Você está na página 1de 5

Closed-Loop Response

Controlled variable Manipulated variable Controlle r type /scheme PI/ Feedback gProms model/process name LevelControl (BasicProcessCt rl)

Liquid level in sump tank

Sump tank outlet liquid flowrate (Stream 8) Inlet liquid flowrate (Stream 6)

Ratio between liquid and gas feed flowrates

P/ Ratio control

RatioControl (BasicProcessCt rl)

S1

Column liquid level (measured in sump tank)

High Level Switch (HLS), (ESDV)

ESDV (SafetyCtrl)

The objective of this closed loop assessment is to observe the models dynamic response towards disturbances identified. From Section 2, Control Configuration 2 is recommended and simplified as shown above in Figure __ with only 3 control loops being modelled as per Table ___. Two separate process simulations were executed; a) basic process control with 2 control loops and b) safety control with 1 loop which is independent and assumes failure of the basic process control. The following disturbances were introduced to the simulation: 1. 20% step change increase in inlet gas flowrate, Vin for 500 seconds (Basic Process Control) 2. 20% step change decrease in inlet gas flowrate, Vin for 500 seconds (Basic Process Control) 3. 500% step change increase in inlet liquid flowrate, Lin for 500 seconds (Safety Control) Since the inlet liquid flowrate is being manipulated in the basic process simulation for control, disturbances were introduced in the inlet gas flowrate instead and 20% magnitude was chosen which is reasonable for normal column operation as mentioned before. Whereas under special conditions in the safety control simulation, the inlet liquid flowrate was manipulated and a high step change was introduced to determine if the safety loop can withstand the extreme condition. The setpoint height for the level controller was obtained from the steady-state value obtained in the open-loop simulation and the setpoint ratio from Literature. A PI controller was used for the liquid level control as it eliminates offsets, has a reasonable response speed, are computationally cheaper and easier to tune compared to PID controllers (Stephanopolous, 2004). A proportional (P) controller was used for the ratio control as it is (Stephanopolous, 2004). Ideally, a Pcontroller would be used for the level control to avoid fluctuations in the manipulated stream flowrate as it will be entering another unit as feed. A PI ratio is also normally used for ratio control to eliminate offsets. However, the fluctuations and offsets produced by these controllers are found to be minimal in this simulation which will be discussed in detail below. shows the closed loop response of the mole fraction acetic acid in the distillate against time where the distillate purity set point is 0.9879 mol/mol which is the purity achieved in steady state simulations. From Figure 4.9, the system runs at steady state for 200 minutes before disturbance 1 which is a 20 % positive step change of feed is introduced. The disturbance is run for 5000 minutes so that the duration taken for the response to return to steady state can be investigated. It can be seen that as feed flow rate increases there is no observable effect on distillate purity. The column is then returned to steady state for 200 minutes before disturbance 2 which is a 20% positive step change of condenser duty is introduced for 5000 minutes. When condenser duty increases, the liquid returning back into the column is sub-cooled so the column temperature decreases thus less energy is available to vapourise the heavy components and the distillate purity increases. The integral action of the PI controller results in an oscillatory response to reduce the reflux flow rate so less sub-cooled liquid returns into the column. The

column returns to steady state in roughly 4500 minutes. The system is reinitialised at steady state for a further 200 minutes before disturbance 3 which is a 20% positive step change in reboiler duty is introduced for 5000 minutes. As reboiler duty increases, more heavy components are vapourised and propagate up the column thus reducing the distillate purity. Again an oscillatory response by the integral action of the PI controller returns the distillate purity to the set point of 0.9879 mol/mol in approximately 4600 minutes by increasing the reflux flow rate. This is because increasing reflux flow rate causes the acetic acid to be reprocessed and thus becomes purer. Hence, based on the closed loop analysis, it is shown that the set point of 0.9879 meets the specification set so the control system proposed is robust and able to reject the disturbances identified Based on the open loop response developed earlier, controllers are added to reject the three main disturbances which are inlet feed flow rate, condenser duty and reboiler duty. A proportional-integral (PI) controller has been proposed to reject the effect of disturbances on acetic acid distillate purity, xD. A proportional (P) controller is used not used because an offset in distillate purity set point is not tolerable so the integral portion of the PI controller is needed to eliminate this offset. On the other hand, a proportional-integral-derivative (PID) controller was deemed unnecessary because the response is relatively quick and an inverse response was not noticed. Furthermore, PID controllers require more computationally expensive and are harder to tune. The effects of condenser and reboiler duty on the distillate purity was chosen to be modelled because the aim of this column is meet the design specifications so it is essential to verify that the controls are robust enough to meet this aim. The same three disturbances and magnitudes are simulated as shown in Figure 4.7. The gPROMS closed loop models and processes can be found in Appendix 4.5 and Appendix 4.6 respectively. The magnitudes and duration of the 3 main disturbances modelled in this report are as follow: 1. 50% step change increase in the inlet solvent flowrate, Lin for 500 minutes (black line) 2. 50% step change increase in the inlet gas feed flowrate, Vin for 500 minutes (red line) 3. 500% step change increase in the inlet gas feed flowrate, Vin for 500 minutes (yellow line)

The first 2 are to demonstrate the response of the basic process control (liquid level and ratio control) and the third for the safety loop (ESDV trigger loop which measures the liquid level). A closed loop simulation was then executed to examine the dynamic responses of the inlet ratios, height of liquid on the tray. Closed loop and Open loop responses are plotted below. From Figure __, when Disturbance 1 is introduced at 500 seconds, the liquid height in the sump tank increases to 0.0728m before the oscillatory control action kicks in and reduces it back to its set-point value of 0.066m. This is because when the gas inlet flowrate was increased, the ratio control is put into action which increases the liquid inlet flowrate into the column, thus increasing the liquid height. This then activates the level control which increases the liquid outlet flowrate to 2.20kmol/s as shown on the graph thus bringing the liquid height back down. At 1500seconds, Disturbance 2 was introduced and as shown previously, the ratio control will act by reducing the liquid inlet flowrate which in turn, reduces the liquid height in the tank to 0.0592m initially as shown. The oscillatory level control then responds by reducing the liquid outlet flowrate to 1.46kmol/s which returns the liquid height back to the initial set-point within. The oscillation in the PI control action in the manipulated liquid outlet flowrate is shown to be small (with a maximum amplitude difference of ~2.5% from the set-point) and stabilises very quickly (within 100 seconds) and thus will only introduce small fluctuations in the next units inlet flowrate with minimal effects. Furthermore, in the closed-loop response, due to the interactions between the level and ratio control loops, the liquid height will change when the gas inlet flowrate is changed because the ratio controller will act by changing the liquid inlet flowrate. In contrast, in the open-loop response, the liquid height does not change at all when the gas inlet flowrate is changed. Also, the closed loop response was slower than the open-loop even though the opposite should be expected due to the reduction of the time constant (Stephanopoulous, 2006). This is because as mentioned before, the hydraulic time constant used is too small, leading to an almost instantaneous response in the open loop. From Figure 3, when Disturbance 3 is introduced at 50 seconds, the liquid height in the sump tank increases gradually to reach 0.3m which is the maximum liquid level set, hmax. This triggers the High Level Switch which then closes the emergency shutdown valves, changing their positions from 1.0 to 0.0. This immediately stops the gas and liquid flowrates into the column as shown by reducing both flowrates to 0.0 kmol/s and column operation is then stopped. As a sharp increase in the liquid holdup and thus the liquid height is a good indicator for flooding (Silvey &

Keller, 1966), this safety loop will be effective in preventing flooding in case of failure in the basic process control system. However, a step change of equal magnitude in the gas inlet flowrate will not trigger this safety loop as it will have no effect on the liquid height in the tank due to the assumption of no vapour holdup. In reality, such a large increase in the gas inlet flowrate would lead to column flooding and this false response may be hazardous if a great disturbance such as this occurs during operation. Nevertheless, since the column is assumed to undergo dynamic operation close to nominal steady-state conditions and far from the upper capacity limit, such a great disturbance in either liquid or gas inlet flowrates is rather unlikely.

As shown from Figure 2, when Disturbance 1 was introduced at 500seconds, the liquid inlet flowrate increased from 1.83kmol/s to 2.200kmol/s due to the ratio control action to match the increased inlet gas flowrate. The ratio initially decreased due to more gas coming in but then increased to the value of 1.493 due to the controller action. The closed-loop liquid molar fraction of CO2 leaving the column initially increased from 0.00799 to reach a maximum value of 0.00916 as more CO2 is coming into the column and being removed, but the controller action effects soon took place and more solvent flows into the column, diluting and reducing the CO2 molar fraction to 0.00797, which is slightly lower than the initial steady-state value (0.00799). At 1500seconds, Disturbance 2 was introduced and the liquid inlet flowrate was reduced to 1.46kmol/s in response by the controller. The ratio initially increased but was then reduced by the controller action and stabilised at 1.505. CO2 molar fraction dropped to 0.0686 initially as less CO2 is now coming into the column and less is absorbed into the liquid phase. The controller action then kicks in by reducing the solvent flowrate and increases the molar fraction to 0.00803 which is slightly higher than the steadystate value. As evident, there is a slight offset in the new steady-state values for both the ratio and CO2 molar fractions since a P-controller was used but as the offset is miniscule (change in the 3rd and 5th decimal respectively), the unit is still meeting its specification range (0.00770 - 0.0085 mol/molCO2 coming out in the liquid phase which corresponds to 90 - 92% CO2 removal). This also proves that the P-controller is sufficient to control the feed flowrates ratio for this particular process and unit.

As the inlet liquid flowrate was increased, the liquid height on the tray increased from 0.092m to a higher new steady-state at 0.113m in the open loop. In the closed loop, the P controller implemented to control the liquid level reduced the outlet liquid flowrate and reached a much lower steady-state value of 0.094m. There is an offset from the original steady-state value since a P controller was used but it is small and within acceptable range. However, the closed loop response was slower than open-loop even though the opposite should be expected due to the reduction of the time constant (Stephanopoulous, 2006). This is because the change in the open loop is instantaneous which is not realistic. When the inlet gas flowrate was increased by 500%, there was no change in the liquid height on the tray which did not trigger the ESDV safety loop. This response is unrealistic as such a big increase in gas flowrate would make the column susceptible to flooding and this would be indicated by a sharp increase in liquid holdup and thus the liquid height (Silvey & Keller, 1966). The lack of response is due to the assumption of no molar vapour holdup which holds under normal operating conditions due to the low pressure (1atm). However, if a great disturbance such as this occurs during operation this false response may be hazardous. A correlation between the differential pressure across the column, which is another good flooding point indicator (Kister, 1992), and the inlet gas flow could also be implemented and measured instead for the safety loop instead of the liquid height. Nevertheless, since the column is assumed to undergo dynamic operation close to nominal steady-state conditions and far from the upper capacity limit, such a great disturbance is rather unlikely. From Figure 3, when the inlet liquid flowrate is increased, the ratio between the inlet flowrates (Lin/Vin) increased from the optimum value of 1.5 and reached a new steady-state at 2.1 in the open-loop response. In the closed-loop, the ratio controller implemented reduced the valve position value and thus the flowrate of the inlet solvent, returning the ratio back to the original steady-state value. When Vin is increased, the ratio decreased and reached a new steady-state value of 1.0 in the open loop. In the closed loop, the valve position was increased to increase the inlet solvent flowrate to match the inlet gas flow thus returning the ratio back to 1.5. Again, the closed-loop response to both disturbances was slower than the open-loop which is not realistic. Nevertheless, the ratio controller was effective in maintaining the desired ratio for separation purposes between the inlets within a reasonable response time and resembles the response of a feedforward controller (Stephanopoulous, 2006). This loop is implemented to meet the unit specification but this disturbance is not particularly hazardous.

Technical: design P and T in addition to operating values. (This will determine when your unit blows up for instance.) For absorbers, why have you chosen your particular solvent? If a solvent mixture, why that particular ratio? For all units, what is the specification in terms of outlet product purity of your particular unit? Saying 99% removal over two columns tells us nothing about which composition to expect out of the first one for instance. Also, the requirement should be measurable (recovery generally isnt). For control structures, there is no one-to-one match between disturbance and control loop, ie you don't have one loop per disturbance. The control structure is in place to deal with all disturbances, although a single disturbance may affect one aspect of the operation more than others. For steps, does your chosen step size make sense, and why? If is useful to also plot the disturbance (so we can see how much it is changing), the manipulated variable (so we can see how much this needs to change) and the set point (so we know where you want to be) in addition to the controlled/measured variable. It must be completely clear when the disturbance is taking place. For absorbers, if you have solvent flow rate as a disturbance - and at the same time used it for control - this is not possible (unless under special conditions). For the responses, you must relate to safety as this is the major concern, not just product compositions. For the final report, we will want to know how you implemented your controllers and how you tuned them. (You should have briefly described your control in Report 4 as it is otherwise very hard to consider your closed loop responses.) General report writing (keep in mind also for research project): Always refer to tables and figures in your main text - if you don't mention them, your reader will not look at them.

When presenting a table or figure, say (in the text) what is plotted. In some cases you also need to say why you have chosen the variables you have chosen. Figure captions go below the figure but table captions goes above the table. Don't ask me why, but this is the convention. Make sure your figures with captions can be understood stand-alone, without having to refer to the main text. Use a consistent referencing convention, e.g Harvard. Always start a main chapter on a new page. Never start a chapter or subchapter with a table or figure, always start with text explaining what follows. Always proof-read. Read your report backwards paragraph by paragraph. That way there is no meaning and you concentrate on typos and grammar. Better still, get someone else to read it for you. Final note: We are expecting you to take our feedback into account, ie if you make the same mistakes again, we will deduct more marks the second time around.

Você também pode gostar