KR20000045523A - Method for controlling rolling of continuous roller using neural circuit network - Google Patents

Abstract

PURPOSE: A method for controlling a continuous roller is provided to accurately obtain final thickness of a rolled board using a neuron circuit network. CONSTITUTION: A condition of a rolling such as the gap and velocity of a roll is set in advance. Once reaching a rolling material to a fourth stand(4), rolling values(rolling force, a roll gap, thickness between stands and so on) of a first to a fourth stand(1-4) are obtained and a roll gap of a fifth to a seventh stand(5-7) is obtained. Based on the values, a device dynamically corrects and resets the roll gap and roll driving velocity of the fifth to the seventh stand in order to make the rolling condition of each stand be most suitable.

Description

Rolling Control Method of Continuous Rolling Mill Using Neural Network

The present invention relates to the improvement of the rolling control method in a continuous rolling machine in which a rolled material is successively rolled through a plurality of tandem rolling stands arranged in series to obtain a steel plate having a target thickness. Neural network that can further increase the degree of rolling control by predicting the setting error from the rolling performance and rolling conditions of the upstream stand and the rolling conditions of the downstream stand to dynamically reset the rolling conditions of the downstream stand during the rolling process. The rolling control method of the continuous rolling mill using the present invention.

In the continuous rolling mill according to the present invention, the rolling conditions at each rolling stand are calculated in advance according to the thickness, width, temperature, steel grade, and rolling specifications before rolling on the rolled material, that is, rolling force, rolling speed, rolling position, and the like. Before execution, the rolling work is started after setting for each rolling stand. If the rolling conditions of each stand set as described above are appropriate, a product having excellent thickness accuracy at the distal end portion can be obtained. However, if an error is included in the pre-set rolling conditions, a large plate thickness deviation occurs at the tip of the steel sheet coming out through the final rolling stand, and a part whose plate thickness is out of the management tolerance cannot be used as a product. This will cause a big drop in the real rate. For this reason, in a continuous rolling mill, it is an important technical subject to set the rolling conditions of each rolling stand suitably anyway.

The rolling conditions at each rolling stand of the continuous rolling mill, in particular, the roll gap Sd, can be obtained in order to obtain the target exit thickness h based on the gauge metric equation as shown in the following equation (1) in the state of predicting the rolling load. Is set.

h = Sd + F / M + GME -------------------------------- (1)

Where F is a rolling load, M is a stiffness coefficient, and GME is a gauge meter correction coefficient. In order to accurately predict the plate thickness by the above equation (1), the dense stiffness coefficient (M) should be corrected according to the rolling force and the plate width, etc., and the gauge meter correction coefficient (GME) is mainly used as a correction factor for the actual gap variation due to the roll temperature rise. Since it is influenced by the thermal expansion of a roll, it is necessary to grasp | ascertain the change of GME exactly with time, including cooling at the time of non-rolling. On the other hand, the rolling load (F) is calculated according to the general formula, such as the following formula (2).

F = KmWd (R '(Hh)] ¹ / 2Qp ---------------------- (2)

Where Km is material mean strain resistance, Wd is sheet width, H is grain thickness, (Hh) is rolling reduction, R 'is flat roll radius, and Qp is rolling reduction function (friction coefficient, F, H, h, Km , Tension, and so on). Therefore, when the plate thickness (H), the plate thickness (h), the plate width (Wd), the tension, and the material of the rolled material are determined from the process plan and the product specifications, the friction coefficient between the roll and the rolled material, the plate temperature, etc. According to the results of the above, based on Equation (1) and (2) above, a roll gap suitable for setting rolling conditions in each rolling stand can be calculated and determined.

The plate thickness and plate temperature vary with each rolling, and the coefficient of friction is mainly determined by the surface condition of the roll and the lubrication conditions and gradually changes as the rolling is repeated. For this reason, in the actual rolling condition setting calculation, it is necessary to estimate the side plate thickness and the plate temperature immediately before rolling, and the friction coefficient is predicted based on the rolling performance up to the previous time. The measured error included in the measured entry plate thickness and plate temperature is a disturbance factor of the set error with respect to the roll gap. In addition, when there is no plate thickness meter between each rolling stand, especially in the case of Tandem Mill, it is generally difficult to install the plate thickness meter between the rolling stand, so the estimated plate thickness calculated from one stand is used as it is. It can only be used as a side plate thickness. For this reason, when an error is included in the predicted calculated board thickness, this error becomes the side plate thickness error of the next stand as it is to be a roll gap setting error factor of the next stand. In addition, it is also difficult to correct the preset roll gap before the mail order in the stand.

In recent years, the rolling error (rolling load, roll gap, etc.) at the upstream stand of a continuous rolling mill is estimated by the setting error in the stand, and the rolling thickness is set dynamically by setting the roll gap of the downstream stand based on this setting error. Attempting to increase the degree of control. For example, Japanese Patent Laid-Open No. 3-3241 discloses a method for correcting a roll gap set on a downstream stand based on a load error (expected load-measured load) calculated using rolling results from a plurality of stands on the upstream side. Is disclosed. However, in this case, since the roll gap correction is performed using a regression equation, it is difficult to make a dynamic setting suitable for various rolling conditions. In addition, Japanese Patent Laid-Open No. 9-276915 estimates the entry speed and gauge meter correction coefficient of the rolled material based on the rolling load performance at the first stand when the rolled material is inserted into the first stand, and the estimated gauge meter correction is performed. The exit plate thickness is predicted using the coefficient, and the exit material temperature of the first stand is estimated based on the exit plate thickness and the entry speed thus obtained, and the entrance material temperature of the second stand calculated based on the exit material speed. The method of dynamic setting is introduced by calculating the roll gap and the roll speed which minimize the rolling error in each stand below the second stand. Here, the rolling set equation is used to predict the entry speed of the rolled material, gauge measurement coefficient, exit plate thickness, temperature of the rolled material, etc. from the rolling load performance. . Therefore, when the degree of the rolling setting equation is low, the degree of dynamic setting is also inevitably lowered.

In order to solve the above problems, the present invention utilizes a neural network having characteristics that can describe complex rolling phenomena through learning without relying on rolling setting equations. The study aims to provide a method for accurately obtaining the final target plate thickness by predicting the set error (plate thickness error) from the rolling condition and dynamically resetting the rolling condition of the downstream stand.

1 is a flowchart of a rolling control process according to the method of the present invention;

2 is a structural diagram of a neural network for predicting a set thickness error in the present invention;

3 is a structural diagram of a neural network for calculating a set thickness error correction coefficient according to the present invention;

Figure 4 is a general configuration diagram of a continuous rolling mill to which the present invention is applied.

Explanation of symbols on the main parts of the drawings

A: Predicted thickness error prediction neural network B: Correction factor calculation neural network

W: Rolling material 1-7: Rolling stand 8: Rolling roll

9: Backup roll 10: Rolling controller 11: Control system

12: sensor

Hereinafter, with reference to the accompanying drawings, the technical details of the present invention for achieving the above object will be described in more detail.

The rolling control method of the continuous rolling mill according to the present invention is upstream when the tip of the rolled material passes through the intermediate stand (for example, the fourth stand in the case of the continuous rolling mill in which the first to seventh stands are installed in series) as shown in FIG. The final stand exit thickness error Δhnn ₁ is predicted by the neural network A from the rolling load of the side stands (first to fourth stands) and the roll gap of the downstream stands (five to seventh stands), The correction coefficient β for correcting the thickness error in consideration of the steel grade is calculated by using the neural network B, and then the thickness error obtained is multiplied by the correction coefficient to calculate the set thickness error Δh ₁ . And the roll gap and the roll speed of the downstream stands (5th to 7th stands) are dynamically set so that the thickness error becomes zero.

In the present invention, as shown in Figure 2, the neural network (A) is the width (wd) of the rolled material, the rolling force (F) of the first to fourth stands, the roll gap (Sd ₁ ) of all the stands and the target The difference between the thickness (Ht ₁ ), the number of rolls after roll change (Nr), the temperature of the rolled material before rolling (FTo), the target thickness of the rolled material (ht ₇ ) and the temperature (FTt) at the exit of the final stand are used as input variables. To output the final stand exit thickness error (Δhnn ₁ ).

In addition, the neural network (B) is a constituent chemical element (C, Mn, Si, Ni, Cr, Nb, V, Cu, Ti, P, S, Al, Mo, B, Sn, N, Ca) is used as an input variable to output a correction coefficient β representing the ratio of the actual thickness error to the predicted thickness error Δhnn ₁ of the neural network A.

EXAMPLE

Figure 4 shows the overall configuration of a continuous rolling mill to which the present invention is applied. Herein, a case in which the present invention is applied to a continuous rolling mill having a series of seven stages 1 to 7 arranged in series is shown. Each stand 1 to 7 includes a rolling roll 8 and a backup roll 9, and the pressure load is controlled under the control of the rolling controller 10 to allow the pressure to enter between the rolling rolls 8, respectively. The softening material W is rolled. The rolled material W is sequentially rolled from the first stand 1 to the seventh stand 7 to obtain a plate thickness of a desired product specification at the exit side of the final seventh stand 7. 11 is a general control system mainly composed of microprocessors and the like. In accordance with the control of this control system 11, the operation of the rolling controller 10 with respect to each stand 1-7 is controlled, and adjustment of the rolling load etc. by a rolling roll is made by this. In addition, 12 is a sensor, such as a plate thickness meter, provided in the exit side of the 7th stand 7. As shown in FIG. The thickness and the plate temperature of the to-be-rolled material rolled through the rolling mill by the sensor 12, that is, after rolling, are measured.

In the continuous rolling mill configured as described above, the dynamic setting of the rolling conditions is basically determined in advance the rolling conditions at each stand 1 to 7 based on the process plan and the product specifications, and the roll gap Sdi and the roll speed ( Vi) (where i = 1 to 7) is set. After that, rolling starts and when the rolled material W reaches the fourth stand 4, the rolling performance (rolling force) in the first to fourth stands 1, 2, 3, and 4 which are upstream stands. , Roll gap, thickness measurement between stands, etc.) and the roll gaps of the fifth to seventh stands (5) (6) and (7) which are the downstream stands, and based on these values, the roll gap and the roll driving speed of the downstream stand are dynamically determined. By modifying and resetting, the rolling conditions of each stand are optimized. The dynamic setting of the rolling conditions in the downstream stands 5, 6 and 7 is made by the control system 11 which controls the rolling mill as a whole.

Accordingly, the present invention reduces the plate thickness deviation at the tip of the rolling material by dynamically correcting and resetting the error of the initial rolling condition of the rolling mill in the downstream stand by using the dynamic setting method of the rolling conditions as described above. It is to provide an effect that can greatly contribute to the improvement.

Claims (1)

In a continuous rolling machine in which a plurality of rolling stands are arranged in series to perform continuous rolling, the rolling load of the upstream stands and the exit thickness at the final stand from the roll gap of the upstream stands and the roll gap of the downstream stands when the leading end of the rolled material passes through the intermediate stand. Predicting an error Δhnn ₁ , correcting the thickness error with a correction coefficient β calculated in consideration of steel grade, calculating a set thickness error, and a roll gap of a downstream stand such that the thickness error is zero. And dynamically setting the roll speed, The prediction of the thickness error of the final stand exit side includes the width of the rolled material, the rolling force of the upstream stand, the difference between the roll gap and the target thickness of all stands, the number of rolls after the roll change, the temperature of the rolled material before rolling, and the rolled material at the exit of the final stand. It is made by the neural network that outputs the final stand exit thickness error by using the target thickness and temperature as input variables. The rolling control method for a continuous rolling mill using a neural network, characterized in that the calculation of the correction coefficient for correcting the thickness error is made by a neural network for outputting a correction coefficient using the constituent chemical elements of the material to be rolled.