluctuation width after switching is improved and when it is worse than before switching, and the control method that increases the level fluctuation Create a decision tree.

When the operating condition at a certain point of time is judged by these decision trees, two control methods are provided: a control method for suppressing the level fluctuation and a control method for increasing the level fluctuation. If the two control methods obtained here are the same, the same control is continued without switching the control method, and if different control methods are obtained, the control method is switched to the control method for suppressing the level fluctuation.

To explain the present invention in more detail, operating conditions such as the amount of inert gas blown in, the casting speed, the weight of the tundish, the flow rate of the secondary cooling water, the mold width, the molten steel temperature and the level fluctuation range are collected online. The operational information such as analog data is divided into items and the values are divided by thresholds. Also, the components of molten steel are similarly divided into several groups.

For example, the amount of the inert gas blown in is set to A, and the condition that the flow rate is 10 liters / minute or more is A
The condition of 1, 10 to 8 liters / minute is divided into A2, and the condition of 8 to 6 liters / minute is divided into A3 and designated. Similarly, by setting the casting speed as item B and the tundish weight as item C, the operating condition can be changed to A1, B3, C.
It can be expressed as a combination of conditions of each item such as 2.D6.E3 ....

In order to select the optimum control method for the operating conditions expressed by this combination, the control method determined to be optimum during the operation is selected and switched. By switching in this way, when the level fluctuation range before switching is improved after switching, the combination of the operating conditions and the switched control method is collected as a level fluctuation suppression case. When the level fluctuation range increases after switching the control method,
A combination of the operating conditions and the switched control method is collected as an example of level fluctuation increase.

From the data thus collected, the entropy is calculated from the occurrence frequency of each item when switching to a certain control system, and the operating conditions are extracted in the ascending order of entropy to create a control system decision tree. It is possible to determine the operating conditions as described above using the control method decision tree thus created and automatically switch to the control method that further suppresses level fluctuations.

FIG. 1 conceptually shows an example of the above decision tree.

In FIG. 1, similarly to the above, A is the amount of inert gas blown, B is the casting speed, C is the tundish weight, D is the molten steel temperature, and E is the tundish switch. Of each item, and the attached numbers represent the difference in each state and condition. However, here, in A1, the amount of the inert gas blown is in the proper range (8 to 10 l).
/ Min), A3 means too much.

The flow in which the control method is determined by the above decision tree will be briefly described. First, at A1, it is determined whether the amount of clogging in the nozzle is in the amount of the inert gas blown within the appropriate range. If there is no problem with the degree of clogging, it is determined in B2 whether the casting speed is in the stable region, and if there is no steel type switching in F2, the disturbance observer control is performed to estimate and compensate for a relatively small disturbance or fluctuation. If A1 is NO, similarly, D5 indicates that the molten steel temperature is appropriate, and C3 indicates that the molten steel weight in the tundish is stable.
It is determined that there is no tundish switching, etc., and the control is switched to the H∞ control that estimates and compensates for periodic disturbances and fluctuations. Although details are omitted, the adaptive control is also switched in the same manner. The data collected to apply to this decision tree indicate casting conditions and states, and attributes are determined by the operator.

[0030]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 2 is an explanatory view showing a schematic configuration of an optimum control system switching system for the molten steel level in the continuous casting machine mold which is an embodiment according to the present invention.

The optimum control system switching system of this embodiment is
A continuous casting machine substantially the same as that shown in FIG. 4 (however,
Instead of the stopper 28, a sliding nozzle driven by a cylinder 28A is used), and the entire system is controlled by the computer 30.

The operation data of this continuous casting machine is collected in the operation data collection unit 32, and the collected operation data is generated from the operation data collection unit 32 by the level fluctuation suppression decision tree created by the method described in detail later. 34 and level fluctuation increase decision tree 3
6 and the operating condition determination unit 38.

Further, when the result processed by the level fluctuation suppression decision tree 34 and the level fluctuation increase decision tree 36 is input to the control method switching judgment section 40 and the optimum control method is judged based on the result, The determination result is the optimum control method switching unit 4
2, the optimum method is selected from among the three control methods of adaptive control, disturbance observer control, and H infinity control based on the determination result, and the method is switched to that method as necessary. When the control system is switched to the optimum control system in this manner, the actuator command value required for the control is output to the cylinder 28A based on the deviation between the current molten steel level obtained from the input level signal and the target value. It has become.

FIG. 3 is an explanatory diagram showing a schematic configuration of a system for automatically learning the decision tree of the optimum control method from the operation data. The level fluctuation suppression decision tree 34 and the level fluctuation increase decision tree 36 are the system. Is created as follows.

In the operation data collecting section 32, the gas flow rate,
The operation data such as casting speed, tundish weight, level fluctuation range, molten steel temperature, mold width, secondary cooling water flow rate and steel type is collected online, and the collected operation data is the operation condition collection unit 44, and the above three controls are performed. It is converted to a combination of predetermined conditions as the current operating conditions taking the system into consideration.

Next, during normal operation, the control system is switched when the operator determines that it is necessary to switch the control system from the state of the mold level. By changing the control method in this way, the level fluctuation determination unit 46
If it is determined that the level fluctuation is suppressed more after the switching than when the control method is switched, the operating condition before the switching and the control method after the switching are compared with the level fluctuation suppression decision tree 3
4 to the case collection unit 48A for creating. On the contrary, when it is determined that the level fluctuation is deteriorated by switching the control method, the operating condition at that time and the control method after the switching are similarly input to the case collection unit 48B for creating the level fluctuation increase decision tree 36. hand over.

After collecting about 200 to 300 cases of the control method for suppressing the level fluctuation and about 200 to 300 cases of the control method for exacerbating the level fluctuation, the case collecting sections 48A and 48B collect the case and the decision tree creating section 50A. Then, the frequency of occurrence of each operating condition is obtained, the entropy is calculated from this frequency of occurrence, and the operating conditions with smaller entropy are sequentially extracted.

By repeating this until all the cases disappear, a decision tree 34 for determining the level control method for suppressing the level fluctuation is obtained from the case where the level fluctuation is suppressed, and the case where the level fluctuation is aggravated. A decision tree 36 that determines the level control method that aggravates the level fluctuation is obtained from. The above two decision trees 34 thus obtained,
36 is applied to the optimum control method switching system of FIG.

In this embodiment, after the two types of decision trees 34 and 36 created as described above are incorporated into the system of FIG. 2, the operating data collecting unit 28 collects operating conditions online, for example, periodically. The operating conditions are input to the control method decision trees 34 and 36, respectively.

The two decision trees 34 and 36 determine a control method for suppressing the level fluctuation and a control method for aggravating the level fluctuation with respect to the input operating conditions.
The two control methods determined in this way are input to the control method switching determination unit 40, where it is determined whether or not to switch the control method.

That is, when the control methods respectively determined by the two decision trees 34 and 36 are exactly the same, it cannot be determined whether the level fluctuation is suppressed or deteriorated by switching to the control method. Do not switch the method. On the contrary, when a different control method is obtained, the control method obtained from the level fluctuation suppression decision tree 34 is selected,
Switch to it.

By switching in this way, it becomes possible to automatically switch to the optimum level control system for the operating conditions, and the level fluctuation can be suppressed regardless of the operating conditions, and stable operation can be performed. It will be possible.

In this embodiment, the operating condition judging section 38
Is a non-steady state (when changing the pan,
In order to avoid (when the pouring speed is changed, when the target level is changed, when the tundish is lowered, etc.), it has a function of checking whether or not the operating conditions are in these unsteady states. Control switching during an emergency is an unstable state, so
In order to avoid such a state, the operating condition judging section 38 is added.

As described in detail above, according to the present embodiment, by learning the data when the molten steel level control in the mold of the continuous casting machine is selected by human judgment, the optimum condition is automatically obtained from the operating conditions. It is possible to create a decision tree that determines the control method, automatically determine the optimum control method by using the obtained decision tree, and switch to that method. Can also suppress the level fluctuation.

In this embodiment, the learning tree is divided into a case in which the level fluctuation is suppressed and a case in which the level fluctuation is deteriorated, and a decision tree is created for each case. Therefore, the latter decision tree exacerbates the level fluctuation in advance. Since the control method can be determined, it is possible to prevent the level fluctuation from becoming worse by switching the control method.

Although the present invention has been specifically described above, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the invention.

For example, the selection of the optimum control method is not limited to the one using the above decision tree. Another method to switch to the optimum control method is to use a neutral network for the learning logic, that is, a neural net that derives the optimum control method from the operating conditions is configured in advance by learning, and the neural network is used to optimize. A method of selecting a control method can be mentioned.

Further, even when the decision tree is adopted, it is not limited to the case of using the two kinds of decision trees shown in the above embodiment, and only the level fluctuation suppressing decision tree may be adopted.

[0050]

As described above, according to the present invention, the optimum control method at that time is selected from among a plurality of optimum control methods for different types of molten steel level fluctuations, based on the operation information at the time of operation. Can be selected. Therefore, under any operating condition, the fluctuation of the molten steel level can be suppressed and minimized, so that stable operation can be always performed.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing an outline of an example of a control method decision tree.

FIG. 2 is an explanatory diagram showing a schematic configuration of an optimum control method switching system according to an embodiment of the present invention.

FIG. 3 is an explanatory diagram showing a learning system for creating a decision tree for selecting an optimum level control method applied to this embodiment.

FIG. 4 is an explanatory diagram showing a schematic configuration of a normal continuous casting machine.

[Explanation of symbols]

 10 ... Molten steel 12 ... Ladle 14 ... Tundish 16 ... Immersion (immersion) nozzle 18 ... Mold 20 ... Pinch roll 22 ... Cutter 24 ... Cast slab 26 ... Level gauge 28 ... Stopper 28A ... Actuator cylinder 30 ... Calculator 32 ... Operating data Collection unit 34 ... Level fluctuation suppression decision tree 36 ... Level fluctuation increase decision tree 38 ... Operating condition determination unit 40 ... Control method switching determination unit 42 ... Optimal control method switching unit

Claims (2)

[Claims] 1. When controlling the level of molten steel in a mold of a continuous casting machine, adaptive control that adapts to large fluctuations in the time constant and updates the optimum control gain, and estimates disturbances and fluctuations with relatively small time constants. Based on the operating information, the disturbance observer control that compensates for the disturbance and the H infinity control that suppresses disturbances and fluctuations in the frequency band of the molten steel level fluctuation in the mold are suppressed based on the operating information during operation. An optimum control system switching system for a molten steel level in a continuous casting machine mold, comprising a control system switching means for selecting a control system determined to be possible. 2. The continuous control system according to claim 1, wherein the control system switching means has a function of selecting an optimum control system using a control system decision tree created based on past operation information. Casting machine Optimal control system switching system for molten steel level in mold.