Controller Performance Tests

The integrated controller was implemented on the ETBE column within the SpeedUp™ dynamic simulation environment. The control loops were all tuned using recommended recommended settings for approximate first-order-plus-dead-time (FOPDT) models which were fitted to step responses (Ogunnaike and Ray, 1994). Although these models were a poor fit in some cases and there was a high degree of non-linearity in most cases (as judged by the differences in the responses to positive and negative perturbations), the resulting values of gain and integral time (no derivative component was included in any of the controllers) were adequate to achieve satisfactory tuning. For example, some increase in the control action was tolerable before the controller became unstable. Aggressive tuning is not appropriate for this system because of its inherent non-linearity and bidirectionality.

The disturbances that are most likely to affect the process are feed rate changes and feed composition changes. The impact of other disturbances (e.g. the temperature of the heating medium to the reboiler) can be attenuated using appropriate cascade controllers and other standard control techniques. On this basis, three performance tests were characterised for this system:

(1) 10% feed rate increase at time = 0 minutes, followed by a 20% feed rate decrease at time = 120 minutes;

(2) 5% increase in the stoichiometric ratio at time = 0 minutes, followed by a 10% increase in non-reactive components at time = 120 minutes;

(3) 5°C increase in the set-point of the mid-stripping section temperature controller at time = 0 minutes, followed by a 1°C decrease in the set-point of the reaction zone temperature controller (where applicable) at time = 120 minutes.

In each case, the changes were scheduled over 5 minutes (rather than using pure step changes) to preserve continuity and to more closely simulate a real process disturbance.

The open-loop effects of the feed rate and feed composition disturbances are shown in Figures 10.18 and 10.19. Significant deviations from the initial operating conditions are seen in both cases. The processes are open-loop stable but steady state was not reattained before the second feed rate disturbance even though an additional hour was allowed. The response shown in Figure 10.18 emphasises the non-linearity and bidirectionality of the process. An increase and a decrease in the feed rate reduce both the ETBE purity and the isobutylene conversion. Thus, an optimum feed rate exists for the base case combination of reflux rate and reboiler duty and the process gain changes sign (i.e. the process is bidirectional). The shapes of the responses are also different, indicating strong non-linearity.

Figure 10.18 - Open-Loop Response to Feed Rate Disturbances (Performance Test 1)

Time (hrs)

Figure 10.19 - Open-Loop Response to Composition Disturbances (Performance Test 2)

Time (hrs)

Figure 10.19 - Open-Loop Response to Composition Disturbances (Performance Test 2)

Interestingly, the feed rate disturbance has a larger effect on the system than the feed composition change. The LV control configuration is often favoured for non-reactive distillation because it provides good open-loop disturbance rejection of feed rate disturbances. This property arises as the feed rate usually affects fractionation only and not the feed split. This is clearly not the case for reactive distillation and closed-loop control is required to counter feed rate disturbances (planned or otherwise).

Figures 10.18 and 10.19 demonstrate that open-loop control is inadequate for this column due to the large deviations from the initial conditions, thereby confirming earlier results (Section 9.3 and 10.2.1). Thus, the need for closed-loop control is established but the effectiveness of the three control modes proposed above must be evaluated against these open-loop responses (using performance tests 1 and 2 for consistency). Only substantially better disturbance rejection should be considered adequate, and set-point sensitivity (performance test 3) must be introduced as an additional requirement.

Each control mode was independently evaluated by subjecting the controller combination (per Table 10.5) to the relevant performance tests using dynamic simulations. The LV control configuration was used in each case to provide a consistent reference. The simulation results are presented below in Figures 10.20-10.27. The comparison with open-loop operation is favourable. Although the key process parameters are not being controlled directly (due to the measurement and control difficulties described previously), the control responses indicate that satisfactory to excellent control is possible in each mode. In all cases, the disturbances result in some deviation from the initial values of ETBE purity and isobutene conversion but the changes are generally not appreciable.

Control mode 1 is a two-point control scheme and the interactions between the two loops are clearly evident (especially in Figures 10.20 and 10.22). This reduces the ability of the controller to reach steady-state quickly. However, both the ETBE purity and reactant conversion are maintained close to their original values following feed rate and feed composition disturbances. Figure 10.21 indicates that the composition affects the temperature set-point required to produce a given value of purity and the set-point might need to be manually updated to control the purity tightly. This cannot be done automatically (via a cascade loop using an analyser signal or otherwise) as such a loop would only be stable over a narrow range of operating conditions. The process gain changes sign so that it is not immediately evident whether the master controller in the cascade system should increase or decrease the set-point of the slave controller in order to achieve a particular ether purity.

The set-point responsiveness of this control mode is compromised by the strong control loop interactions which exist between the purity loop (controlling a mid-stripping section temperature) and the conversion loop (controlling the reaction zone temperature difference).

However, Figure 10.22 shows that set-point changes can be implemented without destabilising the process, although the process settling time is significant (greater than two hours).

Figure 10.20 - C losed-Loop Response to Feed Rate Disturbances in Control Mode 1
Figure 10.21 - Closed-Loop Response to Feed Composition Disturbances in Control Mode 1

Time (hrs)

Figure 10.22 - Closed-Loop Response to Set-Point Changes in Control Mode 1

Time (hrs)

Figure 10.22 - Closed-Loop Response to Set-Point Changes in Control Mode 1

Control mode 2 is a simpler control structure that offers faster dynamic responsiveness than mode 1. The reflux rate is maximised to equipment constraints, which removes the control loop interactions in maximising the isobutene conversion for the target ETBE purity. The purity is maintained close to the original value for feed rate disturbances (Figure 10.23) but allows some variation for feed composition disturbances (Figure 10.24). As with control mode 1, manual adjustment of the temperature set-point could be required to keep the ETBE purity at a target value at all times. Only one set-point is used and Figure 10.25 shows that the controlled variable responds quickly to changes in its value. However, its steady-state sensitivity is very high so that relatively large changes are required to produce a substantive change in the desired process objective. In using the complete, integrated control scheme, it is important to realise that this sensitivity changes between modes (i.e. depending on whether the secondary control loop is in the manual or automatic condition).

Time (hrs)

Figure 10.23 - Closed-Loop Response to Feed Rate Disturbances in Control Mode 2

Time (hrs)

Figure 10.23 - Closed-Loop Response to Feed Rate Disturbances in Control Mode 2

Boiler For Distillation

Time (hrs)

Figure 10.24 - Closed-Loop Response to Feed Composition Disturbance in Control Mode 2

Time (hrs)

Figure 10.24 - Closed-Loop Response to Feed Composition Disturbance in Control Mode 2

]

1

u

1 1 ! 1 1 1

-J

-4

4

1 1

Conversion 1 t i i

1 i

1 i

1 i

1 1 1 1 1 1 !

1

i l l ftjrity | | |

l 1

!

i 1

1 [ 1 1 1 1

[

1

. 1 . . . i j .

. i

I .

, 1 , ,

Time (hrs)

Time (hrs)

hgure I0.25 - Closed-Loop Response to Set-Poinl Chanties in Control Mode 2

Only two responses are shown to characterise control mode 3 as no set-points are used directly in the control of the process so that set-point responsiveness is not a consideration. Figure 10.26 shows that the fluctuations following a feed rate change are larger than for the other modes but that the ETBE purity stabilises to a higher value after each perturbation. Although the isobutene conversion is maximised in this mode (by maintaining the reflux rate at its maximum value), the final values of the conversion are lower than in control mode 2.

This attribute of the control system is unavoidable as the optimum reboiler duty with respect to purity and the optimum with respect to conversion do not coincide exactly. Near the operating point for maximum purity, the conversion may decrease slightly following some disturbances. In other cases, both the conversion and purity will increase together. This property is not an indication of any inadequacy in the control structure but reflects the achievable process results and the control priorities (i.e. the first control objective, maximum purity, takes precedence over the second objective, maximum conversion). The comparison between Figures 10.24 and 10.27 is similar to Figures 10.23 and 10.26. The final ETBE purity is higher after each composition perturbation but the conversion is decreased slightly in each case.

Distillation Tray Downcomer
figure 10.26 - Closed-Loop Response to Feed Rate Disturbances in Control Mode 3
Distillation Tray Downcomer
Figure 10.27 - Closed-Loop Response to Feed Composition Disturbances in Control Mode 3

Was this article helpful?

0 0

Responses

  • sinikka kiiskinen
    What reboiler duty adjustment?
    7 years ago

Post a comment