JP4354978B2 - Plating adhesion amount control device and control method - Google Patents

Description

  The present invention relates to a plating adhesion amount control apparatus and a control method for depositing a plating of a desired thickness on a steel sheet in a continuous plating line of a steel plant, and particularly compensates for the influence of various disturbances when changing the target of the plating adhesion amount. The present invention relates to a control method for calculating an appropriate control command value.

  As a conventional method for performing plating adhesion amount control, for example, Patent Document 1 describes a preset control unit that is performed when a target of plating adhesion amount is changed. An absolute value preset means that directly calculates the operation amount for obtaining the target plating adhesion amount using the plating adhesion amount prediction model, and a variation of the operation amount corresponding to the change amount of the target adhesion amount Use to calculate. This method includes a relative value presetter that calculates the new manipulated variable by adding or subtracting this value to the current manipulated variable, and shows a technique for selecting and presetting appropriate means focusing on the amount of change in the plating target value. Has been.

  Also, in Patent Document 2, the difference between the output of the plating adhesion amount prediction model and the actual adhesion amount is calculated, stored in the model error accumulation means, and the information of the model error accumulation means is used when calculating the operation amount. A method for correcting the plating adhesion amount prediction model is shown.

JP 2004-013393 A Japanese Patent No. 3291201 (Japanese Patent Laid-Open No. 10-18014)

  However, in the method of Patent Document 1, no consideration is given to compensating for the difference in behavior between the plating adhesion amount prediction model and the actual plating plant. Therefore, if there is a discrepancy between the plating adhesion amount prediction model and the actual plating plant characteristics, the appropriateness of the control command value will be impaired depending on the magnitude of the discrepancy, and the accuracy of plating adhesion amount control will be reduced. There was a problem of lowering.

  Further, in the method of Patent Document 2, consideration is given to compensating for the difference in behavior between the plating adhesion amount prediction model and the actual plating plant. However, whether this divergence occurs universally due to insufficient expression ability of the plating adhesion prediction model or changes in the characteristics of the plating plant over time, such as steel plate temperature, nozzle zero-adjustment accuracy, hot-dip plating viscosity, etc. It is not considered separately whether it is temporarily generated due to the state of the latest plating plant due to the variation of the above. For this reason, there are cases where the model error used when the target value of the plating adhesion amount greatly changes is accumulated in the operation with the similar target value in the past that is largely separated in time. At this time, the compensation of the model error becomes inappropriate, and there is a problem that the control accuracy of the plating adhesion amount is lowered.

  In both cases, when a plurality of preset means are provided, an appropriate combination of the selection of the preset means and the compensation method of the plating adhesion amount prediction model used for the preset calculation has not been disclosed.

  The problem to be solved by the present invention is to store the difference in behavior between the plating adhesion amount prediction model and the actual plating plant separately according to the universal model error and the temporarily generated model error. To do. Furthermore, it is to provide highly accurate preset control by appropriately selecting a plurality of preset means and a compensation method for a plating adhesion amount prediction model according to a plating adhesion amount target value and a change amount thereof.

  In order to solve the above problems, the present invention immerses a continuously fed strip in a hot dipping bath, blows a high-pressure gas from a nozzle to a lifting plate, and drops unnecessary plating to a desired thickness on the strip. In a plating adhesion amount control device that receives a performance signal from a plating plant for depositing a plating and transmits a control signal for controlling the plating adhesion amount to the plating plant, the strip speed, gas pressure, nozzle and strip at the nozzle position The plating adhesion amount prediction model describing the relationship between the distance (nozzle gap) etc. and the resulting adhesion amount of the plating, and the results of strip speed, gas pressure, nozzle gap, etc. An adhesion amount estimation means for estimating the adhesion amount of plating and the output of the adhesion amount estimation means and the latest operation To determine the degree of deviation between the control model and the actual plating adhesion behavior from the actual plating adhesion amount, and to use the degree of deviation to bring the plating adhesion prediction model closer to the state of the most recent plating plant The difference between the output value of the adhesion amount estimation means and the actual value of the plating adhesion amount actually adhered to the operating state of the plating plant. Learning means for learning for each layer focusing on similarity, learning result storage means for storing a learning amount as a learning result, and preset means for outputting a control command value of a plant based on the plating adhesion amount prediction model It is characterized by providing.

  According to the present invention, the preset means pays attention to the amount of change of the plating adhesion target value, etc., and appropriately selects the adaptation result and the learning result which are parameters for compensating for the difference between the plating adhesion prediction model and the actual plating adhesion behavior. At the same time, a preset calculation process is appropriately selected to calculate a highly accurate control command value. Furthermore, the validity of the learning result and the adaptation result can be enhanced by selectively using only the stable state data of the plating adhesion plant and performing the adaptation process and the learning process.

  According to the best mode for carrying out the plating adhesion amount control apparatus of the present invention, in the plating adhesion amount control of the steel process line, the plating adhesion amount can be made highly accurate and quality defects can be reduced. Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a configuration diagram of a plating adhesion amount control apparatus according to an embodiment of the present invention. The plating adhesion amount control device 100 controls the plating plant 150 to deposit a plating having a desired thickness on the strip (steel plate) 151.

First, the plating plant 150 will be described. The hot-dip plating is stored in the pot 152 of the plating plant 150, and the strip 151 that is continuously fed is supported by various rolls 154. The strip 151 sent at the plate speed V is once immersed in hot dipping, and then a high-pressure gas is blown from the nozzle 153 to the lifting plate, and unnecessary hot dipping is scraped off to reduce the amount of adhering plating. Control to the desired value. The amount of plating deposited on the strip 151 is generally determined by the speed of the strip 151 (plate speed), the pressure of the gas blown from the nozzle 153, and the distance between the nozzle 153 and the strip 151 at the nozzle blowing position (nozzle gap). These relationships are expressed by, for example, (Equation 1).
ln (W) = a0 + a1 · ln (P) + a2 · ln (V) + a3 · ln (D) (1)
Here, W: plating adhesion amount, P: nozzle gas pressure, V: plate speed, D: nozzle gap, a0 to a3: constant.

  In this embodiment, Equation 1 is referred to as a plating adhesion amount prediction model. The front and rear strips are connected by welding at a welding point 156, and the welding point 156 usually corresponds to a switching position of the plating adhesion amount target value. The adhesion meter 155 measures the amount of plating actually attached, but is attached at a distance of several tens to several hundreds m from the nozzle 153. Further, since the strip is traversed in the width direction and the average value is output, it takes one to two minutes before the plating adhesion amount corresponding to the current P, V, and D can be measured.

  Next, the configuration of the plating adhesion amount control device 100 will be described. The plating adhesion amount control device 100 fetches the plating target value from the manufacturing information table 101 storing the manufacturing information such as the target value of the plating adhesion amount of each strip connected at the welding point and the length of each strip, Preset means 102 is provided for calculating the control command value at the operating end so that a new target value is realized at the transition of the strip. Further, it includes preset timing generation means 103 that takes in the strip speed from the plating plant 150 and tracks the welding point position to generate the start timing of the preset means 102. Furthermore, a stability determining means 104 is provided for determining that the gap, pressure, and plate speed of the nozzle 153 maintain stable values, and that the adhesion amount detected by the adhesion amount meter 155 is a value corresponding to these values. Further, it has a plating adhesion amount prediction model 105 shown in (Equation 1), and has an adhesion amount estimation means 105 for estimating the amount of plating adhered to the strip from the gap, pressure, plate speed, etc. of the nozzle 153. .

  Further, the difference Wd between the output of the adhesion amount estimation means 105 and the detected value from the adhesion amount meter 155 is taken in, and the degree of deviation between the plating adhesion amount prediction model in the current plating plant state and the plating adhesion behavior in the plating plant 100 is calculated. An adaptation means 106 for calculating and an adaptation result storage means 107 for accumulating the calculation results of the adaptation means 106 are provided. Similarly, it has a model error accumulation means 110 that takes in the difference between the output of the adhesion amount estimation means 105 and the value detected from the adhesion amount meter 155 together with the corresponding plating adhesion amount and strip speed and accumulates it over a long period of time. ing. Using the data accumulated in the model error accumulating means 110, the learning means 108 for acquiring a degree of deviation over a long period between the plating adhesion amount prediction model in a similar control state and the plating adhesion behavior in the plating plant 100; A learning result storage means 109 is provided for storing the calculation results of the learning means 108 in association with each control state. In this embodiment, the determination of the similar control state in the learning means 108 is an index that the amount of plating adhesion and the strip speed are both similar, but there are various indexes such as adding a nozzle gap and gas pressure. Combinations are possible. In the present embodiment, the case where the operation end is the nozzle pressure P is shown, but the present invention can also be applied to the case where the operation end is the nozzle gap D or both the nozzle pressure P and the nozzle gap D based on the same concept.

  FIG. 2 is a flowchart showing the processing executed by the stability determination means. The stability determination means 104 initializes a tracking value L indicating the distance at which the stable state of control continues in the longitudinal direction of the steel sheet in S2-1. In S2-2, it is determined whether either the plate speed or the nozzle gap has changed. If changed, the process returns to S2-1 to initialize the tracking value L. If not changed, a value obtained by multiplying the plate speed V by the activation period Δt of the stability determination means 104 is added to L in S2-3, and is newly set to L. In S2-4, is L greater than the product of the distance L1, the plate speed V, and the adhesion amount meter 155 that scans the plate in the width direction, plus the margin M, to the nozzle 153 and the adhesion amount meter 155? Determine if. If it is large, it indicates that the plating plant 150 is stable for a sufficiently long time, and the plating adhesion amount detection value from the adhesion meter 155 corresponds to the current P, V, D, etc. Therefore, a stability determination signal is output in S2-5. If not, the process returns to S2-2 and repeats the processes of S2-2 to S2-4.

  FIG. 3 shows processing executed by the adhesion amount estimation means. The adhesion amount estimation means 105 fetches the results of P, V, and D from the plating plant 150 in S3-1. Next, by substituting P, V, and D into (Equation 1) which is a plating adhesion amount prediction model, ln (W) is obtained, and from this value, the adhesion amount expected to be adhered corresponding to P, V, and D. Calculate the estimated value W ~.

FIG. 4 shows processing executed by the adaptation means. The adaptation means 106 takes in the detected value W of the adhesion amount meter and the output W˜ of the adhesion amount estimation means 105 in S 4-1. In S4-2, the adaptation amount ΔW stored in the adaptation result storage means 107 is taken in, and a new adaptation amount is calculated from (Equation 2).
ΔW ← (1−α) ・ ΔW + α ・ (WW−) (2)
Where α is an adaptive gain. In other words, ΔW is updated by adding, at a fixed rate, the effect of the difference between the output value W ~ of the plating adhesion prediction model generated in the plating operation and the actually measured plating adhesion W to ΔW. To do. Although the adaptive gain α is an adjustment parameter, by making it relatively large, ΔW becomes an adaptive amount that emphasizes the latest operation. In S4-3, ΔW in the adaptation result storage means 107 is updated to the new adaptation amount obtained in (Equation 2). As an adaptive amount calculation method, (Equation 2) is a typical equation, but other methods are also conceivable.

  FIG. 5 shows the data structure of the model error accumulating means 110. The model error accumulating means 110 stores, in addition to the model error Wd, the plating adhesion amount W and the strip speed V when the model error Wd is calculated as a set. Each time a new Wd is calculated, a corresponding combination of three data is accumulated, but data that has passed a certain period is erased from the model error accumulating means 110. As data erasure, a method of erasing from old data after reaching a certain number of data may be used.

FIG. 6 shows processing executed by the learning means. The learning result storage means 109 has a data configuration shown in FIG. 7, for example. That is, the plating adhesion amount and the plate speed are stratified approximately every 10 g / m 2 and 10 mpm in the example of the figure, and the learning amount δWij is accumulated for each layer. The learning means 108 extracts from the model error accumulating means 110 a model error corresponding to each layer having the adhesion amount W and the plate speed V in S6-1. In S6-2, the learning amount δWij is calculated according to (Equation 3).
δWij = (ΣWd) / n (3)
Here, n is the number of Wd for each layer extracted from the model error accumulating means 110. As δWij, it is possible to calculate a difference between a plating adhesion amount prediction model in a similar layer and a plating adhesion behavior in an actual plating plant during a period in which data is accumulated. In S6-3, δWij in the learning result storage means 109 is updated to the new learning amount obtained in (Equation 3). In S6-4, it is confirmed that the learning amount has been updated for all the stratifications of the adhesion amount W and the plate speed V. If not completed, S6-1 to S6-3 are repeated.

  In the example of FIG. 6, the learning results accumulated in the learning result storage unit 109 are stratified according to the plating adhesion amount and the plate speed, but other combinations such as the plating adhesion amount and the nozzle gap, or the nozzle position, the gas pressure value, etc. You can also stratify by. Further, the learning means 108 may be activated in various timings such as a timing when the plating plant stops operation, a fixed period, and activation at a timing manually given by the user.

  FIG. 8 shows processing executed by the preset timing generation means. The preset timing generation means 103 takes in the plate speed V in S8-1. In S8-2, the value obtained by multiplying the passage length L by the plate speed V and the activation period Δt of the preset timing generation means 103 is added to L to obtain a new L. In S8-3, it is determined whether L is larger than the length of the strip fetched from the manufacturing information table 101. If it is larger, it is the timing at which the plating adhesion amount target value is changed instead of the next strip, so the preset timing is output to the preset means 102 in S8-4. Further, the passage length L is cleared. If it is smaller, the process returns to S8-1, and the processes of S8-1 to S8-3 are repeated.

  FIG. 9 shows processing executed by the preset means. The preset means 102 receives the activation timing from the preset timing generation means 103, and starts a process of calculating the nozzle pressure P that realizes the plating adhesion amount target value of the next strip. First, in S9-1, V and D are taken from the plating plant 150, and the plating adhesion amount target value W * of the next strip and the current plating adhesion amount target value Wpre * of the current strip are taken in from the manufacturing information table 101. In S9-2, it is determined whether or not the difference between W * and Wpre * is equal to or smaller than a predetermined value (Wth).

In the case of Wth or less, there is no significant difference between the current and next plating adhesion amount target values, and plating control can be continued at similar operating points, so the plating adhesion amount prediction model in the current plating plant state and the plating plant 100 By compensating for the deviation from the plating adhesion behavior, highly accurate control can be performed. Therefore, the adaptation amount ΔW is fetched from the adaptation result storage means 107 in S9-3. In step S9-4, the plating adhesion amount prediction model is compensated with ΔW, and the operation amount P of the preset control is calculated. That is, by solving (Equation 1) for the nozzle pressure and using W-ΔW instead of W, the model that the plating adhesion amount prediction model is expected to have for the current plating adhesion characteristics Compensate the error and calculate ln (P) by (Equation 4).
ln (P) = {ln (W * −ΔW) − (a0 + a2 ・ ln (V) + a3 ・ ln (D)) / a1 (4)
Here, W *: target value of plating adhesion of the next strip, ΔW: adaptive amount, V: line speed, D: nozzle gap, P: nozzle gas pressure, a0 to a3: constant. A value is substituted into the right side, and ln (P) is obtained, and then converted into nozzle pressure P.

On the other hand, if the difference between W * and Wpre * is not less than or equal to the predetermined constant value Wth in S9-2, there is a large difference between the current and next plating adhesion amount target values. For this reason, significant compensation cannot be performed even if attention is paid to the discrepancy between the plating adhesion amount prediction model in the current plating plant state and the plating adhesion behavior in the plating plant 100, and it is better to control focusing on the universal discrepancy. It is often effective. In step S9-5, the learning result δWij corresponding to the target value of the plating adhesion amount and the current plate speed layer is fetched from the learning result storage unit 108. In S9-6, the plating adhesion amount prediction model is compensated by δWij, and the preset control operation amount P is calculated by (Equation 5).
ln (P) = {ln (W * −δWij) − (a0 + a2 · ln (V) + a3 · ln (D)) / a1 (5)
Here, W *: target value of the plating adhesion amount of the next strip, δWij: learning amount, V: line speed, D: nozzle gap, P: nozzle gas pressure, a0 to a3: constants. A value is substituted into the right side, and ln (P) is obtained, and then converted into nozzle pressure P. The P calculated in S9-6 is output to the plating plant 150.

  FIG. 10 shows the configuration of the plating adhesion amount control apparatus according to the second embodiment of the present invention. In this embodiment, the model error accumulating means 110 of the first embodiment is not provided, and the learning means 1101 calculates δWij by sequential calculation and updates the stored contents of the learning result storage means 109 every time Wd is taken.

FIG. 11 shows processing executed by the learning means. In this embodiment, the learning unit 1001 is sequentially executed every time the adhesion amount estimation unit 105 estimates a new adhesion amount. In S11-1, the detected value W of the adhesion amount meter, the output W ~ of the adhesion amount estimation means 105, and the plate speed V are fetched. In this example, the plating adhesion amount and the plate speed are stratified approximately every 10 g / m 2 and 10 mpm, and a learning amount δWij is accumulated for each layer. In S11-2, the learning amount δWij stored in each layer of the learning result storage means 109 corresponding to the values V and W of the adhesion amount and the plate speed is taken in, and a new learning amount δWij is calculated in (Equation 6). To do.
δWij ← {η · δWij + β · (W−W ~)} / (η + β) (6)
Where β is a learning gain and η is a forgetting factor. The learning gain β is a smaller value than α in (Equation 2). In this way, by reducing the influence of the latest Wd, the difference between the output value of the plating adhesion amount prediction model generated in the same layer and the actually measured plating adhesion amount can be obtained by learning over a long period of time. As a result, it is possible to extract a universal difference between the output value of the plating adhesion amount prediction model and the actually measured plating adhesion amount. Also, η is a forgetting factor, and adjusts how much time constant the information of the current learning amount is lost. The learning speed is determined by the relative relationship between β and η. In S11-3, δWij in the learning result storage means 109 is updated to the new learning amount obtained in (Equation 6). Equation (6) is an example of a learning amount calculation formula, and various other methods are conceivable.

  FIG. 12 shows the configuration of the plating adhesion amount control apparatus according to the third embodiment. In this embodiment, instead of the preset means 102 of the second embodiment, a preset means 1200 having a first preset means 1201 and a second preset means 1202 for calculating a control command with different mathematical expressions is provided.

  FIG. 13 shows processing executed by the preset means. In S13-1, the preset means 1200 first adds the plate speed V and the nozzle gap D from the plating plant 150, the current nozzle pressure setting value Pcur, and the manufacturing information table 101 from the plating adhesion target value of the next strip. W * and current strip plating amount target value Wpre * are taken in. In S13-2, it is determined whether or not the difference between W * and Wpre * is equal to or less than a predetermined value Wth.

If it is equal to or less than Wth, the plating control can be continued at a similar operating point. Therefore, the first preset means 1201 is activated in S13-3, and the adaptation amount ΔW is fetched from the adaptation result storage means 107. In S13-4, based on the current nozzle pressure command value Pcur, the command value for nozzle pressure is relatively calculated by adding or subtracting the pressure change corresponding to the change in the adhesion amount target value. Furthermore, the control command value is made highly accurate by compensating the plating adhesion amount prediction model used for the calculation using the adaptive amount ΔW. Specifically, in S13-4, the control command value is calculated according to (Equation 7).
P = Pcur + f ⁻¹ (W * −ΔW, V, D) −f ⁻¹ (Wpre * −ΔW, V, D) (7)
Where ln {f ⁻¹ (W, V, D)} = {ln (W) − (a0 + a2 · ln (V) + a3 · ln (D)) / a1, W *: Plating adhesion amount target value, Wpre *: Current strip plating adhesion amount target value, ΔW: Adaptation amount, V: Line speed, D: Nozzle gap, P: Nozzle gas pressure, a0 to a3: Constants.

On the other hand, if the difference between W * and Wpre * is not less than a predetermined value Wth in S13-2, there is a large difference between the current and next plating adhesion amount target values, and the control operating point is separated. Therefore, effective compensation cannot be performed even if attention is paid to the current state of the plating plant, and the plating adhesion amount prediction model is focused on the universal divergence between the plating adhesion amount prediction model and the plating adhesion behavior of the plating plant 100. Need to compensate. Therefore, the learning amount δWij corresponding to each layer of the target value of the plating adhesion amount and the current plate speed is fetched from the learning result storage means 108 in S13-5. Then, in S13-6, the plating adhesion amount prediction model is compensated with δWij to increase the accuracy of the control command value. Specifically, in S13-6, the control command value is calculated by (Equation 8).
P = f ⁻¹ (W * −δWij, V, D) (8)
Where ln {f ⁻¹ (W, V, D)} = {ln (W) − (a0 + a2 · ln (V) + a3 · ln (D)) / a1, W *: Plating adhesion amount target value, δWij: learning amount, V: line speed, D: nozzle gap, P: nozzle gas pressure, a0 to a3: constant. The finally calculated gas pressure control command value is output to the controlled object 150 in S13-7.

  Next, Example 4 of the present invention will be described. FIG. 14 shows processing executed by the preset means of this embodiment. The difference from Example 1 is that the pressure calculated by compensating the plating adhesion amount prediction model with δWij in S14-6 is used as a reference pressure command value, and the change in pressure corresponding to the change in adhesion amount target value is added or subtracted. Thus, the command value of the nozzle pressure in the preset control is relatively calculated. Furthermore, the control command value is made highly accurate by compensating the plating adhesion amount prediction model used for the calculation of the pressure change by using the adaptive amount ΔW.

Specifically, in S14-6, the control command value is calculated according to (Equation 9). The first term on the right side is the reference pressure command value, and the pressure change is calculated by subtracting the third term from the second term.
P = f ⁻¹ (W * −δWij, V, D) + f ⁻¹ (W * −ΔW, V, D) −f ⁻¹ (Wpre * −ΔW, V, D) (9)
Where ln {f ⁻¹ (W, V, D)} = {ln (W) − (a0 + a2 · ln (V) + a3 · ln (D)) / a1, W *: Plating adhesion amount target value, Wpre *: Current plating plating amount target value, ΔW: Adaptation amount, δWij: Learning amount, V: Line speed, D: Nozzle gap, P: Nozzle gas pressure, a0 to a3: It is a constant. The finally calculated gas pressure control command value is output to the controlled object 150 in S14-7.

  Next, Example 5 of the present invention will be described. FIG. 15 shows processing executed by the preset means of this embodiment. In this embodiment, when the preset processing is selected in S15-2, whether the difference between the plating adhesion target value W * of the next strip and the current strip Wpre * is equal to or larger than a predetermined constant value Wth1. In addition, it is determined whether the plating adhesion target value W * of the next strip is equal to or greater than a certain value Wth2. The process proceeds to S15-5 only when both are satisfied, and otherwise proceeds to S15-3. Thereafter, the same processing as in the first embodiment is performed. As described above, the selection criteria for the preset processing include various differences such as the difference between the plating adhesion target value of the next strip and the plating adhesion target value of the current strip, the absolute value of the plating adhesion target value, the strip speed, and the nozzle position. Conditions can be considered.

Next, Example 6 of the present invention will be described. FIG. 16 shows the data structure of the learning result storage means. The learning result of the present embodiment is acquired as parameters a0 to a3 of the plating adhesion amount prediction model of (Equation 1). The parameters may be updated by the successive least squares method. Calculations for calculating a0 to a3 of the plating adhesion amount prediction model are performed each time at the same timing as in the second embodiment, and stored in the learning result storage unit 1601. In this embodiment, the preset means 102 uses a0 to a3 stored for each layer of the learning result storage means 1601 instead of correcting the plating adhesion amount prediction model with δWij, and the gas pressure according to (Equation 10). The control command value is calculated.
ln (P) = {ln (W *) − (a0 + a2 · ln (V) + a3 · ln (D)) / a1 (10)
The sequential least square method can be referred to in many literatures. For example, it is described in detail in “Theory and Application of Least Square Method (Akira Tajima, Kazuo Komaki, Toyo Shoten (1986))”.

  As described above, the present invention described in each of the embodiments can be widely applied to control of the amount of plating adhesion in the steel process line.

The block diagram which showed Example 1 of the plating adhesion amount control apparatus of this invention. The flowchart which shows the process of a stability determination means. The flowchart which shows the process of an adhesion amount estimation means. The flowchart which shows the process of an adaptation means. The data block diagram of a model error storage means. The flowchart which shows the process of a learning means. The data block diagram of a learning result storage means. The flowchart which shows the process of a preset timing production | generation means. The flowchart which shows the process of a preset means. The block diagram which showed Example 2 of the plating adhesion amount control apparatus. FIG. 9 is a flowchart showing processing of learning means according to the second embodiment. The block diagram which shows Example 3 of a plating adhesion amount control apparatus. FIG. 10 is a flowchart showing processing of preset means of the third embodiment. FIG. 10 is a flowchart showing processing of preset means according to the fourth embodiment. FIG. 10 is a flowchart showing processing of preset means according to the fifth embodiment. The data block diagram of the learning result memory | storage means of Example 6. FIG.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 100 ... Plating adhesion amount control apparatus, 101 ... Manufacturing information table, 102 ... Preset means, 103 ... Preset timing generation means, 104 ... Stability determination means, 105 ... Adhesion amount estimation means, 106 ... Adaptation means, 107 ... Adaptation result storage means , 108 ... learning means, 109 ... learning result storage means, 150 ... control target, 151 ... strip (steel plate), 153 ... nozzle, 155 ... adhesion amount meter, 1200 ... preset means, 1201 ... first preset means, 1202 ... Second preset means, 1601... Learning result storage means.

Claims (9)

The continuous strip is immersed in a hot dipping bath, and high pressure gas is blown from a nozzle to the lifting plate, and unnecessary plating is dropped, and a performance signal is sent from a plating plant that attaches plating of the desired thickness to the strip. In the plating adhesion amount control device that receives and transmits a control signal for controlling the plating adhesion amount to the plating plant,
A plating adhesion amount prediction model describing the relationship between the strip speed, gas pressure, nozzle-to-strip distance (nozzle gap) at the nozzle blowing position (nozzle position), and the amount of plating adhesion deposited as a result,
Adhesion amount estimation means for taking in the results of strip speed, gas pressure, nozzle gap, etc., and estimating the plating adhesion amount using the plating adhesion amount prediction model;
From the output of the adhesion amount estimation means and the plating adhesion amount actually adhered in the latest operation, the degree of deviation between the plating adhesion amount prediction model and the actual plating adhesion behavior is obtained, and the degree of deviation is used. An adaptive means for calculating an adaptive amount for approximating the plating adhesion amount prediction model to the state of the latest plating plant; an adaptive result storage means for storing the calculated adaptive amount;
Model error accumulation means for accumulating the deviation between the output of the adhesion amount estimation means and the actual value of the amount of adhesion of plating that has actually adhered ,
Learning means for learning the deviation accumulated in the model error accumulation means for each stratification focusing on the similarity of the operation state of the plating plant, and a learning result storage means for storing a learning amount as a learning result;
When the target value of the plating adhesion amount is changed, the plating adhesion amount prediction model is corrected using at least one of the contents of the adaptive result storage unit and the learning result storage unit, and the corrected plating adhesion amount prediction model is obtained. A plating adhesion amount control device comprising preset means for calculating a control command value for obtaining a desired plating adhesion amount .   The control command value is at least one of a command value of gas pressure or a command value of nozzle position, and the similarity of the operation state of the plant includes plating deposition amount, strip speed, gas pressure, nozzle position or nozzle The plating adhesion amount control device according to claim 1, wherein at least one of the gaps is similar. The preset means calculates a difference between the target amount of plating adhesion of the current strip and the target amount of plating adhesion of the next strip, and stores the adaptation result according to the magnitude relationship between the difference and a predetermined value. It means amount and the learning result coating weight control apparatus of claim 1, wherein determining whether to correct the coating weight prediction model by using any of the learning amount memory means. The preset means calculates a difference between the target amount of plating adhesion of the current strip and the target amount of plating adhesion of the next strip, and when the difference is small, the plating amount is calculated using the adaptation amount of the adaptation result storage means. The adhesion amount prediction model is corrected, and a control command value for obtaining a desired plating adhesion amount is calculated using the corrected plating adhesion amount prediction model. When the difference is large, the learning amount of the learning result storage unit is used. the corrected coating weight prediction model, according to any one of claims 1 to 3, characterized in that calculates a control command value for obtaining the desired coating weight by using a coating weight prediction model correction Plating adhesion amount control device. The preset means adds a control command value calculated using the plating adhesion target value of the next strip to the current control command value, and further calculates a control command value calculated using the plating adhesion target value of the current strip And a second preset means for calculating the control command value from the plating adhesion amount target value of the next strip using the plating adhesion amount prediction model. Have
When the difference between the target amount of plating adhesion before the change and the target amount of plating adhesion after the change is small, the plating amount prediction model is corrected by correcting the plating amount prediction model using the adaptive amount of the adaptation result storage means. A control command value for obtaining a desired plating adhesion amount is calculated by the first preset means using a model, and when the difference is large, the plating adhesion amount prediction model is calculated using the learning amount of the learning result storage means. corrected, according to any one of claims 1 to 3, characterized in that calculates a control command value for obtaining the desired coating weight by the second preset means using coating weight prediction model correction Plating adhesion amount control device. The preset means adds a control command value calculated using the plating adhesion target value of the next strip to the current control command value, and further calculates a control command value calculated using the plating adhesion target value of the current strip The first preset means for calculating the final control command value by subtracting the value and the plating adhesion amount prediction model obtained by correcting the plating adhesion amount target value of the next strip with the learning amount of the learning result storage means. The control command value, the control command value calculated using the plating adhesion amount prediction model corrected by the adaptive amount of the adaptive result storage means to the plating adhesion amount target value of the next strip, and the plating adhesion amount target value of the current strip A second preset for calculating a final control command value from three values of the control command value calculated using the plating adhesion amount prediction model corrected by the adaptive amount of the adaptation result storage means We have the means,
When the difference between the plating adhesion amount target value before the change and the plating adhesion amount target value after the change is small, a control command value for obtaining a desired plating adhesion amount is calculated by the first preset means, and the difference is large. sometimes, coating weight control apparatus according to any one of claims 1 to 3, characterized in that calculates a control command value for obtaining the desired coating weight in the second preset means. The preset means pays attention to at least one of the difference between the current plating plating target value and the next strip plating target value, the next strip plating target value, strip speed or nozzle position. 4. The method according to claim 1, further comprising: determining whether to correct the plating adhesion amount prediction model using an adaptation amount of the adaptation result storage unit or a learning amount of the learning result storage unit . coating weight control apparatus according to.   The preset means pays attention to at least one of the difference between the current plating plating target value and the next strip plating target value, the next strip plating target value, strip speed or nozzle position. Determining whether to correct the plating adhesion amount prediction model using the adaptation amount of the adaptation result storage unit or the learning amount of the learning result storage unit, and the first preset unit and the second preset 8. The plating adhesion amount control device according to claim 7, wherein which of the means is selected is determined. In a plating plant where a strip that is continuously fed is immersed in a hot dipping bath, a high pressure gas is blown from a nozzle to a lifting plate, and unnecessary plating is dropped to adhere a plating of a desired thickness to the strip.
The results of strip speed, gas pressure, nozzle gap, etc., using a plating adhesion prediction model that describes the relationship between strip speed, gas pressure, nozzle-to-strip distance at the nozzle position (nozzle gap), etc. From this estimation result and the actual plating adhesion amount in the most recent operation, the degree of deviation between the plating adhesion prediction model and the actual plating adhesion behavior is obtained. Using the degree, the adaptive amount for approximating the plating deposition amount prediction model to the state of the latest plating plant is calculated and stored, and in parallel, the estimated result and the actual value of the actual coating adhesion amount are in a long period of time. When the average deviation is stored as a learning amount for each layer focusing on the similarity of the operation state of the plating plant, and the target value of the plating adhesion amount is changed, The plating characterized in that the plating adhesion amount prediction model is corrected using at least one of the response amount and the learning amount, and a control command value for obtaining a desired plating adhesion amount is calculated using the corrected control model Amount control method.

Images