CN101004594A - Data correction method of chemical engineering procedure based on dual thread module - Google Patents

Abstract

A method for calibrating data of chemical industry process based on bilinear model includes calculating and obtaining measurement data of total flow rate and composition content of certain chemical industrial process unit according to relevant chemical balance equation, setting up calibration model of material flow in chemical industrial process unit according to said basic relation and measurement data table, decomposing bilinear problem in calibration operation to be two linear constraints for making derivation be more easier.

Description

Chemical process data correcting method based on bilinear model

Technical field

The present invention relates to the bearing calibration of a kind of industrial process control data, particularly a kind of chemical process data correcting method based on bilinear model belongs to chemical engineering industry process control technology field.

Background technology

Chemical process data unavoidable error in measuring process, measured value can not accurately meet chemical reaction quantitative relation, material balance and heat balance of chemical process or the like.The fundamental purpose of the adjustment of data is for the accuracy that improves measurement data, balance and integrality, is beneficial to the work such as design, simulation and control of process.In the adjustment of data, a class bilinearity restricted problem is arranged, as in product that total flow and concentration are arranged in the material balance equation, the heat balance equation product of quality and heat content etc. being arranged.The criterion of proofreading and correct is: under the condition that satisfies material balance and heat balance, require the quadratic sum minimum of the corrected value deviation of measuring value corresponding with it.In most chemical enterprises, the error ubiquity of measurement data, its reason is many-sided: because also there is more fault error in the logistics instrument in some chemical plant, make the adjustment of data can not obtain abundant " fulcrum "; Set up one and can reflect the technological process actual state, be beneficial to again and carry out relatively difficulty of adjustment of data calculation mathematic model; Variance one covariance (variance-covariance) matrix that needs each measurement data during the adjustment of data is calculated.But after the factory measurement instrument was installed, the variance of stochastic error that can not measurement instrument was checked the statistics performance that instrument does not go to check instrument yet; The theoretical comparative maturity of the adjustment of data of the problem of linear stable state, the enforceable algorithm of computing machine is practical, and non-linear, and dynamic theory problem is still in research, and algorithm is also among exploitation.Wish in the chemical engineering industry process control that the data after the calibrated processing of logistics data must satisfy chemical physics rules such as material balance, heat balance, chemical reaction quantitative relation, these relations are unified to be expressed as: $F(\hat{x},U)=0$ 。Obviously, relational expression above the estimated value that not only will make the corrected value of measurement data and not survey data satisfies, and to make surveyed data on the occasion of with the deviation of measuring value minimum, that is: $\underset{{\hat{x}}_i}{\min }\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{({\hat{x}}_i-x_i)}^2/{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="A20071003678700151.png"\; state="\{\{state\}\}">i</figure-callout>}}^2$ 。How to make data value after the correction more near actual data value, this is a key issue that needs to be resolved hurrily during the chemical engineering industry process data is proofreaied and correct.Find through literature search prior art, using more bilinearity bound data bearing calibration at present still is traditional method, people such as Chen are at Optimal implementation of on-lineoptimization (Computers and Chemical Engineering, 1998,22, pp.S435-S442) a kind of method for on-line optimization based on the Robust Estimation method is proposed in the literary composition, utilize the Robust Estimation method of Lorentzian distribution function structure more effective to the bigger appreciable error of amplitude, its deficiency is: this method does not provide the data correcting method of bilinearity and trilinear system.In the material balance equation of shunt equipment, need to introduce and split the shunting index that variation per minute α represents respective components in the output stream thigh, thereby material balance equation becomes trilinear problem (not being bilinearity).

Summary of the invention

The objective of the invention is at the deficiencies in the prior art, a kind of chemical process data correcting method based on bilinear model is proposed, make it can be under the situation that does not increase operand, according to material balance relationship, automatically correcting measuring value and estimate not measured value, make and survey adjustment of data value and measured value deviation minimum, reach better control effect.

The present invention is achieved by the following technical solutions, the present invention on the basis of existing data correcting method, proposed a kind of with the trilinear equation of constraint be converted into the bilinearity equation of constraint method, bilinearity restricted problem resolved into two linear problems find the solution.Find the solution the bilinear data correction, adopting two-step approach respectively total flow rate data, component concentration to be proofreaied and correct.In logistic data correction modelling to the chemical engineering industry process apparatus, according to the fundamental relation of chemical plant mixer, shunt, separation vessel and reactor four big classes, the topological structure of this device and the measurement data table that provides, set up its adjustment of data model.The user makes roughly data calibration model error condition according to the situation of actual chemical industry industry spot and judging, by from the variance yields of a large amount of sequence real time datas through filtering, denoising, estimation measured value, estimate best correction data, make corrected value more near actual value.The user also can be by improving the technological process of production line, and the balance of the scope of the pattern of the foundation of the structure of flow process, balance, equation of constraint, correction and the degree of depth or the like aspect is held, and obtains the adjustment of data effect that requires.

The present invention includes following steps:

The first step, from the real-time data base of factory, obtain the stream number of share of stock certificate of chemical process, satisfy the relation of material balance and the balance equation of flow process topological structure according to complicated chemical system, calculate the total flow rate and the component concentration measurement data of trying to achieve certain chemical process device or certain several device.Main chemical unit according to actual condition is selected following classification:

The A mixer: two inlet flows 1 and 2 provide output stream 3, X ₃=X ₁+ X ₂

The B shunt: an inlet flow 1 is divided into two or more output streams 2 and 3, $\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="A20071003678700151.png"\; state="\{\{state\}\}">X</figure-callout>}}_2={\mathrm{\&lambda;X}}_1\\ {\mathrm{<figure-callout\; id="3"\; label="X"\; filenames="A20071003678700151.png"\; state="\{\{state\}\}">X</figure-callout>}}_3=(1-\mathrm{\&lambda;}){\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="A20071003678700151.png"\; state="\{\{state\}\}">X</figure-callout>}}_1\end{array}\right.;$

C reactor: inlet flow 1 and output stream 2, X ₂=X ₁+ υ ^Tζ, wherein υ is a chemical reaction stoichiometric number matrix, its element υ _IjRepresent the stoichiometric coefficient of j component in i the reaction, ζ is the transforming degree vector that respectively reacts metering-type, its formula parameter, or variable;

The D separation vessel: an inlet flow 1 and two or more output streams 2 and 3, $\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="A20071003678700151.png"\; state="\{\{state\}\}">X</figure-callout>}}_2={\mathrm{\&lambda;X}}_1\\ {\mathrm{<figure-callout\; id="3"\; label="X"\; filenames="A20071003678700151.png"\; state="\{\{state\}\}">X</figure-callout>}}_3=(1-\mathrm{\&lambda;}){\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="A20071003678700151.png"\; state="\{\{state\}\}">X</figure-callout>}}_1\end{array}\right.$ , wherein parameter lambda is a diagonal matrix, the segregation index of element representation respective components on each diagonal line, and it is a parameter matrix, I is a unit matrix;

In second step,, set up the logistic data correction model according to this chemical process device classification results that the first step obtains:

$\begin{array}{c}\min \mathrm{\&Sigma;}{({\hat{x}}_i-x_i)}^T{Q}^{-1}({\hat{x}}_i-x_i)\\ \mathrm{st}.\mathrm{AX}+\mathrm{BU}+C=0\\ f\left(x_i\right)\&le;0\\ \mathrm{\&phi;}\left(x_i\right)=0\end{array}$ , x wherein _iBe measured value,

Be corrected value, Q is the variance covariance matrix of measuring error, its diagonal entry σ _i ²It is the mean square deviation of its measuring error.

The 3rd step, with the trilinear equation of constraint be converted into the bilinearity equation of constraint method, bilinearity restricted problem resolved into two linear problems find the solution.

The 4th step, according to the adjustment of data model that second step set up measurement data is proofreaied and correct, must satisfy chemical physics rules such as material balance, heat balance, chemical reaction quantitative relation after the calibrated processing of measurement data, these relations are unified to be expressed as: $F(\hat{x},U)=0$ , wherein U does not survey data or waits to estimate the vector that parameter is formed, and F is a functional vector, all physics chemistry relational expression that representative should be satisfied.In the data of having proofreaied and correct, find out objective function, i.e. ε ^T∑ ^-1That value of ε minimum.

The 5th step, according to the 4th objective function that obtain of step, detect the corresponding adjustment of data value of that value of objective function minimum whether within actual range, determine promptly whether following inequality is set up: $x_l\&le;\hat{x}\&le;x_{\mathrm{\&mu;}}$ , x wherein _μ, x _lBe respectively the upper and lower limit of actual measurement data,, then the variable of its correspondence be used as the appreciable error variable if above-mentioned inequality is set up.Otherwise find out objective function time that little value, the variable of its correspondence is regarded as the appreciable error variable.Measured value to the appreciable error variable carries out error compensation, i.e. x%=x%B _Rmm _b, B wherein _RmThe position of expression appreciable error, m _bThe amplitude of expression appreciable error.

The 6th step, the appreciable error variable result according to the 5th step obtained, carry out following judgement: if some variable that exists any one variable and appreciable error to concentrate in the data after proofreading and correct constitutes the loop, then that variable in the data of back is proofreaied and correct in deletion; If have any two variablees and appreciable error to concentrate arbitrary variable to form the loop in the data after proofreading and correct, then one of two variablees in the data of back are proofreaied and correct in deletion; If it is not empty proofreading and correct the back data, then turned back to for second step, circulation is proofreaied and correct again; If it be empty proofreading and correct the back data, then obtain the equal set of the concentrated variable of appreciable error, then obtained the data after all possible correction.

Before system enters the adjustment of data, the error variance σ of the measurement data in the Q matrix (Q is the variance-covariance matrix of measuring error) _i ²Provide in advance, the sample that it can be taken from the precision of instrument or get one group of measurement data estimates that the variance of real time data measured value can be handled by following easy mode:

Control the data correcting method of the bilinear model of field by using the present invention proposition at process industrial, maximum characteristics be exactly guaranteed unbiasedness, the appreciable error variable in the corrected value estimated disallowable, find the solution simple, operand is little, has avoided the calculating of projection matrix.By adding bound constraint, avoided the erroneous judgement of appreciable error to the process measurement variable.When the method for using the present invention to propose, the work that engineering technical personnel will do is exactly: according to the node topology figure of desired data correction, main chemical unit is classified, other work is finished automatically by adjustment of data module: system can calculate correction data automatically, realization is estimated partially to the nothing of data, simultaneously the data that have appreciable error is implemented effectively to reject.By being compiled into corresponding adjustment of data module, existing MATLAB TOOLBOX non-linear have the constraint sub-piece environment under, after above-mentioned model programming (command file and function file), after parameters suitably is provided with, machine makes by loop iteration and surveys adjustment of data value and deviation of measuring value minimum, provide adjustment of data result, whole chemical engineering industry procedures system has just been realized effective control.

Take the chemical engineering industry procedures system of data correcting method of the present invention can be widely used in the production run control of all kinds of enterprises in the industries such as the energy, metallurgy, petrochemical industry, light industry, medicine, building materials, weaving.

Description of drawings

Fig. 1 is the adjustment of data node process flow diagram of certain process units that adopts the chemical process data correcting method of the inventive method.

Embodiment

Below in conjunction with accompanying drawing one embodiment of the invention are elaborated: present embodiment is being to implement under the prerequisite with the technical solution of the present invention; provided detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.

The data correcting method that the present invention is proposed is used for certain certain device of chemical engineering industry process, its objective is to make and surveyed adjustment of data value and deviation of measuring value minimum, promptly, eliminate imbalance and incomplete process data and brought difficulty for activities in production such as the scheduling of chemical industry, accounting, examination, row's product, optimization, operation more near actual value.

As shown in Figure 1, be the adjustment of data node process flow diagram of certain process units that adopts the chemical process data correcting method of the inventive method, have 11 nodes, 7 kinds of components, 19 stream strand.Stream thigh 18 and 19 is value data, does not participate in the adjustment of data, and other stream thigh all can be measured.Certain mixed gas in the stream thigh 1 (a, b, c, d, e, f g) enters shunt cell node 1, is divided into output stream thigh 2,3 and 4.The function of node unit 2 is that this gas goes waste heat recovery, and stream thigh 3,4 goes a conversion respectively.Each node device can be classified as table 1 according to the conversion of its generation in this routine node flow process.Wherein be divided into a plurality of output stream thighs on the node unit 1,2,4,5, on the node 3,6,8,9,11 chemical reaction take place, node 7 is separation vessels, and node 10 is mixers.Other several process units can be translated into similar node process flow diagram, adopt the adjustment of data algorithm that proposes previously then, and it is proofreaied and correct.

Before system enters the adjustment of data, utilize the correction node process flow diagram of this process units that the node unit type is classified earlier: shunt (1,2,4,5), reactor (3,6,8,9,11), separation vessel (7), mixer (10).Each stream thigh is gas in the process flow diagram, and its measurand is divided into total flow rate and component concentration two parts.When only considering the constraint of total flow rate balance, this adjustment of data problem contains 11 equation of constraint.When considering the component concentration balance, it should have 77 equation of constraint mutually; When considering the total flow rate balance, its corresponding 119 equation of constraint.On this basis, enter adjustment of data process,

Present embodiment is to adopt adjustment of data module to move realization on computers, and concrete implementation step has following a few step:

The first step, from the database of factory, obtain the stream number of share of stock certificate of chemical process, satisfy the relation of material balance and the balance equation of flow process topological structure according to complicated chemical system, calculate the total flow rate and the component concentration measurement data of trying to achieve certain chemical process device or certain several device.For trilinear adjustment of data problem is described, we carry out total flow rate balance and component balanced the analysis to node unit 1, and node unit 1 is a shunt, and inlet flow thigh 1 is divided into stream thigh 2,3 and 4.

Component stream balance equation:

F _jx _jk-α _jF ₁x _1k＝0，j＝2，3，4；k＝1，Λ，7

Total flow rate normalization balance equation:

$\underset{j=2}{\overset{4}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_j=1$

The normalization equation of inlet flow:

$\underset{k=1}{\overset{7}{\mathrm{\&Sigma;}}}{x}_{1k}=1$

Split branch coefficient equation of constraint:

F _j-α _jF ₁＝0，j＝2，3，4

After the branch coefficient variation was split in elimination from component stream balance equation, the independent equation that can obtain node unit 1 was as follows:

x _jk-x _1k＝0， j＝2，3，4；k＝1，Λ，7

This shows that the material balance of node unit 1 is convertible into component constraint equation, and obviously available column constraint equation is down represented, thereby the trilinear problem is changed into bilinear problem.

$\underset{j=2}{\overset{4}{\mathrm{\&Sigma;}}}{F}_j-{\mathrm{<figure-callout\; id="1"\; label="F"\; filenames="A20071003678700151.png"\; state="\{\{state\}\}">F</figure-callout>}}_1=0$

Second step, set up this chemical process rig flow data calibration model,, utilize two-step approach, surveyed the reduced data correction problem that variable constitutes and be by correctable according to the fundamental relation of step 1, the topological structure of this device and the measurement data table that provides:

$\underset{\hat{X}\hat{Y}}{\min }{(\hat{X}-X)}^T{Q}_X^{-1}(\hat{X}-X)+{(\hat{Y}-Y)}^T{Q}_Y(\hat{Y}-Y)$

$\begin{array}{cc}s& t\end{array}$

$A\hat{X}=0$

$(A\&CircleTimes;E)\hat{C}\hat{X}=(A\hat{G}\&CirclePlus;E)\hat{Y}=0$

Wherein:

X＝x _1，1?Λ?x _1，7?x _2，1?Λ?x _2，7?Λ?Λ?Λ?x _17，1?Λ?x _17，7

Y＝x _1，1?Λ?y _1，7?y _2，1?Λ?y _2，7?Λ?Λ?Λ?y _11，1?Λ?y _11，7

Wherein: total stream number of share of stock is 19 in the chemical-process, and a stream strand interior number of components is 7.Q _XAnd Q _YCorrespond respectively to the variance-covariance matrix of total flow rate and composition; Corrected value for total flow rate X;

Corrected value for component stream Y; G and C are the Neudeker operator; E is 7 rank unit matrixs.Balance equation equation 2 is the balance equation of constraint of total flow rate, and formula 2 is the balance equation of constraint of component stream.

Data after the treatment for correcting must satisfy chemical physics rules such as material balance, heat balance, chemical reaction quantitative relation, and these relations are unified to be expressed as: $F(\hat{x},U)=0$ , wherein

Be to have surveyed the vector that adjustment of data value is formed, U does not survey data or waits to estimate the vector that parameter is formed, and F is a functional vector, all physics chemistry relational expression that representative should be satisfied.Believe that most measured values have certain reliability (always having stochastic error) so, the estimated value that not only will make the corrected value of measurement data and not survey data satisfies relational expression, and to make surveyed data on the occasion of with the deviation of measuring value minimum.Objective function is asked least square solution, linear equality constraints equation, and the U in the formula in fact also is included into X, comes after the variable to be corrected, f (x _i)≤the 0th, linear inequality constraint equation, φ (x _iThe non-linear equality constraint equation of splitting branch coefficient formation of)=0 shunt.Total flow rate and component concentration are proofreaied and correct respectively for twice as the variable of two groups of independent measurements, and at stream thigh total flow rate in the material balance equation and the constraint of component concentration balance bilinearity, correction work can adopt two step layerings to realize.At first the total flow rate data are proofreaied and correct, with its balance equation, the component concentration data are proofreaied and correct then, thereby bilinear problem can be resolved into two linear restrictions, find the solution and become simple as exact value substitution component concentration.The present invention is by being compiled into corresponding adjustment of data module, existing MATLAB TOOLBOX non-linear have the constraint sub-piece environment under, after above-mentioned model programming (command file and function file), after parameters is provided with, the general operation of machine 5～10 minutes, iteration provides the result about 10000 times.The present embodiment machine makes by loop iteration and surveys adjustment of data value and deviation of measuring value minimum, and it is as shown in table 1 to provide adjustment of data result, and whole chemical engineering industry procedures system has just been realized effective control.Table 1 is the adjustment of data result of calculation based on distinct methods in the present embodiment, adopts this device to be in DCS measurement data continuous under the same stable state, respectively to NLP method, San Xianxing ﹠amp; The result of the Bilinear Method adjustment of data compares, and wherein the data with * are the data of tracking down out that contain human error.

Table 1. is based on the adjustment of data result of calculation of distinct methods

Just because of the present invention considers in the process that design data is proofreaied and correct is the worst condition that is corrected the issuable human error of data, and this worst condition has the greatest impact to the data accuracy of system, the designed data correcting method of the present invention can make system be issued to the error minimum at worst condition, so when measurement data and actual value mismatch degree are not worst condition, what the result of this bearing calibration will be than expection is not bad.What therefore guaranteed the present invention's acquisition is that optimum data is proofreaied and correct the result.

Described above is the content that the preferred embodiment of the present invention is considered, obviously the present invention is not only limited to the foregoing description, can do all distortion to it under the prerequisite of the related scope of flesh and blood of the present invention and is implemented not departing from essence spirit of the present invention and do not exceed.And, although the illustrated data correcting method of the foregoing description is to carry out with the MATLABTOOLBOX making software on the basis of existing software package, purpose is to make interest of clarity clear, but understand that significantly they can realize in other occasions with related hardware according to same principle by adopting chip.

Claims (3)

1, a kind of chemical process data correcting method based on bilinear model, it is characterized in that, on the basis of data correcting method, the trilinear equation of constraint is converted into the method for bilinearity equation of constraint, bilinearity restricted problem is resolved into two linear problems to be found the solution, finding the solution the bilinear data correction, adopt two-step approach respectively to the total flow rate data, component concentration is proofreaied and correct, in logistic data correction modelling to the chemical engineering industry process apparatus, according to the chemical plant mixer, shunt, the fundamental relation of separation vessel and reactor four big classes, the topological structure of this device and the measurement data table that provides are set up its adjustment of data model; The user estimates best correction data according to by from the variance yields of a large amount of sequence real time datas through filtering, denoising, estimation measured value, makes corrected value more near actual value; Perhaps the user is by improving the technological process of production line, and the balance of the scope of the pattern of the foundation of the structure of flow process, balance, equation of constraint, correction and the degree of depth or the like aspect is held, and obtains the adjustment of data effect that requires.

2, the chemical process data correcting method based on bilinear model according to claim 1 is characterized in that, may further comprise the steps:

The first step, from the real-time data base of factory, obtain the stream number of share of stock certificate of chemical process, satisfy the relation of material balance and the balance equation of flow process topological structure according to complicated chemical system, calculate the total flow rate and the component concentration measurement data of trying to achieve certain chemical process device or certain several device; Main chemical unit according to actual condition is selected following classification:

The A mixer: two inlet flows 1 and 2 provide output stream 3, X ₃=X ₁+ X ₂

The B shunt: an inlet flow 1 is divided into two or more output streams 2 and 3, $\left\{\begin{array}{c}{X}_2={\mathrm{\&lambda;X}}_1\\ X_3=(1-\mathrm{\&lambda;})X_1\end{array};\right.$

C reactor: inlet flow 1 and output stream 2, X ₂=X ₁+ υ ^Tζ, wherein υ is a chemical reaction stoichiometric number matrix, its element υ _IjRepresent the stoichiometric coefficient of j component in i the reaction, ζ is the transforming degree vector that respectively reacts metering-type, its formula parameter, or variable;

The D separation vessel: an inlet flow 1 and two or more output streams 2 and 3, $\left\{\begin{array}{c}{X}_2={\mathrm{\&lambda;X}}_1\\ X_3=(1-\mathrm{\&lambda;})X_1\end{array}\right.,$ Wherein parameter lambda is a diagonal matrix, the segregation index of element representation respective components on each diagonal line, and it is a parameter matrix, I is a unit matrix;

In second step,, set up the logistic data correction model according to this chemical process device classification results that the first step obtains:

$\min \mathrm{\&Sigma;}{({\hat{x}}_i-x_i)}^T{Q}^{-1}({\hat{x}}_i-x_i)$

St.AX+BU+C=0, wherein x _iBe measured value,

Be corrected value, Q measures

f(x _i)≤0

φ(x _i)＝0

The variance covariance matrix of error, its diagonal entry σ _i ²It is the mean square deviation of its measuring error;

The 3rd step, with the trilinear equation of constraint be converted into the bilinearity equation of constraint method, bilinearity restricted problem resolved into two linear problems find the solution;

The 4th step, according to the adjustment of data model that second step set up measurement data is proofreaied and correct, must satisfy material balance, heat balance, chemical reaction quantitative relation after the calibrated processing of measurement data, these relations are unified to be expressed as: $F(\hat{x},U)=0$ , wherein U does not survey data or waits to estimate the vector that parameter is formed, and F is a functional vector, and all physics chemistry relational expression that representative should be satisfied in the data of having proofreaied and correct, is found out objective function, i.e. ε ^T∑ ^-1That value of ε minimum;

The 5th step, according to the 4th objective function that obtain of step, detect the corresponding adjustment of data value of that value of objective function minimum whether within actual range, determine promptly whether following inequality is set up: $x_l\&le;\hat{x}\&le;x_{\mathrm{\&mu;}}$ , x wherein _μ, x _iAnd be respectively the upper and lower limit of actual measurement data, if above-mentioned inequality is set up, then the variable of its correspondence is used as the appreciable error variable; Otherwise find out objective function time that little value, the variable of its correspondence is regarded as the appreciable error variable, the measured value of appreciable error variable is carried out error compensation, i.e. x%=x%-B _Rmm _b, B wherein _RmThe position of expression appreciable error, m _bThe amplitude of expression appreciable error;

The 6th step, the appreciable error variable result according to the 5th step obtained, carry out following judgement: if some variable that exists any one variable and appreciable error to concentrate in the data after proofreading and correct constitutes the loop, then that variable in the data of back is proofreaied and correct in deletion; If have any two variablees and appreciable error to concentrate arbitrary variable to form the loop in the data after proofreading and correct, then one of two variablees in the data of back are proofreaied and correct in deletion; If it is not empty proofreading and correct the back data, then turned back to for second step, circulation is proofreaied and correct again; If it be empty proofreading and correct the back data, then obtain the equal set of the concentrated variable of appreciable error, then obtained the data after all possible correction.

3, the chemical process data correcting method based on bilinear model according to claim 2 is characterized in that, before system enters the adjustment of data, and the error variance σ of the measurement data in the Q matrix _i ²Provide in advance, the sample that it is taken from the precision of instrument or gets one group of measurement data estimates that the variance of real time data measured value is handled by following easy mode:

Images