CN102363159B - Thickness control method for single precision cold-rolled sheet thickness measuring system - Google Patents

Abstract

The invention discloses a thickness control method for a single precision cold-rolled sheet thickness measuring system, and belongs to the field of mechanical automation control. Because a single sheet is short and requires short rolling time, a thickness feedback value acquired by a thickness gauge cannot be used for thickness closed-loop control. A thickness measured value is used for self-learning of a sheet thickness computation model of a process control system, the deformation resistance and a friction coefficient under the actual working condition of a material are corrected, and the thickness control accuracy is improved fundamentally by improving the set computational accuracy of the sheet thickness computation model. By an advanced support vector machine modeling method, the deformation resistance and the friction coefficient of the single sheet are subjected to statistical modeling, and an analytic model and statistical model combined model structure is formed; and a plurality of deformation resistance support vector machines are established according to different cumulative reduction ratios, and a plurality of friction coefficient support vector machines are established according to different rolling linear speeds, so that the action range of each support vector machine is narrowed, the computational accuracy of the model is effectively improved and the thickness control accuracy for single cold-rolled products can be greatly improved.

Description

A kind of method for controlling thickness of spiral-plate precision cold-rolled thickness of slab measuring system

Technical field

The application belongs to the mechanical automation control field, relates in particular to a kind of method for controlling thickness of spiral-plate precision cold-rolled thickness of slab measuring system.

Background technology

In present cold rolling production, more realized the tension force coil rolling, and rolling for special kinds such as titanium alloys, because single batch of demand is little, perhaps because resistance of deformation is difficult than macrobending, be difficult to coil rolling, more adopt without the tension force spiral-plate rolling.

Yet the spiral-plate precision rolling but has more technological difficulties:

(1) spiral-plate precision rolling specification is flexible, single batch of output is little, and supplied materials thickness and section configuration have larger uncertainty;

(2) spiral-plate is cold rolling owing to lack tension force, so that resistance of deformation is large, thereby causes that rolling load is large, the relative deformation rate is little, and rolling process is many, rolling pass is many;

(3) with respect to coil rolling, the cold rolling because spiral-plate length of spiral-plate is shorter, lacks online thickness, plate shape is regulated automatically, and the assurance of thickness of slab, strip shape quality very relies on the setting model precision;

(4) (the downstream passage is set thickness error may reach same magnitude with pass deformation to spiral-plate for cold rolling thickness≤6mm, finished product thickness 0.2～2.0mm), and thickness of slab control difficulty is large without a precision rolling thin thickness;

Comprehensive above characteristics as can be known, spiral-plate precision cold-rolled resistance of deformation is large, rolling pass is many, it is the most cold rolling mode of production of difficult assurance of gauge and shape precision, yet, in spiral-plate is cold rolling, because spiral-plate length is shorter, can't as the tension force coil rolling, realize online closed loop thickness control, and can only guarantee the finished product thickness precision by the thickness setting precision that improves model.

Improve the thickness setting precision of model, must solve two key links, namely obtain thickness measurements accurately and method for controlling thickness based on thickness measurements.

At first should solve the thickness measure problem.Present spiral-plate is cold rolling, can finish at the spiral-plate milling train of special use, also can finish at the dual-purpose milling train of coiled sheet.The milling train of the two itself is as broad as long, is four rollers or six roll reversing rollers, mainly be machine before, the difference of the configuration of the roller-way behind the machine and coiling machine configuration.The personality board milling train is installed roller-way before and after the single chassis reversable mill, realize the transportation of spiral-plate and sting steel that the dual-purpose milling train of coiled sheet has both been installed roller-way before and after milling train, can realize that spiral-plate is rolling, slewing rollers and coiling machine also are installed in both sides simultaneously, can be realized coil rolling.

Present special-purpose spiral-plate milling train, general automatic control accuracy is lower, and calibrator seldom is installed, and band special rolling machine is then more installed X ray or gamma thickness gauge in the milling train both sides, and is equipped with high-precision automatic gauge control system, i.e. AGC.The dual-purpose milling train of coiled sheet, actual is on roll coil of strip milling train basis, and roller way system is installed, and has increased the function of rolling spiral-plate.

For special-purpose spiral-plate milling train and since roller-way stop that calibrator can't record belt steel thickness; And for the dual-purpose milling train of coiled sheet, although in the milling train both sides calibrator is installed, owing to stopping of roller-way, when spiral-plate is rolling, can't record sheet metal thickness, and when the roll coil of strip is rolling, remove roller-way behind interior machine front roller-way and the interior machine, thereby avoid roller-way to stop, realize thickness measure.

The present patent application people has designed special segmentation roller-way structure before rolling machine, behind the machine in working before, realized when spiral-plate is rolling by the Measurement accuracy of calibrator to thickness of slab.The present invention will be cold rolling according to spiral-plate characteristics, be designed with targetedly method for controlling thickness.

Summary of the invention

Key technical problem to be solved by this invention is, for exist in the cold rolling THICKNESS CONTROL of spiral-plate can't real-time closed-loop control, thickness and precision guarantee the problems such as difficult, developed the method for controlling thickness with Parameter Self-learning function that combines with the SVMs statistical model based on analytic modell analytical model.

Technical scheme of the present invention is: a kind of method for controlling thickness of spiral-plate precision cold-rolled thickness of slab measuring system.

The thickness measurement system of spiral-plate precision cold-rolled and control system general structure (as shown in Figure 1) comprise following three parts:

(1) executing agency and measurement mechanism

(2) Basic automation control system

(3) Process Control System

On the basis of special roller-way structural design, the strip thickness information that X ray (or gamma-rays) calibrator measures, by the THICKNESS CONTROL PLC of Profibus-DP Internet Transmission to Basic automation level, THICKNESS CONTROL PLC depresses closed loop control mode in order to finish two kinds of roll gap closed-loop control, pressure closed loop controls, namely respectively by magnetic scale feedback signal and pressure head feedback signal, realize ring function in the thickness control system execution.

Because spiral-plate length is short, rolling time is few, the thickness value of feedback that can't utilize calibrator to obtain is carried out closed loop thickness control.Therefore, the present invention proposes following control method: the thickness of slab computation model that thickness measurements is used for Process Control System carries out Model Self-Learning, material deformation drag and actual condition coefficient of friction are revised, by improving the precision of thickness of slab computation model, fundamentally improve thickness control accuracy.For finding the source of the model error of calculation, analyze each factor to the impact of model accuracy, the mode that this control method adopts analytic modell analytical model to combine with statistical model is carried out modeling, and overall structure is as shown in Figure 2.

This control method is divided into thickness setting to be calculated and thickness self study calculating two parts content, forms the thickness Learning Control Method, and detailed content is as follows:

The thickness setting calculating section:

The cold rolling production of spiral-plate is to begin to final products from the hot rolling material, according to specification requirement, and generally need to be through repeatedly annealing.Between hot rolling material and the 1st time annealing or the spiral-plate between the twice annealing rolling, be called a rolling process.In the rolling process, comprise again a plurality of rolling passes.Spiral-plate is carried out thickness setting when calculating in its single rolling process, at first distribute each passage thickness, determine that namely each passage entrance side is set thickness and outlet side is set thickness, then in accordance with the following steps since the 1st passage, carry out every time thickness setting and calculate:

(1) reads the rolling inlet of rolling mill side of spiral-plate and set thickness H, milling train outlet side setting thickness h, spiral-plate width W and spiral-plate roll line speed v;

(2) calculate rolling this passage resistance of deformation calculated value of spiral-plate Kf according to corresponding duty parameter by deformation resistance model, corresponding duty parameter comprises the accumulated deformation rate ε since the 1st passage _a(its computation model is ε _a=0.4 (H-h _J-1)/H+0.6 (H-h _j)/H, j is passage number), chemical composition Che (mass percent that refers to each constitutional chemistry element of spiral-plate), the hot rolling finishing temperature FTe as front operation, coiling temperature CTe, annealing temperature ATe, annealing time ATm, deformation resistance model is the model that the resistance of deformation analytic modell analytical model combines with the resistance of deformation statistical model, according to accumulated deformation rate ε _aObtain its output Kf by the resistance of deformation analytic modell analytical model _B, the resistance of deformation analytic modell analytical model is Kf _B=f ₁(ε _a);

The resistance of deformation statistical model is different according to the accumulated deformation rate, difference corresponding accumulated deformation rate 0%, 20%, 40%, 60%, 80%, 100%, formed by 6 SVMs, each SVMs be input as chemical composition Che, hot rolling finishing temperature FTe, coiling temperature CTe, annealing temperature Ate and annealing time ATm, obtain each SVMs output Kf according to input _Ci(i=1,2 ..., 6), 6 SVMs output Kf _CiBy accumulated deformation rate ε _aCarry out Gauss interpolation and obtain total statistical model output Kf _C, the Gauss interpolation computation model is:

$K{f}_C=\underset{i=1}{\overset{6}{\mathrm{\Σ}}}K{f}_{\mathrm{Ci}}\·g{k}_i\left({\mathrm{\ϵ}}_a\right)$

${\mathrm{gk}}_i\left({\mathrm{\ϵ}}_a\right)=\frac{e^{-\frac{{({\mathrm{\ϵ}}_a-{\mathrm{\ϵ}}_{\mathrm{ai}})}^2}{{\mathrm{\σ}}^2}}}{\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{e}^{-\frac{{({\mathrm{\ϵ}}_a-{\mathrm{\ϵ}}_{\mathrm{ai}})}^2}{{\mathrm{\σ}}^2}}}$

In the formula, ε _AiBe 6 the corresponding standard accumulated deformation of SVMs rates, be respectively 0%, 20%, 40%, 60%, 80%, 100%, σ is resistance of deformation statistical model Gauss interpolation width, is made as 14%, gk _i(ε _a) be accumulated deformation rate interpolation weights;

With resistance of deformation analytic modell analytical model output Kf _BWith resistance of deformation statistical model output Kf _CCarry out product, just obtain rolling this passage resistance of deformation calculated value of spiral-plate Kf, that is:

Kf＝Kf _B·Kf _C

(3) calculate rolling this passage calculation of friction coefficient value of spiral-plate μ according to corresponding duty parameter by the coefficient of friction model, corresponding duty parameter comprises gloss level of roll R, lubricant medium LTy, mill length L, spiral-plate roll line speed v, the coefficient of friction model is the model that the coefficient of friction analytic modell analytical model combines with the coefficient of friction statistical model, obtains its output μ according to gloss level of roll R, lubricant medium LTy, mill length L, spiral-plate roll line speed v by the coefficient of friction analytic modell analytical model _B, the coefficient of friction analytic modell analytical model is μ _B=f ₂(R, LTy, L, v); The coefficient of friction statistical model is different according to spiral-plate roll line speed, the corresponding roll line speed 0m/s of difference, 1m/s, 2m/s, 3m/s, 4m/s, 5m/s, formed by 6 SVMs, each SVMs is input as gloss level of roll R, lubricant medium LTy, mill length L, obtains each SVMs output μ according to input _Ci(i=1,2 ..., 6), 6 SVMs output μ _CiCarry out Gauss interpolation by spiral-plate roll line speed v and obtain total statistical model output μ _C, the Gauss interpolation computation model is:

${\mathrm{\μ}}_C=\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{\mathrm{\μ}}_{\mathrm{Ci}}\·g{\mathrm{\μ}}_i\left(v\right)$

$g{\mathrm{\μ}}_i\left(v\right)=\frac{e^{-\frac{{(v-v_i)}^2}{{\mathrm{\λ}}^2}}}{\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{e}^{-\frac{{(v-v_i)}^2}{{\mathrm{\λ}}^2}}}$

In the formula, v _iBe 6 corresponding standard speeds of SVMs, be respectively 0m/s, 1m/s, 2m/s, 3m/s, 4m/s, 5m/s, λ is coefficient of friction statistical model Gauss interpolation width, is made as 0.7m/s, g μ _i(v) be the speed interpolation weights;

With coefficient of friction analytic modell analytical model output μ _BWith coefficient of friction statistical model output μ _CCarry out product, just obtain rolling this passage calculation of friction coefficient value of spiral-plate μ, that is:

μ＝μ _B·μ _C

(4) on the basis that obtains spiral-plate rolling this passage resistance of deformation calculated value Kf and calculation of friction coefficient value μ, read current rolling duty parameter and milling equipment parameter, respectively by the roll-force computation model, the Calculating Torque during Rotary model, the power calculation model calculates rolling force setup value P, torque setting value T and set value of the power N, and further obtain fixed value of roller slit S by the roll gap computation model, rolling duty parameter is that the inlet of rolling mill side is set thickness H, the milling train outlet side is set thickness h, spiral-plate width W and spiral-plate roll line speed v, the milling equipment parameter comprises work roll diameter D _wWith milling train longitudinal rigidity C _P, the roll-force computation model is P=f ₃(H, h, W, Kf, μ, D _w), the roll torque computation model is T=f ₄(H, h, P, D _w), the power calculation model is N=f ₅(T, H, h, D _w, μ, v), gap Set Model is S=f ₆(P, h, C _p);

So far, obtained the fixed value of roller slit of rolling this passage of spiral-plate, the thickness setting of having finished this passage calculates, and comes back to step (1) until extreme trace;

After finishing a rolling process of a spiral-plate rolling, carry out in accordance with the following steps the thickness self study and calculate according to setting hit situation, can triggering:

(1) reads each passage entrance side thickness measurements H of the rolling milling train of spiral-plate ^*, milling train outlet side thickness measurements h ^*

(2) read the measurement of rolling force value P of each passage in the spiral-plate operation of rolling ^*, rolling power measured value N ^*And spiral-plate roll line velocity measurement v ^*

(3) according to every time inlet of rolling mill side thickness measurements H ^*, milling train outlet side thickness measurements h ^*, measurement of rolling force value P ^*, power measurement values N ^*, spiral-plate roll line velocity measurement v ^*And work roll diameter D _w, set up equation group by roll-force computation model, roll torque computation model and power calculation model, calculate every time resistance of deformation actual value Kf ^*With coefficient of friction actual value μ ^*

(4) calculate every time resistance of deformation actual value Kf ^*With the deviation ek of resistance of deformation calculated value Kf, computation model is ek=Kf-Kf ^*, calculate each resistance of deformation SVMs output bias ek according to the Gauss interpolation computation model on this basis _i(i=1,2 ..., 6), computation model is:

$e{k}_i=\∂\mathrm{ek}/\∂g{k}_i\left({\mathrm{\ϵ}}_a\right)$

Then obtain the new output Kfc of each SVMs _i ^*=Kfc _i-ek _iEach rolling pass, 6 new SVMs are exported with corresponding duty parameter chemical composition Che, hot rolling finishing temperature FTe, coiling temperature CTe, annealing temperature Ate and annealing time ATm and are formed one group of sample data, sample data by a plurality of passages in the single rolling process, be used for the weights of 6 SVMs are trained, thus the self study of implementation model resistance of deformation;

(5) calculate every time coefficient of friction actual value μ ^*With the deviation eu of calculation of friction coefficient value μ, computation model is eu=μ-μ ^*, calculate each resistance of deformation SVMs output bias eu according to the Gauss interpolation computation model on this basis _i(i=1,2 ..., 6), computation model is:

${\mathrm{eu}}_i=\∂\mathrm{eu}/\∂g{u}_i\left(v\right)$

Then obtain the new output μ of each SVMs _Ci ^*=μ _Ci-eu _iEach rolling pass, 6 new SVMs are exported and corresponding one group of sample data of the L shaped one-tenth of duty parameter gloss level of roll R, lubricant medium LTy, mill length, sample data by a plurality of passages in the single rolling process, be used for the weights of 6 SVMs are trained, thus the self study of implementation model coefficient of friction;

So far, by the self study to resistance of deformation and coefficient of friction, realized the self study to the thickness setting model.

The present invention adopts advanced model construction of SVM method that spiral-plate resistance of deformation and coefficient of friction are carried out statistical modeling, form the model structure that analytic modell analytical model combines with statistical model, and accumulate reduction ratio according to difference and set up a plurality of resistance of deformation SVMs, and set up a plurality of coefficient of friction SVMs according to different roll line speed, thereby dwindle the sphere of action of each SVMs, Effective Raise the model computational accuracy.

Description of drawings

Fig. 1 is the thickness measure of spiral-plate precision cold-rolled and control system overall construction drawing;

Fig. 2 is thickness of slab computation model overall structure figure.

Symbol description

Symbol	Title	Unit
			H	The inlet of rolling mill side is set thickness	mm
h	The milling train outlet side is set thickness	mm
			W	Width	mm
Kf	Resistance of deformation	Mpa
			ε a	The accumulated deformation rate	％
Che	Chemical composition	％
			FTe	Hot rolling finishing temperature	℃
CTe	The hot rolling reeling temperature	℃
			ATe	Annealing temperature	℃
ATm	Annealing time	min
			Kf B	The output of resistance of deformation analytic modell analytical model	Mpa
Kf C	The output of resistance of deformation statistical model	-
			σ	Resistance of deformation statistical model Gauss interpolation width	％
μ	Coefficient of friction	-
			R	Gloss level of roll	μm
LTy	Lubricant medium	-
			L	Mill length	km
v	Roll line speed	m/s
			μ B	The output of coefficient of friction analytic modell analytical model	-
μ C	The output of coefficient of friction statistical model	-
			λ	Coefficient of friction statistical model Gauss interpolation width	m/s
P	The rolling force setup value	kN
			T	The torque setting value	kNm
N	Set value of the power	kW
			S	Fixed value of roller slit	mm
H *	Inlet of rolling mill side thickness measurements	mm
			h *	Milling train outlet side thickness measurements	mm
P *	The measurement of rolling force value	kN
			N *	The rolling power measured value	kW
v *	The roll line velocity measurement	m/s
			Kf *	The resistance of deformation actual value	Mpa
μ *	The coefficient of friction actual value	-

The specific embodiment

A kind of thickness Learning Control Method for spiral-plate precision cold-rolled thickness of slab measuring system, this control method are divided into thickness setting and calculate and thickness self study calculating two parts calculation content; Spiral-plate is carried out thickness setting when calculating in its single rolling process, at first distribute each passage thickness, determine that namely each passage entrance side is set thickness and outlet side is set thickness, for example rolling for the TA1 spiral-plate, a rolling process thickness rolls 1.5mm by 3.5mm, the wide 1090mm of plate, by 8 rolling finishing of passage, passage thickness distributes as shown in table 1.

(3.5～1.5 * 1090mm) rolling pass thickness distribute table 1TA1 spiral-plate

The road number of times	1	2	3	4	5	6	7	8
									Inlet thickness (mm)	3.5	2.859	2.425	2.134	1.936	1.776	1.66	1.58
Exit thickness (mm)	2.859	2.425	2.134	1.936	1.776	1.66	1.58	1.51

Then in accordance with the following steps since the 1st passage, carry out every time thickness setting and calculate:

(1) reads the rolling inlet of rolling mill side of spiral-plate and set thickness H, milling train outlet side setting thickness h, spiral-plate width W and spiral-plate roll line speed v;

(2) calculate rolling this passage resistance of deformation calculated value of spiral-plate Kf according to corresponding duty parameter by deformation resistance model, corresponding duty parameter comprises the accumulated deformation rate ε since the 1st passage _a(its computation model is ε _a=0.4 (H-h _I-1)/H+0.6 (H-h _i)/H, i is shelf number or passage number), chemical composition Che (mass percent that refers to each constitutional chemistry element of spiral-plate), chemical composition Che, the hot rolling finishing temperature FTe as front operation, coiling temperature CTe, annealing temperature ATe, annealing time ATm, deformation resistance model is the model that the resistance of deformation analytic modell analytical model combines with the resistance of deformation statistical model, according to accumulated deformation rate ε _aObtain its output Kf by the resistance of deformation analytic modell analytical model _B, the resistance of deformation analytic modell analytical model is Kf _B=f ₁(ε _a);

Wherein, the detailed expression formula of resistance of deformation analytic modell analytical model is:

Kf _B＝b ₀+b ₁ε _a+b ₂ε _a ²+b ₃ε _a ³

In the formula, b ₀, b ₁, b ₂, b ₃Be constant coefficient, be made as respectively 446,9.5 ,-0.07,0.00069 for TA1 trade mark titanium alloy (also being industrially pure titanium), be made as respectively 901,18 ,-0.27,0.0017 for TC4 trade mark titanium alloy;

The resistance of deformation statistical model is different according to the accumulated deformation rate, difference corresponding accumulated deformation rate 0%, 20%, 40%, 60%, 80%, 100%, formed by 6 SVMs, each SVMs be input as chemical composition Che, hot rolling finishing temperature FTe, coiling temperature CTe, annealing temperature Ate and annealing time ATm, obtain each SVMs output Kf according to input _Ci(i=1,2 ..., 6), 6 SVMs output Kf _CiBy accumulated deformation rate ε _aCarry out Gauss interpolation and obtain total statistical model output Kf _C, the Gauss interpolation computation model is:

$K{f}_C=\underset{i=1}{\overset{6}{\mathrm{\Σ}}}K{f}_{\mathrm{Ci}}\·g{k}_i\left({\mathrm{\ϵ}}_a\right)$

${\mathrm{gk}}_i\left({\mathrm{\ϵ}}_a\right)=\frac{e^{-\frac{{({\mathrm{\ϵ}}_a-{\mathrm{\ϵ}}_{\mathrm{ai}})}^2}{{\mathrm{\σ}}^2}}}{\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{e}^{-\frac{{({\mathrm{\ϵ}}_a-{\mathrm{\ϵ}}_{\mathrm{ai}})}^2}{{\mathrm{\σ}}^2}}}$

In the formula, ε _AiBe 6 the corresponding standard accumulated deformation of SVMs rates, be respectively 0%, 20%, 40%, 60%, 80%, 100%, σ is resistance of deformation statistical model Gauss interpolation width, is made as 14%, gk _i(ε _a) be accumulated deformation rate interpolation weights;

With resistance of deformation analytic modell analytical model output Kf _BWith resistance of deformation statistical model output Kf _CCarry out product, just obtain rolling this passage resistance of deformation calculated value of spiral-plate Kf, that is:

Kf＝Kf _B·Kf _C

(3) calculate rolling this passage calculation of friction coefficient value of spiral-plate μ according to corresponding duty parameter by the coefficient of friction model, corresponding duty parameter comprises gloss level of roll R, lubricant medium LTy, mill length L, spiral-plate roll line speed v, the coefficient of friction model is the model that the coefficient of friction analytic modell analytical model combines with the coefficient of friction statistical model, obtains its output μ according to gloss level of roll R, lubricant medium LTy, mill length L, spiral-plate roll line speed v by the coefficient of friction analytic modell analytical model _B, the coefficient of friction analytic modell analytical model is μ _B=f ₂(R, LTy, L, v);

Wherein, the detailed expression formula of coefficient of friction analytic modell analytical model is:

${\mathrm{\μ}}_B=({\mathrm{\μ}}_0\left(\mathrm{LTy}\right)+d{\mathrm{\μ}}_v\left(\mathrm{LTy}\right)\·e^{-\frac{v}{v_0}})\·(1+c_R\·(R-R_0))\·(1+\frac{c_W}{1+L/L_0})$

μ in the formula ₀(LTy)---by the coefficient of friction a reference value of lubricant medium decision;

D μ _v(LTy)---the friction coefficient velocity variations that is determined by lubricant medium concerns;

v ₀---reference speed, m/s;

R ₀---roughness a reference value, μ m;

c _R---roughness value, μ m ^-1

c _W---the mill length coefficient;

L ₀---basic mill length, km;

The coefficient of friction statistical model is different according to spiral-plate roll line speed, the corresponding roll line speed 0m/s of difference, 1m/s, 2m/s, 3m/s, 4m/s, 5m/s, formed by 6 SVMs, each SVMs is input as gloss level of roll R, roll material RTy, lubricant medium LTy, mill length L, obtains each SVMs output μ according to input _Ci(i=1,2 ..., 6), 6 SVMs output μ _CiCarry out Gauss interpolation by spiral-plate roll line speed v and obtain total statistical model output μ _C, the Gauss interpolation computation model is:

${\mathrm{\μ}}_C=\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{\mathrm{\μ}}_{\mathrm{Ci}}\·g{\mathrm{\μ}}_i\left(v\right)$

$g{\mathrm{\μ}}_i\left(v\right)=\frac{e^{-\frac{{(v-v_i)}^2}{{\mathrm{\λ}}^2}}}{\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{e}^{-\frac{{(v-v_i)}^2}{{\mathrm{\λ}}^2}}}$

In the formula, v _iBe 6 corresponding standard speeds of SVMs, be respectively 0m/s, 1m/s, 2m/s, 3m/s, 4m/s, 5m/s, λ is coefficient of friction statistical model Gauss interpolation width, is made as 0.7m/s, g μ _i(v) be the speed interpolation weights;

With coefficient of friction analytic modell analytical model output μ _BWith coefficient of friction statistical model output μ _CCarry out product, just obtain rolling this passage calculation of friction coefficient value of spiral-plate μ, that is:

μ＝μ _B·μ _C

(4) on the basis that obtains spiral-plate rolling this passage resistance of deformation calculated value Kf and calculation of friction coefficient value μ, read current rolling duty parameter and milling equipment parameter, respectively by the roll-force computation model, the Calculating Torque during Rotary model, the power calculation model calculates rolling force setup value P, torque setting value T and set value of the power N, and further obtain fixed value of roller slit S by the roll gap computation model, rolling duty parameter is that the inlet of rolling mill side is set thickness H, the milling train outlet side is set thickness h, spiral-plate width W and spiral-plate roll line speed v, the milling equipment parameter comprises work roll diameter D _wWith milling train longitudinal rigidity C _P, the roll-force computation model is P=f ₃(H, h, W, Kf, μ, D _w), the roll torque computation model is T=f ₄(H, h, P, D _w), the power calculation model is N=f ₅(T, H, h, D _w, μ, v), gap Set Model is S=f ₆(P, h, C _p);

Wherein the detailed expression formula of roll-force computation model is:

$P=W\·\mathrm{Kf}\·\sqrt{{D_w}^{\′}\·(H-h)}\·Q_P$

W in the formula---rolled piece width, mm;

Q _P---the stress state coefficient;

D _w' be the roll flattening radius, mm, its computing formula is:

D _w′＝D _w(1+A ₀·P/(W·(H-h)))

A in the formula ₀---the roll flattening coefficient;

Stress state coefficient Q _PComputation model is as follows:

$Q_P=e_0+e_1\mathrm{\ϵ\μ}\sqrt{{D_w}^{\′}/h}+e_2\mathrm{\ϵ}$

ε in the formula---be this pass deformation rate, its computation model is:

$\mathrm{\ϵ}=\frac{H-h}{H}$

e ₀, e ₁, e ₂---constant coefficient, Hill formula commonly used is taken as respectively 1.08,1.79 ,-1.02 at present;

The detailed analytical expression of roll torque computation model is:

$T=2P\sqrt{{D_w}^{\′}\·(H-h)}{\mathrm{\ξ}}_M/1000$

ξ in the formula _M---arm of force coefficient;

The detailed expression formula of power calculation model is:

$N=\frac{2\mathrm{\π}\×{10}^3}{60\×102}M\·n=0.103M\·n$

In the formula

M---be applied to the total torque on the motor shaft, kN.m, its computation model is:

$M=\frac{T}{r}+M_f+M_x\±M_d$

T in the formula---roll torque, kN.m;

M _f---additional friction moment, kN.m;

M _x---idling torque, kN.m;

M _d---kinetic moment, kN.m;

R---main transmission speed ratio;

N---motor rotational shaft speed, rpm, its computation model is:

n＝v/(1+Sl)/(π·D _w)

Wherein, Sl is advancing slip amount, and its computation model is:

$\mathrm{Sl}=\frac{D_w}{h}{{\mathrm{\γ}}_M}^2$

γ in the formula _M---be neutral angle, its computation model is:

${\mathrm{\γ}}_M=\frac{H-h}{2h}{(1-\frac{1}{2\mathrm{\μ}}\sqrt{\frac{H-h}{D_w}})}^2$

Take six-high cluster mill as example, the detailed expression formula of gap Set Model is:

S＝h-(P-P ₀)/C _P-(F _W-F _W0)/C _FW-(F _M-F _M0)/C _FM-G

H in the formula---exit thickness, mm;

P ₀---acyclic homologically trioial roll-force, kN;

C _P---roll-force longitudinal rigidity, kN/mm;

F _W, F _M---set working roll, intermediate calender rolls bending roller force, kN;

F _W0, F _M0---the working roll during acyclic homologically trioial, intermediate calender rolls bending roller force, kN;

C _FW, C _FM---working roll, intermediate calender rolls bending roller force longitudinal rigidity, kN/mm;

G---the roll gap zero-bit variable quantity that the factors such as thermal expansion of rollers, wearing and tearing cause, mm;

If for four-high mill, this model only needs the 3rd is removed about the part of intermediate calender rolls roller;

So far, obtained the fixed value of roller slit of rolling this passage of spiral-plate, the thickness setting of having finished this passage calculates, and comes back to step (1) until extreme trace;

After finishing a rolling process of a spiral-plate rolling, carry out in accordance with the following steps the thickness self study and calculate:

(1) reads by each measured passage inlet of rolling mill side thickness measurements H of the arbitrary described spiral-plate precision cold-rolled thickness of slab measuring system of claim 1～6 ^*, milling train outlet side thickness measurements h ^*

(2) read the measurement of rolling force value P of each passage in the spiral-plate operation of rolling ^*, rolling power measured value N ^*And spiral-plate roll line velocity measurement v ^*

(3) according to every time inlet of rolling mill side thickness measurements H ^*, milling train outlet side thickness measurements h ^*, measurement of rolling force value P ^*, power measurement values N ^*, spiral-plate roll line velocity measurement v ^*And work roll diameter D _w, set up equation group by roll-force computation model, roll torque computation model and power calculation model, calculate every time resistance of deformation actual value Kf ^*With coefficient of friction actual value μ ^*

(4) calculate every time resistance of deformation actual value Kf ^*With the deviation ek of resistance of deformation calculated value Kf, computation model is ek=Kf-Kf ^*, calculate each resistance of deformation SVMs output bias ek according to the Gauss interpolation computation model on this basis _i(i=1,2 ..., 6), computation model is:

$e{k}_i=\∂\mathrm{ek}/\∂g{k}_i\left({\mathrm{\ϵ}}_a\right)$

Then obtain the new output Kfc of each SVMs _i ^*=Kfc _i-ek _iEach rolling pass, 6 new SVMs are exported with corresponding duty parameter chemical composition Che, hot rolling finishing temperature FTe, coiling temperature CTe, annealing temperature Ate and annealing time ATm and are formed one group of sample data, sample data by a plurality of passages in the single rolling process, be used for the weights of 6 SVMs are trained, thus the self study of implementation model resistance of deformation;

(5) calculate every time coefficient of friction actual value μ ^*With the deviation eu of calculation of friction coefficient value μ, computation model is eu=μ-μ ^*, calculate each resistance of deformation SVMs output bias eu according to the Gauss interpolation computation model on this basis _i(i=1,2 ..., 6), computation model is:

${\mathrm{eu}}_i=\∂\mathrm{eu}/\∂g{u}_i\left(v\right)$

Then obtain the new output μ of each SVMs _Ci ^*=μ _Ci-eu _iEach rolling pass, 6 new SVMs are exported and corresponding one group of sample data of the L shaped one-tenth of duty parameter gloss level of roll R, lubricant medium LTy, mill length, sample data by a plurality of passages in the single rolling process, be used for the weights of 6 SVMs are trained, thus the self study of implementation model coefficient of friction;

So far, by the self study to resistance of deformation and coefficient of friction, realized the self study to the thickness setting model.

Claims (4)

1. the method for controlling thickness of a spiral-plate precision cold-rolled thickness of slab measuring system, because spiral-plate length is short, rolling time is few, be installed in the thickness value of feedback of the calibrator acquisition of spiral-plate reversable mill front and back, can't be used for carrying out closed loop thickness control, the thickness of slab computation model that can only be used for Process Control System carries out Model Self-Learning, material deformation drag and actual condition coefficient of friction are revised, by improving the control accuracy of thickness of slab computation model, fundamentally improved thickness control accuracy; It is characterized in that this control method is divided into thickness setting and calculates and thickness self study calculating two parts content, detailed content is as follows:

The thickness setting calculating section:

Spiral-plate is carried out thickness setting when calculating in its single rolling process, at first carrying out each passage thickness distributes, determine that namely each passage entrance side is set thickness and outlet side is set thickness, then in accordance with the following steps since the 1st passage, carry out every time thickness setting and calculate:

(1) reads the rolling inlet of rolling mill side of spiral-plate and set thickness H, milling train outlet side setting thickness h, spiral-plate width W and spiral-plate roll line speed v;

(2) calculate rolling this passage resistance of deformation calculated value of spiral-plate Kf according to corresponding duty parameter by deformation resistance model, corresponding duty parameter comprises the accumulated deformation rate ε since the 1st passage _a, chemical composition Che, the hot rolling finishing temperature FTe as front operation, coiling temperature CTe, annealing temperature ATe, annealing time ATm, deformation resistance model is the model that the resistance of deformation analytic modell analytical model combines with the resistance of deformation statistical model, according to accumulated deformation rate ε _aObtain its output Kf by the resistance of deformation analytic modell analytical model _B, the resistance of deformation analytic modell analytical model is Kf _B=f ₁(ε _a);

The resistance of deformation statistical model is different according to the accumulated deformation rate, corresponding accumulated deformation rate is 0%, 20%, 40%, 60%, 80%, 100% respectively, formed by 6 SVMs, each SVMs be input as chemical composition Che, hot rolling finishing temperature FTe, coiling temperature CTe, annealing temperature Ate and annealing time ATm, obtain each SVMs output Kf according to input _Ci(i=1,2 ..., 6), 6 SVMs output Kf _CiBy accumulated deformation rate ε _aCarry out Gauss interpolation and obtain total statistical model output Kf _C, the Gauss interpolation computation model is:

${\mathrm{Kf}}_C=\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{\mathrm{Kf}}_{\mathrm{Ci}}\·{\mathrm{gk}}_i\left({\mathrm{\ϵ}}_a\right)$

${\mathrm{gk}}_i\left({\mathrm{\ϵ}}_a\right)=\frac{e^{-\frac{{({\mathrm{\ϵ}}_a-{\mathrm{\ϵ}}_{\mathrm{ai}})}^2}{{\mathrm{\σ}}^2}}}{\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{e}^{-\frac{{({\mathrm{\ϵ}}_a-{\mathrm{\ϵ}}_{\mathrm{ai}})}^2}{{\mathrm{\σ}}^2}}}$

In the formula, ε _AiBe 6 the corresponding standard accumulated deformation of SVMs rates, be respectively 0%, 20%, 40%, 60%, 80%, 100%, σ is resistance of deformation statistical model Gauss interpolation width, is made as 14%, gk _i(ε _a) be accumulated deformation rate interpolation weights;

With resistance of deformation analytic modell analytical model output Kf _BWith resistance of deformation statistical model output Kf _CCarry out product, just obtain rolling this passage resistance of deformation calculated value of spiral-plate Kf, that is:

Kf=Kf _B·Kf _C

(3) calculate rolling this passage calculation of friction coefficient value of spiral-plate μ according to corresponding duty parameter by the coefficient of friction model, corresponding duty parameter comprises gloss level of roll R, lubricant medium LTy, mill length L, spiral-plate roll line speed v, the coefficient of friction model is the model that the coefficient of friction analytic modell analytical model combines with the coefficient of friction statistical model, obtains its output μ according to gloss level of roll R, lubricant medium LTy, mill length L, spiral-plate roll line speed v by the coefficient of friction analytic modell analytical model _B, the coefficient of friction analytic modell analytical model is μ _B=f ₂(R, LTy, L, v); The coefficient of friction statistical model is different according to spiral-plate roll line speed, corresponding roll line speed 0m/s, 1m/s, 2m/s, 3m/s, 4m/s, 5m/s are comprised of 6 SVMs respectively, each SVMs is input as gloss level of roll R, lubricant medium LTy, mill length L, obtains each SVMs output μ according to input _Ci(i=1,2 ..., 6), 6 SVMs output μ _CiCarry out Gauss interpolation by spiral-plate roll line speed v and obtain total statistical model output μ _C, the Gauss interpolation computation model is:

In the formula, v _iBe 6 corresponding standard speeds of SVMs, be respectively 0m/s, 1m/s, 2m/s, 3m/s, 4m/s, 5m/s, λ is coefficient of friction statistical model Gauss interpolation width, is made as 0.7m/s, g μ _i(v) be the speed interpolation weights;

With coefficient of friction analytic modell analytical model output μ _BWith coefficient of friction statistical model output μ _CCarry out product, just obtain rolling this passage calculation of friction coefficient value of spiral-plate μ, that is:

μ=μ _B·μ _C

(4) on the basis that obtains spiral-plate rolling this passage resistance of deformation calculated value Kf and calculation of friction coefficient value μ, read current rolling duty parameter and milling equipment parameter, respectively by the roll-force computation model, the Calculating Torque during Rotary model, the power calculation model calculates rolling force setup value P, torque setting value T and set value of the power N, and further obtain fixed value of roller slit S by the roll gap computation model, rolling duty parameter is that the inlet of rolling mill side is set thickness H, the milling train outlet side is set thickness h, spiral-plate width W and spiral-plate roll line speed v, the milling equipment parameter comprises work roll diameter D _wWith milling train longitudinal rigidity C _P, the roll-force computation model is P=f ₃(H, h, W, Kf, μ, D _w), the roll torque computation model is T=f ₄(H, h, P, D _w), the power calculation model is N=f ₅(T, H, h, D _w, μ, v), gap Set Model is S=f ₆(P, h, C _p);

So far, obtained the fixed value of roller slit of rolling this passage of spiral-plate, the thickness setting of having finished this passage calculates, and comes back to step (1) until extreme trace;

Thickness self study calculating section:

After finishing the single rolling process of spiral-plate rolling, carry out in accordance with the following steps the thickness self study and calculate:

(1) reads each measured passage inlet of rolling mill side thickness measurements H of spiral-plate precision cold-rolled thickness of slab measuring system ^*, milling train outlet side thickness measurements h ^*

(2) read the measurement of rolling force value P of each passage in the spiral-plate operation of rolling ^*, rolling power measured value N ^*And spiral-plate roll line velocity measurement v ^*

(3) according to every time inlet of rolling mill side thickness measurements H ^*, milling train outlet side thickness measurements h ^*, measurement of rolling force value P ^*, power measurement values N ^*, spiral-plate roll line velocity measurement v ^*And work roll diameter D _w, set up equation group by roll-force computation model, roll torque computation model and power calculation model, calculate every time resistance of deformation actual value Kf ^*With coefficient of friction actual value μ ^*

(4) calculate every time resistance of deformation actual value Kf ^*With the deviation ek of resistance of deformation calculated value Kf, computation model is ek=Kf-Kf ^*, calculate each resistance of deformation SVMs output bias ek according to the Gauss interpolation computation model on this basis _i(i=1,2 ..., 6), computation model is:

${\mathrm{ek}}_i=\∂\mathrm{ek}/{\∂\mathrm{gk}}_i\left({\mathrm{\ϵ}}_a\right)$

Then obtain the new output Kf of each SVMs _Ci ^*=Kf _Ci-ek _i(i=1,2,6), each rolling pass, the new output of 6 SVMs forms one group of sample data with corresponding duty parameter chemical composition Che, hot rolling finishing temperature FTe, coiling temperature CTe, annealing temperature Ate and annealing time ATm, by the sample data of a plurality of passages in the single rolling process, be used for the weights of 6 SVMs are trained, thus the self study of implementation model resistance of deformation;

(5) calculate every time coefficient of friction actual value μ ^*With the deviation eu of calculation of friction coefficient value μ, computation model is eu=μ-μ ^*, calculate each resistance of deformation SVMs output bias eu according to the Gauss interpolation computation model on this basis _i(i=1,2 ..., 6), computation model is:

${\mathrm{eu}}_i=\∂\mathrm{eu}/{\∂\mathrm{gu}}_i\left(v\right)$

Then obtain the new output μ of each SVMs _Ci ^*=μ _Ci-eu _iEach rolling pass, the new output of 6 SVMs and corresponding one group of sample data of the L shaped one-tenth of duty parameter gloss level of roll R, lubricant medium LTy, mill length, sample data by a plurality of passages in the single rolling process, be used for the weights of 6 SVMs are trained, thus the self study of implementation model coefficient of friction;

So far, by the self study to resistance of deformation and coefficient of friction, realized the self study to the thickness setting model.

2. the method for controlling thickness of spiral-plate precision cold-rolled thickness of slab measuring system according to claim 1, it is characterized in that: described spiral-plate reversable cold-rolling machine is four rollers or six roller coiled sheet dual-purpose type cold-rolling mills, before and after the milling train roller-way and coiling machine are arranged all, possess the rolling and two kinds of modes of production of coil rolling of spiral-plate.

3. the method for controlling thickness of spiral-plate precision cold-rolled thickness of slab measuring system according to claim 1, it is characterized in that: the length maximum of described spiral-plate is less than 10 meters.

4. the method for controlling thickness of spiral-plate precision cold-rolled thickness of slab measuring system according to claim 1, it is characterized in that: described spiral-plate is titanium plate or titanium alloy sheet.

Images