CN105023191A - Online recursive calculation method for fume occurrence rate of metallurgical furnace - Google Patents

Abstract

The present invention provides an online recursive calculation method for fume occurrence rate of a metallurgical furnace. The method comprises: according to independent historical data in an uncertain interval period, constructing a nonlinear regression constraint model; performing nonlinear transformation and normalization processing on sample data; and calculating a coefficient matrix of a recursive model with a least square partial differential method. According to the method, the data is stored by using a fixed-length time sequence queue to facilitate updating of the historical data, and the regression coefficient matrix of the recursive model is calculated with the least square partial differential method, so that online prediction of the fume occurrence rate is implemented and data support is provided for online control and production decision of the metallurgical furnace.

Description

A kind of online recurrence calculation method of metallurgical furnace kiln flue dust incidence

Technical field

The present invention relates to the recurrence calculation method of metallurgical furnace kiln flue dust incidence, belong to metallurgical furnace kiln On-line Control field.

Background technology

What carry out in metallurgical furnace kiln is the complicated physical-chemical reaction entering the storerooms such as stove concentrate, fuel and air under high temperature fused state.For controlling metallurgical furnace kiln output object quality, the online control model of the built vertical metallurgical furnace kiln in domestic and international most of large-scale smeltery, carry out metallurgical furnace kiln metal balance and heat load balance calculating in real time, and using the feedforward setting of result of calculation as online control model and the foundation of feedback modifiers, for wind, oxygen, loop, wet goods basis provide Optimal Decision-making data.Flue dust incidence is the required argument carrying out above-mentioned two large EQUILIBRIUM CALCULATION FOR PROCESS, is generally set to constant.

Be subject to the impact into factors such as stove material, body of heater reaction tower temperature, reactant mixture homogeneity, uptake flue changes, flue dust incidence can change, and is set as that constant carries out the result that metal balance and heat load balance calculate and cannot reflect metallurgical furnace kiln flue dust generation in actual production process.Existing documents and materials are only simply introduced flue dust incidence and computing method thereof, and simply show the static calculation method of flue dust incidence, this computing method deviation is relatively large, can not meet the requirement of existing large-scale metallurgical stove On-line Control parameter computational accuracy.

Summary of the invention

The present invention is directed in existing metallurgical furnace kiln flue dust Incidence calculus and calculate shortcoming low with precision not in time, a kind of method in line computation metallurgical furnace kiln flue dust incidence is provided, the method is according to metallurgical furnace kiln flue dust incidence and enter stove material, oxygen unit consumption, produce restriction relation between the control elements such as the quantity of slag, derivation regression model, adopt the regression coefficient of least square fitting Recursive Solution flue dust incidence recurrence model, recursion goes out metallurgical furnace kiln flue dust incidence, for metallurgical furnace kiln heat balance and mass balance optimizing control models provides Data support.

The present invention is achieved by the following technical solutions:

The online recurrence calculation method of a kind of metallurgical furnace kiln of the present invention flue dust incidence, comprises the steps:

1) selected element affect metallurgical furnace kiln flue dust incidence, according to flue dust incidence data and element data in independently variable interval cycle, the recurrence sample data of structure nonlinear model;

2) set up fixed length sequential queue to flue dust incidence data and element data, the principle of foundation first in first out, to sample data storage update and pre-service, checks fixed length sequential queue S (i)={ (x _ij, y _i) | [x _i1, x _i2..., x _{i (2p)}], y _ii=1 ..., the validity of the data in n, flue dust incidence data and element data bound in setting fixed length sequential queue, reject, if x abnormal out-of-limit data ^* _ij> x _jU∪ x ^* _ij< x _jL∪ y ^* _i> y _u∪ y ^* _i< y _l, then if i>=n in fixed length sequential queue S (i), then regression model carries out linear transformation and obtains flue dust incidence recurrence model;

3) the sample data element stored in fixed length sequential queue is normalized;

4) adopt least square differential partial differentiation to solve the regression coefficient matrix of recurrence model, obtain flue dust incidence recursion formula;

5) production run element data is inputted described recursion formula, calculate flue dust incidence;

6) when Data Update in fixed length sequential queue, 1 is carried out) ~ 4) calculation procedure;

Wherein:

S (i) is fixed length sequential queue;

(x _ij, y _i) be fixed length sequential queue element;

X _ijit is independent variable;

Y _iit is dependent variable;

I is the position sequence of fixed length sequential queue element, and n is fixed length sequential queue length, i=1 ..., n;

J is fixed length sequential queue independent variable element position;

P is fixed length sequential queue independent variable number of elements;

X _jUit is the maximal value of a jth independent variable in fixed length sequential queue;

X _jLit is the minimum value of a jth independent variable in fixed length sequential queue;

it is pre-service independent variable in fixed length sequential queue;

Y _uit is the maximal value of dependent variable in fixed length sequential queue;

Y _lit is the minimum value of dependent variable in fixed length sequential queue;

Y ^*it is pre-service dependent variable in fixed length sequential queue;

it is element to be deleted in fixed length sequential queue.

The sample data stored in described fixed length sequential queue is from the data of production leadtime value and metallurgical furnace kiln Distributed Control System (DCS) (DCS control system) Real-time Collection.

Described recurrence calculation method, adopt the two computation model of redundancy backup, when recurrence model carries out matrix of coefficients calculating, backup model puts into operation and calculates flue dust incidence, after completing the calculating of recurrence model matrix of coefficients, recurrence model matrix of coefficients writes backup model automatically.

Computation process is as follows:

Under metallurgical furnace kiln high temperature fused state, flue dust incidence is subject to the impact into multiple independent variable elements such as stove concentrate amount, oxygen utilization, assuming that its nonlinear regression model (NLRM) is expressed as

$y=a_0+a_1{x}_1+...+a_p{x}_p+a_{p+1}{x}_1^2+a_{p+2}{x}_2^2+...+a_{2p}{x}_p^2---\left(1\right)$

Y is regression model dependent variable value;

X ₁, x ₂..., x _pit is regression model argument value;

Order then non-linear regression expression formula (1) can be converted into following linear regression model (LRM) expression formula

y＝a ₀+a ₁x ₁+a ₂x ₂+…+a _px _p+a _p+1x _p+1+a _p+2x _p+2+…+a _2px _2p(2)

If raw sample data is

$X={\left[\begin{array}{cccc}{x}_{11}& x_{12}& \mathrm{...}& x_{1p}\\ x_{21}& x_{22}& \mathrm{...}& x_{2p}\\ & \mathrm{...}& \mathrm{...}& \\ x_{n1}& x_{n2}& \mathrm{...}& x_{np}\end{array}\right]}_{n\&times;p}---\left(3\right)$

To x _ij, i=1 ..., p; J=1 ..., n.Then Augmented Data matrix is such as formula (4)

$\stackrel{\&OverBar;}{X}={\left[\begin{array}{cccccccc}{x}_{11}& x_{12}& \mathrm{...}& x_{1p}& {\left(x_{11}\right)}^2& {\left(x_{12}\right)}^2& \mathrm{...}& {\left(x_{1p}\right)}^2\\ x_{21}& x_{22}& \mathrm{...}& x_{2p}& {\left(x_{12}\right)}^2& {\left(x_{22}\right)}^2& \mathrm{...}& {\left(x_{2p}\right)}^2\\ & \mathrm{...}& \mathrm{...}& & \mathrm{...}& \mathrm{...}& \mathrm{...}& {\left(x_{3p}\right)}^2\\ x_{n1}& x_{2n}& \mathrm{...}& x_{np}& {\left(x_{n1}\right)}^2& {\left(x_{n2}\right)}^2& \mathrm{...}& {\left(x_{np}\right)}^2\end{array}\right]}_{n\&times;2p}---\left(4\right)$

Wherein first sample data is normalized, namely

$\begin{array}{cc}if& j\&le;p,x_{ij}^{\&prime;}=\frac{x_{ij}-x_{i\min }}{x_{i\max }-x_{i\min }},\end{array}i=1,...,n;j=1,...,p$

$\begin{array}{cc}if& j>p,x_{ij}^{\&prime;}\end{array}=\frac{{\left(x_{ij}\right)}^2-{\left(x_{i\min }\right)}^2}{{\left(x_{i\max }\right)}^2-{\left(x_{i\min }\right)}^2},i=1,...,n;j=p+1,...,2p$

Y＝[y ₁y ₂… y _n] ^T

X is the historical data matrix be made up of fixed length sequential queue;

it is the augmented matrix be made up of formula (3) historical data matrix and formula (1);

it is the transformation matrix after augmented matrix normalization;

P is influence factor;

N is historical data sequence;

X _{i max}it is the maximal value of the i-th independent variable factor in fixed length sequential queue;

X _{i min}it is the minimum value of the i-th independent variable factor in fixed length sequential queue;

X ' _iji-th historical record in fixed length sequential queue, the normalized value of a jth independent variable;

Y is dependent variable in fixed length sequential queue.

By from historical data observation station (i=1,2 ..., n; J=1,2 ..., 2p) and carry out the linear model of matching represented by above formula (2)

y _i＝a ₀+a ₁x _i1+a ₂x _i2+…+a _px _ip+a _p+1x _i(p+1)+a _p+2x _i(p+2)+…+a _2px _i(2p)+ε

(6)

Y _ireturn dependent variable value in i-th historgraphic data recording in fixed length sequential queue sample data; x _ij, j=1 ..., 2p is regressor value in the i-th historgraphic data recording in fixed length sequential queue sample data;

ε is removing 2p independent variable x _ijafterwards to y _ithe stochastic error of impact.

To from production process data sequence set x _m=(x ₁, x ₂..., x _p), m=1 ..., p, independent variable expands conversion can obtain matching argument data combined sequence i.e. x _m=(x ₁, x ₂..., x _p, x _p+1, x _p+2..., x _2p), m=1,2 ..., 2p, wherein meets formula (7) for part independent variable

$x_{p+k}=x_p^2,k=1,2,...,p---\left(7\right)$

For production process data sequence set x _m=(x ₁, x ₂..., x _p), m=1 ..., the flue dust incidence corresponding to p is such as formula shown in (8).

y＝a ₀+a ₁x ₁+a ₂x ₂+…+a _px _p+a _p+1x _p+1+a _p+2x _p+2+…+a _2px _2p(8)

X ₁..., x _pit is production process data sequence;

X _p+1..., x _2pby x ₁..., x _pthe variable of deriving;

A ₀, a ₁..., a _p, a _p+1, a _p+2..., a _2ptreat regression model coefficient;

Y is the mathematical model regressand value of flue dust incidence.

According to the data that formula (4) provides, fit mathematics model can be expressed as

$Y=\stackrel{\&OverBar;}{X}A---\left(9\right)$

Y is historical data, a dimensional vector, Y _{n × 1}=[y _1,y ₂..., y _n] ^t

A is fitting coefficient, one dimension row vector, A _{1 × 2p}=[a ₀, a ₁..., a _p, a _p+1, a _p+2..., a _2p]

By independent observation data acquisition least square determination fitting coefficient a in the production procedure shown in formula (5) ₀, a ₁..., a _p, a _p+1, a _p+2..., a _2p, namely in guarantee

$\min Q(a_0,a_1,...,a_p,a_{p+1},a_{p+2},...,a_{2p})=\underset{i=0}{\overset{n}{\&Sigma;}}{\&lsqb;y_i-(a_0+a_1{x}_1+...+a_{2p}{x}_{2p})\&rsqb;}^2---\left(10\right)$

Respectively a is asked to formula (10) ₀, a ₁..., a _p, a _p+1, a _p+2..., a _2pthe local derviation of coefficient, can obtain sample coefficient estimated value

$\begin{array}{c}-2\underset{i=1}{\overset{n}{\&Sigma;}}\&lsqb;y_i-(a_0+a_1{x}_1+\mathrm{...}+a_{2p}{x}_{2p})\&rsqb;=0\\ -2\underset{i=1}{\overset{n}{\&Sigma;}}\&lsqb;y_i-(a_0+a_1{x}_1+\mathrm{...}+a_{2p}{x}_{2p})\&rsqb;x_1=0\\ -2\underset{i=1}{\overset{n}{\&Sigma;}}\&lsqb;y_i-(a_0+a_1{x}_1+\mathrm{...}+a_{2p}{x}_{2p})\&rsqb;x_2=0\\ .\\ .\\ .\\ -2\underset{i=1}{\overset{n}{\&Sigma;}}\&lsqb;y_i-(a_0+a_1{x}_1+\mathrm{...}+a_{2p}{x}_{2p})\&rsqb;x_p=0\\ -2\underset{i=1}{\overset{n}{\&Sigma;}}\&lsqb;y_i-(a_0+a_1{x}_1+\mathrm{...}+a_{2p}{x}_{2p})\&rsqb;x_{p+1}=0\\ .\\ .\\ .\\ -2\underset{i=1}{\overset{n}{\&Sigma;}}\&lsqb;y_i-(a_0+a_1{x}_1+\mathrm{...}+a_{2p}{x}_{2p})\&rsqb;x_{2p}=0\end{array}---\left(11\right)$

Formula (11) can obtain after arranging

According to the regression estimates value that the partially micro-differentiate of formula (12) draws be

From formula (13), the estimated value of flue dust incidence is

$\hat{y}={\hat{a}}_0+{\hat{a}}_1{x}_1+{\hat{a}}_2{x}_2+...+{\hat{a}}_p{x}_p+{\hat{a}}_{p+1}{x}_{p+1}+{\hat{a}}_{p+2}{x}_{p+2}+...+{\hat{a}}_{2p}{x}_{2p}---\left(14\right)$

that recursion formula exports dependent variable predicted value;

it is the recurrence model coefficient drawn by fixed length sequential queue sample data.

The online recurrence calculation method of flue dust incidence of the present invention builds non-linear regression restricted model according to historical data in the variable interval cycle, and nonlinear transformation and normalized are carried out to sample data, adopt least square differential partial differentiation to solve the matrix of coefficients of recurrence model.Regression model adopts the two computation model of redundancy backup, when recurrence model carries out flue dust incidence recurrence calculation, calculates backup model coefficient matrix.After completing recurrence model recurrence calculation, backup model coefficient writes backup model automatically.

Tool of the present invention has the following advantages:

1. the present invention uses fixed length sequential queue storing sample data, historical data is facilitated to upgrade, adopt least square differential partial differentiation online Recursive Solution recurrence model coefficient, realize the on-line prediction of flue dust incidence, for the vehicle air-conditioning of metallurgical furnace kiln and production decision provide Data support.

2. in the present invention, flue dust incidence forecast model adopts redundancy backup pattern, and a road calculates for the recurrence of flue dust incidence, and another road is used for the real-time update of recurrence model.This redundancy backup computation schema improves the safe reliability of recurrence model.

3. field conduct of the present invention is convenient and simple, easily realizes in existing control system, is conducive to large-scale promotion application.

Accompanying drawing explanation

Fig. 1 is embodiment recursion result figure;

Fig. 2 is the two computation model structural drawing of redundancy backup in the present invention;

Embodiment

Below in conjunction with embodiment, the present invention is described in further detail:

Flash smelting furnace for copper segregation is example, the flue dust that fusion process produces mainly with enter stove mixed copper concentrate amount, slag concentrate amount, sludging flux silica sand amount, oxygen unit consumption, produce the quantity of slag, output matte amount is relevant.Therefore, flue dust incidence regressing fitting model is totally 12 independents variable, wherein main Variable selection enters stove mixed copper concentrate, slag concentrate, sludging flux silica sand, oxygen unit consumption, produces the quantity of slag, output matte amount totally 6, for improving the precision of model prediction, 6 corresponding augmentation independents variable are drawn through nonlinear transformation by main independent variable.

Recursion step write computer software in the present embodiment, and dock with output material statistical analysis system, DCS control system, metal balance, heat load balance model.Before production scene is implemented, oxygen unit consumption calculates according to output matte amount and oxygen utilization, the output quantity of slag and output matte amount are calculated by production leadtime, the label that above-mentioned data are corresponding with entering stove bulk concentrate amount, slag concentrate amount, sludging flux silica sand amount, flue dust incidence typing control system.Flue dust incidence recurrence model adopts DCS control system configuration interface and script instrument, and creation state display, data inputting window and action button, set up corresponding relation with already present data label respectively.Fixed length sequential queue returns sample data and is stored in the historical data base of DCS control system.

Choosing fixed length sequential queue length n is 25, to pick up the car copper concentrate amount in a copper flash smelting stove production scene DCS control system database, slag concentrate amount, sludging flux silica sand amount, the data of oxygen unit consumption, and the generation quantity of slag in this spacer, output matte amount statistics and the flue dust incidence record 30 groups calculated thus, reject after process through data, obtain 27 groups of valid data records, fixed length sequential queue is pressed into successively by the time order and function order produced, calculate flue dust incidence regression model coefficient, 10 groups of production process data are utilized to predict flue dust incidence respectively, as shown in Figure 1, M1 is the flue dust incidence predicted by the 1st recurrence model, M2 is the flue dust incidence predicted by the 2nd recurrence model, M3 is the flue dust incidence predicted by the 3rd recurrence model, recurrence calculation result and statistical computation result substantially identical, relative error is less than 2%, may be used for metallurgical process metal balance and heat load balance model online rapid calculation.

Concrete computation process is as follows:

1) flue dust incidence data record y is inputted _iand corresponding initial time t _bwith intermission t _e, the historical data of copper concentrate amount, slag concentrate amount, sludging flux silica sand amount, oxygen unit consumption in index control system DCS database, and the generation quantity of slag, the output matte amount statistics in this spacer thus, forms least square fitting sample data record

X _i＝{(x _ij,y _i)|[x _i1,x _i2,x _i3,x _i4,x _i5,x _i6],y _i}。Wherein,

X _i1it is copper concentrate semi-invariant;

X _i2it is slag concentrate semi-invariant;

X _i3it is sludging flux silica sand semi-invariant;

X _i4it is the average consumption of oxygen unit consumption;

X _i5it is output slag semi-invariant;

X _i6it is output matte semi-invariant;

Y _iit is flue dust incidence statistical computation amount;

T _bnew samples data record X _icorresponding initial time;

T _enew samples data record X _ithe corresponding intermission;

I=1 ..., n is that this is recorded in the particular location in fixed length sequential queue.

According to the main independent variable of selected input, determine fit non-linear mathematical model

$y=a_0+a_1{x}_1+...+a_6{x}_6+a_7{x}_1^2+a_8{x}_2^2+...+a_{12}{x}_6^2---\left(15\right)$

A ₀it is regression model constant coefficient;

A ₁it is the regression model coefficient that independent variable copper concentrate is corresponding;

A ₂it is the regression model coefficient that independent variable slag concentrate is corresponding;

A ₃it is the regression model coefficient that independent variable slag making quartz is corresponding;

A ₄regression model coefficient corresponding to independent variable oxygen unit consumption;

A ₅it is the regression model coefficient that independent variable output slag is corresponding;

A ₆it is the regression model coefficient that independent variable output matte is corresponding;

A ₇it is the regression model coefficient that augmentation independent variable copper concentrate is corresponding;

A ₈it is the regression model coefficient that augmentation independent variable slag concentrate is corresponding;

A ₉it is the regression model coefficient that augmentation independent variable slag making quartz is corresponding;

A ₁₀regression model coefficient corresponding to augmentation independent variable oxygen unit consumption;

A ₁₁it is the regression model coefficient that augmentation independent variable output slag is corresponding;

A ₁₂it is the regression model coefficient that augmentation independent variable output matte is corresponding;

2) make then non-linear regression expression formula (15) can

Be converted into as Linear Model with Side expression formula

y＝a ₀+a ₁x ₁+a ₂x ₂+…+a ₆x ₆+a ₇x ₇+a ₈x ₈+…+a ₁₂x ₁₂(16)

After conversion ${\stackrel{\&OverBar;}{X}}_i=\left\{(x_{ij},y_i)\right|\&lsqb;x_{i1},x_{i2},x_{i3},x_{i4},x_{i5},x_{i6},x_{i7},x_{i8},x_{i9},x_{i10},x_{i11},x_{i12}\&rsqb;,y_i\}$ For flue dust incidence fitting data record.I=1 ..., n is that this is recorded in the particular location in fixed length sequential queue.

3) to above-mentioned sample data is normalized by formula (17).

$\begin{array}{cc}if& j\&le;6,x_{ij}^{\&prime;}=\frac{x_{ij}-x_{i\min }}{x_{i\max }-x_{i\min }},i=1,...,n;j=1,...,6\end{array}$

$\begin{array}{cc}if& j>6,x_{ij}^{\&prime;}\end{array}=\frac{{\left(x_{ij}\right)}^2-{\left(x_{i\min }\right)}^2}{{\left(x_{i\max }\right)}^2-{\left(x_{i\min }\right)}^2},i=1,...,n;j=7,...,12---\left(17\right)$

If, k < n, n=k+1, and by the recording pushed fixed length sequential queue of data after process;

If k >=n, then perform

n＝k-1

$\begin{array}{c}{\stackrel{\&OverBar;}{X}}_{i+1}={\stackrel{\&OverBar;}{X}}_i,i=1,2,\mathrm{...},k,k\&le;n\\ {\stackrel{\&OverBar;}{X}}_1=\stackrel{\&OverBar;}{X}\end{array}---\left(18\right)$

N is fixed length sequential queue length;

K is current record number in fixed length sequential queue;

be in fixed length sequential queue i-th, an i+1 record;

wait to be pressed into the new data records in fixed length sequential queue.

4) according to independent variable with dependent variable Y corresponding data fixed length sequential queue, bring formula (19) into and solve fitting coefficient $A_{1\&times;12}=\&lsqb;{\hat{a}}_0,{\hat{a}}_1,\mathrm{...},{\hat{a}}_6,{\hat{a}}_7,{\hat{a}}_8,\mathrm{...},{\hat{a}}_{12}\&rsqb;.$

5) host computer flue dust incidence recurrence model interface input output quantity of slag x is logged in ₅with output matte amount x ₆, select corresponding initial time t _bwith intermission t _e, copper concentrate amount x in index control system DCS database thus ₁, slag concentrate amount x ₂, sludging flux silica sand amount x ₃, oxygen unit consumption x ₄historical data, order be normalized by formula (20).

$\begin{array}{cc}if& j\&le;6,{\hat{x}}_j^{\&prime;}=\frac{x_j-x_{i\min }}{x_{i\max }-x_{i\min }},i=1,...,n;j=1,...,6\end{array}$

$\begin{array}{cc}if& j>6,{\hat{x}}_j^{\&prime;}\end{array}=\frac{{\left(x_j\right)}^2-{\left(x_{i\min }\right)}^2}{{\left(x_{i\max }\right)}^2-{\left(x_{i\min }\right)}^2},i=1,...,n;j=7,...,12---\left(20\right)$

By recursion independent variable j=1,2 ..., 12 bring formula (21) into, draw the recurrence estimation value of flue dust incidence.

Flue dust incidence recurrence estimation value send into metallurgical furnace kiln Optimized model parameter database, carry out metal balance and heat load balance calculating.

The two computation model of redundancy backup is adopted in the present embodiment, as shown in Figure 2, when in fixed length sequential queue, data X changes, then recurrence model coefficient matrices A starts to calculate, and models switching switch β is False; Production process data after normalized pass through recurrence model coefficient matrices A _rthe recurrence carrying out flue dust incidence calculates, and namely backs up model and drops into prediction and calculation.In recurrence model after matrix of coefficients recurrence calculation completes, models switching switch β is True, and recurrence model coefficient matrices A writes backup model A automatically _r.

6) if new flue dust incidence data record y _iinput fixed length sequential queue, repeats step 1 to 4 after sample data upgrades.

Claims (3)

1. the online recurrence calculation method of a kind of metallurgical furnace kiln of the present invention flue dust incidence, comprises the steps:

1) selected element affect metallurgical furnace kiln flue dust incidence, according to flue dust incidence data and element data in independently variable interval cycle, the recurrence sample data of structure nonlinear model;

2) set up fixed length sequential queue to described flue dust incidence data and element data, the principle of foundation first in first out, to sample data storage update and pre-service, checks described fixed length sequential queue S (i)={ (x _ij, y _i) | [x _i1, x _i2..., x _{i (2p)}], y _ii=1 ..., the validity of the data in n, flue dust incidence data and element data bound in setting fixed length sequential queue, reject, if x abnormal out-of-limit data ^* _ij> x _jU∪ x ^* _ij< x _jL∪ y ^* _i> y _u∪ y ^* _i< y _l, then if i>=n in fixed length sequential queue S (i), then regression model carries out linear transformation and obtains flue dust incidence recurrence model;

3) the sample data element stored in described fixed length sequential queue is normalized;

4) adopt least square differential partial differentiation to solve the regression coefficient matrix of described recurrence model, obtain flue dust incidence recursion formula;

5) production run element data is inputted described recursion formula, calculate flue dust incidence;

6) when Data Update in fixed length sequential queue, 1 is carried out) ~ 4) calculation procedure;

Wherein:

S (i) is fixed length sequential queue;

(x _ij, y _i) be fixed length sequential queue element;

X _ijit is independent variable;

Y _iit is dependent variable;

I is the position sequence of fixed length sequential queue element, and n is fixed length sequential queue length, i=1 ..., n;

J is fixed length sequential queue independent variable element position;

P is fixed length sequential queue independent variable number of elements;

X _jUit is the maximal value of a jth independent variable in fixed length sequential queue;

X _jLit is the minimum value of a jth independent variable in fixed length sequential queue;

it is pre-service independent variable in fixed length sequential queue;

Y _uit is the maximal value of dependent variable in fixed length sequential queue;

Y _lit is the minimum value of dependent variable in fixed length sequential queue;

Y ^*it is pre-service dependent variable in fixed length sequential queue;

it is element to be deleted in fixed length sequential queue.

2. the online recurrence calculation method of a kind of metallurgical furnace kiln flue dust incidence as claimed in claim 1, is characterized in that: the sample data stored in described fixed length sequential queue is from the data of production leadtime value and metallurgical furnace kiln DCS control system Real-time Collection.

3. the online recurrence calculation method of a kind of metallurgical furnace kiln flue dust incidence as claimed in claim 1 or 2, it is characterized in that: described recurrence calculation method, adopt the two computation model of redundancy backup, when described recurrence model carries out the calculating of matrix of coefficients, backup model puts into operation and calculates flue dust incidence, after completing the calculating of described recurrence model matrix of coefficients, recurrence model matrix of coefficients writes backup model automatically.