Control Configuration

The control configuration designations refer to the two manipulated variables that are not used for inventory control. In one-point control, one of these variables is used to control a product composition and one is fixed or adjusted only intermittently (normally to manage an equipment constraint). The distillate and bottoms product draw rates directly affect the feed split (i.e. the column material balance) while the reflux rate and reboiler duty only affect fractionation. The relative magnitudes of these effects differ by approximately an order of magnitude so that the manipulated variable selected for composition control should be one of the product draw rates where a choice exists. With this restriction, it is possible to ascertain the variable pairings that are implied by each control configurations. These are indicated in Table 9.2 for several common control schemes, including some ratio schemes. Figure 9.4 depicts the control connections for the LV configuration. The other configurations can be setup similarly. The DB control structure is excluded because the two manipulated variables are not independent at steady state and, therefore, satisfactory one-point control is not realisable.

Table 9.2 - Distillation Control Configurations

Compositio

n Control

Inventory Control

Reflux

Reboiler

Configuration

Varied

Fixed

Accumulator

Sump

LV

reflux rate or

reboiler duty or

distillate rate

bottoms rate

reboiler duty

reflux rate

LB

bottoms rate

reflux rate

distillate rate

reboiler duty

DV

distillate rate

reboiler duty

reflux rate

bottoms rate

(L/D)V

reboiler duty

reflux ratio

distillate rate

bottoms rate

(L/D)(V/B)

reflux ratio

boilup ratio

distillate rate

bottoms rate

or boilup ratio

or reflux ratio

(V/B)L

reflux rate

boilup ratio

distillate rate

bottoms rate

(D/F)V

distillate yield

reboiler duty

reflux rate

bottoms rate

Some general process control rules can expedite the selection of the best control structure for a particular column. Firstly, the control scheme should combine steady state sensitivity and dynamic responsiveness (Neisenfeld and Seeman, 1981). Since the sensing point for the 10 stage ETBE column that is examined here has been specified to be near the middle of the stripping section, the reboiler duty (V) and bottoms draw rate (B) are favoured as the principal manipulated variables although other choices might also show a good sensitivity for the controlled variable. Secondly, the level control is favoured by the use of the larger outlet stream when more than one exists from a given vessel (Luyben, 1992). This implies using reflux rate to control the reflux accumulator level on columns where the reflux ratio is higher than unity and the reflux rate when the converse applies. This consideration favours the DV configuration for columns with high reflux ratio and the LV, LB and ratio configurations for columns with low reflux ratio

Stripper With Reboiler
Figure 9 4 - LV Control Configuration

The ratio control schemes have been strongly recommended by some authors (e.g. Skogestad and Morari, 1987; Shinskey, 1984) but are generally more suited to two-point control. These schemes effectively result in one or more non-diagonal 2x2 controllers (i.e. a MIMO controller instead of several SISO controllers). For example, schemes involving L/D or the reflux ratio L/(L+D) lead to a controller that manipulates both L and D for reflux drum level control, regardless of how the composition ratio is configured. The advantages of ratio control schemes generally centre on the implicit decoupling that is achieved (Sandelin et al., 1991). For example, the reflux ratio and boilup ratio are essentially independent while the reflux rate and the reboiler duty are closely related. This feature of ratio control schemes is highly advantageous for two-point composition control but less important for one point composition control where control loop interactions are less significant. The disadvantages of ratio schemes are that they often make consistent operation at equipment constraints harder to achieve and they increase the complexity of the control problem.

No clear preference for any control configuration is evident for the 10 stage ETBE column. However, dynamic simulation provides a means of quickly assessing each scheme. Perfect level and pressure control was assumed and a simple PI controller was implemented in the simulation to control the liquid temperature on stage 7 (middle of the stripping section). The tuning constants were determined empirically to approximately produce quarter decay responses. The attainable control performance was tested against two common disturbances (a feed rate step increase of 8% and a feed composition step increase of 2% to the ratio of ethanol to isobutene) and a set point change of 5°C. The results are presented in Figures 9.5-9.13 and summarised in Table 9.3 in terms of the integrated absolute error (IAE) and the integrated time-weighted absolute error (ITAE) of the controlled variable.

Table 9.3 - Control Scheme Performance

Feed Rate

Feed Composition

Set Point Change

Step Increase (+8%)

Step Increase (+2%)

(+5°C)

Scheme

IAE

ITAE

IAE

ITAE

IAE

ITAE

LV

0.3

0.6

0.05

0.3

1.3

2.2

LB

0.4

0.8

0.05

0.1

0.6

1.0

DV

42

430

1.4

7.8

15

73

(L/D)V

0.8

2.3

0.07

0.6

1.6

3.4

(L/D)(V/B)

0.1

0.6

0.1

0.6

1.5

2.5

(D/F)V

11

400

0.1

0.7

1.5

2.5

If only the basic (non-ratio) control structures are considered, Table 9.3 suggests that either the LV or the LB configuration is most suitable but this is less evident from the simulation responses shown in Figures 9.5-9.13. The LV (Figures 9.5-9.7) and LB (Figures 9.8-9.10) configurations produced a faster response to a feed rate change than the DV scheme (Figures 9.11-9.13) and the deviation trom the initial conditions was less. The latter point is significant since the product composition is not being measured or controlled directly. However, the dynamic responses of the ether product to changes in the feed composition and set point appeared similar for all configurations, despite the much higher IAE and ITAE for the DV configuration.

These initial results suggest that the presence of a reaction (or a reactive section) within the column has a significant effect on the performance of the control system. The DV configuration is very widely used in industry for conventional distillation columns, including those where the primary product is the column bottoms, and normally yields veiy satisfactory results based on controlling the feed split. However, in this reactive distillation column, manipulating the feed split does not necessarily ensure that satisfactory reaction conditions are maintained and a control configuration that manipulates the reboiler energy input (either directly with the LV configuration or indirectly with the LB configuration) is preferred.

The performances of the ratio schemes were varied, and usually slightly inferior to the LV and LB configurations, but the double ratio configuration appears to have potential. The dynamic simulation responses are similar to those shown in Figures 9.5-9.13 but have not been included here due to space limitations. These results provide no clear incentive to install a ratio scheme on this column and, given the increased complexity and the disadvantages discussed previously, a non-ratio scheme should be implemented in this column.

100%

Boiler For Distillation

20 30 40

Time (min)

Figure 9.5 - Step Increase in Feed with the LV Configuration

20 30 40

Time (min)

Figure 9.5 - Step Increase in Feed with the LV Configuration

100%

I Conversion

>

Bottoms Temperature

I I

I i I l I Ether Purity

t I I I I I

0 0

Responses

  • Adaldrida
    What is stripping column?
    6 years ago

Post a comment