JPH0691186A - Method for controlling roller mill - Google Patents

Abstract

PURPOSE:To provide a method for controlling roller mills, by which the operating condition of a roller mill not relatively easy to operate is automatically and/or optimally controlled. CONSTITUTION:A mill characteristics model showing the relationship between the variation of operating condition data detected by a group of detectors 2 and the manipulated variable from the operation set values of control factors corresponding to the variation, a restraining condition about the operating condition data and operation set values and an evaluation function for evaluating the operating condition data and operated set values are previously stored. Then, a simplex computing element 5 of an optimum control computing element 3 processes the operating condition data detected by the group of detectors 2 using the mill characteristics model (in the figure, 6), restraining condition in the figure 6 and evaluation function (in the figure, 6) by a simplex algorithm and calculates the manipulated variable of each operation set value so that the evaluation function 6 may be optimized under conditions meeting the restraining condition 6 to output the each operation set value to a main controller 4.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a cement raw material,
The present invention relates to a roller mill control method for optimally controlling a vertical roller mill that grinds cement clinker, granulated slag, and the like.

[0002]

2. Description of the Related Art The above-described roller mill has a very great merit that it has a high grinding efficiency (that is, a low power consumption) as compared with a ball mill known as a production facility. However, since the control of the roller mill is very sensitive to disturbance such as pulverization characteristics of the raw material, the roller mill is a system that is difficult to operate. Therefore, at present, as shown in FIG. 5, the roller mill is operated depending on the qualitative know-how of the skilled operator. For example, a skilled operator, while monitoring operating conditions such as product quality and mill vibration in the roller mill, sets operational values such as separator rotation speed setting value, tension hydraulic pressure setting value, power setting value, etc. so that the operating status is suitable. I was adjusting the value.

[0003]

By the way, when the operation set value of the roller mill is not properly set, the following abnormal phenomenon occurs. For example, when the vibration of the roller mill increases and the vibration exceeds a preset vibration trip set value, the roller mill is forcibly stopped. After that, even if it is restarted, about 1 immediately after the restart
Over time, the product quality will be poor and defective products will be produced. In the worst case, the vibration may cause damage to the roller mill body. Therefore, in order to prevent the failure of the roller mill body and the reduction of the yield, it is necessary to reduce the vibration of the roller mill and operate the roller mill so as not to stop it as much as possible. In addition, it may not be possible to maximize the ability of the roller mill, which is evaluated by the amount of raw material crushed per unit time and the amount of electric power used. Further, the product quality may be deteriorated due to the inappropriate operation setting value as described above. In order to avoid such abnormal phenomenon,
While a skilled operator who constantly monitors the operating condition of the roller mill is required, a great deal of time and experience are required to train the skilled operator who can suitably control the operating condition. Therefore, there has been a demand for suitable and automated control of the roller mill. Therefore, fuzzy control has been developed in which the know-how of a skilled operator is expressed by fuzzy rules (Japanese Patent Laid-Open No. 122848/1990).
issue). However, even in this case, it was difficult to integrally correct many setting values such as the pressing force, the crushing amount, the separator rotation speed, the air flow, and the table rotation speed. Therefore, an object of the present invention is to provide a roller mill control method capable of automatically and / or optimally controlling the operating state of a roller mill, which is relatively easy to operate.

[0004]

In order to achieve the above object, the main means adopted by the first invention is to adjust the operation set value of the control factor of the roller mill based on the operation state data of the roller mill. In the roller mill control method for controlling the operating state of the roller mill according to the above, the operating state data is repeatedly detected, and the change amount of the operating state data detected before and after is calculated as the change amount of the operating state data and the operation corresponding thereto. The roller mill control method is characterized in that a new operation set value is obtained by applying it to a mill characteristic model that represents a preset relationship with an operation amount from a set value. Further, in the roller mill control method for controlling the operation state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operation state data of the roller mill, the operation state data is repeatedly detected and detected. An evaluation function for evaluating the operation state data and / or the operation set value under the condition that a preset constraint condition is satisfied with respect to the performed operation state data and / or the operation set value corresponding to the operation state data. It is a roller mill control method characterized by selecting a new operation set value so as to optimize. Furthermore, based on the operation status data of the roller mill,
In the roller mill control method for controlling the operation state of the roller mill by adjusting the operation set value of the control factor of the roller mill, the operation state data is repeatedly detected, and the change amount of the operation state data detected before and after A new operation set value is obtained by applying it to a mill characteristic model that represents a preset relationship between a change amount of operation state data and a corresponding operation amount from the operation set value, and the detected operation state described above. Under the condition that the preset condition regarding the data and / or the new operation setting value is satisfied, the new evaluation function for optimizing the operation state data and / or the operation setting value is optimized. A roller mill control method characterized by selecting an operation set value.

The main means adopted by the second invention is a roller mill control method for controlling the operating state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operating state data of the roller mill. In the above, the operation state data is repeatedly detected, and the change amount of the operation state data detected before and after is set in advance by the change amount of the operation state data and the operation amount from the operation set value corresponding thereto. A new operation set value is obtained by applying it to the mill characteristic model representing the relationship, and the operation state data detected when the operation state of the roller mill is controlled by the obtained new operation set value and the obtained new operation set value. It is configured as a roller mill control method characterized by updating the mill characteristic model based on an operation set value. Further, in the roller mill control method for controlling the operating state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operating state data of the roller mill, the operating state data is repeatedly detected, The amount of change in the detected operating state data is
A new operation set value is obtained by applying it to a mill characteristic model representing a preset relationship between a change amount of the operation state data and a corresponding operation amount from the operation set value, and the detected operation Under the condition that the preset condition is satisfied with respect to the state data and / or the new operation set value, the new evaluation function is optimized to evaluate the operation state data and / or the operation set value. Of the mill characteristic model based on the operating state data detected when the operating state of the roller mill is controlled by the selected new operating set value and the selected new operating set value. Is a method for controlling a roller mill. Furthermore,
The roller mill control method according to claim 4 or 5, wherein the mill characteristic model is updated by performing a state estimation using a Kalman filter. Furthermore, the rolling mill control method according to claim 4 or 5, wherein the mill characteristic model is updated by performing a multiple regression analysis while partially updating the held data.

[0006]

In the roller mill control method according to the first aspect of the present invention, the mill characteristic model representing the relationship between the change amount of the control state data of the roller mill and the operation amount corresponding to the change amount of the control factor of the roller mill is preset. It is set. Therefore, the operating state data of the roller mill is repeatedly detected, and the change amount of the operating state data detected before and after is applied to the mill characteristic model to obtain a new operation set value. Therefore, the operating state of the roller mill is automatically controlled. Further, the operating condition of the roller mill is repeatedly detected, and the constraint condition regarding the detected operating condition data and / or the operation set value corresponding to the operating condition data is preset. Therefore, in order to optimize the evaluation function for evaluating the operation state data and / or the operation set value under the condition that the above constraint condition is satisfied,
From the above operation set values, new operation set values for appropriately controlling the operating state of the roller mill are selected. Therefore,
The operating condition of the roller mill is optimally controlled. Furthermore,
A mill characteristic model representing the relationship between the amount of change in the control state data of the roller mill and the operation amount from the operation set value of the corresponding control factor of the roller mill is preset. Therefore,
The operating state data of the roller mill is repeatedly detected, and the change amount of the operating state data detected before and after is applied to the mill characteristic model to obtain a new operation set value. On the other hand, the detected operating condition data and / or the constraint condition regarding the new operation set value is also preset. Therefore, under the condition that the above constraint condition is satisfied, from the new operation set value obtained from the mill characteristic model, in order to optimize the evaluation function for evaluating the operation state data and / or the operation set value, , A new operation set value is selected to suitably control the operating state of the roller mill. Therefore, in this case, by using an appropriate model, it is possible to greatly speed up the optimization. As a result, it is possible to comprehensively correct many operation set values, which were difficult with conventional fuzzy control and the like, and the operating state of the roller mill is automatically and optimally controlled.

Further, in the roller mill control method according to the second aspect of the present invention, a mill characteristic model representing a relationship between the amount of change in the control state data of the roller mill and the corresponding operation amount from the operation set value of the roller mill control factor. Is preset. Therefore, the operating state data of the roller mill is repeatedly detected, and the change amount of the operating state data detected before and after is applied to the mill characteristic model to obtain a new operation set value. The mill characteristic model is updated based on the operating state data detected when the operating state of the roller mill is controlled by the obtained new operating set value and the obtained new operating set value. Therefore, the mill characteristic model is updated in accordance with changes over time in the operating state of the roller mill and changes in the properties of the raw materials, so that the operating state of the roller mill can always be kept optimal. As a result, even with this method, the operating condition of the roller mill can be automatically and optimally controlled. Furthermore, a mill characteristic model representing the relationship between the amount of change in the control state data of the roller mill and the corresponding operation amount from the operation set value of the control factor of the roller mill is set in advance. Therefore, the operating state data of the roller mill is repeatedly detected, and the change amount of the operating state data detected before and after is applied to the mill characteristic model to obtain a new operation set value. On the other hand, the detected operating condition data and / or the constraint condition regarding the new operation set value is also preset. Therefore, under the condition that the above constraint condition is satisfied, from the new operation set value obtained from the mill characteristic model, in order to optimize the evaluation function for evaluating the operation state data and / or the operation set value, , A new operation set value is selected to suitably control the operating state of the roller mill. The mill characteristic model is updated based on the operating state data detected when the operating state of the roller mill is controlled by the selected new operating set value and the selected new operating set value. Therefore, in this case as well, the mill characteristic model is updated in accordance with changes over time in the operating state of the roller mill and changes in the properties of the raw material. It is possible to greatly speed up the process to the operating state. As a result, the operating state of the roller mill can be controlled automatically, optimally and quickly. Further, the mill characteristic model is updated by performing a state estimation using a Kalman filter or performing a multiple regression analysis while partially updating the held data. Whichever updating method is used, it is possible to reliably update the mill characteristic model according to changes in the operating state of the roller mill.

[0008]

The embodiments of the present invention will be described below with reference to the accompanying drawings to provide an understanding of the present invention. The following embodiments are examples of embodying the present invention and are not intended to limit the technical scope of the present invention. here,
FIG. 1 is a schematic configuration diagram showing a roller mill to which a roller mill control method according to an embodiment of the present invention (first and second inventions) is applied, and FIG. 2 is an outline of a first roller mill control device 1 for controlling the roller mill. FIG. 3 is a block diagram showing a configuration, FIG. 3 is a block diagram showing a schematic configuration of a second roller mill control device 1'for controlling the roller mill, and FIG. 4 is a flow chart showing an example of logical operation of the second roller mill control device 1 '. . In the roller mill 10 to which the roller mill control method according to the first and second embodiments of the present invention is applied, as shown in FIG. 1, the material is introduced from the raw material introduction port 14 on the side of the mill casing 11 onto the rotating table 12. The raw material is crushed and crushed between the roller 12 which rotates in cooperation with the table 12 and the table 12. Then, the particles which have been made finer than a predetermined particle size from the above raw material are supplied from the hot air generating furnace 16 to the air supply port 21.
It is blown upward by the hot air blown into the mill casing 11 through. The particles blown up are separated by the separator 19 that is rotated by the drive of the motor 20.
Is classified into coarse particles and fine particles, and the fine particles are introduced together with hot air from the mill outlet 15 at the top of the mill casing through the feed duct 17 to the bag filter 22 and collected as a product. On the other hand, the coarse particles classified by the separator 19 fall in the mill casing 11, are returned to the table 12 again, and are crushed. Then, the particles that have not been blown up from the vicinity of the table 12 by the hot air fall to the bottom of the mill casing 11, and are returned from the bottom to the raw material side by a basket elevator or the like (not shown). In addition, the product particle size distribution is obtained for the fine-grained product collected by the bag filter 22. When the particle size of the product is too fine in view of this product particle size distribution, the motor 20 is driven so that the rotation speed of the separator 19 is decreased, and conversely, when the product particle size is too coarse, the rotation speed of the separator 19 is increased. Controlled. If the value of the mill vibration detected by the vibration detection sensor (not shown) attached to the speed reducer between the table 12 and the motor 18 for rotating the table 12 becomes excessive, the roller mill 1
Since mechanical damage to each part is caused, if the value of the mill vibration exceeds, for example, 6 mm / sec, the operation of the roller mill 10 is urgently stopped by the action of a trip circuit (not shown). That is, the operating state of the roller mill 10 is controlled by adjusting the operation set value such as the separator rotation speed which is a control factor of the roller mill 10 based on the operating state data such as the mill vibration of the roller mill 10.

Here, the roller mill control method according to the first aspect of the present invention has the following or configuration. The operation state data of the roller mill is repeatedly detected, and the change amount of the operation state data detected before and after is determined by a preset relationship between the change amount of the operation state data and the operation amount from the corresponding operation set value. This is a method of applying the present mill characteristic model to obtain a new operation set value. The operation state data of the roller mill is repeatedly detected, and the operation state data and / or the operation state data is detected under the condition that preset constraint conditions are satisfied with respect to the detected operation state data and / or the operation set value corresponding to the operation state data. This is a method of selecting the new operation set value so as to optimize the evaluation function for evaluating the operation set value. The operation state data of the roller mill is repeatedly detected, and the change amount of the operation state data detected before and after is determined by a preset relationship between the change amount of the operation state data and the corresponding operation amount from the operation set value. Is applied to a mill characteristic model that represents a new operating set value, and the operating state is detected under the condition that the detected operating state data and / or the new operating set value satisfy a preset constraint condition. It is a method of selecting the new operation set value so as to optimize data and / or an evaluation function for evaluating the operation set value.

In the first invention, the above method is embodied by the first roller mill controller 1 for controlling the roller mill 10 of this embodiment. That is, as shown in FIG. 2, the apparatus 1 includes, as shown in FIG. 2, raw material properties such as particle size and hardness, the crushed layer thickness on the table 12, the mill vibration, the mill power (electric power) required for the motor 18, the mill differential pressure, Detector group 2 consisting of multiple detectors (not shown) that detect product quality such as particle size and particle size distribution, and operation data of roller mill 10 (mill differential pressure, product particle size distribution, mill power, mill vibration, etc.) Of the control factor of the roller mill 10 (roller pressing force, raw material input amount, separator rotation speed, blowing hot air amount) corresponding to these changes, the mill characteristics representing a preset relationship. Using a model, the detected operating condition data, and preset constraint conditions for the new operating setting values obtained based on the operating condition data, and the evaluation function for evaluating the operating condition data and operating setting values. And an optimal control arithmetic unit 3 for selecting an operation amount of the operation set value so as to optimize the evaluation function under the condition that the constraint condition is satisfied by using a mathematical programming method such as a simplex method. Based on the operation set value selected by the device 3, the operation set values of the above-mentioned control factors (separator rotation speed, tension hydraulic pressure that determines the roller pressure, raw material supply amount, hot air blowing amount) are adjusted, and the roller mill 10 It comprises a main controller 4 for controlling the operating state. The method described above can be embodied by a similar device, and thus the description thereof will be omitted.

An example of the logical operation of the device 1 will be described below. Here, the following specifications were used as the operation setting values of the above control factors. x ₁ ; roller pressure (tension oil pressure) (%) ... “Lower is preferable” x ₂ ; Raw material input amount (t / h)… “Higher is preferable” x ₃ ; Separator rotation speed (rpm) x ₄ ; Blow-in hot air flow rate (m ³ / min) As the operation state data of the roller mill 10, the following specifications were used. y ₁ ; Mill differential pressure (mm aq.) ... “Predetermined set value” y ₂ ; Product particle size distribution (%)… “Preset set value” y ₃ ; Mill power (kw) ... "it is preferably small" y _4; mill vibration (mm / sec) ... "lesser preferably" Then, showing the following equation (1) as the mill characteristic model.

[Equation 1] Here, Δx _{1 to} Δx ₄ are respective operation amounts corresponding to the above operation set values x _{1 to} x ₄ .

Further, Δy _{1 to} Δy ₄ represent respective amounts of change corresponding to the operating state data y _{1 to} y ₄ . Further, each element a _{11 to} a ₄₄ of the coefficient matrix in the equation (1) is a time average value of each operation set value and each operation state data sampled at each constant sampling time within a predetermined period, These are the coefficients obtained by applying the multiple regression analysis method to these time average values. Further, as the constraint conditions, the conditional expressions (2) to (9) are preset as shown below. 0 ≦ x ₁ + Δx ₁ ≦ 200 (2) 0 ≦ x ₂ + Δx ₂ (no upper limit) ... (3) 0 ≦ x ₃ + Δx ₃ (no upper limit) (4) 0 ≦ x ₄ + Δx ₄ ≦ 270… ( 5) 250 ≦ y ₁ + Δy ₁ ≦ 300 (6) 18 ≦ y ₂ + Δy ₂ ≦ 22 (7) 0 ≦ y ₃ + Δy ₃ ≦ 175 (8) 0 ≦ y ₄ + Δy ₄ ≦ 6 (9) Further, the evaluation function can be expressed by a function P shown by the following expression (10). P = 30 [(y ₃ -175) / 175] _{+50 (y 4/6) +30} (x 1/100) +30 [(7-x ₂₎ / 7] ... (10) to the evaluation function P to minimize As above y ₃ , y ₄ , x
₁ and x ₂ are determined. P → MIN (minimization) ... (11) In this example, the constraint conditions of the mill characteristic model shown in the above equations (1) to (10) and the evaluation function P are both preset and stored in the memory of the optimum control computing unit 3. 6 is stored.

The simplex computing unit 5 of the optimum control computing unit 3 is mainly composed of a central processing unit (CPU) (not shown), and stores a symptom algorithm, which is an example of mathematical programming, in a ROM (not shown). The change amounts Δy _{1 to} Δy ₄ of the operation state data from the detector group 2 from the operation state data at the previous sampling time are obtained, and these change amounts are applied to the mill characteristic model to set each operation set value. Change amounts Δx _{1 to} Δx ₄ are calculated. Various combinations are conceivable as new operation set values for obtaining a suitable operating state of the roller mill 10. The simplex calculator 5 selects new operation set values x ₁ + Δx _{1 to} x ₄ + Δx ₄ from the combinations thereof so as to optimize the evaluation function P under the condition that the constraint condition is satisfied. Output to the output device 7. Therefore, the manipulated variable output device 7 obtains the manipulated variables Δx _{1 to} Δx ₄ for each of the selected new manipulated set values and outputs them to the main controller 4. Then, the main controller 4 adjusts the current operation set value based on each operation amount from the operation amount output device 7. As a result, the operating state of the roller mill device 10 is optimally controlled.

Next, another example of the logical operation of the device 1 will be described. In this example, the handling amount is made dimensionless by using the rate of change of these values instead of the above-mentioned manipulated variable and change amount, and a value that can be taken as the central value of the permissible range is brought. Here, the following parameters were used as the operation setting values of the control factor. x ₁ ; Roller pressure (tension hydraulic pressure) (kgf / cm ² ) ... “Lower is preferable” x ₂ ; Raw material input (t / h)… “Higher is preferable” x ₃ ; Separator rotation speed (rpm) x ₄ ; Blow-in hot air flow rate (m ³ / min) Further, as the operation data of the roller mill 10, the following specifications were used. y ₁ ; 90% pass particle size (μ) ... “The center value of the allowable range is good” y ₂ ; Mill power (kw)… “It is preferable that it is small” y ₃ ; Mill vibration (mm / sec)… “It is small” Then, the following equation (12) is shown as a mill characteristic model.

[Equation 2]Where Δ′x₁~ Δ'x_FourIs the operation setting value x₁~ X
_FourEach manipulated variable Δx corresponding to₁~ Δx_FourAnd the operation
It is the ratio with the work setting value, and Δ′y₁~ Δ'y ₃Is above
Operation status data y₁~ Y_FourChange amount corresponding to
Δy₁~ Δy₃And the corresponding operation status data
ing. In addition, each element a of the coefficient matrix in equation (12)₁₁~
a₃₄At a fixed sampling time within a predetermined period
Each operation set value and each operation status data collected
Calculate the time averaged values and perform multiple regression analysis on these time averaged values
It is a coefficient obtained by applying the method. Furthermore, the above constraint conditions
As shown below, the conditions of equations (13) to (19) are
The expression is preset. 100 ≤ x₁+ X₁Δ'x₁ ≤ 150 (13) 100 ≤ x₂+ X₂Δ'x₂ ≤ 140 (14) 60 ≤ x₃+ X₃Δ'x₃ ≤ 80 ... (15) 4200 ≤ x_Four+ X_FourΔ'x_Four ≤ 4800… (16) 39.42 ≤ y₁+ (Y₁-B) Δ'y₁≦ 41.41… (17) 2400 ≦ y₂+ Y₂Δ'y₂ ≤ 3000 (18) 0 ≤ y₃+ Y₃Δ'y₃ ≤ 6 (19) (b: constant)

The evaluation function can be expressed by the function P ₁ shown in the following equation (20).

[Equation 3] Here, C represents the central value of the allowable range of y ₁ (C = (3
9.42 +41.41) / 2). The above y ₂ , y ₃ and x ₂ are determined so as to minimize this evaluation function P ₁ . P ₁ → MIN (minimization) ... (21) In this example, the constraint condition of the mill characteristic model shown in the above equations (12) to (20) and the evaluation function P ₁ are both set in advance and the optimum control calculator 3 Stored in the memory 6. The simplex calculator 5 of the optimum control calculator 3 is a ratio of the change amount of the operation state data from the detector group 2 from the operation state data at the previous sampling time to the operation state data as in the above-described embodiment. Δ'y ₁ ~ Δ'y ₃
Then, the ratios Δ′y _{1 to} Δ′y ₃ are applied to the mill characteristic model to obtain the ratios Δ′x _{1 to} Δ′x ₄ of the change amount of each operation set value and the operation set value. . The simplex computing unit 5 optimizes the evaluation function P ₁ under the conditions satisfying the above-mentioned constraints, so that new operation set values x ₁ (1 + Δ′x ₁ ) to x ₄ (1 + Δ ′) are obtained from the combinations thereof.
x ₄ ) are selected and output to the manipulated variable output device 7. Therefore, the manipulated variable output device 7 controls the respective manipulated variables Δx _{1 to} Δx with respect to the selected new manipulated set values.
₄ is obtained and output to the main controller 4. Then, the main controller 4 adjusts the current operation set value based on each operation amount from the operation amount output device 7. Thereby, the roller mill device 1
The operating state of 0 is optimally controlled, and in this example, the manipulated variable and the variable amount are made dimensionless, which facilitates the calculational handling of these amounts and brings the desired value to the center value of the allowable range. be able to. Therefore, stable control can be performed without the risk of deviation from the constraint conditions.
In the above two examples, among the methods of the first invention, the above method, that is, the method of performing optimization by the evaluation function under the constraint condition has been described. The roller mill 10 may be controlled using only the mill characteristic model without performing such optimization. In that case, basic automatic control is performed, but when fluctuations in the operating state of the roller mill 10 are not so large, such a simple method is sufficient.

Further, the roller mill 10 may be controlled by only optimizing the operation set value by the evaluation function under the constraint condition without using the mill characteristic model as in the above method. In that case, since optimization control can be performed, there is an advantage that power consumption can be reduced, for example. The case where such methods are combined corresponds to the method, but in the method, more excellent controllability is obtained by the synergistic effect of the methods as described above. In other words, by using an appropriate model, it is possible to greatly speed up the process of optimization. As a result, the pressing force that was difficult with conventional fuzzy control,
It is possible to comprehensively correct the operation setting values such as the crushing amount, the rotation speed of the separator, the air flow amount, etc., and it is possible to automatically and optimally control the operating state of the roller mill. In this way, the mill characteristic model applied to any one of the roller mill control methods according to the first aspect of the present invention is obtained in advance using multiple regression analysis and is fixed. However, the operating condition of the roller mill 10 is such that the crushing part is worn and the crushing characteristics change with the lapse of operating time. In addition, the properties of the raw materials have changed. Therefore, if the mill characteristic model is updated in accordance with these changes, the operating state of the roller mill 10 can always be kept in an optimum state. The following second invention focuses on this point.

The roller mill control method according to the second invention will be described below. The roller mill control method according to the second aspect of the invention is configured as described below or as follows. The operation state data of the roller mill is repeatedly detected, and the change amount of the operation state data detected before and after is determined by a preset relationship between the change amount of the operation state data and the operation amount from the corresponding operation set value. A new operation set value is obtained by applying to the mill characteristic model represented, and the operation state data detected when the operation state of the roller mill is controlled by the obtained new operation set value and the obtained new operation setting This is a method of updating the mill characteristic model based on the value. The operation state data of the roller mill is repeatedly detected, and the change amount of the operation state data detected before and after is determined by a preset relationship between the change amount of the operation state data and the operation amount from the corresponding operation set value. A new operation set value is obtained by applying it to the mill characteristic model represented, and the operation state data is obtained under the condition that the detected operation state data and / or the new operation set value satisfies a preset constraint condition. And / or when the new operation set value is selected so as to optimize the evaluation function for evaluating the operation set value, and the operation state of the roller mill is controlled by the selected new operation set value. A method for updating the mill characteristic model based on the detected operating state data and the selected new operation set value. In the second invention, the above method is embodied by the second roller mill controller 1 '. That is, as shown in FIG. 3, this apparatus 1'has a configuration in which a model updating unit 8 for updating a mill characteristic model is newly added to the first roller mill control apparatus 1 according to the first aspect of the present invention. Since the above method can be embodied by a similar device, the description thereof will be omitted.

Next, an example of the logical operation of the device 1'will be described. In this example, the mill characteristic model is updated by performing state estimation using a so-called Kalman filter. Here, the following parameters were used as the operation setting values of the control factors. x ₁ ; Roller pressure (tension hydraulic pressure) x ₂ ; Grinding amount x ₃ ; Separator rotation speed x ₄ ; Blow hot air flow rate x ₅ ; Table rotation speed Also, the following parameters are used as the operating status data of the roller mill 10. I was there. y ₁ ; Blaine value y ₂ ; AJS 30μ residue y ₃ ; basic unit (= grinding amount / mill power) y ₄ ; mill differential pressure y ₅ ; layer thickness y ₆ ; layer thickness fluctuation range Expression (22) is shown.

[Equation 4]

Here, Δx _j is the operation set value x _j (x
Each operation amount corresponding to _{1 to} x ₅ ) and Δy _i represent each change amount corresponding to the operation state data y _i (y _{1 to} y ₆ ). Further, each element a _ij of the coefficient matrix in the equation (22) obtains a time average value of each operation set value and each operation state data sampled at every constant sampling time within a predetermined period, and obtains these time average values. Is a coefficient obtained by applying the multiple regression analysis method to. Furthermore, as the constraint conditions, as shown below, (23) to (25)
The conditional expression of the expression is set in advance. Grain size constraint c _i + ε _i ≤y _i + Δy _i ≤d _i −ε _i (i = 1 to 2) (23) ・ Basic unit constraint y ₃ + Δy ₃ ≥e ₃ + ε ₃ (24) ・ Vibration constraint g _k ≦ f _k (x _k + Δx _k ) ≦ h _k (k = 1 to n) (25) where c _i , d _i , e ₃ , g _k , and h _k are user set values, and ε ₃ is a model. The error, f _k, is a function expression for the nonlinear optimization method with inequality constraints (iterative calculation).

Further, as the above evaluation function, the following (2
It can be expressed by the function p ₂ shown by the equation (6).

[Equation 5] Here, w _i represents a weight, which is also a user setting value. The manipulated variable Δx _j (j = 1 to 6) is determined so as to maximize the evaluation function P ₂ . P ₂ → MAX (maximization) (27) In this example, the constraint conditions of the mill characteristic model shown in the above equations (23) to (25) and the evaluation function P ₂ are both set in advance and the optimum control computing unit is set. 3 is stored in the memory 6. Simplex calculator 5 of the optimum control calculator 3
Is the variation Δy of the operating state data from the detector group 2 from the operating state data at the previous sampling time.
_i is obtained, and these variations are applied to the mill characteristic model to obtain the variation Δx _j of each operation set value. The simplex calculator 5, so as to optimize the evaluation function p ₂ above satisfy a constraint _condition, the new operation setting combinations thereof values ^{_{x j New = x j Old +}} α j Δx
Select _j (j: gain) and output it to the manipulated variable output device 7.
Therefore, the manipulated variable output device 7 determines each manipulated variable α _j Δx _j for the selected new manipulation set value x _j ^New and outputs it to the main controller 4. However, the gain α _j is a user setting value. Then, the main controller 4 adjusts the current operation set value based on the operation amount α _j Δx _j from the operation amount output device 7, but the operation state data y _i ^New and the new operation set value x _j at this time are adjusted. Update the mill characteristic model based on ^New and.

As the updating method, for example, the parameter identification by the known method of least squares is performed by the sequential calculation method. This is equivalent to performing state estimation by applying a standard Kalman filter to a system equation in which unknown parameters are regarded as state variables. Therefore, the solution can be expressed by the following equation as is well known. a '(N) = a' (N-1) -K (N) {x T (N) a '(N-1) -y (N)} ... (28) K (N) = P (N- 1) x (N) {1 + x T (N) P (N-1) x (N)} -1 ... (29) P (N) = {1-K (N) x T (N)} P (N −1) (30) where N = 1, 2, ... a ′ (N); Estimated value of coefficient a _ij of mill characteristic model x (N); New operation set value x _j ^New y (N) Operation state data y _i ^New P (N); covariance of estimation error initial values a ′ (0), P (0) = βI (β >> 0) The estimated value a ′ (N) of the coefficient a _ij of can be sequentially updated.

That is, a '(0), P (0) In this way, by controlling the operating state of the roller mill 10 while updating the mill characteristic model by the model updating unit 8, it is possible to perform control that follows changes over time in the operating state of the roller mill 10 and changes in the properties of raw materials. The operation state of the roller mill 10 can always be kept in an optimum state. Furthermore, due to the synergistic effect with the optimization using the evaluation function, it is possible to significantly speed up the operation of the roller mill 10 to the optimum operating state. As a result, the operating state of the roller mill 10 can be controlled automatically, optimally and quickly.

Next, another example of the logical operation of the device 1'will be described. In this example, the mill characteristic model is updated by repeating the multiple regression analysis while partially updating the retained data. That is, in this example, the basic logic itself uses the well-known multiple regression analysis used when initializing the coefficient a _ij of the mill characteristic model, but since its use is characteristic, refer to the flow chart shown in FIG. I will explain. As shown in FIG. 4, in this example, first, the number of data items to be subjected to the multiple regression analysis is input from, for example, a keyboard (not shown) (S1). Operating state data y detected by the detector group 2 and a new set value x corresponding thereto
Is sampled as operation data D ₁ (S2). After the operation data D ₁ is sampled, the data group ΣD _i accumulated in the memory (not shown) for multiple regression analysis when determining the initial value is further shifted by one set, and the oldest operation data is discarded. The data group ΣD _i ′ for this multiple regression analysis is created by adding the operation data D ₁ (S3). Using this data group .SIGMA.D _{i ',} performing a multiple regression _analysis, it calculates the partial regression coefficient (S4). The calculated new partial regression coefficient is output as the coefficient a _ij of the mill characteristic model (S
5). The mill characteristic model is updated by repeating S2 to S5 among these steps. The series of operations is controlled or executed by the model updating unit 8, and in this case also, the same effect as when the Kalman filter is used can be obtained. However, in this case, as described, there is an advantage that the multiple regression analysis which is the logic for initial setting of the coefficient a _ij of the mill characteristic model can be used as it is.

In the above two examples, the method of the second invention, that is, the method of performing optimization using the evaluation function under the constraint condition has been described. The mill characteristic model may be simply updated without using the function. In that case, as the new operation set value x corresponding to the operation data y of the roller mill 10, instead of the "selected new operation set value" by the optimization using the evaluation function, the mill characteristic before optimization is used. The "calculated new operation set value" using the model is set to ₁ of the operation data D ₁ .
Used as one. In this case, the mill characteristic model is updated in accordance with changes over time in the operating state of the roller mill 10 and changes in the properties of the raw materials, so that the operating state of the roller mill 10 can always be kept in an optimum state. As a result, the operating state of the roller mill can be automatically and optimally controlled. It should be noted that these methods differ from the method of the first aspect of the present invention in that the operation set value itself is optimized in that optimum control is performed by eliminating the offset in control. Therefore, when these are applied in combination, it is expected that even greater controllability will be obtained due to the synergistic effect. In the embodiments of the first and second inventions, when the constraint condition and the evaluation function are used, the operation state optimization calculation is performed using both the operation state data and the operation set value. However, it is also possible to perform an operation state optimization calculation using either the operation state data or the operation set value for the above constraint conditions and evaluation functions.

In the embodiments of the first and second aspects of the invention, the manipulated variable of the operation set value is determined so that the evaluation function is either minimized or maximized. This is because the evaluation function was created so that the evaluation function becomes the minimum or maximum when it becomes, and the operation amount of each operation setting value is calculated so that the evaluation function is transformed and the convergence value of this evaluation function is reversed. May be. Each element of the first and second roller mill control devices 1 and 1'according to the above first and second inventions may be configured as separate hardware, or each of software components of a computer. It may be configured as a process. As described above, the present invention (first, second
Inventions), first and second roller mill control devices 1,
According to 1 ', even when the operating state of the roller mill changes due to changes in the material to be crushed, etc., many operation set values that were difficult to achieve with conventional fuzzy control without relying on a skilled operator are comprehensive. It is possible to automatically operate the above operating state in an optimal state by correcting the above. For example, the vibration of the roller mill 10 is suppressed as much as possible, and continuous operation is possible without causing an operation stop due to a vibration trip. Further, it is possible to maximize the raw material processing amount of the roller mill 10, and it is possible to minimize the power consumption of the mill power required at this time. Further, it is possible to obtain the optimum product quality while ensuring the above-mentioned suitable operating condition.

[0026]

According to the first invention, in the roller mill control method for controlling the operating state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operating state of the roller mill, A mill that detects the state data repeatedly and represents the amount of change in the operating state data detected before and after the preset amount of change in the operating state data and the corresponding operation amount from the operation set value. A roller mill control method is provided which is applied to a characteristic model to obtain a new operation set value. As a result, the operating state of the roller mill, which is relatively easy to operate, can be automatically controlled. Furthermore, for evaluating the operating state data and / or the operating set value under the condition that the preset constraint condition is satisfied with respect to the detected operating state data and / or the operating set value corresponding to the operating data. A roller mill control method is provided which is characterized by selecting the new operation set value so as to optimize the evaluation function. This makes it possible to optimally control the operating state of the roller mill, which is relatively easy to operate. Further, the operation state data is repeatedly detected, and the change amount of the operation state data detected before and after is set in advance by the change amount of the operation state data and the operation amount from the operation set value corresponding thereto. A new operation set value is obtained by applying it to the mill characteristic model representing the relationship, and the above-mentioned operation is performed under the condition that preset constraint conditions are satisfied with respect to the detected operation state data and / or the new operation set value. A roller mill control method is provided, characterized in that the new operating setpoint is selected so as to optimize the state data and / or the evaluation function for evaluating the operating setpoint. In this method, more excellent controllability is obtained due to the synergistic effect obtained by combining the above two methods. In other words, by using an appropriate model, it is possible to greatly speed up the optimization process. This makes it possible to comprehensively correct a large number of operation set values, which were difficult with conventional fuzzy control, etc., and to automatically and optimally control the operating state of the roller mill.

According to the second invention, in the roller mill control method for controlling the operating state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operating state data of the roller mill, the operating state data Is repeatedly detected, and the change amount of the operating condition data detected before and after is expressed as a mill characteristic model showing a preset relationship between the change amount of the operating condition data and the corresponding operation amount from the operation set value. Based on the operation state data detected when the operation state of the roller mill is controlled by the obtained new operation set value and the obtained new operation set value. A roller mill control method is provided, which is characterized by updating the mill characteristic model. Therefore, since the mill characteristic model is updated in accordance with changes over time in the operating state of the roller mill and changes in the properties of the raw material, the operating state of the roller mill can always be kept in an optimum state. As a result, even with this method, the operating condition of the roller mill can be automatically and optimally controlled. Furthermore, the above-mentioned detected operational status data and / or
Alternatively, the new operation set value is optimized so as to optimize the operation state data and / or the evaluation function for evaluating the operation set value under the condition that the preset constraint condition for the new operation set value is satisfied. In this case, the operation state data detected when the operation state of the roller mill is controlled by using the selected new operation set value instead of the obtained new operation set value, and the selection A method for controlling a roller mill is provided, which is characterized by updating the mill characteristic model based on the new operation set value. Therefore, in this case as well, the mill characteristic model is updated by following changes over time in the operating state of the roller mill and changes in the properties of the raw material. In this case, however, the optimum characteristics of the roller mill are further optimized by the synergistic effect with the optimization by the evaluation function. It is possible to greatly speed up the process to the operating state. As a result, the operating state of the roller mill can be controlled automatically, optimally and quickly. Furthermore, by updating the above-mentioned mill characteristic model by performing state estimation using a Kalman filter or performing multiple regression analysis while partially updating the retained data, the mill characteristic according to changes in the operating state of the roller mill It can be a model. Therefore, even if there is a change with time in the roller mill, it is possible to always obtain an optimum model. As described above, according to the present invention (first and second inventions), it is possible to obtain a controller for a roller mill that can automatically and / or optimally control the operating state of the roller mill.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram showing a roller mill to which a roller mill control method according to an embodiment of the present invention is applied.

FIG. 2 is a block diagram showing a schematic configuration of a first roller mill control device 1 for controlling the roller mill.

FIG. 3 is a block diagram showing a schematic configuration of a second roller mill control device 1 ′ for controlling the roller mill.

FIG. 4 is a flowchart showing an example of a logical operation of a second roller mill controller 1 ′.

FIG. 5 is a state explanatory view showing a control state of a conventional roller mill by a skilled operator.

[Explanation of symbols]

1 ... 1st roller mill control apparatus 1 '... 2nd roller mill control apparatus 2 ... Detector group 3 ... Optimal control arithmetic unit 4 ... Main controller 5 ... Simplex arithmetic unit 6 ... Memory 8 ... Model update part 10 ... Roller mill x _{1 to} x ₄ ... operation set value x _j ... operation set value α _j Δx _j ... operation amount Δx _{1 to} Δx ₄ ... operation amount y _{1 to} y ₄ ... operation state data y _j ... operation state data Δy _{1 to} Δy ₄ ... Change amount Δy _i ... Change amount P ... Evaluation function P ₂ ... Evaluation function

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Masaaki Hamaguchi 1-3-18 Wakihama-cho, Chuo-ku, Kobe-shi, Hyogo Kobe Steel Works, Kobe Head Office

Claims (7)

[Claims] 1. A roller mill control method for controlling the operating state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operating state data of the roller mill, wherein the operating state data is repeatedly detected, The amount of change in the operating state data detected before and after is applied to a mill characteristic model that represents the preset relationship between the amount of change in the operating state data and the corresponding operation amount from the operation set value. A method for controlling a roller mill, which is characterized by obtaining various operation set values. 2. A roller mill control method for controlling the operating state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operating state data of the roller mill, wherein the operating state data is repeatedly detected, Evaluation for evaluating the operating condition data and / or the operation setting value under the condition that a preset constraint condition is satisfied with respect to the detected operating condition data and / or the operation setting value corresponding to the operating condition data. A roller mill control method comprising selecting a new operation set value so as to optimize a function. 3. A roller mill control method for controlling the operating state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operating state data of the roller mill, wherein the operating state data is repeatedly detected, The amount of change in the operating state data detected before and after is applied to a mill characteristic model that represents the preset relationship between the amount of change in the operating state data and the corresponding operation amount from the operation set value. The operation state data and / or the operation state value under the condition that preset constraint conditions for the detected operation state data and / or the new operation state value are satisfied. A method for controlling a roller mill, characterized in that the new operation set value is selected so as to optimize an evaluation function for performing. 4. A roller mill control method for controlling the operation state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operation state data of the roller mill, wherein the operation state data is repeatedly detected, The amount of change in the operating state data detected before and after is applied to a mill characteristic model that represents the preset relationship between the amount of change in the operating state data and the corresponding operation amount from the operation set value. Of the mill characteristic model based on the operation state data detected when the operation state of the roller mill is controlled by the obtained new operation set value and the obtained new operation set value. A roller mill control method characterized by updating. 5. A roller mill control method for controlling the operation state of the roller mill by adjusting the operation set value of the control factor of the roller mill based on the operation state data of the roller mill, wherein the operation state data is repeatedly detected, The amount of change in the operating state data detected before and after is applied to a mill characteristic model that represents the preset relationship between the amount of change in the operating state data and the corresponding operation amount from the operation set value. The operation state data and / or the operation state value under the condition that preset constraint conditions for the detected operation state data and / or the new operation state value are satisfied. The new operation set value is selected so as to optimize the evaluation function for A roller mill control method characterized in that the mill characteristic model is updated based on the operation state data detected when the operation state of the roller mill is controlled and the new operation set value selected. 6. The roller mill control method according to claim 4, wherein the mill characteristic model is updated by performing state estimation using a Kalman filter. 7. The roller mill control method according to claim 4, wherein the mill characteristic model is updated by performing multiple regression analysis while partially updating the held data.