CN101173952B - Soft measuring method for confirming 4-carboxybenzaldehyde content - Google Patents

Abstract

The invention discloses a soft meter technology for determining the content of carboxybenzaldehyde in a crude terephthalic acid product on line. Measurable process variables chosen, namely, the oxygen content of air (or oxygen enriched air) entering into an oxidation reactor, the oxygenic content in the tail gas of the reactor, and the carbon oxygen compound content in the tail gas are taken as the input variables of a soft meter; based on the oxygen content of the oxygen enriched air, the oxygenic content in the tail gas and the carbon oxygen compound content in the tail gas, the corrected value of the carbon oxygen compound content in the tail gas is calculated under the standard condition; furthermore, based on the corrected value of the carbon oxygen compound content in the tail gas and the nonlinear correlation model of the content of carboxybenzaldehyde in the crude terephthalic acid product, the content of carboxybenzaldehyde in the crude terephthalic acid product are calculated on line and in real time. The soft meter on-line calculated value of the content of carboxybenzaldehyde in the crude terephthalic acid product can be used for guiding the operation and used as the basis of the further control according to the quality index.

Description

A kind of flexible measurement method of confirming carboxyl benzaldehyde content

Technical field

The invention belongs to petrochemical complex and process control crossing domain; Relate to pure terephthalic acid (PureTerephthalic Acid; PTA) industrial installation oxidation unit product---online definite method of crude terephthalic acid (TA) product quality, promptly in the TA product to online definite method of carboxyl benzaldehyde (4-CBA) content.

Background technology

PTA is the important source material of synthetic polyester fibers and plastics, mainly is used for the intermedium ethylene glycol terephthalate (PET) of synthesizing polyester.Whole PTA production technology is oxidation unit and refined unit.Wherein the oxidation unit flow process is as shown in Figure 1: be raw material with the P-xylene, acetic acid is solvent, at cobalt acetate, under the effect of manganese acetate catalyst, is promoter and airborne oxygen reaction with hydrogen bromide (or tetrabromoethane), generates terephthalic acid (TPA).A large amount of reaction heat of emitting in the reaction are taken away through the evaporation of solvent, and reclaim this part heat through byproduct steam.Oxidation liquid is through the crystallizer decrease temperature and pressure of series connection, and through filtering, drying obtains intermediate product TA again.4-CBA is an important intermediate of reaction, and it can be separated out with the TA product with oxidation unit product TA cocrystallization.Because 4-CBA can influence the growth of PTA chain when PET transforms, it is become p-methylbenzoic acid (p-toluic acid, PT acid) by hydrogenating reduction and removes basically in refined unit.Because the refined unit hydrogenation is limited in one's ability, and over oxidation (being that 4-CBA content is too low in the TA product) can cause the combustible loss of oxidation reactor to increase the weight of, so 4-CBA content must strict control among the TA, and 4-CBA content is TA product important quality index.

Because in P-xylene liquid-phase catalytic oxidation process, after oxidation reaction finished, 4-CBA content was definite, but does not have suitable online instrument measurement, can only sampling analysis.Generally the manual analysis sampling spot of 4-CBA is obtained in discharging place of dryer; Need at least 1.5 hours from first crystallizer to dryer; The manual analysis time of 4-CBA (off-line stratographic analysis) needs about 40 minutes, sampling and about 30 minutes of sample preparation time.Therefore, in conventional 4-CBA control, current manual analysis value is to be determined by preceding 2.5～3.5 hours reaction condition, is not the real-time measurement values of 4-CBA, and operating personnel can only by virtue of experience adjust, and product quality is difficult to guarantee and fluctuates big.Especially when reaction condition is unstable, in order to obtain suitable 4-CBA content, whole adjustment process even need reach several days time.The 4-CBA content analysis lags behind in the TA product, and 4-CBA content brings difficulty in the TA product in real time, effectively controlling.

Jap.P. (JP101739) discloses the method for 4-CBA content in a kind of TA of prediction product: through carbon dioxide content in the detection reaction device tail gas; Return carbon dioxide content (x in the acquisition tail gas based on production data; %) with the TA product in 4-CBA content (y; One-variable linear regression equation y=a*x+b %) (wherein a, b is a regression coefficient) obtains 4-CBA content real-time estimate value in the TA product.Because the complicacy of oxidation reaction process, one-variable linear regression equation are difficult to describe accurately in the tail gas relation between the 4-CBA content in the carbon dioxide content and TA product.Simultaneously; The oxygen content of the air of entering oxidation reactor is often adjusted with climatic condition according to the production load and (is promptly added pure oxygen; Form oxygen-enriched air), the tail oxygen content fluctuation range of reactor tail gas is bigger, causes under identical oxidation depth and combustible loss degree, 4-CBA content is constant in the TA product; But carbon dioxide content changes in the corresponding tail gas, thereby the model prediction precision is further reduced.

Chinese patent (ZL01113517.4) discloses the method for 4-CBA content in the another kind of prediction TA product: set up the neural network model that gets into 4-CBA content in oxidation reactor P-xylene flow, the air flow process that gets into the secondary oxidation crystallizer and secondary oxidation tail oxygen content and the TA product, realize the real-time estimate of 4-CBA content in the TA product.This patent only can be applicable to the production run with secondary oxidation technology.Simultaneously; Because it is numerous and be the high nonlinearity characteristic to influence the factor of p xylene oxidation course of reaction; Each reaction factor is complicated to once oxidation and secondary oxidation influence, only through above-mentioned three measurands, is difficult to disclose the influence of each factor to 4-CBA content in the TA product.

Influence in the TA product that in the 4-CBA content and tail gas there be the main technique operating parameter of hydrocarbon content (it can be used as the sign of reactor burning reaction degree): the Co catalysts concentration in the reaction feed; Manganese catalyst concentration; The bromine promoter concentration; The reactor reaction temperature, reactor is extracted discharge, reactor residence time etc. out.Table 1 has provided the rule that influences of these process operation parameters.Provide in the oxygen content that gets into the reactor air and the tail gas tail oxygen content in the table 2 to the rule that influences of hydrocarbon content in 4-CBA content and the tail gas in the TA product.

Each process operation parameter of table 1 influence rule

Process operation parameter	Hydrocarbon content in the tail gas	4-CBA content in the TA product
			Co catalysts concentration rises	Rise	Descend
The Co catalysts density loss	Descend	Rise
			Manganese catalyst concentration rises	Rise	Descend
Manganese catalyst concentration descends	Descend	Rise
			The bromine promoter concentration rises	Rise	Descend
The bromine promoter concentration descends	Descend	Rise
			Temperature of reaction rises	Rise	Descend
Temperature of reaction descends	Descend	Rise
			Reactor is extracted discharge out and is risen	Rise	Descend
Reactor is extracted discharge out and is descended	Descend	Rise
			Reactor residence time rises	Rise	Descend
Reactor residence time descends	Descend	Rise

Table 2 gets into the rule that influences of tail oxygen content in oxygen content and the tail gas of reactor air

Process operation parameter	Hydrocarbon content in the tail gas	4-CBA content in the TA product
			The oxygen content of air rises	Rise	Constant
The oxygen content of air descends	Descend	Constant
			The tail oxygen content rises in the tail gas	Descend	Constant
The tail oxygen content descends in the tail gas	Rise	Constant

Can find out by table 1 and table 2: under the situation that the tail oxygen content remains unchanged in oxygen content that gets into the reactor air and tail gas; In the tail gas in hydrocarbon content and the TA product 4-CBA content have opposite Changing Pattern; And, so exist a kind of high nonlinearity to concern between the 4-CBA content in hydrocarbon content and the TA product in the tail gas because each process operation parameter all is the height nonlinear characteristic to oxidation reaction and subsidiary reaction process influence.Simultaneously, consider to produce load adjustment and climatic condition and change, the air that gets into reactor often adds pure oxygen, forms the oxygen-enriched air of different oxygen; And the tail oxygen content often fluctuates in the tail gas.For this reason; Choose in tail oxygen content and the tail gas of oxygen content, reactor tail gas of the air (or oxygen-enriched air) that gets into oxidation reactor three measurable variables such as hydrocarbon (being that carbon monoxide or carbon dioxide or carbon monoxide and carbon dioxide are total) content; As the input variable of 4-CBA content soft instrument in the TA product, will set up the soft instrument of 4-CBA content in effective, the reliable TA product.Realize in the TA product 4-CBA content in real time, on-line prediction exactly, be the timely adjustment of production operating conditions, the optimal control of product quality provides the basis.

Summary of the invention

The object of the invention provides the soft instrument technology of 4-CBA content in a kind of online definite TA product.Choose in tail oxygen content and the tail gas of oxygen content, reactor tail gas of the air (or oxygen-enriched air) that gets into oxidation reactor hydrocarbon (being that carbon monoxide or carbon dioxide or carbon monoxide and carbon dioxide are total) content as the input variable of soft instrument model; Then, hydrocarbon content in the tail gas is modified to the value under the standard condition; At last; Through based on the commercial plant production process data, adopt in the tail gas that Nonlinear Modeling technology (like nerual network technique etc.) sets up the correlation model of 4-CBA content in the hydrocarbon content modified value and TA product, online in real time is confirmed 4-CBA content in the TA product.

Content of the present invention is mainly following:

A kind of flexible measurement method of confirming carboxyl benzaldehyde content is characterized in that this flexible measurement method comprises following steps:

At first, with hydrocarbon content in the tail oxygen content of the oxygen content of the air that gets into oxidation reactor or oxygen-enriched air, reactor tail gas and the tail gas as the independent variable of soft instrument, this independent variable of input in the robotization treating apparatus;

Then, to be dependent variable to carboxyl benzaldehyde content in the terephthalate product, based on independent variable, in, the real-time estimate terephthalate product online through soft instrument to carboxyl benzaldehyde content.

Above-mentioned independent variable and dependent variable are the neural network models of non-linear correlation.Hydrocarbon content in the tail gas is modified to the content under the standard condition

To said

And to carboxyl benzaldehyde content y _4-CBACarry out normalization and handle, obtain sx and sy respectively; Through gathering industrial process data, adopt nerual network technique to set up the neural network model between sx and the sy; The input layer of said neural network model has 1 node, and hidden layer has 1～30 node, and output layer has 1 node.

The sy of said neural network model tries to achieve y through anti-normalization _4-CBAPredicted value

1. independent variable and dependent variable chooses

If the composition of turnover oxidation reactor gas is as shown in Figure 2, each symbol description is following among the figure:

V ₁---advance the volume of air (or oxygen-enriched air) under standard state of oxidation reactor, m ³

V ₂---go out the volume of tail gas under standard state of oxidation reactor, m ³

---oxygen content in the air, %

---nitrogen content in the air, %

---oxygen content in the tail gas, %

---nitrogen content in the tail gas, %

---hydrocarbon content in the tail gas, %

Wherein, V ₁,

There is in-line meter to measure.Keeping weighing apparatus, production status and nitrogen according to material is inert gas, has following relational expression to set up (or approximate establishment):

$\left\{\begin{array}{c}{x}_{O_2}^1+x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1=100\%\\ x_{O_2}^2+x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2+x_{{\mathrm{CO}}_x}^2=100\%\\ V_1{x}_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1=V_2{x}_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2\end{array}\right.---\left(1\right)$

The independent variable of 4-CBA content soft instrument model in the TA product, promptly input variable is chosen:

(1) oxygen content of air (or oxygen-enriched air) (

%)

(2) the tail oxygen content of reactor tail gas (

%)

(3) hydrocarbon content ( %) in the tail gas

The dependent variable of 4-CBA content soft instrument model in the TA product, promptly output variable is chosen:

(1) 4-CBA content (y in the TA product _4-CBA, %)

2. the correction of hydrocarbon content in the tail gas

The accurate operating mode of bidding is: the oxygen content that gets into the air of oxidation reactor does

Wherein $100\%>x_{O_2}^{s,1}\&GreaterEqual;21\%,$ Usually get $x_{O_2}^{s,1}=21\%;$ The tail oxygen content of reactor tail gas does

Wherein $8\%>x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^{s,2}\&GreaterEqual;1\%,$ Usually get $x_{O_2}^{s,1}=3.5\%.$ Then the correction of hydrocarbon content is following in the tail gas: (1) is to the oxygen content of the air of entering oxidation reactor

With standard condition content

The correction of deviation

If after revising for the first time, the air mass flow that gets into oxidation reactor is V ₁ ¹, the total flow of tail gas is V ₂ ¹, the tail oxygen content of tail gas does The hydrocarbon content of tail gas does

And the amount of oxygen of establishing the air that gets into oxidation reactor remains unchanged, then: $V_1^1=\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1}{x_{O_2}^{s,1}}{V}_1.$

Its addition is: $V_1^1-V_1=\frac{x_{O_2}^1}{x_{O_2}^{s,1}}{V}_1-V_1=\frac{x_{O_2}^1-x_{O_2}^{s,1}}{x_{O_2}^{s,1}}{V}_1,$ And be nitrogen.

Then be through the total flow of revising back tail gas for the first time: ${\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">V</figure-callout>}}_2^1={\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">V</figure-callout>}}_2+\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1-x_{O_2}^{s,1}}{x_{O_2}^{s,1}}{V}_1=(\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}+\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1-x_{O_2}^{s,1}}{x_{O_2}^{s,1}})V_1$

The tail oxygen content of tail gas is: $x_{O_2}^{\mathrm{2,1}}=\frac{V_2{x}_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^2}{{\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">V</figure-callout>}}_2^1}=\frac{x_{O_2}^2\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}{V}_1}{(\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}+\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1-x_{O_2}^{s,1}}{x_{O_2}^{s,1}})V_1}=\frac{x_{O_2}^2{x}_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{(x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1+\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}{x_{O_2}^{s,1}}(x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1-x_{O_2}^{s,1}))}$

The hydrocarbon content of tail gas is:

$x_{C{O}_x}^{\mathrm{2,1}}=\frac{V_2{x}_{C{O}_x}^2}{{\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">V</figure-callout>}}_2^1}=\frac{x_{C{O}_x}^2\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}{V}_1}{(\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}+\frac{x_{O_2}^1-x_{O_2}^{s,1}}{x_{O_2}^{s,1}})V_1}=\frac{x_{{\mathrm{CO}}_x}^2{x}_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{(x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1+\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}{x_{O_2}^{s,1}}(x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1-x_{O_2}^{s,1}))}$

(2) on the oxygen content of the exhaust tail

content with the standard conditions

deviations,

If after revising for the second time, the air mass flow that gets into oxidation reactor is V ₁ ¹¹, the total flow of tail gas is V ₂ ¹¹, the hydrocarbon content of tail gas does

The tail oxygen content of tail gas does

And establishing the oxygen of revising for the second time in the back increase air capacity is not consumed.

Then the total flow of tail gas does ${\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">V</figure-callout>}}_2^{11}=(V_1^{11}-V_1^1)+{\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">V</figure-callout>}}_2^1$

By $\frac{(V_1^{11}-V_1^1)x_{O_2}^{s,1}+x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^2\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}{V}_1}{(V_1^{11}-V_1^1)+{\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">V</figure-callout>}}_2^1}=x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^{s,2},$

: $V_1^{11}=\frac{(x_{O_2}^{s,2}-x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^2)\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}-x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^{s,2}+x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1}{(x_{O_2}^{s,1}-x_{O_2}^{s,2})}{V}_1$

: $V_2^{11}=(V_1^{11}-V_1^1)+{\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">V</figure-callout>}}_2^1=\frac{(x_{O_2}^{s,1}-x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^2)\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}+(x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1-x_{O_2}^{s,1})}{(x_{O_2}^{s,1}-x_{O_2}^{s,2})}{V}_1$

Then the hydrocarbon content of tail gas is:

$x_{{\mathrm{CO}}_x}^{\mathrm{2,11}}=\frac{V_2{x}_{{\mathrm{CO}}_x}^2}{V_2^{11}}=\frac{x_{{\mathrm{CO}}_x}^2\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}{V}_1}{\frac{(x_{O_2}^{s,1}-x_{O_2}^2)\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}+(x_{O_2}^1-x_{O_2}^{s,1})}{(x_{O_2}^{s,1}-x_{O_2}^{s,2})}{V}_1}=\frac{x_{{\mathrm{CO}}_x}^2\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}(x_{O_2}^{s,1}-x_{O_2}^{s,2})}{(x_{O_2}^{s,1}-x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^2)\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^1}{x_{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}^2}-x_{O_2}^{s,1}+x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1}$

The final modified value of hydrocarbon content that is tail gas is:

$x_{{\mathrm{CO}}_x}^{\mathrm{2,11}}=\frac{x_{{\mathrm{CO}}_x}^2(x_{O_2}^{s,1}-x_{O_2}^{s,2})}{(x_{O_2}^{s,1}-x_{O_2}^2)+(x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1-x_{O_2}^{s,1})\frac{100\%-x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^2-x_{{\mathrm{CO}}_x}^2}{100\%-x_{{\mathrm{<figure-callout\; id="2"\; label="O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">O</figure-callout>}}_2}^1}}---\left(2\right)$

Then pass through three measured values of

of commercial plant measurement instrument; And the oxygen content

that gets into the air of oxidation reactor under the standard condition of setting just can be tried to achieve through (2) formula with the tail oxygen content

of reactor tail gas; The final modified value of hydrocarbon content

that is converted into tail gas under the standard condition has been eliminated the air that gets into reactor and has often been added pure oxygen, forms the oxygen-enriched air of different oxygen; And in the tail gas tail oxygen content often fluctuation to the influence of the hydrocarbon relative content of tail gas.

3. the correlation model of 4-CBA content in the modified value of hydrocarbon content and the TA product in the tail gas

Based on the modified value of hydrocarbon content in the tail gas (

%) with the TA product in 4-CBA content (y _4-CBA, there is the height nonlinear relationship between %), can adopt Nonlinear Modeling technology arbitrarily to set up in modified value and the TA product of hydrocarbon content in the tail gas correlation model between the 4-CBA content, like neural network modeling approach etc.

The modified value of hydrocarbon content in the tail gas (

%) with the TA product in 4-CBA content (y _4-CBAIt is the Sigmoid function that neural network model %) adopts activation functions, and three layers of forward pass neural network are set up, and adopts error back propagation algorithm (BP; Back Propagation) network is trained 1 node of input layer, 1～30 node of hidden layer, 1 node of output layer.The input variable of network model is utilized formula (3) to carry out normalization and is handled:

$\mathrm{sx}=\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="CO\; x"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">CO</figure-callout>}}_x}^{\mathrm{2,11}}-x_{\min }}{x_{\max }-x_{\min }}(b-a)+a,---\left(3\right)$

(3) in the formula, sx representes after the normalization value as the neural network input, [x _Min, x _Max] represent to collect

Variation range, after the normalization input independent variable variation range be [a, b].

The output variable of network model is utilized formula (4) to carry out normalization and is handled:

$\mathrm{sy}=\frac{y_{4-\mathrm{CBA}}-y_{\min }}{y_{\max }-y_{\min }}(b-a)+a,---\left(4\right)$

(4) in the formula, sy representes after the dependent variable normalization desired value as neural network output, [y _Min, y _Max] expression collects the variation range of dependent variable, the variation range of neural network output is [a, b] after the normalization.

Collect n and organize representational commercial plant data; Wherein every group data comprise

through forming

after (2) formula correction through (3), (4) formula normalization after is [sx; Sy], form training sample.To neural network model, with the input of sx as network, corresponding sy is as desired value, training network.When reaching certain accuracy requirement, stop training, obtain the neural network model between sx and the sy, be made as:

sy＝f(sx) (5)

Wherein, f () is the neural network model between sx and the sy.Then through anti-normalization to neural network model output st, just can be promptly in the hope of the predicted value

of 4-CBA content in the TA product

${\tilde{y}}_{4-\mathrm{CBA}}=y_{\min }+\frac{y_{\max }-y_{\min }}{b-a}(\mathrm{sy}-a)---\left(6\right)$

4.TA 4-CBA content soft instrument in line computation in the product

The online calculation process of 4-CBA content soft instrument is as shown in Figure 3 in the TA product; Record oxygen content in the air

through the oxygen analysis appearance that gets into the reactor air and record tail oxygen content

through reactor tail gas oxygen analysis appearance and record hydrocarbon content in the tail gas

based on three measured values of

through reactor tail gas carbon monoxide content analyser and carbon dioxide content analyser, and the standard condition that passes through Xia pass through oxidation reactor air oxygen content

Yu the tail oxygen content

of reactor tail gas pass through

(2) formula is tried to achieve, and the final modified value of hydrocarbon content

that is converted into tail gas under the standard condition is passed through

(3) formula is tried to achieve the value sx after

normalization; Through (5) formula, try to achieve neural network output valve sy; Through (6) formula; Computer distributed control system (the Distributedcontrol system that the soft instrument of 4-CBA content in predicted value

the TA product of 4-CBA content in the TA product can put into operation at the TA process units is tried to achieve in anti-normalization; DCS) in; Based on above-mentioned three real-time measurable variables; Through soft instrument in line computation, realize 4-CBA content in the TA product in real time, on-line prediction.

Description of drawings

Fig. 1 is a PTA production technology oxidation unit flow process;

Fig. 2 is the gaseous species and the composition thereof of turnover oxidation reactor;

Fig. 3 is the online calculation process of 4-CBA content soft instrument in the TA product.

Embodiment

Explanation through following examples will help to understand the present invention, but not limit content of the present invention.

Below through embodiment the present invention is described further:

1.TA the foundation of 4-CBA content soft instrument in the product

Gather in 100 groups of production runes under the different representative operating modes, the oxygen content of air (or oxygen-enriched air) (

%), the tail oxygen content of reactor tail gas ( %), hydrocarbon content in the tail gas ( %), with corresponding TA product in 4-CBA content (y _4-CBA, %) form sample data.

The accurate operating mode of bidding is:

(1) the oxygen standard content of the air of entering oxidation reactor $x_{O_2}^{s,1}=21\%$

(2) the tail oxygen standard content of reactor tail gas $x_{O_{}}^{}$ ${{}_2}_{s,2}^{}=3.5\%$

Through (2) formula, the final modified value of hydrocarbon content of tail gas is under the standard condition:

${\mathrm{<figure-callout\; id="2"\; label="x\; CO\; x"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{{\mathrm{CO}}_x}^{{}_{}}=\frac{x_{{\mathrm{CO}}_x}^2(21-3.5)}{(21-{\mathrm{<figure-callout\; id="2"\; label="x\; O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{O_{}}^{{}_2})+({\mathrm{<figure-callout\; id="2"\; label="x\; O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{O_{}}^{{}_2}-21)\frac{100\%-{\mathrm{<figure-callout\; id="2"\; label="x\; O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{O_{}}^{{}_2}-x_{{\mathrm{CO}}_x}^2}{100\%-{\mathrm{<figure-callout\; id="2"\; label="x\; O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{O_{}}^{{}_2}}}---\left(7\right)$

Then, form 100 groups of

sample datas by original 100 groups of sample datas.Right

y _4-CBACarrying out normalization handles:

Variation range [1803,2.688], y _4-CBAVariation range [0.13,0.42], get a=0.2, b=0.8, carry out normalization and calculate:

$\mathrm{sx}=\frac{x_{{\mathrm{<figure-callout\; id="2"\; label="CO\; x"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">CO</figure-callout>}}_x}^{\mathrm{2,11}}-1.803}{2.688-1.803}(0.8-0.2)+0.2---\left(8\right)$

$\mathrm{sy}=\frac{y_{4-\mathrm{CBA}}-0.13}{0.42-0.13}(0.8-0.2)+0.2---\left(9\right)$

If the Nonlinear Modeling technology is chosen nerual network technique, network structure is: the input layer number is 1, and the hidden layer node number is 1, and output layer node number is 1.With 100 groups after the normalization [sx, sy] sample datas is training sample, adopts the BP algorithm that network is trained; During network convergence, obtain following one group of weights and threshold value:

$w_{11}^{\left(1\right)}=-5.7367$

$w_{11}^{\left(2\right)}=8.1739$

$b_1^{\left(1\right)}=0.35383$

$b_1^{\left(2\right)}=-1.213$

W wherein _Ij ⁽¹⁾Be the weights of i node of input layer to j node of hidden layer; w _Ij ⁽²⁾Be the weights of i node of hidden layer to j node of output layer; b _i ⁽¹⁾Be i node threshold value of hidden layer; b _i ⁽²⁾Be i node threshold value of output layer.

Then, the neural network model between sx and the sy is:

$\mathrm{net}1=w_{11}^{\left(1\right)}*\mathrm{sx}+b_1^{\left(1\right)},---\left(10\right)$

$\mathrm{out}1=\frac{1}{1+\exp (-\mathrm{net}1)},---\left(11\right)$

$\mathrm{net}2=w_{11}^{\left(2\right)}*\mathrm{out}1+b_1^{\left(2\right)},---\left(12\right)$

$\mathrm{sy}=\frac{1}{1+\exp (-\mathrm{net}2)},---\left(13\right)$

The sy value is handled through anti-normalization just can be in the hope of 4-CBA content (y in the TA product _4-CBA, predicted value %)

When the normalization of training sample dependent variable, y _4-CBAVariation range [0.13,0.42], a=0.2, b=0.8, then

${\tilde{y}}_{4-\mathrm{CBA}}=0.13+\frac{0.42-0.13}{0.8-0.2}(\mathrm{sy}-0.2)---\left(14\right)$

Above-mentioned through case description, how to set up 4-CBA content soft instrument model in the TA product based on the industrial processes data.

2.TA 4-CBA content soft instrument in line computation in the product

Computer distributed control system (the Distributed control system that the soft instrument of 4-CBA content can put into operation at the TA process units in the TA product; DCS) in; Based on the oxygen analysis appearance that gets into the reactor air record oxygen content in the air

reactor tail gas oxygen analysis appearance record tail oxygen content and reactor tail gas carbon monoxide content analyser and carbon dioxide content analyser record hydrocarbon content in the tail gas

through soft instrument in line computation, realize real-time, the on-line prediction of 4-CBA content in the TA product.

If the on-line measurement value of three process measurement variablees is:

$\left(1\right),{\mathrm{<figure-callout\; id="2"\; label="x\; O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{O_{}}^{{}_2}=24\%$

$\left(2\right){,\mathrm{<figure-callout\; id="2"\; label="x\; O"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{O_{}}^{{}_2}=3.8\%$

$\left(3\right),x_{{\mathrm{CO}}_x}^2=2.8\%$

Then get through (7) formula correction:

${\mathrm{<figure-callout\; id="2"\; label="x\; CO\; x"\; filenames="H200710047413520080121E000021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{{\mathrm{CO}}_x}^{{}_{}}=\frac{2.8\&times;(21-3.5)}{(21-3.8)+(24-21)\frac{100-3.8-2.8}{100-24}}\%=2.346\%$

Get through (8) formula normalization:

$\mathrm{sx}=\frac{2.346-1.803}{2.688-1.803}(0.8-0.2)+0.2=0.5681$

Through the neural network model between sx and the sy, promptly formula (10), (11), (12), (13) get sy=0.3215

Get the predicted value of 4-CBA content in the TA product through the anti-normalization of (14) formula:

${\tilde{y}}_{4-\mathrm{CBA}}=0.13+\frac{0.42-0.13}{0.8-0.2}(0.3215-0.2)=0.189\%$

Above-mentioned through case description; Based on 4-CBA content soft instrument in the TA product, through the oxygen analysis appearance that gets into the reactor air record that oxygen content in the air

reactor tail gas oxygen analysis appearance records that tail oxygen content

and reactor tail gas carbon monoxide content analyser and carbon dioxide content analyser record that hydrocarbon content in the tail gas

is real-time, 4-CBA content in the on-line prediction TA product.

Claims (2)

1. confirm flexible measurement method for one kind, it is characterized in that this flexible measurement method comprises following steps carboxyl benzaldehyde content:

At first, with hydrocarbon content in the tail oxygen content of the oxygen content of the air that gets into oxidation reactor or oxygen-enriched air, reactor tail gas and the tail gas as the independent variable of soft instrument, at this independent variable of robotization treating apparatus input;

Then, to be dependent variable to carboxyl benzaldehyde content in the terephthalate product, based on independent variable, in, the real-time estimate terephthalate product online through soft instrument to carboxyl benzaldehyde content y _4-CBA

Said independent variable and dependent variable are the neural network models of non-linear correlation;

Through following formula (1) hydrocarbon content in the tail gas is modified to the content under the standard condition:

$x_{{\mathrm{CO}}_x}^{2,\mathrm{II}}=\frac{x_{{\mathrm{CO}}_x}^2(x_{O_2}^{s,1}-x_{O_2}^{s,2})}{(x_{O_2}^{s,1}-x_{O_2}^2)+(x_{O_2}^1-x_{O_2}^{s,1})\frac{100\%-x_{O_2}^2-x_{{\mathrm{CO}}_x}^2}{100\%-x_{O_2}^1}}---\left(1\right)$

Wherein,

---oxygen content in the air, %

---tail oxygen content, %

---hydrocarbon content in the tail gas, %

---standard condition gets into the oxygen content of the air of oxidation reactor, % down

---the tail oxygen content of reactor tail gas under the standard condition, %

---the modified value of hydrocarbon content in the tail gas under the standard condition, %;

Through following formula (2) said being carried out normalization handles:

$\mathrm{sx}=\frac{x_{{\mathrm{CO}}_x}^{2,\mathrm{II}}-x_{\min }}{x_{\max }-x_{\min }}(b-a)+a;---\left(2\right)$

Sx---to the value after

normalization processing;

[x _Min, x _Max]---collect

Variation range;

[a, b]---the variation range of independent variable after the normalization;

Through following formula (3) to said y _4-CBACarrying out normalization handles:

$\mathrm{sy}=\frac{y_{4-\mathrm{CBA}}-y_{\min }}{y_{\max }-y_{\min }}(b-a)+a;---\left(3\right)$

y _4-CBA---in the terephthalate product to the content of carboxyl benzaldehyde;

Sy---to y _4-CBAValue after normalization is handled;

[y _Min, y _Max]---the y that collects _4-CBAVariation range;

Through gathering industrial process data, adopt nerual network technique to set up the correlation model between sx and the sy;

The output sy of said neural network model tries to achieve y through the anti-normalization of following formula (4) _4-CBAPredicted value

${\tilde{y}}_{4-\mathrm{CBA}}=y_{\mathrm{Min}}+\frac{y_{\mathrm{Max}}-y_{\mathrm{Min}}}{b-a}(\mathrm{Sy}-a)---\left(4\right).$

2. flexible measurement method according to claim 2 is characterized in that, the input layer of said neural network model has 1 node, and hidden layer has 1～30 node, and output layer has 1 node.

Images