Method for controlling a rectification column

Abstract

The present invention relates to a method of controlling a concentration of a first component of a rectification column for separating a binary mixture of the first component with a second component on the basis of temperature measurements, wherein a control path defined by temperature sensors (T3, T2, T6) arranged in the longitudinal direction of the column is linearized with the aid of an estimated temperature profile, wherein a real temperature profile T*(h), determined by means of the temperature sensors, is approximated by a function T(h) in dependence on a column height h, wherein the column id divided into two sections along the column height h and the function T(h) is defined section by section on the basis, in each case, of a logistical function.

Claims

1. A method of controlling a concentration of at least one first component in a rectification column for separation of a binary mixture of the first component with a second component based on temperature measurements, the method comprising: (a) linearizing a control zone defined by temperature sensors arranged in the longitudinal direction of the rectification column with the aid of an estimate of a temperature profile, and (b) approximating a real temperature progression T*(h) determined by means of the temperature sensors by a function T(h) as a function of a column height h, wherein: (i) the column is divided into two sections over the column height h, and (ii) the function T(h) is defined in sections on the basis of one logistic function in each section.

2. The method of claim 1, in which the logistic functions are formed on the basis of a logistic function of the following form: $f\left(x\right)=\frac{e^x}{1+e^x}.$

3. The method of claim 1, wherein: $T\left(h\right)=\{\begin{array}{cc}{T}_{\mathrm{ab}}\left(h\right),& h<h_{\mathrm{feed}}\\ T_v\left(h\right),& h\ge h_{\mathrm{feed}}\end{array}$ where T.sub.ab(h) and T.sub.v(h) are defined on the basis of one logistic function each and h.sub.feed defines the column height at which a solution consisting of 40% by volume to 60% by volume of the first component and 40% by volume to 60% by volume of the second component is fed to the column.

4. The method of claim 3, in which: $\begin{array}{c}{T}_v\left(h\right)=T_{0,v}+v_v.Math.\frac{e^{\left(h-h_{0,v}\right)k_v}}{1+e^{\left(h-h_{0,v}\right)k_v}}\\ T_{\mathrm{ab}}\left(h\right)=T_{0,\mathrm{ab}}+v_{\mathrm{ab}}.Math.\frac{e^{\left(h-h_{0,\mathrm{ab}}\right)k_{\mathrm{ab}}}}{1+e^{\left(h-h_{0,\mathrm{ab}}\right)k_{\mathrm{ab}}}}\end{array}$ where T is normalized to boiling temperatures T_1 and T_2 of the corresponding first and second components, with 0%=T_1 and 100%=T_2 and h is normalized to an absolute column height H and T(h)=T.sub.v(h) for h ϵ [x %, 100%] in a rectifying section v, T(h)=T.sub.ab(h) for h ϵ [0%, x %] in a stripping section ab of the column height h with 0<x<100, and where T.sub.0,v and T.sub.0,ab are respective support vectors for the temperature, v.sub.v and v.sub.ab are a respective range for the temperature, k.sub.v and k.sub.ab are a respective shrinkage in the direction of the column height h, and h.sub.0,v and h.sub.0,ab are a respective support vector of the height h.

5. The method of claim 4, wherein x is 24.

6. The method of claim 4, in which the respective shrinkage k.sub.v or k.sub.ab in the direction of the respective height is ascertained from a comparison of a slope of the logistic function f(x) at its inflection point and the measured temperature progression T*(h) at its inflection point, wherein the inflection point of T*(h) is calculated from an average of respective slopes at a respective inflection point of temperature progressions ascertained from disturbance tests.

7. The method of claim 6, in which h.sub.0,v and h.sub.0,ab are each determined by a parameter estimate.

8. The method of claim 7, in which the Gauss-Newton method is employed for estimation.

9. The method of claim 4, in which a value range of T.sub.v(h) is between the boiling point of the first component and the feed temperature (T5) and hence T.sub.0,v(h)=0 and v.sub.v=T.sub.feed and a value range of T.sub.ab (h) is between the feed temperature (T5) and the boiling point of the second component and hence T.sub.0,ab(h)=T.sub.feed and v.sub.ab=1−T.sub.feed, where 1 corresponds to the boiling point of the second component.

10. The method of claim 4, in which a value range of T.sub.v (h) is between the boiling point of the first component and the feed temperature (T5) and hence T.sub.0,v(h)=0 and v.sub.v=T.sub.feed.

11. The method of claim 3, in which: $T_v\left(h\right)=T_{0,v}+v_v.Math.\frac{e^{\left(h-h_{0,v}\right)k_v}}{1+e^{\left(h-h_{0,v}\right)k_v}}$ $T_{\mathrm{ab}}\left(h\right)=T_{0,\mathrm{ab}}+v_{\mathrm{ab}}.Math.\frac{e^{\left(h-h_{0,\mathrm{ab}}\right)k_{\mathrm{ab}}}}{1+e^{\left(h-h_{0,\mathrm{ab}}\right)k_{\mathrm{ab}}}}$ where T is normalized to boiling temperatures T_1 and T_2 of the corresponding first and second components, with 0%=T_1 and 100%=T_2 and h is normalized to an absolute column height H and T(h)=T.sub.v(h) for h ϵ [x %, 100%] in a rectifying section v, T(h)=T.sub.ab(h) for h ϵ [0%, x %] in a stripping section ab of the column height h with 0<x<100, and where T.sub.0,v and T.sub.0,ab are respective support vectors for the temperature, v.sub.v and v.sub.ab are a respective range for the temperature, k.sub.v and k.sub.ab are a respective shrinkage in the direction of the column height h, and h.sub.0,v and h.sub.0,ab are a respective support vector of the height h, and in which a transition of the logistic functions is determined by a temperature (T5) at a feed to the column.

12. The method of claim 11, in which a value range of T.sub.v(h) is between the boiling point of the first component and the feed temperature (T5) and hence T.sub.0,v(h)=0 and v.sub.v=T.sub.feed.

13. The method of claim 11, in which a value range of T.sub.ab(h) is between the feed temperature (T5) and the boiling point of the second component and hence T.sub.0,ab(h)=T.sub.feed and v.sub.ab=1−T.sub.feed, where 1 corresponds to the boiling point of the second component.

14. The method of claim 1, in which a transition of the logistic functions is determined by a temperature (T5) at a feed to the column.

15. The method as claimed of claim 1, in which the first component comprises 1,2-dichlorobenzene (ODB) and the second component comprises COCl.sub.2 (phosgene).

16. A system comprising a rectification column that executes the method of claim 1.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 is a schematic diagram of a rectification column that can be controlled in one embodiment of the method of the invention; and

(2) FIG. 2 shows illustrative temperature profiles of a rectification column after performance of disturbance tests, in each case by alteration of a coolant rate in each case as abruptly altered manipulated variable.

DETAILED DESCRIPTION

(3) The present invention relates to a method of controlling a concentration of at least one first component in a rectification column for separation of a binary mixture of the first component and a second component based on temperature measurements. In this method, a control zone defined by temperature sensors arranged in longitudinal direction of the column is linearized with the aid of an estimate of a temperature profile, wherein a real temperature progression T*(h) determined by means of the temperature sensors is approximated by a function T(h) as a function of a column height h, wherein the column is divided into two sections over the column height h and the function T(h) is defined in sections on the basis of one logistic function in each section.

(4) In the approximation of the temperature progression, it is also possible to determine or estimate a position of a mass transfer zone. The position of a mass transfer zone in turn serves as controlled variable in order to establish a desired position of the mass transfer zone connected to a desired product concentration. There is a defined relationship between product concentration and mass transfer zone, which has to be determined separately taking account of the substances used as components of the binary mixture. Product concentration may be understood to mean the concentration of the first component (for example top product), the concentration of the second components (for example bottom product) or both components. By means of the estimated temperature progression, it is possible, for example, to estimate the temperature at the top and use this as a basis to ascertain the concentration of the low boiler or the top product. The inventive use of logistic functions in the respective sections results in a linear relationship between each section of a mass transfer zone and a manipulated variable. The binary mixture results in two mass transfer zones, namely a first in the lower region of the column, i.e. below a feed or in the stripping section, and a second in the upper region of the column, i.e. above the feed or in the rectifying section. The manipulated variable for the rectifying section is a coolant rate; the manipulated variable for the stripping section is a steam rate. Controllers used may be standard PI/PID controllers.

(5) In a possible configuration, the method of the invention is employed in a rectification column for separation of 1,2-dichlorobenzene (ODB) and COCl.sub.2 (phosgene).

(6) The advantage of the invention is firstly the distinctly lesser modeling complexity compared to modeling approaches known from the prior art, since only relatively few physical data (boiling temperatures) and column data (packing type etc.) are required, and secondly the lesser computing power required as a result. A rigorous process model as known from the prior art can easily contain several hundred differential equations, whereas the simplified process model presented here, depending on the number of measurement points or temperature sensors installed, is one to two orders of magnitude below that. This enables use in a productive system without using dedicated hardware. Rigorous models represent a technical mechanism with exact scientific methodology. They are created on the basis of physical, chemical or chemical engineering relationships. For example, a rigorous model recreates and simulates the rectification column or the process that proceeds therein, in that all known processes and reactions in the plant or in the column are simulated bit by bit physically and kinetically by characteristic lines, differential equations and balance equations. However, rigorous models are very costly to implement—too costly for most operations.

(7) This aspect can have a major influence on the decision for or against such a productive system, since looking after an additional system in a plant which is operated for 50 years generates additional labor (operating system migration, maintenance, updates, lack of compatibility with new operating system) that can lead to shutdown of the system.

(8) Under particular conditions, the estimated temperature at the top or bottom of the column can be used to calculate the concentration. It is a prerequisite that the concentration of secondary components is “small”. But the concentration to be determined can have a lower concentration than that of the secondary components. In a configuration, the temperature profile can be visualized in a control panel assigned to the plant or to the column, such that, for example, effects on the top temperature by perturbations of any increases in temperature in the middle of the column can be better estimated.

(9) There follows a description of the method of the invention using the example of a rectification column for separation of a mixture of phosgene and ODB, wherein phosgene is the first component and the low boiler in the mixture and ODB is the second component and the high boiler in the mixture. In the case of performance of rectification, phosgene accordingly collects preferentially in the upper region of the column, i.e. in the top of the column, and is accordingly the top product. By contrast, ODB as high boiler is enriched in the bottom of the column and is accordingly the bottom product. The estimate of the temperature profile, using the example of the rectification column for separation of phosgene and ODB, originates from observations during disturbance tests in a simulated system. A disturbance test is understood to mean an abrupt change in a manipulated variable, for example coolant rate at the top or steam rate at the bottom. In a temperature profile along the column which is obtained therefrom, for example along the column height, in particular, characteristic zones (or fronts, i.e. points with a high temperature gradient—cf. “Jörg Raisch: Mehrgrößenregelung im Frequenzbereich [Multiparameter Control in the Frequency Range], Oldenbourg Wissenschaftsverlag, Berlin, 1994.”, chapter 9.18) with a high temperature gradient (based on height) are observed. At these points, there is significant mass transfer, and therefore these points are also referred to as mass transfer zones. If what is called a Wiener model, i.e. a linear system in which static nonlinearity occurs at the output, is assumed, the temperature characteristic can be represented in the form of static nonlinearity. A characteristic feature of the progression is two S-shaped temperature progressions. The first covers the rectifying section between the top of the column (100% of the column height) and feed (about 24% of the column height), and the other the stripping section between the feed and bottom of the column (0%). The characteristic S shapes are maintained during the disturbance tests and have merely been moved along the column height. The result of this is the approach of approximating the behavior of the column during the disturbance tests, i.e. more particularly the temperature characteristics along the column height, by a shift in a static characteristic. Observations of the disturbance tests mentioned show the following: the breadth of variation of the steady-state gain in temperature, i.e. the breadth of variation of the temperature at a particular column height, is greater than the breadth of variation in the shift in the characteristic along the column height. For example, there is a significant change (i.e. by about 20%) at a temperature measurement point at about 40% of the column height only in the last disturbance. If, by contrast, in the disturbance tests, a particular temperature is monitored as a function of column height, an approximately linear progression is apparent. What is first being sought is the height (level) at which a particular temperature occurs. In a formal sense, the temperature for which the height is to be monitored must be between the boiling points of the first and second components, i.e. between the boiling points of the pure substances. One possible temperature is that which describes the position of the fronts. The real temperature progression T*(h) is known from a static simulation. This progression is now approximated in accordance with the invention by a function T(h).

(10) In a configuration of the invention, a transition of logistic functions is determined by a temperature at a feed to the column.

(11) In one embodiment of the method of the invention, the respective logistic function chosen is a logistic function of the following form:

(12) $\begin{array}{cc}f\left(x\right)=\frac{e^x}{1+e^x}& \left(1\right)\end{array}$

(13) The logistic function from equation (1) has an inflection point at x=0 and a range of values of [0,1]. In order to fit the logistic function to the temperature profile, the logistic function has to be shrunk and shifted. This gives rise to the following function with the given parameters:

(14) $\begin{array}{cc}T\left(h\right)=T_0+v\frac{e^{\left(h-h_0\right)k}}{1+e^{\left(h-h_0\right)k}}& \left(2\right)\end{array}$

(15) h.sub.0 is a support vector that shifts the inflection point of the logistic function from 0 to the point h.sub.0. The parameter h.sub.0 ascertained online can be used to control the column since it shows a lesser degree of nonlinear behavior. k describes a shrinkage in the direction of the height of the column, T.sub.0 describes a support vector of the temperature (T.sub.min) and v a range of the temperature (T.sub.max−T.sub.min).

(16) In a further embodiment of the method of the invention, the function T(h) is defined as follows:

(17) $\begin{array}{cc}T\left(h\right)=\{\begin{array}{cc}{T}_{\mathrm{ab}}\left(h\right),& h<h_{\mathrm{feed}}\\ T_v\left(h\right),& h\ge h_{\mathrm{feed}}\end{array}& \left(3\right)\end{array}$

(18) where T.sub.ab(h) and T.sub.v(h) are each defined on the basis of a logistic function and h.sub.feed defines the column height at which the binary mixture, especially a solution consisting of equal parts of the first and second components, is fed to the column.

(19) In a further configuration of the method of the invention:

(20) $\begin{array}{cc}{T}_v\left(h\right)=T_{0,v}+v_v.Math.\frac{e^{\left(h-h_{0,v}\right)k_v}}{1+e^{\left(h-h_{0,v}\right)k_v}}& \left(4\right)\\ T_{\mathrm{ab}}\left(h\right)=T_{0,\mathrm{ab}}+v_{\mathrm{ab}}.Math.\frac{e^{\left(h-h_{0,\mathrm{ab}}\right)k_{\mathrm{ab}}}}{1+e^{\left(h-h_{0,\mathrm{ab}}\right)k_{\mathrm{ab}}}}& \left(5\right)\end{array}$

(21) where the temperature has been normalized to the respective boiling temperatures T_1 and T_2 of the corresponding first and second components, with 0%=T_1 and 100%=T_2 and h normalized to an absolute column height H, where T(h)=T.sub.v(h) for h ϵ [x %, 100%] in a rectifying section v, T(h)=T.sub.ab(h) for h ϵ [0%, x %] in a stripping section ab of the column with 0<x<100, and where T.sub.0,v and T.sub.0,ab are respective support vectors for the temperature, v.sub.v and v.sub.ab are a respective range for the temperature, k.sub.v and k.sub.ab are a respective shrinkage in the direction of the column height h, and h.sub.0,v and h.sub.0,ab are a respective support vector of the height h. h.sub.0,v and h.sub.0,ab correspond to the respective mass transfer zones in the rectifying section and in the stripping section of the column.

(22) In one configuration of the method of the invention, x is chosen as 24. This corresponds to the relative height at which the feed valve is disposed. The specific feed height depends on a design of a respective column and can be calculated accurately with knowledge of the column design. The feed height defines the transition of the “part-functions” T.sub.ab(h) and T.sub.v(h).

(23) The feed temperature determines the transition between the logistic functions.

(24) In a further configuration, it is assumed that a range of values of T.sub.v(h) is between the boiling point of the first component and the feed temperature and hence T.sub.0,v(h)=0 and v.sub.v=T.sub.feed.

(25) The range of values of T.sub.v(h) is therefore to include only function values between the boiling point of the first component, i.e. the low boiler, and the feed temperature, i.e. the temperature at the feed. These give rise to the support vector and the range:
T.sub.0,v=0  (6)
v.sub.v=T.sub.feed  (7)

(26) where T.sub.feed is normalized to the boiling temperature of the second component and the boiling temperature of the first component is set to “0”.

(27) In addition, in a further configuration, it is assumed that a range of values for T.sub.ab(h) is between the feed temperature T.sub.feed and the boiling point of the second component. The logistic function in the stripping section T.sub.ab(h) is to have a range of values between T.sub.feed and the boiling point of the second component, i.e. the high boiler. This gives:
T.sub.0,ab=T.sub.feed  (8)
v.sub.ab=1−T.sub.feed  (9)

(28) where 1 corresponds to the boiling point normalized to the real boiling temperature of the second component or to the normalized boiling temperature of the second component, and T.sub.feed has been normalized to the boiling temperature of the second component.

(29) In a further embodiment of the method of the invention, a respective shrinkage k.sub.v or k.sub.ab in the direction of the respective height is ascertained from a comparison of a slope of the logistic function f(x) at its inflection point and the measured temperature progression T*(h) at its inflection point, wherein the inflection point of T*(h) is calculated from an average of respective slopes at a respective inflection point of temperature progressions ascertained from disturbance tests. The shrinkage in the direction of the height k.sub.v or k.sub.ab is thus found from a comparison of the slope of the logistic function and the real temperature profile at the inflection point.

(30) Thus, all parameters except h.sub.0,v and h.sub.0,ab have been determined. h.sub.0,v and h.sub.0,ab can be calculated via any temperature point T at a point h in the respective region or section of the column, i.e. in the rectifying section or in the stripping section, by solving a respectively corresponding equation of those above. Since, however, there are multiple temperature measurement points available, this is an overrepresented equation system. A disadvantage of the method mentioned is that measurement inaccuracies, such as measurement noise and drift, are amplified when the temperature sensor is far removed from h.sub.0,v or h.sub.0,ab. Rather than using just one equation for solution, the present invention proposes, in a further configuration, supplementing the above-described method with a parameter estimate in order to compensate for the disadvantages mentioned, where h.sub.0,v and h.sub.0,ab are each determined by a parameter estimate.

(31) There follows an elucidation of the parameter estimate envisaged in accordance with the invention in general terms based on equation (2). Application is identical for the determination of h.sub.0,v and h.sub.0,ab.

(32) This involves first minimizing the least mean squares of the temperatures in relation to the parameter h.sub.0:

(33) $\begin{array}{cc}{\min }_{h_0}={.Math.\stackrel{.fwdarw.}{f}-\stackrel{.fwdarw.}{T}.Math.}_2^2\mathrm{with}\stackrel{.fwdarw.}{T}=\left[T_{1\mathrm{.Math.}}{T}_i\right]\mathrm{with}i=1,\mathrm{.Math.},n& \\ \stackrel{.fwdarw.}{f}=\left(\begin{array}{c}{f}_1\left(h_1,h_0\right)\\ \mathrm{.Math.}\\ f_i\left(h_i,h_0\right)\end{array}\right)=\left(\begin{array}{c}{T}_o+v.Math.\frac{e^{\left(h_1{h}_0\right)k}}{1+e^{\left(h_1-h_0\right)k}}\\ \mathrm{.Math.}\\ T_o+v.Math.\frac{e^{\left(h_1-h_0\right)k}}{1+e^{\left(h_i{h}_0\right)k}}\end{array}\right)\stackrel{.fwdarw.}{h}=\left(\begin{array}{c}h\left(T_1\right)\\ \mathrm{.Math.}\\ h\left(T_i\right)\end{array}\right)& \left(10\right)\\ \stackrel{.fwdarw.}{f^{\prime }}=\frac{\partial \stackrel{.fwdarw.}{f}}{\partial h_0}=\left(\begin{array}{c}-v.Math.\frac{{\mathrm{ke}}^{\left(h_o+h_1\right)k}}{{\left(e^{h_{1}k}+e^{h_{0}k}\right)}^2}\\ \mathrm{.Math.}\\ -v.Math.\frac{{\mathrm{ke}}^{\left(h_o+h_i\right)k}}{{\left(e^{h_{i}k}+e^{h_{0}k}\right)}^2}\end{array}\right)& \left(11\right)\end{array}$

(34) The function f(h) corresponds to equation (2), i.e. f(h)=T(h). The serial parameter i indicates the temperature measurement points present in each case in the rectifying region or section or in the stripping region or section. The parameter h.sub.0, i.e. h.sub.0,v for the rectifying section and h.sub.0,ab for the stripping section is to be estimated. This is a nonlinear minimization problem since the parameter is in the exponent.

(35) In one possible configuration of the method of the invention, the Gauss-Newton method is employed for estimation.

(36) A residual vector {right arrow over (r)} describes the current deviation of estimated from measured temperature and is formed as follows:
{right arrow over (r)}={right arrow over (f)}({right arrow over (h)})−{right arrow over (T)}  (12)

(37) The Gauss-Newton method also requires the partial derivative of {right arrow over (r)} in h.sub.0

(38) $\begin{array}{cc}\frac{\partial \stackrel{.fwdarw.}{r}}{\partial h_0}=\frac{\partial \stackrel{.fwdarw.}{f}}{\partial h_0}=\stackrel{.fwdarw.}{f^{\prime }}& \left(13\right)\end{array}$

(39) This gives the Jacobi matrix

(40) $\begin{array}{cc}D=\left(\begin{array}{c}\frac{\partial r_1}{\partial h_0}\\ \mathrm{.Math.}\\ \frac{\partial r_n}{\partial h_0}\end{array}\right)& \left(14\right)\end{array}$

(41) The step width s is then calculated as follows:
s=(D.sup.TD).sup.−1D.sup.T{right arrow over (r)}  (15)

(42) In the case of a parameter to be determined, D is a vector. Therefore, inter alia, the inversion from equation (15) with (13) is simplified to give:

(43) $\begin{array}{cc}{\left(D^{T}D\right)}^{-1}=\frac{1}{{.Math.\stackrel{.fwdarw.}{f^{\prime }}.Math.}_2^2}& \left(16\right)\\ s=\underset{i=1}{\overset{n}{.Math.}}\frac{f_i^{\prime }{f}_i}{{.Math.\stackrel{.fwdarw.}{f^{\prime }}.Math.}_2^2}& \left(17\right)\end{array}$

(44) The next iteration value is
h.sub.0(t)=h.sub.0(t−1)−s  (18)

(45) The estimate of h.sub.0 is used as controlled variable both for the rectifying section h.sub.0,v=H1=y.sub.1 as mass transfer zone and for the stripping section h.sub.0,ab=H2=y.sub.2 as mass transfer zone. Manipulated variables used are the coolant rate and the steam rate.

(46) Further advantages and configurations of the invention are apparent from the description and the appended drawings.

(47) It will be apparent that the features mentioned above and yet to be elucidated hereinafter are usable not just in the combination specified in the respective case but also in other combinations or on their own without leaving the scope of the present invention.

(48) The invention is shown in schematic form in the drawings by means of a working example, and is described schematically and in detail hereinafter with reference to the drawings.

(49) FIG. 1 shows a schematic diagram of a rectification column that can be controlled in one embodiment of the method of the invention.

(50) FIG. 2 shows illustrative temperature profiles of a rectification column after performance of disturbance tests, in each case by alteration of a coolant rate in each case as abruptly altered manipulated variable.

(51) The rectification column shown in FIG. 1 is incorporated, for example, into a plant for preparation of toluene diisocyanate (TDI). The column shown has the task of supplying TDI reaction lines with vaporous phosgene. The column has an upper region, i.e. a top region or top, and a lower region, i.e. a bottom region or bottom. On performance of the rectification, phosgene as top product is enriched in the top, and ODB as bottom product accumulates in the bottom. Unlike in a standard column, the top product formed is not totally condensed, but drawn off in vaporous form. Therefore, the column shown in FIG. 1 does not have a condensate vessel. Instead, an inserted condenser is used. As a result, a reflux rate and a reflux ratio can be controlled only indirectly via a coolant rate. The phosgene is introduced firstly in the form of phosgene solution via a feed F2 having a feed valve Y3. Secondly, it is applied in pure form as liquid phosgene at the top F1 of the column. The phosgene solution consists in roughly equal parts of phosgene and the ODB solvent. Via a valve Y1 at the top, cooling liquid or coolant is fed in via a coolant feed F3. The cooling liquid can be removed again from the column via an outlet G3. A manometer P1 measures the pressure, which should be kept essentially constant. There are also multiple points or stages in the column, at each of which there are disposed temperature sensors T1 to T7. Via a valve Y2 and a feed F4, a heating medium or steam is supplied in the bottom of the column. A corresponding bottom product formed in the separation in the column is removed via a valve Y4. The feed F2 divides the column into two sections, namely an upper rectifying section between feed F2 and the top of the column, in the range from 24% to 100% of the column height in the case shown here, and a lower stripping section between feed F2 at the bottom of the column, i.e. between 0% and 24% in the case shown here. In the rectifying section, phosgene is enriched as low boiler. In the stripping section, the phosgene is parted from the ODB solvent. For control of concentration, the position of the temperature sensors in particular is important. The standard method is generally that neither the top temperature nor the bottom temperature is used directly for control, but rather temperatures between feed F2 and the respective product outlet. Useful temperatures in the rectifying section are the temperatures from temperature sensors T2 to T4 for the control of the top product concentration, and in the stripping section the temperature from temperature sensor T6 for the control of the bottom product concentration.

(52) FIG. 2 shows a temperature profile T*(h) of the rectification column from FIG. 1 before and after performing disturbance tests, in each case by abruptly altering the coolant rate as manipulated variable.

(53) On an abscissa 10 is a column height h in %, i.e. normalized to an absolute column height H. 100% corresponds here to a top of the column; 0% corresponds to a bottom of the column. On an ordinate 20 is plotted a temperature in %, where the temperature is standardized the respective boiling temperatures of phosgene and ODB. 0% corresponds here to the boiling temperature of phosgene and 100% corresponds to the boiling temperature of ODB.

(54) A characteristic feature of the respective progressions shown is two S-shaped temperature progressions for each disturbance test per temperature profile. The first covers the rectifying section between the top of the column (100%) and feed F2 (24%), the other the stripping section between feed F2 and bottom of the column (0%). The characteristic S shapes are maintained during the disturbance tests and have merely been shifted along the column height, i.e. along the abscissa. This gives rise to the approach of approximating the behavior of the column in the disturbance tests by a shift in a static characteristic. The following becomes apparent from the observations of the disturbance tests: the range of variation of the steady-state gain in temperature (along the ordinate 20) is greater than the range of variation of the shift in the characteristic (along the abscissa 10). A temperature measurement point at 40% of the column height changes significantly by about 20% only in the last disturbance. If, by contrast, a temperature is monitored (for example 40% on the abscissa 10), an approximately linear progression is apparent. What is being sought is firstly the height (step) at which a particular temperature occurs. In a formal sense, the temperature for which the height is to be monitored must be between the boiling points of the pure materials, i.e. of phosgene and ODB. One option is the temperature which describes the position of the fronts in FIG. 2.