CA2066533A1

CA2066533A1 - Production of ethylene/vinyl acetate copolymers

Info

Publication number: CA2066533A1
Application number: CA 2066533
Authority: CA
Inventors: Franz Ferdinand Rhiel; Helmut Fischer; Gunther Weymans; Dieter Kuhlmann; Erhard Asch; Rudolf Plummer
Original assignee: Individual
Current assignee: Bayer AG
Priority date: 1991-04-24
Filing date: 1992-04-21
Publication date: 1992-10-25
Also published as: JPH05148307A; DE4113291A1; EP0510478A1

Abstract

Production of ethylene/vinyl acetate copolymers A b s t r a c t The present invention relates to a process for the production of polymers, preferably ethylene/vinyl acetate copolymers, in reactors with a special control system. By the process according to the invention, polymers of uniform quality and free from specks can be produced.

Le A 28 167-Foreign Countries 1

Description

BAYER AXTIENGESET~r~SCHAFT 5090 Leverkusen, Bayerwerk ~roup Management RP
Group Patents Ki/bo/c36 22.4.91 Production of ethylene/vinyl acetate copolymers The invention starts from a process for the quasi-continuous production of polymers, especially ethylene/vinyl acetate copolymers with vinyl acetate contents of 20 to 98 weight % in tert-butanol at temperatures between 3~ C and 15~ C and pressures of 100 bar to 1000 bar, in which th~ feed materials are metered continu~usly ko a reactor or reactor cascade and the polymer formed i~ discharged intermittently from a let-down vessel coupled downstream, and in which the pres~ur changes caused by the discharge propagate into the reactor or reactor cascade.

In the production of polymers by exothermic reactions at elevated pressures and tempPratures and at long raaction t.imes r in many o cases a cascade of 2 to 7 autoclaves in which the polymerization takes place is used. The product quality depends to a high degree on the temperatures at which the series-coup~ed autoclaves are operated during the polymerization. A faultless temperature control in the autoclaves is crucial for a constant, high level 2$ of quality. ~here has ~een no lack of previou3 attempts to meet ~hese quality conditio~s.
,, .
: In the first place, it has been attempted by means oE special chemical r~actions to compensate the pressure or temperature fluctuations caused by the production process or to make the product insensitive during the course of the reaction to such luctuations. Thus for example DE-l 126 613 describes for , ethylenejvinyl acetate copol~mers a process for the production i of these mixed polymers, which are produced in homogeneous liquid phase at special proportions of tert-butanol and monom~rs in the ~ f~ed.

:~ Le A 28 167 2 .~....

2~6~

According to DE-3 815 946, these copolymers are produced in improved (gel-free) for~ by meeting specific pressure-temperature minimum values, preferably over the whole course of the polymerization.

According to DE~3 731 054 such copolymers are produced by feeding special polymerization initiator~ and using speci~ic polymerization temperatures so that the reaction preferably takes place in the first part of the autoclave cascade.
In general the probl~m of constant control of reaction in the production of polymers relates to all polymerization, ; polycondensation and polyaddition processes in which process-determined pressure or temperature fluctuations and pressure or temperature disturbances influence the constancy of the polymer : guality.

These are ~or example polymexizations o~ vinyl chloride in emulsions (DE-3 627 287), the production of polycarbonates by the two phase interfacial process at ele~ated pressures and temperatures (phosgene pressure above 5 bar, temperatures above 5~ C), the production processes known in principle for polymers from the class of polybutadienes, polyolefin~, polystyrenes, . polyhaloolefins, polyvinyls, polyethers, polyacrylates, ; 25 poly(alkyl methacrylates)l polydienes, polyesters, polyestercarbonates, poiyamides, polyarylenes such as e.g.
polya~ylene sulphides, polyarylene ketones and polyarylene sulphones. Furthex polymer~ for which production processes at elevated temperatures and pressures are known are polyketones, polysulphones, polyimides, polyetheresters, and polysiloxanes.

There has been no lack o~ attempts to compensate the temperature : and pressure fluctuations by conventional control of the process.
~ But conventional control conceptæ generally have the disadvantage ; 1 3 5 o~ detecting the process trend too late and with that no longer ~;' enabling intervention in the production process in good time by : means of manipulated variables.
~., ~ Le A 28 167 3 - . . .
. , .
. :~

2 0 ~ 6; 3 t3 ~

In DE-3 627 287, therefore, a process is described in which, especially ~or the continuous production of poly(vinyl chloride), th~ reactor internal temperature i6 controlled by the amount of cooling water, and the amount of activator solution fed by the cooling water recirculation temperature. The polymerization temperature is thereby held constant at a fixed value and the polymerization pressure is always bel~w the satuxation pressure of the vinyl chloride and varied hy fine control o~ the amount of ~iquid vinyl chloride fed.
lû
: This type of control is preferred when the pressure in the polymerization vessel is controlled automatically by changing the feed rate of the vi~yl chloride and/or of the activato:r. But, especially for reactor cascades and there in particular for mixed polymers, this has the disadvantages that the chemical composition o~ the product can be altered by alteration of the proportioning, the pol~mer~ can acguire inhomogeneous molecular weight distributions, and the throughput of the plant cannot always be chosen optimally according to the current operating condition Further reasons for pressure drops in such autoclave reaction~
.~ can be: evacuating the last reactor of the cascade into a let-~` down vessel by a let-down ~ystem that leads to automatically pulsating pressure drops; and manual depressurizations for purging tubing. Also temperature fluctuations arise through temperature peake as a result of polymer agglomerates.

For reasons o~ safety, it has also freguently happened previous~y 30 that manual pressure reductions have been necessary in the course of the process in order to remove excess heats o~ reactiQn in the form o~ heats o~ vaporization.

These fluctuations have previously led to polymers produced by the prior art n euch autr3clave cascades ~orming irreversible microgels which unfavourably affected the properties of the polymers.

Le A 28 167 4 1 ~

r ~

2~3~3 Measures for process control of polymer production processes at elevated pressures a~d temperatures which remove this disadvantage of previous modes of operation are not yet known.
Improved process operation must make it possible to detect changes in ths reaction so early that, without changing the metering o~ the amounts ~ed, an intervention can be made with regard to the temperaturP of the cooling water or its amount, so that the overshooting of given temperature limits and especially a (manual) depressurization which may otherwise be necessary for ; 10 safety reasons are avoided.

Starting out from a process for the production of polymers in ~ which the feed materials are continuously metered into a reactor; or reactor cascade and the polymer formed is d:ischarged intermittently in a fixed cycle from a let-down vessel coupled downstream and the pressure chan~es caused by the discharge propagate in the reactor or reactor ca~cade, a special process control method has been developed, by which according to the invention the temperature in the reactor or reactor aascade is held aonstant by means of the followin~ algorithm:

a) In~roduction o~ a pressure-compensated temperature T~
:~ maXi~g allowance for the actual plant-determined pressure change~ according to the rule (i) ~*(t) = T(t) - b(p(t)-p(t)*) ~ ~T

valid in the linear region of the pressure build-up up to the transition from pressure build-up to pressure decay, the controller always receivlng another signal in place of the pressure-compensated tamperature (T*) when ~ (I) the absolute value of dT*/dt exceeds } Kelvin/min. cr ; .1 (II) the absolute valua of (p - p*) exceeds the pressure ;~l changes caused by evacuation of the product.

,` 1 Le A 28 167 5 Y~

~, , , ~ , ,, : , , ' ' ' ~s' ~

,~,.,.~,, ~ O ~ ~ ~ ~ 3 Here T*(t) is the pressure-compensated temperature at time t; ~ T is a constant general correction value, independent of time, which is determined once for each apparatus by fitting and is usually between 0.01 and 2 Kelvin; T(t) is the temperature at the reactor measured with a temperature sensor at time t; p(t) is the pressure measured at time t, preferably in proximity to the temperature measuring point;
p*(t) is the time-averaged pressure up to time t at the pressure measuring point; b is a coefficient dependent on the chemical composition o~ the substances in the reactor, which $s experimentally detPrmined once ~or the reactor and the given substance system by the following steps before starting the control:

.
- measurement of pressure and temperature over time :
- formation of the pressure-temparature diagram by elimination of the time - b is the ascending gradient in the linear region of the pxessure build-up o~ the diagram.
~: ' b) Smoothing of this temperature aignal T*(t~ by known ~; methods.
~ 25 ;~ c) ~sing this smoothed signal as controlled variable for the adjustment of the inlet temperature o~ the cooling water and/or the amount of the cooling water, so that the reactor temperature T is maintained at the preset target ~alue.
Th~ term "pressure-compensated temperature" is defined by rule ~ (i). The control can be used both for a single reactor and for ,~ several or preferably all reactors of a cascade. Possible controllers are analogue or digital PID controllers or other methods of control technology such as e.g. adaptive controllers.

The method according to the invention is usually applied after ~,~ Le A 28 167 6 ,: .

,::
,:: .
~, ,. ! , ' ., .,.' :

,7J ~

a certain start-up period. ~p to then pressure and temperature have been measured in close proximity on the reactor at suitable intervals - e.g. 15-second intervals. From the pressure values, the arithmetic mean value p*(tl) is formed at the start of the control, where tl is the time at which the control according to the invention star~s. In averaging and tracking the pressure, pxeferably only those pressures are taken into account which fall within the pressure fluctuation range determined ~y the evacuation system at the end of the autoclave system. It has proved to be particularly advantageous in the averaging not to take into aacount also pressure fluctuations caused by p~rg~ng tubing, drops owing to compressor failure etc. For continuous determination of the pressure-compensated temperature T*, the substance and system-dependent coeffici~nt b must be determined once, before the control is started. To this ~nd, on the existing react~r system with usually varying pressures and as a result varying temperatures, temperature and pressure are measured as a function of time during operation over a period of at least one measurement interval and the quantities pressure and temperature plotted against each other by eliminating the time variables. ~ is determined as the ascending gradient in the linear region of the pressure build-up of the pressure-temperature curve.

The temperature is preferably measured at the reactor exit. All possible points for the given reactor are, however, conceivable.
The pressure physically associated with this temperature is the pressure which the polymer mass has in the neighbourhood of the temperature measuring point. For this reason the reaction pressures p(t) and temperatures T~t) necessary for the control are preferably ~easured ad;acently.
I' The control system according to the invention is preferably put int~ operation when the absolute value of the derivative with ,, ~ 35 respeat to time of the pressure~compensated temperature is less ,~; than 0.25 Kelvin/mlnute. Furthermoxe, the method is preferably used when the pressure di~ference& between the time-averaged ":,.~ ' ~ ~ Le A 28 1~7 7 :' .','' ~
,:"

::,. :

2 ~ 6 i6 3 ~5 ~

p*~t) and p(t) do not exceed 70 bar, and pre~erably do not exceed 50 bar. Duxing pressure fluctuations which occur as a result of the depressurization of a reactor ~or the purging of tubing, as well as during pressure upsets, such as e.g. comprei~sor failure, for the period of this upset the process trend derived from the previous pressure-compensat d temperature changes is continued ~: and the control according to the invention by rule (i) suspended.

In accordance with a variant Df the method according to the invention, the contr~l system can als~ only be put into operation when the actual pressure p(t) falls during pressure build-up in a range whose upper limit is formed ~y the last actual value o~
the pressure during the pressure build-up and whose magnitude is : 15 90 %, preferably 70 %, of the amplitude of the pressure . ~ fluctuations caused by the intermittent depressurization system.

~oth the amount of cooling water and the cooling water inlet ~: temperature, or both simultaneously, can be used as the ~ 20 manipulated variable. The preferred manipulated vaxiable is the .~ cooling water inlet temperature.

Outside the range o~ validity o~ (i), the controller can be gi~en normal in6tructions according to the prior art~
~:~ 25 . ~ Preferably the control aaaording to ~escription a) is also supported by the additional step that a) the pressure-compensated temperature according to (i) is introduced, ~' ' a') in case of the time-limited invalidity of control condition ~i , (i), there is defined l 35 (ii) T*(t) = T*(t ~ a(t ,~ j ~ , where c~t - 1) is a coefficient ~ormed by taking the mean of the .
.
.$ ~ Le A 28 167 8 ~,f~
' ' .

,^
`:f, ' `

variations with time of the preceding temperatures T*(t ~ i>o) and at the starting point of the reaction can be chosen at will between -0.7 and +0.7 Kelvin~minuta; and T*(t-l) is the las~ pressure-compensated temperature defined via (i~ or (ii) at time t 1. The time t-l means the time at which the last measurement of T and p takes place before the now current measurement at time t.

The control method according to the invention proceeds especially preferably when for the tra~sition of the control conditions of a') to a) the f~11QWing step is al~o taken into account:

If T* has been determined according to formula ~ii) at a time t' - t - 1, in the next application of formula (i) at time t, b is 1~ changed according to the rule:

~iii) b(new) - b(old) - d x (T~(t') -T*(b(old)lt))f(p~t) p*(t)) ` 20 and maintained until altered again according to this rule for the ;' later time o~ valtdity of (i), d being a damping ~actor betwsen :~ 0.00001 and 1 which can be chosen at will.

.l The smoothing b3 o~ the time-dependent signal T*(t), determined according to rules (i) to (iii), can be carried out by any known mathematical method, especially by the method of exponential smoothing.

The method uses a recursive algorithm to determine the temperature gradient, in which the straight line of ascent is determined by application of the Gaussian law of the least error : sguares method. ~ompared with the e~ual weighting of the series of measurement values which is usual in regression analysis, in this preferred me~hod ~he weighting of preceding values i~
carried out according to a geometrical series. This means that ; the last-measured values enter into the calculation of the :' ~
derivative of temperature with respect to time with a higher ~' Le A 28 167 9 .`' ~
: !

weighting than the values measured earlier. It is possible via an additional damping factor to adjust how strongly the values from the past shall be weighted.
The smooth signal T* transmits to the controller the necessary information for the alteration of the manipulated variables. It is essential that the controller has a differential component.
The derivative which appears in the control algorithm, matched to the reactor concerned, is formed by a known method from the smoothed T* signal. The derivative of the smoothed signal with respect to time is a measure of the process trend.
The method according to the invention is preperably used for carrying out the reaction of copolymerization of ethylene and vinyl acetate, here in particular at vinyl acetate contents of 20 to 98 weight %, at temperatures between 30°C and 150°C and pressures of 100 bar to 1000 bar, the mean pressures in the reactors being chosen, at least briefly and preferably over the whole course of the reaction, according to the following formula, which depends on the material composition: P = 1.1 x (-204.2 +
9.12 x (E)/(VA) + 44 x (E) x (E)/((VA) x (VA)) + 1.12 x T) bar, in which (E) is the amount of copolymerization ethylene in the final product in weight %, (VA) is the vinyl acetate content in the final product in weight % and T is the polymerization temperature in °C. Furthermore the method can be used specifically for ethylene/vinyl acetate copolymers if 0.02 to 1.5 weight % of polymerization initiator - relative to copolymerizable monomers -is used in the polymerization and the polymerization initiator has a decomposition half-life at 70°C of 30 minutes to 6 hours and the polymerization temperatures in the reactors are so chosen that at least 50 % of the initiator decomposes in the first half of the cascade volume.
For most reactors used in commerical production the ratio of reactor surface to reactor volume is small. That means for example that a temperature rise in the neighbourhood of the operating point leads to an increase in the heat of reaction 2~53 ~ ~

which is greater than the increase of the heat dissipated in the cooling medium as a result of the increased temperature gr~dient.
Heat ~f reaction and heat dissipated are accordingly no longer in equilibrium. From this there follows a further temperature rise, which leads to a still greater di~ference ~etween heat of reaction and dissipated heat and results in a renewed temperature rise. ~he operating po~nt o~ such reactors is therefore unstable. In con~rol, therefore, it ls important to detect change~ in the course o~ reaction and temperature as early as possible, in order that they can he counteracted in good time.
The control method according to the invention provides an important contribution with regard ~o the early detection of a change of reaction, preferably for reactors with unstable operating points.
Moreo~er the method according to the invention enables $he product guality to be improved, since temperature ~luctuations in the reaction which spoil the composition and uniformity of the .~
~ product are avoided.
! ~ 20 ~he invention is illustrated in more detail ~elow with re~erence to drawing~ and examples. The figures show the ~ollowing:
,: I
~ Figure 1 a process diagram, ,~, 25 Figure 2 a block diagram ~or the control system, ~igure 3 the reactor pressure p and the reactor temperature T as a function of time for Example 1, Figure 4 the reactor temperature T as a function of the reactor pressure p in the measuring process according to Figure 3, Figure 5 the rising straight line plotted in the linear part of ~ the T(p) diagram of Figure 4, ,~ ~ Figure 6 for Example 1, reactor pressure p, reactor temperature ~; 35 T, and the smoothed pressure-compensated temperature T* used as control signal, each as a function o~ time, Figure 7 for Exampla 2, reactor pressure p and reactor ..:
~ Le A 28 167 11 .
;, ,:, ~ ., ;, .
,, - , .' . - - ~;

:,.. , . , - ~ .

2 ~

temperature T as a Punction of time, Figure 8 the T(p) dependence found from the measurement according to Figure 7, Figure 9 the rising straight line plotted in the linear part of the T(p) diagram o~ Figure 8, Figure 10 ~or Example 2, reactor pressure p, reactor temperature T, and the smoothed pressure-compensated temperature T* used as control signal, each as a function of time.

According to Figure 1, feed materials reguired for the polymerization (monomers, solvent, initiator) are fed continuously from the tanks 1, 2 and via the line 3 by means of metering pumps 4, 5, 6 into a reactor cascade 8 consisting of six reactors 7 connected in ~eries. The polymer formed is collected i~ a let-down vessel 9. The let-down vessel 9 is intermittently evacuated at regular intervals of time (in a fixed cycle) by means of a discharge pump 10. This is thu~ a quasi-continuous production process. The pressure fluctuations arising as a result of the intermittent ~vacuation propagate through the whole reactor cascade and would, withQut control measures, lead to considerable temperature ~luctuations. All reactors 7 are -~ therefore water-cooled. In the previously usual conventional temperature control, the flow rates and/or temperatures of the ; cooling water can be adjusted in order to counteract temperature increases. It has not previously been possible, however, to satisfactorily eliminate temperature fluctuations in the reactors 7 caused ~y reaction, i.e. to run the polymerization reaction with stable operation over fairly long periods of time at a constant temperature.
For the6e reasons, in at least one reactor 7 a special temperature control is used, which is illustrated in detail with the aid of the following figures. Figure 2 shows the control circuit required ~or this, which is connected to Reactor 7. Fl and F2 denote the sensors for indicating the cooling water flow rates; and Tl, ~z and T3 denote the temperature sensors for the cooling water temperatures. Here F1, F2 and T1 relate to cooling ~ ~ Le A 28 167 12 ~
',: , ' .:

:, : , . , . . ., , ~ . , 2 ~ 3 ~

- water inflow lines 11, 12 and 13 and T2 and T3 to cooling water outflow lines 14 and 15. The temperature in the reactor is measured with the sensor T4 and the pressure with the pressure sensor P, All measurement values are fed to a computer 16, which forms the eontrol algorithm described below and delivers an output signal characteristic of the reactor temperature and the process trend, which is compared in the contr~ller 17 with a preset target value for the reactor temperature. The controller 17 is a conventional module with a differential component. The output signal of the controller 17 controls, via an intermediate controller 18, the flow rate for ~he steam heating 19 of the cooling water in~low lines 12, 13 and/or the c~oling water flow rate 20 through these lines~ A part of the cooliny water is recycled by means of the pump 21. The special control algorithm, 1~ on which the computer 16 is based, for generating the control signal from the reactor actual Yalues, is described ~ quantitatively below with the aid of Material Example~ 1 and 2.

:
.
` `:
., ~, . ~

:.
. ~

~j Le A 28 167 13 ,.~, ...'-:
.. ~ .

-.. : , . . .
:; . , , '.'`` ' ' :
.

_a~

A copolymer with a mean molecular weight of 570,000 g/mol and vinyl acetate content o~ 45 weight % has to be producedc The polymerization is carried out in the solvent tert butanol. 2,2'-azobisisobutyronitrile (trade name Porophor N), whose decomposition half life at 7~ C is about 6 hours, is used as the initiator. The polymerization must be run so that the product leaving the last reactor of the cascade has a solids content of lo about 47 %. The polymerization takes place according to Figure 1 in six reac~or~ connected in series, with capacities of 4, 6, 6, 6, 6, and 4 m3. In the xesulting total cascade volume of 32 m3 the mean product xesidence time is 5.4 hours. The throughput is about 1.2 t polymer/h. The last reactor o~ the cascade is evacuated into a let-down vessel according to Figure 1 via a let~
down system which leads to regular pres~ure fluctuations in the ; reactors. The polymerization should occur at a mean reactor pressure of 300 bar. This pressure is the sama for all reactors apart from the pressure losses in the connecting lin0s. The reaction should ~e carried out in such a way that the mean temperatures, measured in each case at the outlet o~ the reactors are 70, 75, 72, 75, 75 and 115~ C. The method vf the invention ~: in accordance with Figure 2 is applied according to Figure 1 on reactors 2 and 3. The remaining reactors are run conventionally.
In the foll~wing the application of the method of the invention i5 ill ustrated in more detail ta~ing reactor 3 as an example.
This reactor has a diameter of about 1 m and a height of about 7.6 m. In this reactor, with volume 6 m~, the product residence time is about one hour. The pressure-compensated temperature T*
is determined from the temperature T actually measured at the reactor exit and the "physically" associated pressure p, whose measuring point is adjacent to the temperature measuring point.

During the start~up period the mean reactor Pressure is determined. This is done by taking the arithmetic mean o~
maximum and minimum pressure during an oscillation cycle, in doing which care must be taken that these extreme values are Le A 28 167 14 . .
:
-- - ~ . - - , .

., . ~ .

~6 determined only ~rom the pressure ~luctuation range due to the evacuation ~ystem. In Figure 3, ~or exampla, the maximum value during a cycla is 318.96 bar and the minimum value 284.36 bar.
This gives a starting value for the mean pressure of (1) p*(t - 0) = (284.36 ~ 318.96)/2 = 301.66 ~ar It follows that the amplitude of the pressure oscillation ap is t2) ap = 318096 - 284.36 = 34.6 bar.

The subsequent adjustme~t with time of the mean pressure to small pressure changes (<10 bar) ~ollows in accordance with an ~ exponential smoothing of the ~irst order from the eguation : 15 (3) p*(t) = 0.00~ x p~t) ~ 0.002) x p*~t - 1), :~ the adaptation of p*(t) according to this rule only taking place when the actual pressure p(t) fall~ in the range p*(t~ 0.6 x ap ~ p(t) S p*(t-l) + 0.6 x ap.

, Qtherwisa (4) p*(t) = p*(t-l) applies.

If the mean pressure change is gxeater than 10 bar - caused for example by ch n~e of the operating condition - the mean pressure p*(t) must be determined anew, starting with eguation (1).

Apart from tha mean reactor pres~ure, the initial value of the coe~icient b for the substance system present in the reactor must be determined during the start-up period. In this case the - 35 measured values o~ pressure p~t) and temperature T~t) for a cycle ; - a~ shown in Figure 3 - must be recorded. By eliminating the .l time, the dlagram in Figure 4 i~ arrived at, in which the Le A 28 167 15 ;:
'-- ' -: , .
:
:::

~ 3 pressure is plotted as the abscissa and the accompanying temperature as the ordinate. In the diagram a linear region is de~ined by the following rule:

- only the pairs of measurements during the pressure build-up are chosen - the range is limited above by the last pair of measured values during the pressure build-up, and the size of this range is 75 % o~ the amplitude of the pressure fluctuation.

In the present example the linear region ~alls between p~ =
318.g6 bar and pu = 318.96 - 0.75 x (318.96 ~ 284.36) = 293.01 bar.
~5) Linear regi~n thus: 293.01 ~ p(t) ~ 318.96 bar For the measurement values falling in the linear region, the straight line vf the regression ~Figure 5) is d~termined. The ascending gradient of the regression line gives the initial value of the coef~icient b. In the pre~ent case, b has the value - (6) b 0.0291 Relvin/bar.
.:
After determination o~ the coef~icient b, the determination of the pressure-compensated temperature T* ~t) is started according to the following rule:
: ~ ~
(7~ T*(t) = T(t) - bx(p(~) - p~(t)) + ~T
valid in the linear region according to (5), as well as which in the present example:

~i) absolute value dT*/dt ~ 0.25 Kelvinfminute and ~ T - ~.1 Kelvin.
. ~
, ~, Le A 28 I67 16 .

:: . :, ,:

3 ~

At time t = 21:33:44 hours (Figure 3), the following values were measured:

actual pressure p(t) = 314.3g bar, actual temperature T(t) = 72.0~ c.

This pair of values falls in the linear region. Consequently at a mean pressure ~f p*(t) = 310.66 bar there follows:
(8) T~ ~t~ - 72.05 - 0.0291 x (314.39 - 301.66) + 0.1 = 71.7~ C

~here was detenmined at t - lOs = 21:33:34 hours (one time interval is 10 seconds):

T*(t - lOs) = 71.7~ C.

:~ From this it follows that dT~/dt - (71.78 - 71.77~/10 x 60 - 0.06 Kelvin/min. The requirement (ii) is accordingly met, and it remains true that ~; 20 t) = 71~7~ C.
~, The measured temperature values are smoothed, a~d the derivative of temperature with r~spect to time determined, ~y the method of : : 25 exponential smoothing of the second order according to the following rule:

;~ (9) T*~l)(tj z ~ ~ T*~t) + (1 - ~) x T*(l)(t - 1) `~. (10) T*(2)~t) - a x T*tl?(t) + (1 - a) x T*~2)(t - 1) ; ~ 30 ~11] C~t) 3 ~ ) X ~T*~ t) - T*~2)~t)), :~ whare is the smoothing constant T~l) is the once-cmoothed temperature signal . 35 T~(2) is th~ twice-smoothed temperatur~ signal and c is the derivative of the temperature with respect to time.

Le A 28 167 17 , .

.

,.~

In the present case, the smoothing constant is chosen according to:

~ 12) ~ = ~.03.

At 21:33:34 hours/ thP following were the calculated values:
T~ (t - lOs) = 71.77~ C
T*(2)(t - lOs) = 71.78~ C

If these values are used in equations (9) to (11), taking ~12 into account, it ~oll~ws that:

T*~1~(t) = 0.03 x 71.78 + ~1 - 0.033 x 71.776 = 71.776r C
T*~2)(t) = 0.03 x 71.7761 + (1 - 0.03) x 71.78~ = 71.785 c(t~ o 0.03j~1 - 0.03) x (71.7761 - 71.7857) = -0.0002502 Kelvin/lOs = -0.0015 Kelvin/min (averaged derivative with respect to time of the pressure-compensated ~0 temperature) This algori~hm is continued up to the last time of the linear - region (= highest value of the pressure during the pressure : 25 build-up) t + 30s 5 21:34:14 hours. At this time the following v~lues were measured:

: p(t + 30s~ - 318.96 bar ~t ~ 30s) = 72.21C
The following existed ag calculated values:
~; ;;
: p*(t ~ 30s) - 301.66 bar .
T*tt + 30s) c 71.7~ C
.
T*~1)(t + 30s) = 71.78~ C

~e A 28 167 18 : !
'"' '' r~

2 (16 6 ~ ~3 T*(2)(t + 30s) = 71.78~ C

c(t + 30s) - -0.0000928 Kelvin/lOs For the subsequent times which do not fall in the linear region, the following now applies:

(13) T*~t) = T*~t ~ c(t - 1) (14) T*~l)(t) = ~ x T*(t) ~ a) x T*~(t - 1) ~' .
(15) T*~ t) - T~1)(t) c(t - 1) x (1 - a~/a (1~) c(t) = c(t - 1) For the subseguent time t~40~ = 21:34:24 hours (i.e. 10 seconds later), the following numerical values are obtained correspondingly:

T*(t ~ 40s) - 71.79 - 0.0000928 K/lOs x lOs ~ 71.7~ C
.
T*~1)(t ~ 4~s) = 0.03 x 71.79 ~ 0.03) x 71.78 = 71.7803~ ~:
. . ~
~ i , ....
T*~2)(t ~ 40s) = 71.7803 - (-0.0000928 K/lOs x lOs~ x 0.03)/0.03 = 71.783~ C
!; ~
-~ c(t ~ 40s) - -0.0000928 K/lOs This calculation is continued up to the last time before the beginning of the linear region. At this time t~llOs = 21:35:34 hours, the following values were obtained:
, T*~t + llOs~ = 71.7~ C
,: I
~ ' T*~l)(t ~ l~Os) = 71.7B2~ C

~ !
I Le A 28 167 19 ~ . 1L.

`1`: , " ~ , ' ~6~'3 T*t2)(t + 1105) - 71.785~ C

With the beginning of the subsequent linear region at the time t ~ 120s = 21:35:44 hours, the ~ollowing wer~ measured:
: 5 T(t + 120s) = 71.5~ C

p(t + 120s) = 295.0S bar.

From this there follows, applying eguations ~8) to (12) and using p*(t + 120s) = 301.66 bar:

T*(t + 120s~ ~ 71.50 - 0.0291 x (295.05 - 301~66) + 0.1 = 7~.7~2~ C
T*~l~(t ~ 120s) = 0.03 x 71.7924 + ~1 - 0~03) x 71.7822 = 71.782~ C

T*~2)~t ~ 120s) = 0.03 x 71.7~25 ~ 0.03) x 71.7852 : = 71.7851 C

. c~t ~ 120s) = 0.03/(1 - OtO3) X ~71.7825 - 71.7851) = -0.0000804 Kel~in/lOs ~: At ~he transition to the linPar region the ~alue of the coefficient is altered according to the foll~wing rule:

~: b~new) - b~ld) - d x ~T*(t 1- llOs) - T*~t + 120s))/
~ptt ~ 12~s) :- p*(t ~ 120~)3 d is set in the present example according to ~17~ d = 0.01.
Thus there is obtained:

:
~ Le A 28 167 20 , .

.

- ~: . . :
, ' : ' ' ~ . ,:~

,. , , : .

Ib'~ 3 ~

b(new) =
0.0291 - 0~01 x (71.79 - 71.7924)~(295.05 - 301.66) = 0.029096 ~ 0.0291 Kelvin/bar b(new) = b(old~ = 0.0291 Kelvin/bar remains vali~ until the next transition into the linear region.

Figure 6 shows~ over a ~airly long psriod of time, the course of the pressure p, temperature T, and first smoothed value of the pressure-compensated temp~rature ~*tl~. ~he value of T*(1) is fed to the computer as th~ actual value of the controlled variable (Figure 2)~
;~- 15 ~esult:

The product had a clearly improved ~uality in accorda~ce with the specifications mentioned over the whole period of action of the ~: control method according to the inve~tion.
20:
, In this example, a copolymer with a mean molecular weight of ~2S 210,000 and vinyl acetate con~ent of 68 weight % has to be :~ produced. ~ere tert-butanol is used again as the solvent and 2,2'-azobisisobutyronitrile as the initiator. The polymerization ~:~: must be run so that the product lea~ing the last reactor of the cascade has a solid content of 45 %. The poly~erization takes place in four reactors connected in seriesf with capacities of 4, 6, 6, and 4 m3. This corresponds to a cascade volume of 20 m3. The mean product residence time is 4.3 hours. The ;~ throughput i~ 1.1 t polymer/h. The same let-down system is used or evacuation of:the last reactor of the cascade as in Example ~ 35 1. The polymerization should occur at a mean reactor pressure :: o~ 295 bar. As in Example 1, this pressure is the same for all ~ reactors apart from the pressure losses in the connecting lines.
~`
: Le A 28 167 21 ,, 1 ,~'`'~',:
.
- , .

:.: :
, ~ ~ . . .
: ~ . . - .

2 ~ 3 3 The reaction ~hould be carried out in such a way that the mean temperatures, measured in each case at the outlet of the reactors, are 69, 70, 71, and 115 C. The method of the invention in accordance with Figure 2 is applied on reactor 2.

5 This reactor has a diameter of about 1 m and a height of about 7.6 m. The reactor volume i~ accordingly 6 m3. The product residence time in this reactor, assuming the above values, is 1.3 h. As in the case of Example 1, the pressure-compensated temperature T* is determined from the temperature T actually measured at the reactor exit and the "physically" associated pressure p, whose measuring point is adjacent to the temperature measuring point.

The procedure in tha calculation of the pressure-compensated temperature corresponds to that in Example 1, so that it is no longer neressary in the following to give in detail all equations contained in Example 1. Only the most important points of the - calculation of the pressure-compensated temperature are considered.
First the pressure and temperature measurements for an oscillation cycle are plotted. From the course of the pre~sure the mean pres~ure an~ the amplitude of the pressure fluctuation are determined. By elimination o~ the time one obtains the p-T
diagram. For the linear region the ~traight r~gression line is determined. The ascending gradient ~f the regression line yives - the value of the coefficient b. For the coefficients a and d the same values as in Example 1 are chosen.

The calculation of the pressure-compensated temperature T* and the smoothed values T*(l~ and T*~2~ is carried out analogously to Example 1. As in Example 1, the value T*(l~ is ~ed to the controller as the actual value of the controlled variable.

Similarly to Example 1, it turned out that here also the product had a clearly improved ~uality in accordance with the pecifications mentioned over the whole period of action of the Le A 28 167 22 ` ' , .
,~ j . : . ~ . . .
" - ~ . ' ' .

~ ~ . . . . . . . .

. , , ,, . ~ .

f-;J~

I control method according to the invention.

' ,~

~ .
, '~

~, ~ ` .
`"' ~
:,`, .
~; Le A 28 167 23 :`' . .
.i :
.., :.:

, !

Claims

1. A process for the production of polymers in which the feed materials are continuously metered into a reactor or reactor cascade and the polymer formed is discharged intermittently in a fixed cycle from a let-down vessel coupled downstream and the pressure changes caused by the discharge propagate in the reactor or reactor cascade, characterized in that the temperature in the reactor or reactor cascade is held constant by means of a control system with the following algorithm:
a) Introduction of a pressure-compensated temperature T*
making allowance for the actual plant- determined pressure changes according to the rule (i) T*(t) = T(t) - b(p(t)-p(t)*) + .DELTA.T

valid in the linear region of the pressure build- up up to the transition from pressure build-up to pressure decay, while the controller always receives another signal in place of the pressure-compensated temperature (i) when (I) the absolute value of dT*/dt exceeds 1 Kelvin/min or (II) the absolute value of (p - p*) exceeds the pressure changes caused by evacuation of the product, where T*(t) is the pressure-compensated temperature at time t;

.DELTA.T is a constant general correction value, independent of time, which is determined once for each apparatus by fitting and is between 0.01 and 2 Kelvin;

Le A 28 167 24 T(t) is the temperature at the reactor measured with a temperature sensor at time t;

p(t) is the pressure measured at time t, preferably in proximity to the temperature measuring point;

p*(t) is the time-averaged pressure up to time t at the pressure measuring point;

b is a coefficient dependent on the chemical composition of the substances in the reactor, which is experimentally determined once for the reactor and the given substance system before the start of the control by the following steps:

- measurement of pressure and temperature over time - formation of the pressure-temperature diagram by elimination of the time - b is the ascending gradient in the linear region of the pressure build-up in the diagram.

b) Smoothing of this temperature signal T*(t) by known methods.

c) Use of this smoothed signal as control signal for the controlled variables.

2. A process according to Claim 1, characterized in that the control according to rule (i) is supplemented by the rule a') in case of the time-limited invalidity of rule (i), there is defined:
(ii) T*(t) = T*(t - 1) + c(t - 1), where c(t - 1) is a coefficient formed by averaging the variations with time of the preceding temperatures T*(t -Le A 28 167 25 i) (i>0) and at the starting point of the reaction can be chosen at will between -0.7 and +0.7 Kelvin/minute, and T*(t-1) is the last pressure-compensated temperature defined either via (i) or according to (ii) at time t-1.

3. Process according to Claim 2, according to which the control rule is in addition supplemented by the step:
If, at a time t' = t - 1, T* has been determined by formula (II), at the next application of formula (i) at time t, b is changed according to the rule (iii) b(new) = b(old) - d x (T*(t') - T*(b(old),t))/
(p(t) - p*(t)) and maintained in accordance with this rule until again altered for the later times of validity of (i), d being a damping factor between 0.00001 and 1.

4. Process according to Claims 1-3, characterized in that the reactors are used for the production of copolymers from vinyl acetate and ethylene.

Le A 28 167 26