CN101927269A - Method for adjusting three-roll mill roll gap controller - Google Patents
Method for adjusting three-roll mill roll gap controller Download PDFInfo
- Publication number
- CN101927269A CN101927269A CN200910191174XA CN200910191174A CN101927269A CN 101927269 A CN101927269 A CN 101927269A CN 200910191174X A CN200910191174X A CN 200910191174XA CN 200910191174 A CN200910191174 A CN 200910191174A CN 101927269 A CN101927269 A CN 101927269A
- Authority
- CN
- China
- Prior art keywords
- roll
- controller
- adjusting
- roll gap
- calculate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Control Of Metal Rolling (AREA)
Abstract
The invention discloses a method for adjusting a three-roll mill roll gap controller. The roll gap controller is characterized in that the adjusting method comprises the steps of: A. deducing controller parameters Kp, Ki and Kd through a neural network and calculating control signals by increment control equations for adjusting the controller; and B. Balancing a rolling force: acquiring corresponding values of a off-line measuring position and pressure, calculating rolling force deviation of the roll to be compensated on a stand according to a curvilinear equation generated from a sampling point, and compensating the control signals through the calculated displacement deviation. The invention uses the neural network to calculate proper controller parameters so as to make the control more accurate. Meanwhile, the roll gap controller adopts a centering compensating method, thus the roll gap controller can avoid position deviation of roll gap resulted under the action of rolling pressure. In addition, the adjusting method for balancing rolling force solves the problem of uneven load distribution in different directions of a pipe so as to achieve uniformity of wall thickness of a product and improve productivity of the product.
Description
Technical field
The present invention relates to a kind of control method of milling train, relate in particular to a kind of mill roll-gap controller control method.
Background technology
The HGC of hot continuous rolling seamless steel tubing mill (hydraulic pressure roll gap control both at home and abroad at present, Hydraulic Gap Control) system all can expose a problem on production and product quality: the thickness of steel pipe fluctuation that the front pony roughing mill comes out is bigger, and the precision influence to last finished frame is very big like this.In addition, because thickness fluctuation is big, the tension force control and the roll-force that are inconjunction with tubing all can produce big fluctuation, and the consequence that causes is exactly the vibration severity of follow-up frame, and ripples line or buckling phenomenon appear in body easily.The reason that thickness fluctuation is big mainly is because the position control method that the front frame adopts is not to be in perfect condition in precision control, is exactly that to influence accuracy factors be through tubular bad, the uneven thickness of hollow billet after the perforation in addition.With regard to the technology of three-roll continuous rolling, controller+servo valve+hydraulic pressure cludy+position sensor, this is typical mechanical electronic hydraulic hybrid system, its characteristic feature is that non-linear, uncertain, time variation, external interference and cross-couplings are disturbed.
Traditional mode is the mode that adopts off-line to regulate, and determines ratio, integration and differential coefficient, and then puts in the controlling unit that produces control signal and go.The shortcoming of this way has: (1) accuracy aspect.The controller parameter that off-line is debugged out is difficult to satisfy actual production and regulates requirement.(2) reliability aspect.Because full dose output, so each output is all relevant with state in the past, when a certain link of output breaks down, corresponding servo valve opening degree will produce sudden change, will inevitably cause the variation significantly of Position of Hydraulic Cylinder, this is unallowed in production practices, also may cause great production accident sometimes.Different with the other types milling train is, three-high mill will be considered the problem of locating in tube rolling, and just the geometric center of three rollers will overlap with rolling center, otherwise the phenomenon that the hollow forging of producing can bend even rupture.So, consider the characteristic of three-roll continuous rolling in the experience of carrying out to indiscriminately imitate simply when roller meets position adjustments sheet material etc.
In addition, when considering the problems referred to above, also be appreciated that the validity of real-time adjustment.Because plant equipment exists large time delay in control, so for electrical control equipment, the output of The whole control system and the location of plant equipment are nonsynchronous.Like this, when roll-force fluctuation ratio in the operation of rolling was serious, the offset fluctuation of instantaneous generation can be very huge, frequently increases concussion because of regulating easily.
Summary of the invention
The purpose of this invention is to provide the high method for adjusting three-roll mill roll gap controller of a kind of control accuracy, to achieve these goals, by the following technical solutions:
A kind of method for adjusting three-roll mill roll gap controller is characterized in that: described control method comprises the following steps:
A, by neutral net derivation controller parameter K
p, K
i, K
dUse the increment type governing equation to calculate control signal, controller is regulated;
Controller parameter K in the described steps A
p, K
i, K
dObtain by following step:
One group of sample that is total to K+1 moment point of rolling data acquisition server picked at random, the K that chooses wherein organizes as training sample, calculates controller parameter K by neutral net
p, K
i, K
dValue;
Use other one group of sample to carry out verification,, then determine controller parameter K if satisfy condition
p, K
i, K
dValue, if do not satisfy, then return resampling and calculate;
Wherein the adjustment expression formula of the network weight coefficient of neutral net is:
In the formula,
Δ w
Il (m)(k)-i neuron of m layer network calculate weights index variation amount about the K point of l input variable;
η: learning efficiency;
α: inertia coeffeicent;
O
l (m)(k): the m layer network is from the K point sampling output of i input;
K: K sampled point;
I: i neuron;
L: l input variable;
E (k): setting value is poor with actual feedback;
η,α∈(0,1);
The increment type governing equation is
u
(k)=u
(k-1)+k
p*[e
(k)-e
(k-1)]+k
ie
(k)+k
d*[e
(k)-2e
(k-1)+e
(k-2)]
In the formula
u
(k): the controlled quentity controlled variable of k sampled point;
E (k): setting value is poor with actual feedback;
k
p: proportionality coefficient;
k
i: integral coefficient;
k
d: differential coefficient;
B, be rolled dynamic balance:
Gather the respective value of off-line measurement position and pressure; According to sampled point formation curve equation, calculate the roll-force deviate of roll to be compensated in this frame, the offset deviation amount that calculates is compensated in the control signal.
In order to increase the lumber recovery of product, also have behind the described step B and three rollers are carried out centering compensate this step: (the fluctuation average Δ A of k ~ k+D), regular back obtains the deviation compensation value M of three roller displacements in D sampling gap to utilize controller cycle to calculate three roller displacements
i, i=1,2,3, the offset deviation amount that itself and the balanced compensated module of roll-force are calculated is done algebraical sum, obtain K+D compensation output constantly;
Beneficial effect of the present invention is: the present invention adopts the rational controller parameter of calculating of neutral net, feasible control is more accurate, adopt the method for fixed middle compensation to solve simultaneously because the effect of draught pressure, roller gap can be offset this problem by occurrence positions, and the control method of employing roll-force balance, solved the problem of pipe, thereby made product reach the uniform purpose of wall thickness, increased the lumber recovery of product at all directions unbalance stress.
Description of drawings
Fig. 1 is a middle controller calculation method of parameters step schematic diagram of the present invention;
Fig. 2 is roll-force balance adjusting method step schematic diagram among the present invention.
The specific embodiment
Below in conjunction with drawings and Examples the present invention is done detailed description:
As shown in Figure 1 and Figure 2:
The present invention adopts is thought derivation controller parameter with neutral net, controller is used the implementation of the double insurance of increment type governing equation again.PID (Proportion Integration Differentiation) control will obtain controls effect preferably, just must be by adjusting three kinds of control actions of ratio kp, integration ki and differential kd, form in the controlled quentity controlled variable and not only to cooperatively interact but also the relation of mutual restriction, this relation is simple " linear combination " not necessarily, can find out best relation from countless changes nonlinear combination.Any non-linear expression's ability that neutral net had can realize that the PID with best of breed controls by the study to systematic function.Adopt BP (Back Propagation) neutral net, can set up the self-regulating controller of parameter kp, ki and kd.
The computing network weight coefficient of this controller parameter is adjusted expression formula:
In the formula,
Δ w
Il (m)(k)-i neuron of m layer network calculate weights index variation amount about the K point of the 1st input variable;
η: learning efficiency;
α: inertia coeffeicent;
O
l (m)(k): the m layer network is from the K point sampling output of i input;
K: K sampled point;
I: i neuron;
L: l input variable;
E (k): setting value is poor with actual feedback;
η,α∈(0,1)。
This computing network weight coefficient is adjusted expression formula and is applicable to all types of milling trains, can be applied in various the needs in hot rolling that milling train participates in and cold rolling or the like the steel rolling process.
Use increment type governing equation (2) to calculate control signal;
u
(k)=u
(k-1)+k
p*[e
(k)-e
(k-1)]+k
ie
(k)+k
d*[e
(k)-2e
(k-1)+e
(k-2)]…(2)
In the formula
u
(k): the controlled quentity controlled variable of k sampled point;
E (k): setting value is poor with actual feedback;
k
p: proportionality coefficient;
k
i: integral coefficient;
k
d: differential coefficient;
The present invention utilizes the increment type governing equation of (2) formula to calculate the size of control signal.Be calculated as output increment because PID participates in,, can also remove adjustings (other instructions as the hydraulic pressure cludy are moved) with the method that logic is judged in case of necessity so influence is minimum during misoperation.
By the sample of K+1 moment point of rolling data acquisition server picked at random, the K that chooses wherein organizes as training sample, is taken at K displacement setting value Rin constantly respectively
(k)(input set), at K Displacement Feedback amount Cout constantly
(k)(output feedback quantity) and Rin
(k)-Cout
(k)(deviation signal) is updated in the neutral net and goes as input parameter, calculates the w of each layer of neutral net
Il (m)(k), with the output valve (k of neutral net
p, k
i, k
d) calculate, and render in the controlling party formula (2).Substitution Rin
(k)And Cout
(k)In the control closed loop, calculate the control output u of this moment
(k), and reading k actual hydraulic pressure cylinder stroke value Cout constantly
T (k)
With Rin
(k)And Cout
T (k)Substitution performance index function E (k)=| Rin
(k)-Cout
T (k)|
2In, obtain current mean square of error difference.The performance index function E (k) here is that self-defining in the system, different system can have different functions.
The error requirements of uniting of setting up departments is A, and so self-regulating end condition is:
|E(k)-A|<ε …………(3)
In the formula
E (k)-performance index function;
The value of ε-be infinitely close to zero.
If the value of E (k) does not satisfy end condition, then proceed the training of next group sample, and come the weights coefficient w of each layer on the regulating networks according to formula (1)
Il (m)(k).When the value of E (k) satisfies condition, regulate stopping the one group of weights coefficient that obtains, just proportionality coefficient k
p, integral coefficient k
i, differential coefficient k
d
Re-use other one group of sample as the verification sample.If this group sample is updated to the network weight coefficient w that preceding step calculates
Il (m)(k) after, make the verification sample restrain equally, then confirm as the optimized parameter that satisfies condition, thereby determined k
p, k
iAnd k
dOtherwise the continuation sample training is till satisfying condition.
The control method of roll-force balance:
Because the technologic restriction of punch, the hollow billet wall thickness of producing is uneven, so after frame was stung steel, the suffered roll-force size of each roll also was unbalanced, the lateral sliding moment that is easy to like this cause steel pipe is to a certain tangential skew, thus the near round property of destruction frame pass.The present invention is according to the thought of dynamic balance, and the deviation of roll-force is converted into the position quantity signal, compensates in the control model again and goes, and makes the stressing conditions of pipe on all directions consistent, thereby makes product reach the uniform purpose of wall thickness.
The control method of roll-force balance divided for four steps carried out in this invention.
The first step: the respective value of gathering off-line measurement position and pressure.
At first under the situation of off-line, serve as rolled piece (reason is not yielding and damages), utilize pressure closed loop to pursue the value that measurement point pressure and location variable are finished in rank with plug or the very strong corner bracket of mechanical stiffness.The handoff procedure of control closed loop from the Position Control closed loop to pressure, guarantee at first that each roll of frame to be measured and hard rolled piece are to be in the actual contact state, the roll-force of just observing every roller is greater than contact thresholding m.This thresholding size is 0<m<F, and wherein, F is for measuring initial pressure.Tandem rolling unit for the multimachine frame, owing to exist correlation between each frame, when measuring particular chassis, simultaneously all frames are adjusted to position to be measured, and all switch to pressure closed loop control, can make influence each other drop to minimum like this.
Second step: according to sampled point formation curve equation.
Then according to the value of each measurement point pressure and location variable, the method for utilizing the substrate in second state space to approach with curve obtains the relation between roll-force and the hydraulic cylinder displacement.
The 3rd step: calculate the displacement roll-force departure of roll to be compensated in this frame.
(sting steel) behind the frame bringing onto load, measure the deviation size of real-time roll-force of each roll and the average roll-force of three rollers, the curvilinear equation that corresponding second step generates obtains the offset deviation amount of roll, and the offset deviation amount that calculates is compensated in the control signal.
The fixed middle compensation method of three rollers among the present invention.
In the process of actual production, be the scope that the present invention considers equally with the intrinsic position deviation of each roll of frame.The present invention utilizes sampling thheorem, mark off rational sampling interval after, calculate the centre of location compensation rate of balance three rollers in D sampling period, to reach the purpose in fixed.
Utilize controller cycle calculate three roller displacements in D sampling gap (fluctuation of k ~ k+D) average Δ A, wherein
P
1, P
2, P
3Be respectively the displacement fluctuation value of three rolls, regular back obtains the deviation compensation value M of three roller displacements
i, (i=1,2,3), wherein M
i=P
i-Δ A.The offset deviation amount that itself and the balanced compensated module of roll-force are calculated is done algebraical sum, just can obtain K+D compensation output constantly.
At last, the output valve of software is imported the PID controlling unit into, through finishing the conversion of control signal after flow gain compensation, drift compensation, filtering and the D/A conversion.
In the production process, running of the present invention is as follows:
At first first hollow billet sampled in the operation of rolling, and drops into balanced compensated function of roll-force and centering function.At first hollow billet and second controller parameter that calculates optimum intermittence according to the sampling situation that hollow billet is rolling.
Adopt the control parameter that obtains according to first sampling to control in second hollow billet operation of rolling, and drop into balanced compensated function of roll-force and centering function, simultaneously second hollow billet sampled.Calculate the control parameter at second hollow billet and the 3rd gap that hollow billet is rolling, balanced compensated function of roll-force and centering function are controlled, dropped into to the parameter value that uses second hollow billet to obtain in the 3rd process that hollow billet is rolling, and sample simultaneously, so repeat, until whole rolling the finishing of hollow billet.
Through above-mentioned steps, in the three-high mill self-regulation cycle of promptly having finished particular chassis, this process all is suitable for each frame of tandem rolling unit.
From above about formation of the present invention and graphic analysis, can draw, this method has been optimized the roll seamless steel tube technological process based on three-high mill, in the process of thereof using PID controller link, taken into full account the real relations of production, basic automatization and process automation have been combined closely.Also taken into account the three roll center lines problem consistent with rolling centerline.In addition, the roll-force based on the three-roll continuous rolling unit that adopts in the program is balanced compensated, can guarantee through rolling hollow billet wall thickness more even.And owing to the self-regulating calculating of controller parameter is carried out separately, and be to utilize carry out the rolling intermittence of facing two hollow billets mutually, get final product so when producing, only need handle follow-up little value calculating.So both guarantee high response speed, guaranteed operational reliability again.
Claims (2)
1. method for adjusting three-roll mill roll gap controller, it is characterized in that: described control method comprises the following steps:
A, by neutral net derivation controller parameter K
p, K
i, K
dUse the increment type governing equation to calculate control signal, controller is regulated;
Controller parameter K in the described steps A
p, K
i, K
dObtain by following step:
One group of sample that is total to K+1 moment point of rolling data acquisition server picked at random, the K that chooses wherein organizes as training sample, calculates controller parameter K by neutral net
p, K
i, K
dValue;
Use other one group of sample to carry out verification,, then determine controller parameter K if satisfy condition
p, K
i, K
dValue, if do not satisfy, then return resampling and calculate;
Wherein the adjustment expression formula of the network weight coefficient of neutral net is:
In the formula,
Δ w
Il (m)(k)-i neuron of m layer network calculate weights index variation amount about the K point of the 1st input variable;
η: learning efficiency;
α: inertia coeffeicent;
O
l (m)(k): the m layer network is from the K point sampling output of i input;
K: K sampled point;
I: i neuron;
L: the 1st input variable;
E (k): setting value is poor with actual feedback;
η,α∈(0,1);
The increment type governing equation is
u
(k)=u
(k-1)+k
p*[e
(k)-e
(k-1)]+k
i?e
(k)+k
d*[e
(k)-2e
(k-1)+e
(k-2)]
In the formula
u
(k): the controlled quentity controlled variable of k sampled point;
E (k): setting value is poor with actual feedback;
k
p: proportionality coefficient;
k
i: integral coefficient;
k
d: differential coefficient;
B, be rolled dynamic balance:
Gather the respective value of off-line measurement position and pressure; According to sampled point formation curve equation, calculate the roll-force deviate of roll to be compensated in this frame, the offset deviation amount that calculates is compensated in the control signal.
2. method for adjusting three-roll mill roll gap controller as claimed in claim 1, it is characterized in that: also have behind the described step B and three rollers are carried out centering compensate this step: (the fluctuation average Δ A of k ~ k+D), regular back obtains the deviation compensation value M of three roller displacements in D sampling gap to utilize controller cycle to calculate three roller displacements
i, i=1,2,3, the offset deviation amount that itself and the balanced compensated module of roll-force are calculated is done algebraical sum, obtain K+D compensation output constantly.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200910191174XA CN101927269A (en) | 2009-10-22 | 2009-10-22 | Method for adjusting three-roll mill roll gap controller |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200910191174XA CN101927269A (en) | 2009-10-22 | 2009-10-22 | Method for adjusting three-roll mill roll gap controller |
Publications (1)
Publication Number | Publication Date |
---|---|
CN101927269A true CN101927269A (en) | 2010-12-29 |
Family
ID=43366852
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN200910191174XA Pending CN101927269A (en) | 2009-10-22 | 2009-10-22 | Method for adjusting three-roll mill roll gap controller |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101927269A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102363160A (en) * | 2011-11-21 | 2012-02-29 | 安阳钢铁股份有限公司 | Method for adjusting zero of rolling mill for producing checkered plates |
CN104932438A (en) * | 2014-03-20 | 2015-09-23 | 南京富士通南大软件技术有限公司 | Industrial process automation control system and control method thereof |
CN105259754A (en) * | 2015-10-16 | 2016-01-20 | 华北理工大学 | Board thickness intelligent control method based on active learning |
CN110227721A (en) * | 2019-05-31 | 2019-09-13 | 宝钢湛江钢铁有限公司 | A kind of control method for mill roll-gap in speed-change process |
CN110424183A (en) * | 2019-08-14 | 2019-11-08 | 浙江华章科技有限公司 | A kind of sizing applicator press nip value control method |
WO2020000264A1 (en) * | 2018-06-27 | 2020-01-02 | Siemens Aktiengesellschaft | Equipment management method, device, system and storage medium |
CN113617855A (en) * | 2021-07-16 | 2021-11-09 | 太原科技大学 | Rolling mill control method and system |
-
2009
- 2009-10-22 CN CN200910191174XA patent/CN101927269A/en active Pending
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102363160A (en) * | 2011-11-21 | 2012-02-29 | 安阳钢铁股份有限公司 | Method for adjusting zero of rolling mill for producing checkered plates |
CN104932438A (en) * | 2014-03-20 | 2015-09-23 | 南京富士通南大软件技术有限公司 | Industrial process automation control system and control method thereof |
CN104932438B (en) * | 2014-03-20 | 2017-11-24 | 南京富士通南大软件技术有限公司 | A kind of industrial process automation control system and its control method |
CN105259754A (en) * | 2015-10-16 | 2016-01-20 | 华北理工大学 | Board thickness intelligent control method based on active learning |
CN105259754B (en) * | 2015-10-16 | 2018-07-10 | 华北理工大学 | A kind of plate thickness intelligent control method based on Active Learning |
WO2020000264A1 (en) * | 2018-06-27 | 2020-01-02 | Siemens Aktiengesellschaft | Equipment management method, device, system and storage medium |
CN110227721A (en) * | 2019-05-31 | 2019-09-13 | 宝钢湛江钢铁有限公司 | A kind of control method for mill roll-gap in speed-change process |
CN110424183A (en) * | 2019-08-14 | 2019-11-08 | 浙江华章科技有限公司 | A kind of sizing applicator press nip value control method |
CN113617855A (en) * | 2021-07-16 | 2021-11-09 | 太原科技大学 | Rolling mill control method and system |
CN113617855B (en) * | 2021-07-16 | 2023-02-17 | 太原科技大学 | Rolling mill control method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101927269A (en) | Method for adjusting three-roll mill roll gap controller | |
CN102632085B (en) | Cold-rolled strip steel plate shape control system and method | |
CN101934292B (en) | Automatic control method for camber and wedge of hot rolled strip roughing mill | |
CN107913911B (en) | It is a kind of to monitor thickness control to the feedforward compensation method and system of entrance tension | |
CN101607264A (en) | A kind of periodic longitudinal variable-thickness strip, longitudinal variable-thickness sheet material and preparation method thereof | |
CN202606511U (en) | Cold rolled steel plate shape feedforward-feedback cooperative control system | |
CN113373295B (en) | Annealing furnace digital twin intelligent alarm system and method based on multilayer perceptron | |
CN103191919B (en) | Optimizing method for on-line control to coefficient model during strip rolling | |
CN103100564A (en) | Novel rolling process self-adaptive control method | |
CN104942019A (en) | Automatic control method for width of steel strips during cold rolling | |
CN105834225A (en) | Thickness controlling method for flying gauge changing rolling of cold tandem mill and controlling system for cold tandem mill | |
CN101618401A (en) | High precision strip rolling thickness control method based on feedback signals by thickness gauge | |
CN104923568B (en) | A kind of control method for preventing thin strip cold rolling process broken belt | |
CN103042042B (en) | Eccentric compensation method of roller based on discrete auxiliary closed loop | |
CN102581024B (en) | Control method for full-length fluctuation of steel rail height | |
CN101927271A (en) | Roll eccentricity compensation method based on on-line recursive parameter estimation and equipment thereof | |
CN101658863A (en) | Compensation method of seamless steel pipe three-roll mill | |
Hameed et al. | Strip thickness control of cold rolling mill with roll eccentricity compensation by using fuzzy neural network | |
CN102303051B (en) | Pipe shape control method for rolling seamless steel pipe by using seven-frame tandem mill | |
CN101745541A (en) | Closed loop speed compensation method for thickness control system of aluminum cold rolling mill | |
CN102350441B (en) | Control method of wall thickness of seamless steel tube | |
CN102873106B (en) | Quick and precise elongation control method for temper mill | |
CN102641896B (en) | Gauge and flatness comprehensive control system of cold rolled steel sheet | |
CN102581032A (en) | Feed-forward control method for cold-rolled steel strip shape | |
CN109174978A (en) | A kind of calculation method overcoming planisher elongation percentage measurement defect |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20101229 |