JP2703865B2 - Plating weight control device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coating amount control apparatus, and more particularly to a coating amount control apparatus for controlling a coating amount to a predetermined uniform thickness in a step of continuously plating a steel sheet. It concerns the device.

[0002]

2. Description of the Related Art In a continuous hot-dip galvanizing line for galvanizing the surface of a steel sheet, the steel sheet is plated by a hot-dip zinc bath. The amount of plating applied at this time is adjusted by spraying a gas from a wiping nozzle (gas injection nozzle) onto the steel sheet from the molten zinc bath and flushing out the applied molten plating by a gas jet. In general, the adjustment of the amount of plating applied is adjusted by performing feedback control of the distance of the wiping nozzle from the steel plate and the gas pressure based on the measured value of the amount of applied plating (JP-A-63-53248). Gazette). However, in the hot-dip galvanizing line, the measuring point of the coating weight is usually set about 100 m away from the molten bath.
In the above-described control method mainly based on feedback control, it takes a long time to respond, and it is not possible to cope with changes in operating conditions, and control accuracy is reduced. In order to solve this problem, Japanese Patent Application Laid-Open No. 3-173
In the control method disclosed in Japanese Patent No. 757, the feedforward control and the feedback control are used in combination, the feedforward control is performed until the operation state of the plating is stabilized, and the feedback control is performed after the operation state is stabilized. An empirical formula or a linear model was used for the feedforward control model.

[0003]

However, the recent trend of high-mix low-volume production is often applied to the continuous galvanizing line described above. If you have a problem. That is, according to the above-mentioned prior art, in a steady state in which the operation is continued under a certain production condition, accurate control of the amount of plating can be achieved, but when the production condition is switched to a different product, the feedforward control is performed. In spite of this, it is not possible to cope with changes in the manufacturing conditions, and the accuracy of the amount of plating may decrease. This is thought to be because the empirical formula or the linear model adopted as the feedforward control model does not provide sufficient model accuracy. The present inventors have studied the improvement of the accuracy of the feedforward control model. As a result, it is difficult to create an accurate single physical model at the time of so-called presetting, in which the continuous hot-dip galvanizing process is performed to replace products having different manufacturing conditions. It was confirmed that sufficient accuracy could not be obtained with the above empirical formula or linear model. In addition, regarding the fact that there is a part where the model accuracy is poor, the accuracy is not uniformly poor over the entire operating condition (target plating adhesion amount, line speed, nozzle height, etc.), but the accuracy is determined by the combination. Data analysis has revealed that there are good parts and parts with reduced accuracy. The present invention has been made in view of the above-described study results, and has as its object the purpose of reducing the controllability of plating when switching between products with different types of steels and different coatings. It is to provide a control device.

[0004]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention provides a method of plating a steel strip formed in a strip shape by continuously passing the steel strip through a metal melting bath and disposing the steel plate downstream of the metal melting bath. The injection pressure of the gas injection nozzle and / or the distance between the gas injection nozzle and the steel plate is controlled by feedforward operation until the operating condition of the plating is stabilized, and after the operating condition is stabilized, the metal melting is performed by performing a feedback operation. In a plating amount control device for controlling a plating amount attached to a steel sheet passing through a bath, the operation data collected after the operation state is stabilized for each product classified by the type of the steel sheet and the plating amount is described. Model identification means for classifying products into similar groups according to products and operating conditions, and identifying a model for each group based on the classified data; Fuzzy inference based on the fitness calculating means for calculating the fitness indicating the degree of belonging to each group, the output from the model identified by the model identifying means, and the fitness calculated by the fitness calculating means. And an operation amount calculation means for calculating the feedforward operation amount. Further, there is provided a plating adhesion amount control device, wherein the model identification means identifies the model by adaptively correcting the parameters of the model.

[0005]

According to the present invention, the operation data collected after the operation condition is stabilized for each product classified by the type of the steel sheet and the coating weight is classified into similar groups according to the product and the operation conditions. A model is identified for each group based on the classified data. Next, the degree of conformity representing the degree to which the operation data of each product belongs to each group is calculated. And
By performing fuzzy inference based on the output from the identified model and the calculated degree of fit, a feedforward manipulated variable is calculated. In this way, it is possible to improve the accuracy of the model, and as a result, to obtain a plating amount control device capable of eliminating a decrease in control accuracy in the case of switching and producing a product having a different type of steel plate and a different coating amount. Can be. Further, in identifying the model, the model is identified by adaptively modifying the parameters of the model. As a result, the parameter learning of each model equation is automatically performed, so that a more robust and accurate model can be obtained with respect to a change in equipment setting conditions.

[0006]

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 embodiment is an example embodying the present invention and is not intended to limit the technical scope of the present invention. Here, FIG. 1 is a schematic diagram showing a schematic configuration of a plating adhesion amount control device A according to one embodiment of the present invention, FIG. 2 is a schematic diagram showing an entire process around the device A, and FIG. FIG. 4 is a diagram illustrating an example of the membership function of FIG. As shown in FIG. 2A, the plating amount control apparatus A according to the present embodiment is configured to continuously pass a steel sheet 4 formed in a belt shape through a metal melting bath 5 to perform plating. The feed pressure of the wiping nozzle 6 (corresponding to a gas jet nozzle) and / or the interval between this nozzle and the steel plate is fed forward until the operating condition of the plating is stabilized, and the operating condition is stabilized. After that, by feedback operation,
This is the same as the conventional example in that the amount of plating applied to the steel sheet 4 passing through the metal melting bath 5 is controlled. However, in the present embodiment, as shown in FIG. 1, the operation data collected after the operation state is stabilized for each product classified by the type of the steel plate 4 and the amount of plating adhesion is determined according to the operation conditions. A model operation mechanism 1 that is classified into groups and includes a model identified for each group based on the classified data; and an adaptation that calculates a degree of fitness indicating the degree to which the operation data of each product belongs to each group. The fuzzy inference is performed based on the degree calculation mechanism 2 (corresponding to the matching calculation means), the output from the model calculated by the model calculation mechanism 1 and the fitness calculated by the fitness calculation mechanism 2. The present embodiment is different from the conventional example in that a preset feedforward control device a1 including an inference operation mechanism 3 (corresponding to an operation amount operation means) for calculating a forward operation amount is provided. Further, the model may be identified by the model operation mechanism 1 adaptively modifying the parameters of the model. For this purpose, a preset model identification device comprising a model type learning device 1 '(corresponding to model identification means) is used. a2 is also different from the conventional example.

Hereinafter, the overall process around the plating adhesion amount control device A will be further embodied and the principle of operation will be described. As shown in FIG. 2A, in the present process, first, the heat-treated steel sheet 4 passes through a zinc bath 5, and the amount of plating applied is adjusted by a wiping nozzle 6. The passing speed is measured by the passing speed meter 7. The plating amount is measured by the plating amount meter 8. Also, as shown in FIG.
The nozzle height h is detected by the nozzle position detector 9. The preset feedforward control device a1 of the plating amount control device A outputs a nozzle pressure change command and / or a nozzle interval change command to the wiping nozzle 6 starting from a production condition change signal (change of product or change of operation condition). After a while, the operation is stabilized by the feedback control. At the stage when the operation is stabilized, the preset model identification device a2 sets each process measurement value (operating time) based on the actual machine data of the material collected from the next signal before the change of the manufacturing condition (changed to the next product). Data) to preset feed forward control device a
The parameters of a plurality of model expressions used for 1 are learned. The production condition change signal and the pre-production condition change signal are separately calculated from the length and speed of each product. Here, first, the operation principle of the preset feedforward control device a1 will be described. Preset feed forward control device a1
Outputs a nozzle pressure p and / or a nozzle interval d using as input each target process value such as a target plating amount w, a line speed v, a nozzle height h, a plating bath temperature, and a steel plate accuracy. Hereinafter, in this embodiment, only the nozzle pressure p is used as the operation amount for convenience of explanation. However, a similar configuration is possible when the nozzle interval d is used as the operation amount. At the time when the product is changed, the fitness value for each group is calculated by the fitness calculation mechanism 2 by using the process measurement value and the set values (target adhesion amount w, line speed v, nozzle interval d) as inputs. Next, the output of the model set corresponding to each group is calculated by the model operation mechanism 1, and the inference operation mechanism 3 calculates the fitness of each group calculated by the adaptation operation mechanism 2 and the output of the model. Fuzzy inference is performed and output calculation is performed. Therefore, it can be said that the model operation mechanism 1 and the inference operation mechanism 3 constitute a kind of fuzzy inference mechanism.

Next, a method of constructing a model will be described.
First, as many actual data as possible are classified into several groups having similar process setting conditions by a well-known clustering algorithm or the like. Here, the model structure of the group i is set as follows. y ⁱ = a ⁱ ₀ + a ⁱ ₁ 1nw + a ⁱ ₂ 1nv + a ⁱ ₃₁
a ⁱ with respect to nd each group data _{^{0, a i 1, a i}} 2, in advance determined by regression Get a ⁱ _3. These parameters are adaptively corrected by the preset model identification device a2. Here, w is a target adhesion amount at the time of control calculation, but is an actual adhesion amount at the time of model identification. a ⁱ _j is the j th parameter of the i group. Now, suppose that M groups are set and there are M model expressions. Hereinafter, the principle of each mechanism will be described assuming that the input data x ^* = [w ^* , v ^* , d ^* ] ^T is given at the time of so-called presetting when products having different manufacturing conditions are used in this process. [Compatibility calculating mechanism 2] The degree to which the input data x ^* (target adhesion amount w ^* , line speed v ^* , nozzle interval d ^* ) belongs to each group i (membership value μ ⁱ , 0 ≦ μ ⁱ ≦ 1). For each of the set M groups, a center c ⁱ of the group i and a width f ^{i of} the group i, which characterize the group, are set. Using these, the input data x is calculated by the following membership function.
^* Calculates the degree of fitness belonging to each group i.

(Equation 1) Here, ‖ · ‖ represents the distance (norm) of the vector.
For example, FIG. 3 shows the degree of conformity to the target adhesion amount w. That fit becomes a conical function of the radius f ⁱ to the center c ⁱ of the group i, the input data x ^* =
takes the maximum value 1 when c ^i. If the input data x ^* is separated from the center c ⁱ by a radius f ⁱ or more, the fitness is 0.
Becomes

[0009] Now with respect to any of the input data x ^0, and has become a decision of the group, such as the following equation holds.

(Equation 2) This means that a group is set that covers the whole so that there is a group corresponding to any input. [Model calculation mechanism 1] Each model output (M) corresponding to input data x ^* (target adhesion amount w ^* , line speed v ^* , nozzle interval d ^* ) is calculated by the following equation. y ⁱ = a ⁱ ₀ + a ⁱ ₁ 1nw ^* + a ⁱ ₂ 1nv ^* + a ⁱ
₃ 1nd ^* [Inference operation mechanism 3] From the M model outputs and the fitness of each group, a preset feedforward control output y ^* is fuzzy inferred by the following equation.

(Equation 3) Next, the operation principle of the preset model identification device a2 will be described. After the preset feedforward control of the specific material is completed, feedback control based on the measured value becomes effective. When the feedback control enters a steady state, actual data is collected, and each model parameter is obtained by the model type learning device 1 '. Make corrections. That is, as shown in FIG. 4, the parameters of the model equation are corrected from the measured value data in the steady state during the feedback control during the manufacture of the material 1, and the steady state during the feedback control is similarly applied to the material 2 as well. Modify the parameters of the model formula from the measured data in.

Here,

(Equation 4) Then, the model output equation is as follows.

(Equation 5) Hereinafter, the subscript N indicates that the time is N. Here, when the above model output equation is expressed as a vector, first, 4M parameters are put together,

(Equation 6) And the corresponding input, including the fitness,

(Equation 7) Then, the following equation is obtained.

(Equation 8) Since this is linear with respect to parameters, the least squares method is applied using actual data. The following sequential algorithm is used so that the parameters can be corrected each time the performance data is obtained (for each material).

(Equation 9)

By updating all the parameters of each model set as described above, it is possible to maintain the model accuracy even when the setting conditions and the operating conditions are changed. As a result of simulating the above off-line and examining the accuracy of this embodiment, the average of the absolute value of the deviation from the target value of the plating adhesion amount at the time of presetting can be reduced to about 2/3, confirming its effectiveness. Was. As described above, according to the present embodiment, a precise model can be constructed using a nonlinear model obtained by combining a plurality of models by fuzzy inference. In addition, by performing parameter learning of each model using a model-type learning device, it is possible to perform robust modeling with respect to changes in setting conditions and operating conditions. As a result, it is possible to obtain a plating adhesion amount control device capable of eliminating a decrease in control accuracy in the case of switching and producing products having different types of steel sheets and different plating adhesion amounts. In the above embodiment, the input variables of the model are reduced for convenience of explanation. For example, bath temperature, nozzle height, heating furnace temperature, steel plate temperature, etc. can be added as input variables of the model. Thus, it is considered that more accurate control can be realized.

According to the present invention, the apparatus for controlling the amount of plating applied is
With the above configuration, it is possible to improve the accuracy of the model, and as a result, it is possible to eliminate the decrease in control accuracy when switching between products with different types of steel sheets and different amounts of plating to produce control. A quantity control device can be obtained. Further, in identifying the model, the model is identified by adaptively modifying the parameters of the model. As a result, the parameter learning of each model equation is automatically performed, so that a more robust and accurate model can be obtained with respect to a change in equipment setting conditions.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing a schematic configuration of a plating amount control device A according to an embodiment of the present invention.

FIG. 2 is a schematic view showing the entire process around the device A.

FIG. 3 is an explanatory diagram showing an example of a membership function of a fitness calculation mechanism.

FIG. 4 is an explanatory diagram showing the operation of the model-type learning device.

[Explanation of symbols]

 A: Plating amount control device 1: Model operation mechanism 2: Fitness calculation mechanism (corresponding to suitability calculation means) 3: Inference calculation mechanism (corresponding to operation amount calculation means) 1 ': Model formula learning device (Model identification means) Equivalent to)

 ──────────────────────────────────────────────────の Continued on the front page (72) Inventor Hiroaki Nakano 1 Kanazawacho, Kakogawa City, Hyogo Prefecture Kobe Steel, Ltd. Inside the Kakogawa Works (72) Inventor Yasutaka Uchiyama 1 Kanazawacho, Kakogawa City, Hyogo Prefecture Kobe Steel, Ltd. Kakogawa Works (56) References JP-A-6-108219 (JP, A) JP-A-6-17218 (JP, A) JP-A-5-171396 (JP, A) JP-A-5-331110 (JP, A A)

Claims (2)

(57) [Claims] 1. A strip-shaped steel sheet is continuously passed through a metal melting bath to be plated, and a gas injection nozzle disposed downstream of the metal melting bath and / or a gas injection nozzle and a steel sheet are plated. The feed distance is controlled until the operating condition of the plating is stabilized, and the feed operation is performed after the operating condition is stabilized to control the amount of the deposited coating on the steel sheet passing through the metal melting bath. In the quantity control device, the operation data collected after the operation condition is stabilized is classified into similar groups according to the product and the operation conditions for each product classified according to the type of the steel sheet and the amount of plating, and the classification is performed. Model identification means for identifying a model for each group based on the obtained data, and fitness calculation for calculating the fitness indicating the degree to which the operation data of each product belongs to each group Means, and an operation amount calculation means for calculating a feedforward operation amount by performing fuzzy inference based on the output from the model identified by the model identification means and the fitness calculated by the fitness calculation means. A plating amount control device, comprising: 2. The apparatus according to claim 1, wherein said model identification means identifies the model by adaptively modifying the parameters of the model.