CN103610227B - Cut tobacco dryer head and tail section process variable optimizing control method - Google Patents

Abstract

The invention discloses a cut tobacco dryer head and tail section process variable optimizing control method. According to historical data of drum temperature, air temperature, moisture removing air door and other process variables of the head section and the tail section in the cut tobacco drying process, a cubic function is adopted to serve as a Cubic-RBF-ARX model of a radial basis function to carry out modeling on cut tobacco drying dynamic features; the built model has a self-adjustment capacity, can reflect influences of changes of inlet flow and inlet moisture on outlet moisture in different modes, and can predict the change situation of the outlet moisture in the future according to the changes of the inlet flow and the inlet moisture of the head section and the tail section in different modes; optimizing setting is carried out on the process variables according to the built model, so that a good effect of controlling cut tobacco outlet moisture on the head section and on the tail section is achieved. According to the method, influences of incoming material flow and moisture changes on the head section and the tail section can be effectively overcome in the cut tobacco drying process, and the method is suitable for head and tail section control of the cut tobacco inlet flow and the inlet moisture in different modes.

Description

A kind of cut-tobacco drier segment process variable optimal control method end to end

Technical field

The present invention relates to cut-tobacco drier segment process variable optimal control method end to end.

Background technology

Cut tobacco drying is most important one manufacturing procedure in cigarette fiber tow production, it is mainly by carrying out heat drying to cut tobacco, reduces the moisture content of cut tobacco, makes the moisture content of cut tobacco after baking, homogeneous temperature consistent, and control in certain number range, to meet manufacturing technique requirent.The technological process of drying silk is mainly divided into preheating, dry head, centre and dry tail process four parts.In the dry head stage, cut tobacco inlet flow rate constantly increases, but without the detected value of cut tobacco moisture content of outlet, is difficult to carry out FEEDBACK CONTROL, easily causes that dry head stage moisture content of outlet Control platform is poor, siccative is many; In the dry tail stage, because cut tobacco inlet flow rate reduces suddenly, and dry silk cylinder and have larger thermal capacitance, wall inner portion temperature is difficult to degradation problem under the speed by regulation, also easily causes that dry tail stage moisture content of outlet control performance is low and siccative is many.Therefore, " over-dried cut tobacco " is the difficult point place that current cut tobacco drying moisture content of outlet controls.

Existing over-dried cut tobacco course control method for use mainly contains:

(1) utilize the material entering and export cut-tobacco drier and drying medium as heat and mass balance object founding mathematical models, in conjunction with the control mode of feedforward PID adjustment barrel temperature.But feedforward Mathematical Modeling only considers moisture content and the flow of charging, do not consider hot blast temperature etc. other have the factor of material impact to moisture content of outlet, can not complete reaction real processes, cause section tobacco-dryer exit moisture fluctuation end to end greatly, need operating personnel to carry out manual intervention, under section different mode end to end, the supplied materials with different inlet flow rate and entrance moisture is difficult to obtain satisfied control effects.

(2) on the basis of above-mentioned feedforward control, increase the correct tailing of steam damping device in section end to end and apply steam water and improve the moisture content of expecting end to end, to reduce siccative amount.But the method only carries out humidification to cut tobacco top layer, improve only cut tobacco top layer humidity, still can cause the reduction of pipe tobacco inherent quality, and add difficulty and the stability of moisture content of outlet control.

(3) by test of many times, seek the best end to end the technological parameter such as stage hot blast temperature value and adjustment humidity discharging valve opening reduce siccative amount.The method lacks self-adjusting ability, cannot ensure that this group technological parameter is optimal setting for when having the supplied materials of different inlet flow rate and entrance moisture under different mode;

(4) on the basis of PID control strategy, the thought of fuzzy control is applied in moisture in cut tobacco dryer control.

Only still cannot obtain optimum processing parameter setting value by the control problem that simple two-dimensional fuzzy controller solves cut tobacco drying moisture content of outlet, and for the change of the inlet flow rate under different mode and entrance moisture, also need to adjust fuzzy control rule table, this makes troubles to industrial production.

Summary of the invention

Technical problem to be solved by this invention is, not enough for prior art, a kind of cut-tobacco drier segment process variable optimal control method is end to end provided, dry head stage cut tobacco moisture content of outlet is made to rise as quickly as possible and arrive stable state fast, dry tail stage cut tobacco moisture content of outlet is declined as far as possible lentamente, thus effectively reduce the siccative amount of section end to end, improve the control performance of cut tobacco drying; More effectively overcome supplied materials flow and moisture change to the impact of cut tobacco drying section end to end, the inconvenience of input state-variable parameter of avoiding manually adjusting.

For solving the problems of the technologies described above, the technical solution adopted in the present invention is: a kind of cut-tobacco drier is segment process variable optimal control method end to end, and the method is:

1) according to the operational process of cut-tobacco drier, set up the sequential relationship of cut tobacco inlet flow rate in cut tobacco drying, entrance moisture, cylinder temperature, wind-warm syndrome, humidity discharging air door, moisture content of outlet, simultaneously do according to cut tobacco drying a stage to divide detected value without cut tobacco inlet flow rate and inlet water feature without cut tobacco moisture content of outlet detected value, dry tail stage, adopt cubic function as the Cubic-RBF-ARX model of RBF, set up the Cubic-RBF-ARX model that cut tobacco drying does a stage and dry tail stage respectively;

2) according to the history data of cut-tobacco drier section end to end, structuring nonlinear parameter optimization method is adopted to optimize the Cubic-RBF-ARX model that cut tobacco drying does a stage and dry tail stage respectively;

3) do the Cubic-RBF-ARX model in a stage and dry tail stage according to the cut tobacco drying optimized, adopt two S type function to describe the optimum input curve of the humidity discharging air door in dry head stage, wind-warm syndrome, cylinder temperature; Jump function is adopted to describe the optimum input curve of the inlet flow rate in dry head stage; Exponential function is adopted to describe the optimum input curve of dry tail stage humidity discharging air door, wind-warm syndrome, the gentle cylindrical shell electric machine frequency of cylinder;

4) row dimension Bouguer Nai Kuier special formula method is adopted, the moisture content of outlet predicted value calculated by the Cubic-RBF-ARX model in the dry head stage and dry tail stage that make optimization and the error of moisture content of outlet setting value minimum, find out the parameter that cut tobacco drying does the optimum input curve in a stage and dry tail stage, to adapt to the change of supplied materials situation, reduce the siccative amount in dry tail stage.

In described step 1), cut-tobacco drier is done head stage Cubic-RBF-ARX model and is:

Wherein:

Wherein, y ^h(t ^h) represent that a moisture content of outlet for stage Cubic-RBF-ARX model done by cut-tobacco drier; represent the humidity discharging throttle opening of dry head stage Cubic-RBF-ARX model, wind-warm syndrome, cylinder temperature, inlet flow rate and entrance moisture respectively; X ^h(t ^h-1) be the state variable of inlet flow rate and entrance moisture; Np ^h, nq ^h, d ^hand m ^hall represent the order of dry head stage Cubic-RBF-ARX model; be respectively the center of the RBF neural of dry head stage Cubic-RBF-ARX model output item and input item; for the scalar weight coefficient of dry head stage Cubic-RBF-ARX model; || || _fthe Frobenius norm of representing matrix; ξ ^h(t ^h) be the modeling error of dry head stage Cubic-RBF-ARX model, be white Gaussian noise; T ₀ ^hfor the Cubic-RBF-ARX model modeling sampling time in head stage done by cut-tobacco drier, T ₁for from having inlet flow rate detected value to the time having inlet water to divide detected value, T ₂for from there being inlet water to divide the time of detected value to there being moisture content of outlet detected value, T ₃for from there being inlet water to divide the time of detected value to drying silk cylinder entrance, T ₄for cut tobacco is in the time of drying the oven dry of silk cylinder.

In described step 1), cut-tobacco drier dry tail stage Cubic-RBF-ARX model is:

Wherein:

Wherein, y ^t(t ^t) represent the moisture content of outlet of cut-tobacco drier dry tail stage Cubic-RBF-ARX model; represent the cylinder temperature of dry tail stage Cubic-RBF-ARX model, hot blast wind-warm syndrome, humidity discharging throttle opening, inlet flow rate, entrance moisture and cylindrical shell electric machine frequency respectively; X ^t(t ^t-1) be the state variable of hot blast wind-warm syndrome and cylindrical shell electric machine frequency; Np ^t, nq ^t, d ^tand m ^tall represent the order of dry tail stage Cubic-RBF-ARX model; be respectively the center of the RBF neural of dry tail stage Cubic-RBF-ARX model output item and input item; for the scalar weight coefficient of dry tail stage Cubic-RBF-ARX model; ξ ^t(t ^t) be dry tail stage Cubic-RBF-ARX model modeling error, be white Gaussian noise; T ₀ ^tfor the Cubic-RBF-ARX model modeling sampling time in cut-tobacco drier dry tail stage.

Described step 2) in, it is as follows that head stage Cubic-RBF-ARX model optimization done by cut-tobacco drier:

$({\widehat{\mathrm{\θ}}}_N^H,{\widehat{\mathrm{\θ}}}_L^H)=\arg \underset{{\mathrm{\θ}}_N^H,{\mathrm{\θ}}_L^H}{\min }\underset{t^{\mathrm{oh}}=1}{\overset{N^H}{\mathrm{\Σ}}}{({\stackrel{\‾}{y}}^H\left(t^{\mathrm{oh}}\right)-{\hat{y}}^H\left(t^{\mathrm{oh}}\right))}^2$

Wherein, that an actual value for stage moisture content of outlet done by cut-tobacco drier, be under actual input action, do a predicted value for the moisture content of outlet that stage Cubic-RBF-ARX model calculates by cut-tobacco drier; ${\widehat{\mathrm{\θ}}}_L^H=\{{\mathrm{\ω}}_0^{H,0},{\mathrm{\ω}}_{i^H,0}^{y^H},{\mathrm{\ω}}_{n,j^H,0}^{u^H},{\mathrm{\ω}}_{k^H}^{H,0},{\mathrm{\ω}}_{i^H,k^H}^{y^H},{\mathrm{\ω}}_{j^H,k^H}^{u^H}|i^H=1,...,{\mathrm{np}}^H;j^H=1,...,{\mathrm{nq}}^H;k^H=1,...,m^H\}$ For a linear dimensions for stage Cubic-RBF-ARX model done by cut-tobacco drier; for a nonlinear parameter for stage Cubic-RBF-ARX model done by cut-tobacco drier; N ^hfor head stage Cubic-RBF-ARX model modeling data length done by cut-tobacco drier.

Cut-tobacco drier dry tail stage Cubic-RBF-ARX model optimization is as follows:

$({\widehat{\mathrm{\θ}}}_N^T,{\widehat{\mathrm{\θ}}}_L^T)=\arg \underset{{\mathrm{\θ}}_N^T,{\mathrm{\θ}}_L^T}{\min }\underset{t^{\mathrm{ot}}=1}{\overset{N^T}{\mathrm{\Σ}}}{({\stackrel{\‾}{y}}^T\left(t^{\mathrm{ot}}\right)-{\hat{y}}^T\left(t^{\mathrm{ot}}\right))}^2$

Wherein, it is the actual value of cut-tobacco drier dry tail process middle outlet moisture; under actual input action, the predicted value of the moisture content of outlet calculated by cut-tobacco drier dry tail stage Cubic-RBF-ARX model; ${\mathrm{\θ}}_L^T=\{{\mathrm{\ω}}_0^{T,0},{\mathrm{\ω}}_{i^T,0}^{y^T},{\mathrm{\ω}}_{n,j^T,0}^{u^T},{\mathrm{\ω}}_{k^T}^{T,0},{\mathrm{\ω}}_{i^T,k^T}^{y^T},{\mathrm{\ω}}_{j^T,k^T}^{u^T}|i^T=1,...,{\mathrm{np}}^T;j^T=1,...,{\mathrm{nq}}^T;k^T=1,...,m^T\}$ For the linear dimensions of cut-tobacco drier dry tail stage Cubic-RBF-ARX model, for the nonlinear parameter of cut-tobacco drier dry tail stage Cubic-RBF-ARX model; N ^tfor head stage Cubic-RBF-ARX model modeling data length done by cut-tobacco drier.

In described step 3):

For describe cut-tobacco drier do a stage humidity discharging air door, wind-warm syndrome, cylinder temperature two S type function expression formulas of optimum input curve be:

$U_{\mathrm{sc}}\left(t^s\right)=\frac{{\mathrm{\λ}}_1}{1+e^{\frac{t^s-{\mathrm{\λ}}_2}{{\mathrm{\λ}}_3}}}+{\mathrm{\λ}}_4+\frac{{\mathrm{\λ}}_5}{1+e^{\frac{t^s-{\mathrm{\λ}}_6}{{\mathrm{\λ}}_7}}}$

Wherein, t ^sfor the time of input, unit is s; λ ₁, λ ₄, λ ₅be respectively the starting point of two S type function, turning point and end point values; λ ₂, λ ₆be respectively two symmetry axis centers of two S type function; λ ₃, λ ₇be respectively the speed that two S type function rises or declines; λ ₃, λ ₇represent when being greater than 0 that S type function rises, λ ₃, λ ₇represent when being less than 0 that S type function declines; C=1,2,3, U _s1(t ^s) be the setting value of humidity discharging air door; U _s2(t ^s) be the setting value of wind-warm syndrome; U _s3(t ^s) be the setting value of cylinder temperature.

The jump function expression formula doing the optimum input curve of a stage inlet flow for describing cut-tobacco drier is:

$U_T\left(t^T\right)=\left\{\begin{array}{cc}\frac{{\mathrm{\κ}}_1{t}^T}{{\mathrm{\κ}}_2}& t^T\∈[1,{\mathrm{\κ}}_2]\\ {\mathrm{\κ}}_1& t^T\∈[{\mathrm{\κ}}_2+1,{\mathrm{\κ}}_3]\end{array}\right.;$

Wherein, t ^tfor the time of input, unit is s; κ ₁, κ ₂, κ ₃be respectively the rate of climb of jump function, rise time and final value.

In described step 4), a moisture content of outlet predicted value that stage Cubic-RBF-ARX model calculates done by cut-tobacco drier for: ${\stackrel{\‾}{y}}^H\left(t^a\right)=f(U_{s1}\left(t^a\right),U_{s2}\left(t^a\right),U_{s3}\left(t^a\right),U_T\left(t^a\right)),$ the input variable of the dry head stage Cubic-RBF-ARX constructed by being substituted into by Optimal Setting curve cut-tobacco drier being done a stage each state-variable in obtain.By the moisture content of outlet predicted value making dry head stage Cubic-RBF-ARX model calculate with moisture content of outlet setting value y _set(t ^a) error e _h(t ^a) minimum, namely adopt row dimension Bouguer Nai Kuier special formula method solving-optimizing problem find out the parameter lambda of input curve of dry head stage humidity discharging air door, wind-warm syndrome, cylinder temperature _xwith the parameter κ of inlet flow rate input curve ₁, κ ₂, κ ₃; Wherein, x=1,2 ..., 7; G=1,2,3; M is the time of dry head phase lasts.

In described step 3), for describe dry tail stage humidity discharging air door, wind-warm syndrome, the gentle cylindrical shell electric machine frequency of cylinder the expression formula of exponential function of optimum input curve be:

$\begin{array}{c}{U}_{\mathrm{zp}}\left(t^z\right)={\mathrm{\α}}_{p1}\×{\left({\mathrm{\α}}_{p2}\right)}^{t^z}+{\mathrm{\α}}_{p3}\\ p=\mathrm{1,2,3,4}\end{array};$

U in formula _z1(t ^z), U _z2(t ^z), U _z3(t ^z), U _z4(t ^z) represent the optimum input curve of the gentle cylindrical shell electric machine frequency of dry tail stage humidity discharging air door, wind-warm syndrome, cylinder respectively.

In described step 4), the moisture content of outlet predicted value that the Cubic-RBF-ARX model in dry tail stage calculates for: ${\stackrel{\‾}{y}}^T\left(t^b\right)=f(U_{z1}\left(t^b\right),U_{z2}\left(t^b\right),U_{z3}\left(t^b\right),U_{z4}\left(t^b\right)),$ by the input variable by the dry tail stage Cubic-RBF-ARX model constructed by the substitution of the Optimal Setting curve of dry for cut-tobacco drier tail stage each state-variable in obtain; By the moisture content of outlet predicted value making the Cubic-RBF-ARX model in dry tail stage calculate with moisture content of outlet setting value y' _set(t ^b) error e _t(t ^b) minimum, namely adopt row dimension Bouguer Nai Kuier special formula method solving-optimizing problem find out the parameter alpha of dry tail stage humidity discharging air door, wind-warm syndrome, the optimum input curve of the gentle cylindrical shell electric machine frequency of cylinder _pg; Wherein, g=1,2,3; M' is dry tail phase duration.

Compared with prior art, the beneficial effect that the present invention has is: the inventive method can make dry head stage cut tobacco moisture content of outlet rise as quickly as possible and arrive stable state fast, dry tail stage cut tobacco moisture content of outlet can be made to decline as far as possible lentamente, thus effectively reduce the siccative amount of section end to end, improve the control performance of cut tobacco drying, there is larger economic worth; The inventive method has considered the dynamic characteristic come between doses and each input variable, more effectively can overcome supplied materials flow and moisture change to the impact of cut tobacco drying section end to end, be applicable to the head tail end control of cut tobacco inlet flow rate and inlet water timesharing under different mode; The inventive method goes out optimum input setting curve based on the model optimization of identification, avoids the inconvenience of input state-variable parameter of manually adjusting.

Accompanying drawing explanation

Fig. 1 is cut-tobacco drier technical process schematic diagram.

Detailed description of the invention

Cut-tobacco drier technical process as shown in Figure 1.Before cut tobacco enters drying process, first detect the inlet flow rate u of cut tobacco ₄with entrance moisture u ₅.Through T ₃time, cut tobacco arrives cut-tobacco drier porch.Cut tobacco is when drying silk cylinder and drying, and system understands the humidity discharging throttle opening u of timing sampling cylindrical shell ₁, wind-warm syndrome u ₂, cylinder temperature u ₃deng state-variable parameter value.Drying course continues T ₄time, the cut tobacco after oven dry is poured out from the outlet of baking silk cylinder, and measures cut tobacco moisture content of outlet value y in exit.From there being inlet flow rate detected value need experience one period of long period to there being moisture content of outlet detected value, such as certain dries silk production line approximately needs 340s.In addition, between the I/O variable of cut-tobacco drier, also there is larger time lag.

When inlet flow rate having been detected, show that cut tobacco drying brings into operation.Initial operating stage cut tobacco drying has cut tobacco inlet flow rate and inlet water to divide detected value, does not have cut tobacco moisture content of outlet detected value, and now cut tobacco drying is done the head stage and started.Do the characteristic in a stage according to cut tobacco drying, set up Cubic-RBF-ARX model structure:

Wherein:

Wherein, y ^h(t ^h) represent that a moisture content of outlet for stage Cubic-RBF-ARX model done by cut-tobacco drier; represent the humidity discharging throttle opening of dry head stage Cubic-RBF-ARX model, wind-warm syndrome, cylinder temperature, inlet flow rate and entrance moisture respectively; X ^h(t ^h-1) be the state variable of inlet flow rate and entrance moisture; Np ^h, nq ^h, d ^hand m ^hall represent the order of dry head stage Cubic-RBF-ARX model; be respectively the center of the RBF neural of dry head stage Cubic-RBF-ARX model output item and input item; for the scalar weight coefficient of dry head stage Cubic-RBF-ARX model; || || _fthe Frobenius norm of representing matrix; ξ ^h(t ^h) be the modeling error of dry head stage Cubic-RBF-ARX model, be white Gaussian noise; T ₀ ^hfor the Cubic-RBF-ARX model modeling sampling time in head stage done by cut-tobacco drier, T ₁for from having inlet flow rate detected value to the time having inlet water to divide detected value, T ₂for from there being inlet water to divide the time of detected value to there being moisture content of outlet detected value, T ₃for from there being inlet water to divide the time of detected value to drying silk cylinder entrance, T ₄for cut tobacco is in the time of drying the oven dry of silk cylinder.

When inlet flow rate becomes 0 from normal value, indicate the beginning of dry tail process, when moisture content of outlet drops to 3%, indicate the end of the whole cut tobacco drying of cut-tobacco drier.Without inlet flow rate detected value in dry tail process, but there is moisture content of outlet detected value.According to the characteristic of the dry tail procedure segment of cut-tobacco drier, set up following Cubic-RBF-ARX model:

Wherein:

Wherein, y ^t(t ^t) represent the moisture content of outlet of cut-tobacco drier dry tail stage Cubic-RBF-ARX model; represent the cylinder temperature of dry tail stage Cubic-RBF-ARX model, hot blast wind-warm syndrome, humidity discharging throttle opening, inlet flow rate, entrance moisture and cylindrical shell electric machine frequency respectively; X ^t(t ^t-1) be the state variable of hot blast wind-warm syndrome and cylindrical shell electric machine frequency; Np ^t, nq ^t, d ^tand m ^tall represent the order of dry tail stage Cubic-RBF-ARX model; be respectively the center of the RBF neural of dry tail stage Cubic-RBF-ARX model output item and input item; for the scalar weight coefficient of dry tail stage Cubic-RBF-ARX model; ξ ^t(t ^t) be dry tail stage Cubic-RBF-ARX model modeling error, be white Gaussian noise; T ₀ ^tfor the Cubic-RBF-ARX model modeling sampling time in cut-tobacco drier dry tail stage.

The present invention adopts structuring nonlinear parameter optimization method (SNPOM) method to estimate model.In order to the Cubic-RBF-ARX model constructed above making can describe the overall dynamic characteristic of cut tobacco drying section end to end, we first adopt SNPOM method come Optimized model, parameter under the minimum situation of one-step prediction error, and using this parameter as the model parameter initial value under long-term forecast optimization aim.Then, row dimension Bouguer Nai Kuier special formula method (LMM) are adopted to carry out the optimization of the model parameter of long-term forecast best performance.

The Parametric optimization problem that head stage Cubic-RBF-ARX model (1) done by cut-tobacco drier is as follows:

$({\widehat{\mathrm{\θ}}}_N^H,{\widehat{\mathrm{\θ}}}_L^H)=\arg \underset{{\mathrm{\θ}}_N^H,{\mathrm{\θ}}_L^H}{\min }\underset{t^{\mathrm{oh}}=1}{\overset{N^H}{\mathrm{\Σ}}}{({\stackrel{\‾}{y}}^H\left(t^{\mathrm{oh}}\right)-{\hat{y}}^H\left(t^{\mathrm{oh}}\right))}^2---\left(5\right)$

Wherein, that an actual value for stage moisture content of outlet done by cut-tobacco drier, be under actual input action, do a predicted value for the moisture content of outlet that stage Cubic-RBF-ARX model calculates by cut-tobacco drier; ${\widehat{\mathrm{\θ}}}_L^H=\{{\mathrm{\ω}}_0^{H,0},{\mathrm{\ω}}_{i^H,0}^{y^H},{\mathrm{\ω}}_{n,j^H,0}^{u^H},{\mathrm{\ω}}_{k^H}^{H,0},{\mathrm{\ω}}_{i^H,k^H}^{y^H},{\mathrm{\ω}}_{j^H,k^H}^{u^H}|i^H=1,...,{\mathrm{np}}^H;j^H=1,...,{\mathrm{nq}}^H;k^H=1,...,m^H\}$ For a linear dimensions for stage Cubic-RBF-ARX model done by cut-tobacco drier; for a nonlinear parameter for stage Cubic-RBF-ARX model done by cut-tobacco drier; N ^hfor head stage Cubic-RBF-ARX model modeling data length done by cut-tobacco drier.

The Parametric optimization problem of cut-tobacco drier dry tail stage Cubic-RBF-ARX model (3) is as follows:

$({\widehat{\mathrm{\θ}}}_N^T,{\widehat{\mathrm{\θ}}}_L^T)=\arg \underset{{\mathrm{\θ}}_N^T,{\mathrm{\θ}}_L^T}{\min }\underset{t^{\mathrm{ot}}=1}{\overset{N^T}{\mathrm{\Σ}}}{({\stackrel{\‾}{y}}^T\left(t^{\mathrm{ot}}\right)-{\hat{y}}^T\left(t^{\mathrm{ot}}\right))}^2---\left(6\right)$

Wherein, it is the actual value of cut-tobacco drier dry tail process middle outlet moisture; under actual input action, the predicted value of the moisture content of outlet calculated by cut-tobacco drier dry tail stage Cubic-RBF-ARX model; ${\mathrm{\θ}}_L^T=\{{\mathrm{\ω}}_0^{T,0},{\mathrm{\ω}}_{i^T,0}^{y^T},{\mathrm{\ω}}_{n,j^T,0}^{u^T},{\mathrm{\ω}}_{k^T}^{T,0},{\mathrm{\ω}}_{i^T,k^T}^{y^T},{\mathrm{\ω}}_{j^T,k^T}^{u^T}|i^T=1,...,{\mathrm{np}}^T;j^T=1,...,{\mathrm{nq}}^T;k^T=1,...,m^T\}$ For the linear dimensions of cut-tobacco drier dry tail stage Cubic-RBF-ARX model, for the nonlinear parameter of cut-tobacco drier dry tail stage Cubic-RBF-ARX model; N ^tfor head stage Cubic-RBF-ARX model modeling data length done by cut-tobacco drier.

Do a stage Cubic-RBF-ARX model to design the optimum input curve of each state-variable according to the cut tobacco drying estimated, to adapt to the change of supplied materials situation, reduce the siccative amount in dry head stage as far as possible.The present invention adopts two S type function to describe the optimum input curve of dry head stage humidity discharging air door, wind-warm syndrome, cylinder temperature, adopts step change type function to describe the optimum input curve of inlet flow rate.

Two S type curve equation is as follows:

$U_{\mathrm{sc}}\left(t^s\right)=\frac{{\mathrm{\λ}}_1}{1+e^{\frac{t^s-{\mathrm{\λ}}_2}{{\mathrm{\λ}}_3}}}+{\mathrm{\λ}}_4+\frac{{\mathrm{\λ}}_5}{1+e^{\frac{t^s-{\mathrm{\λ}}_6}{{\mathrm{\λ}}_7}}}---\left(7\right)$

Wherein, t ^sfor the time of input, unit is s; λ ₁, λ ₄, λ ₅be respectively the starting point of two S type function, turning point and end point values; λ ₂, λ ₆be respectively two symmetry axis centers of two S type function; λ ₃, λ ₇be respectively the speed that two S type function rises or declines; λ ₃, λ ₇represent when being greater than 0 that S type function rises, λ ₃, λ ₇represent when being less than 0 that S type function declines; C=1,2,3, U _s1(t ^s) be the setting value of humidity discharging air door; U _s2(t ^s) be the setting value of wind-warm syndrome; U _s3(t ^s) be the setting value of cylinder temperature.

The step change type function formula describing inlet flow rate input curve is as follows:

$U_T\left(t^T\right)=\left\{\begin{array}{cc}\frac{{\mathrm{\κ}}_1{t}^T}{{\mathrm{\κ}}_2}& t^T\∈[1,{\mathrm{\κ}}_2]\\ {\mathrm{\κ}}_1& t^T\∈[{\mathrm{\κ}}_2+1,{\mathrm{\κ}}_3]\end{array}\right.---\left(8\right)$

Wherein, t ^tfor the time of input, unit is s; κ ₁, κ ₂, κ ₃be respectively the rate of climb of jump function, rise time and final value.

The input variable of the Cubic-RBF-ARX model (1) constructed by the Optimal Setting curve (7-8) of each state-variable is substituted into in, the predicted value of dry head stage moisture content of outlet can be obtained

${\stackrel{\‾}{y}}^H\left(t^a\right)=f(U_{s1}\left(t^a\right),U_{s2}\left(t^a\right),U_{s3}\left(t^a\right),U_T\left(t^a\right))---\left(9\right)$

Adopt row dimension Bouguer Nai Kuierte (Levenberg-Marquardt Method, LMM) method, the moisture content of outlet predicted value calculated by making model and the error of moisture content of outlet setting value minimum, find out the λ of dry head stage humidity discharging air door, wind-warm syndrome, the optimum input curve of cylinder temperature _i(i=1,2 ..., 7) and the κ of the optimum input curve of parameter and inlet flow rate _j(j=1,2,3) parameter.Dry head stage moisture content of outlet setting value and be based on the error between dry head dynamic Model Prediction value (9):

$e_H\left(t^a\right)=y_{\mathrm{set}}\left(t^a\right)-{\stackrel{\‾}{y}}^H\left(t^a\right)---\left(10\right)$

Y _set(t ^a) be moisture content of outlet setting value.

The optimization problem of dry head stage process variable optimum setting is as follows:

$\underset{{\mathrm{\λ}}_x,{\mathrm{\κ}}_g}{\min }J=\underset{t^a=1}{\overset{M}{\mathrm{\Σ}}}{e}_H^2\left(t^a\right)---\left(11\right)$

M is dry head phase duration.The parameter value of optimum setting curve can be obtained by solving above-mentioned optimization problem, thus design the optimum input curve that each state-variable of stage done by cut-tobacco drier.

According to the cut tobacco drying dry tail stage Cubic-RBF-ARX model estimated to design the optimum input curve of each state-variable, to adapt to the change of supplied materials situation, reduce the siccative amount in dry tail stage as far as possible.Adopt exponential type function to describe the optimum input curve of dry tail stage humidity discharging air door, wind-warm syndrome, the gentle cylindrical shell electric machine frequency of cylinder, this exponential type curve formula is as follows:

$\begin{array}{c}{U}_{\mathrm{zp}}\left(t^z\right)={\mathrm{\α}}_{p1}\×{\left({\mathrm{\α}}_{p2}\right)}^{t^z}+{\mathrm{\α}}_{p3}\\ p=\mathrm{1,2,3,4}\end{array}---\left(12\right)$

U in formula _z1(t ^z), U _z2(t ^z), U _z3(t ^z), U _z4(t ^z) represent the optimum input curve of the gentle cylindrical shell electric machine frequency of dry tail stage humidity discharging air door, wind-warm syndrome, cylinder respectively.The input variable of the Cubic-RBF-ARX model (3) constructed by the Optimal Setting curve (12) of each state-variable is substituted into in, the predicted value of dry tail stage moisture content of outlet can be obtained:

${\stackrel{\‾}{y}}^T\left(t^b\right)=f(U_{z1}\left(t^b\right),U_{z2}\left(t^b\right),U_{z3}\left(t^b\right),U_{z4}\left(t^b\right))---\left(13\right)$

Adopt row dimension Bouguer Nai Kuierte (LMM) method, the moisture content of outlet predicted value calculated by making model is minimum with the error of moisture content of outlet setting value, finds out the α of dry tail stage humidity discharging air door, wind-warm syndrome, cylinder gentle cylindrical shell electric machine frequency optimum input curve _pg; Wherein, g=1,2,3.Dry tail stage moisture content of outlet setting value and be based on the error between dry tail dynamic Model Prediction value (13):

$e_T\left(t^b\right)={y^{\′}}_{\mathrm{set}}\left(t^b\right)-{\stackrel{\‾}{y}}^T\left(t^b\right)---\left(14\right)$

Y _sett () is moisture content of outlet setting value.

The optimization problem of dry tail stage process variable optimum setting is as follows:

$\underset{{\mathrm{\α}}_{\mathrm{pg}}}{\min }{J}^{\′}=\underset{k=1}{\overset{M^{\′}}{\mathrm{\Σ}}}{e}_T^2\left(t^b\right)---\left(15\right)$

M is dry tail phase duration.The parameter value of optimum setting curve can be obtained by solving above-mentioned optimization problem, thus design the optimum input curve of cut-tobacco drier dry tail each state-variable of stage.

Claims (3)

1. a cut-tobacco drier segment process variable optimal control method end to end, it is characterized in that, the method is:

1) according to the operational process of cut-tobacco drier, set up the sequential relationship of cut tobacco inlet flow rate in cut tobacco drying, entrance moisture, cylinder temperature, wind-warm syndrome, humidity discharging air door, moisture content of outlet, simultaneously do according to cut tobacco drying a stage to divide detected value without cut tobacco inlet flow rate and inlet water feature without cut tobacco moisture content of outlet detected value, dry tail stage, adopt cubic function as the Cubic-RBF-ARX model of RBF, set up the Cubic-RBF-ARX model that cut tobacco drying does a stage and dry tail stage respectively;

2) according to the history data of cut-tobacco drier section end to end, structuring nonlinear parameter optimization method is adopted to optimize the Cubic-RBF-ARX model that cut tobacco drying does a stage and dry tail stage respectively;

3) do the Cubic-RBF-ARX model in a stage and dry tail stage according to the cut tobacco drying optimized, adopt two S type function to describe the optimum input curve of the humidity discharging air door in dry head stage, wind-warm syndrome, cylinder temperature; Jump function is adopted to describe the optimum input curve of the inlet flow rate in dry head stage; Exponential function is adopted to describe the optimum input curve of dry tail stage humidity discharging air door, wind-warm syndrome, the gentle cylindrical shell electric machine frequency of cylinder;

4) row dimension Bouguer Nai Kuier special formula method is adopted, the moisture content of outlet predicted value calculated by the Cubic-RBF-ARX model in the dry head stage and dry tail stage that make optimization and the error of moisture content of outlet setting value minimum, find out the parameter that cut tobacco drying does the optimum input curve in a stage and dry tail stage, to adapt to the change of supplied materials situation, reduce the siccative amount in dry tail stage;

Described step 1) in, cut-tobacco drier is done head stage Cubic-RBF-ARX model and is:

Wherein:

Wherein, y ^h(t ^h) represent that a moisture content of outlet for stage Cubic-RBF-ARX model done by cut-tobacco drier; represent the humidity discharging throttle opening of dry head stage Cubic-RBF-ARX model, wind-warm syndrome, cylinder temperature, inlet flow rate and entrance moisture respectively; X ^h(t ^h-1) be the state variable of inlet flow rate and entrance moisture; Np ^h, nq ^h, d ^hand m ^hall represent the order of dry head stage Cubic-RBF-ARX model; be respectively the center of the RBF neural of dry head stage Cubic-RBF-ARX model output item and input item; for the scalar weight coefficient of dry head stage Cubic-RBF-ARX model; || || _fthe Frobenius norm of representing matrix; ξ ^h(t ^h) be the modeling error of dry head stage Cubic-RBF-ARX model, be white Gaussian noise; T ₀ ^hfor the Cubic-RBF-ARX model modeling sampling time in head stage done by cut-tobacco drier, T ₁for from having inlet flow rate detected value to the time having inlet water to divide detected value, T ₂for from there being inlet water to divide the time of detected value to there being moisture content of outlet detected value, T ₃for from there being inlet water to divide the time of detected value to drying silk cylinder entrance, T ₄for cut tobacco is in the time of drying the oven dry of silk cylinder;

Described step 1) in, cut-tobacco drier dry tail stage Cubic-RBF-ARX model is:

Wherein:

Wherein, y ^t(t ^t) represent the moisture content of outlet of cut-tobacco drier dry tail stage Cubic-RBF-ARX model; represent the cylinder temperature of dry tail stage Cubic-RBF-ARX model, hot blast wind-warm syndrome, humidity discharging throttle opening, inlet flow rate, entrance moisture and cylindrical shell electric machine frequency respectively; X ^t(t ^t-1) be the state variable of hot blast wind-warm syndrome and cylindrical shell electric machine frequency; Np ^t, nq ^t, d ^tand m ^tall represent the order of dry tail stage Cubic-RBF-ARX model; be respectively the center of the RBF neural of dry tail stage Cubic-RBF-ARX model output item and input item; for the scalar weight coefficient of dry tail stage Cubic-RBF-ARX model; ξ ^t(t ^t) be dry tail stage Cubic-RBF-ARX model modeling error, be white Gaussian noise; T ₀ ^tfor the Cubic-RBF-ARX model modeling sampling time in cut-tobacco drier dry tail stage;

Described step 2) in, it is as follows that head stage Cubic-RBF-ARX model optimization done by cut-tobacco drier:

$({\widehat{\mathrm{\θ}}}_N^H,{\widehat{\mathrm{\θ}}}_L^H)=\arg \underset{{\mathrm{\θ}}_N^H,{\mathrm{\θ}}_L^H}{\min }\underset{t^{\mathrm{oh}}=1}{\overset{N^H}{\mathrm{\Σ}}}{({\stackrel{\‾}{y}}^H\left(t^{\mathrm{oh}}\right)-{\hat{y}}^H\left(t^{\mathrm{oh}}\right))}^2$

Wherein, that an actual value for stage moisture content of outlet done by cut-tobacco drier, be under actual input action, do a predicted value for the moisture content of outlet that stage Cubic-RBF-ARX model calculates by cut-tobacco drier; ${\widehat{\mathrm{\θ}}}_L^H=\{{\mathrm{\ω}}_0^{H,0},{\mathrm{\ω}}_{i^H,0}^{y^H},{\mathrm{\ω}}_{n,j^H,0}^{u^H},{\mathrm{\ω}}_{k^H}^{H,0},{\mathrm{\ω}}_{i^H,k^H}^{y^H},{\mathrm{\ω}}_{j^H,k^H}^{u^H}|i^H=1,...,{\mathrm{np}}^H;j^H=1,...,{\mathrm{nq}}^H;k^H=1,...,m^H\}$ For a linear dimensions for stage Cubic-RBF-ARX model done by cut-tobacco drier; for a nonlinear parameter for stage Cubic-RBF-ARX model done by cut-tobacco drier; N ^hfor head stage Cubic-RBF-ARX model modeling data length done by cut-tobacco drier;

Cut-tobacco drier dry tail stage Cubic-RBF-ARX model optimization is as follows:

$({\widehat{\mathrm{\θ}}}_N^T,{\widehat{\mathrm{\θ}}}_L^T)=\arg \underset{{\mathrm{\θ}}_N^T,{\mathrm{\θ}}_L^T}{\min }\underset{t^{\mathrm{ot}}=1}{\overset{N^T}{\mathrm{\Σ}}}{({\stackrel{\‾}{y}}^T\left(t^{\mathrm{ot}}\right)-{\hat{y}}^T\left(t^{\mathrm{ot}}\right))}^2$

Wherein, it is the actual value of cut-tobacco drier dry tail process middle outlet moisture; under actual input action, the predicted value of the moisture content of outlet calculated by cut-tobacco drier dry tail stage Cubic-RBF-ARX model; ${\mathrm{\θ}}_L^T=\{{\mathrm{\ω}}_0^{T,0},{\mathrm{\ω}}_{i^T,0}^{y^T},{\mathrm{\ω}}_{n,j^T,0}^{u^T},{\mathrm{\ω}}_{k^T}^{T,0},{\mathrm{\ω}}_{i^T,k^T}^{y^T},{\mathrm{\ω}}_{j^T,k^T}^{u^T}|i^T=1,...,{\mathrm{np}}^T;j^T=1,...,{\mathrm{nq}}^T;k^T=1,...,m^T\}$ For the linear dimensions of cut-tobacco drier dry tail stage Cubic-RBF-ARX model, for the nonlinear parameter of cut-tobacco drier dry tail stage Cubic-RBF-ARX model, N ^tfor head stage Cubic-RBF-ARX model modeling data length done by cut-tobacco drier;

Described step 3) in:

For describe cut-tobacco drier do a stage humidity discharging air door, wind-warm syndrome, cylinder temperature two S type function expression formulas of optimum input curve be:

$U_{\mathrm{sc}}\left(t^s\right)=\frac{{\mathrm{\λ}}_1}{1+e^{\frac{t^s-{\mathrm{\λ}}_2}{{\mathrm{\λ}}_3}}}+{\mathrm{\λ}}_4+\frac{{\mathrm{\λ}}_5}{1+e^{\frac{t^s-{\mathrm{\λ}}_6}{{\mathrm{\λ}}_7}}}$

Wherein, t ^sfor the time of input, unit is s; λ ₁, λ ₄, λ ₅be respectively the starting point of two S type function, turning point and end point values; λ ₂, λ ₆be respectively two symmetry axis centers of two S type function; λ ₃, λ ₇be respectively the speed that two S type function rises or declines; λ ₃, λ ₇represent when being greater than 0 that S type function rises, λ ₃, λ ₇represent when being less than 0 that S type function declines; C=1,2,3, U _s1(t ^s) be the setting value of humidity discharging air door; U _s2(t ^s) be the setting value of wind-warm syndrome; U _s3(t ^s) be the setting value of cylinder temperature;

The jump function expression formula doing the optimum input curve of a stage inlet flow for describing cut-tobacco drier is:

$U_T\left(t^T\right)=\left\{\begin{array}{cc}\frac{{\mathrm{\κ}}_1{t}^T}{{\mathrm{\κ}}_2}& t^T\∈[1,{\mathrm{\κ}}_2]\\ {\mathrm{\κ}}_1& t^T\∈[{\mathrm{\κ}}_2+1,{\mathrm{\κ}}_3]\end{array}\right.;$

Wherein, t ^tfor the time of input, unit is s; κ ₁, κ ₂, κ ₃be respectively the rate of climb of jump function, rise time and final value;

Described step 3) in, for describe dry tail stage humidity discharging air door, wind-warm syndrome, the gentle cylindrical shell electric machine frequency of cylinder the expression formula of exponential function of optimum input curve be:

$U_{\mathrm{zp}}\left(t^z\right)={\mathrm{\α}}_{p1}\×{\left({\mathrm{\α}}_{p2}\right)}^{t^z}+{\mathrm{\α}}_{p3};$

p＝1,2,3,4

U in formula _z1(t ^z), U _z2(t ^z), U _z3(t ^z), U _z4(t ^z) represent the optimum input curve of the gentle cylindrical shell electric machine frequency of dry tail stage humidity discharging air door, wind-warm syndrome, cylinder respectively.

2. cut-tobacco drier according to claim 1 segment process variable optimal control method end to end, is characterized in that, described step 4) in, cut-tobacco drier is done head perfecting by stage setting curve U _sc(t ^s), U _t(t ^t) substitute into described cut-tobacco drier and do an input variable for stage Cubic-RBF-ARX model in, obtain cut-tobacco drier and do a moisture content of outlet predicted value that stage Cubic-RBF-ARX model calculates by the moisture content of outlet predicted value making dry head stage Cubic-RBF-ARX model calculate with moisture content of outlet setting value y _set(t ^a) error e _h(t ^a) minimum, namely adopt row dimension Bouguer Nai Kuier special formula method solving-optimizing problem find out the parameter lambda of input curve of dry head stage humidity discharging air door, wind-warm syndrome, cylinder temperature _xwith the parameter κ of inlet flow rate input curve ₁, κ ₂, κ ₃; Wherein, x=1,2 ..., 7; G=1,2,3; M is the time of dry head phase lasts.

3. cut-tobacco drier according to claim 1 segment process variable optimal control method end to end, is characterized in that, described step 4) in, by dry for cut-tobacco drier tail perfecting by stage setting curve U _zp(t ^z) substitute into the input variable of Cubic-RBF-ARX model of described cut-tobacco drier dry tail stage in, the moisture content of outlet predicted value that the Cubic-RBF-ARX model obtaining the cut-tobacco drier dry tail stage calculates by the moisture content of outlet predicted value making the Cubic-RBF-ARX model in dry tail stage calculate with moisture content of outlet setting value y' _set(t ^b) error e _t(t ^b) minimum, namely adopt row dimension Bouguer Nai Kuier special formula method solving-optimizing problem find out the parameter alpha of dry tail stage humidity discharging air door, wind-warm syndrome, the optimum input curve of the gentle cylindrical shell electric machine frequency of cylinder _pg; Wherein, g=1,2,3; M' is dry tail phase duration.