8 replies to this topic

[Kakashi-01]

When Stream 1 flow rate is cut to nearly half of its design value, the flow control valve adjusts its output to maintain the process variable (PV) at the setpoint (SP). It reaches steady state about 6 minutes after the disturbance.

 

Since the circulating oil stream remains at its design flow, but the process stream it is cooling has now a lower flow the heat recovered by the oil reduced. As a result, the temperature rise is lower and the hot process stream is cooled to a lower temperature.

 

In the final HE, the process stream is cooled using cooling water. Now that both the process stream's flow rate and inlet temperature are lower, the temperature controller (TT) for Stream 5 attempts to maintain its outlet temperature at 40 °C by adjusting a valve. However, from the trend plot, the controller appears to respond quite slowly — it took nearly 7 hours to reach setpoint.Why might the temperature control response be so sluggish? Could it be due to poor controller tuning?

 

Another issue I noticed the pressure of Stream 5 decreased following the reduction in Stream 1 flow rate. With a lower pressure entering the flash drum, wouldn’t more volatile components be expected to vaporize, possibly reducing the liquid level instead of increasing it?. Also, the level controller output reaches 100%, but the PV (level fraction) is still only around 0.6 — well below the high limit of the controller range. Why does the controller saturate even though the level hasn’t hit its upper range?

 

For Stream 9 (the syngas stream), steady state was reached approximately 20 minutes after the disturbance, but at a higher temperature. I think this is because the lower flow rate and pressure of Stream 1 lead to a longer residence time in the heat exchanger, allowing more heat recovery. The heat duty of the reactor also decreased due to the reduced feed. Since the shell side of the reactor is simulated as a flash drum (V3), and we set its duty to zero, less energy is available to generate steam.

 

The amount of steam decreases rather if we set the duty of the drum to 0 excluding the heat from the reactor then heat is utilized to increase the temperature of bfw and generate less generate because its supplied at conditions of 240 C and 48 bar. I am not sure why the bfw flow rates is oscillating wildly in response to very small changes in the level fraction.

 

This leads to my final question: the boiler feed water (BFW) flowrate shows significant oscillations even though the level fraction in V3 barely changes. The fluctuations are large and continuous throughout the 12-hour period. Why would such small level changes cause such large swings in BFW flowrate? Could this be caused by overly aggressive controller tuning, or is there something else I might be missing?
 

 

 

Attached Files

•  Plot1.JPG   73.2KB   0 downloads
•  Plot2.JPG   66.18KB   0 downloads
•  Plot3.JPG   61.31KB   0 downloads
•  Plot4.JPG   69.41KB   0 downloads
•  3.pdf   1.67MB   10 downloads

Edited by Kakashi-01, 20 March 2025 - 10:43 PM.

[Pilesar]

Dynamic simulation is an order of magnitude more complicated than steady state simulation. The controller tuning is likely at least some of the cause of the behavior you see. Review your feedback control loop tuning guides. If using only the Proportional band, the PV will not reach the new setpoint. Adding in Integral is needed to close the final gap. Tune each isolated loop individually for the expected disturbance and then again in the larger model to account for interactions. As to the heat exchangers, closer temperature approaches and lower flow rates will transfer heat duty at a slower pace. More heat exchanger area or a different heat exchanger design might help.

[Kakashi-01]

Thanks Pilesar! To be honest, I’m still trying to wrap my head around some of the dynamic behaviors I’m seeing in the system. I took your advice and started exploring how controller tuning affects performance. For the level controller on the shell side of the reactor (simulated as a flash drum), the default values were a proportional gain (Kp) of 1 and an integral time constant (Ti) of 240 seconds.

 

As a test, I disabled the integral action by setting Ti = 0 and reduced Kp to 0.1. I ran the simulation with just the proportional controller, and the results are shown in the attached image labeled “P controller.” The system eventually reached steady state, but rather slowly and with some lag remaining as expected with only proportional control.

 

When I added the integral term and set a small Ti = 0.1 s, the system reached the same steady state faster than with the P-only controller. However, when I tried Ti = 10 s, I saw significant oscillations possibly due to the integral term reacting too aggressively before the system had stabilized.

 

One thing that still confuses me is the behavior of the BFW flow rate. If the reactor heat duty is halved (as the shell-side level fraction drops to its SP of 0.5), then less energy should be available to generate steam. I expected the BFW flow rate to decrease, since less steam is being produced and more of the available heat would go into just raising the temperature of the feedwater. But that wasn’t the case the BFW flow actually spiked at first.

 

Also is bfw oscillating with lower amplitudes because of the flash drum size,  the level fraction (PV) equals the SP at all times despite the changes in bfw which is also shown by the long residence time in the image labelled "Vesseldata" .Could this be why the flowrate fluctuates so much in response to very small level variations?

 

I should note that I don’t have the freedom to change any equipment sizes in the simulation, so I’m trying to understand and work around these dynamics with only control strategy adjustments.

Attached Files

•  KP0.1TI10.JPG   78.71KB   0 downloads
•  VesselData.JPG   52.55KB   0 downloads
•  Pcontroller.JPG   67.29KB   0 downloads

[Pilesar]

https://www.controle...id-controllers/

[breizh]

Hi,

Consider these documents to support your work on controllers.

https://controlguru.com/

Breizh

Attached Files

•  PID controller.pdf   1.15MB   3 downloads
•  PID tuning.pdf   540.19KB   2 downloads
•  tuning.pdf   1.04MB   2 downloads

[Dacs]

I'm trying to come up with something to describe how I understand the system, but it seems that the circuit is intertwined: How heat recovery in E-3 and cooling in E-1 and E-2 behave depends on the relative design duty of those exchangers.

 

Before we start with the tunings, I suggest that using a converged model at initial condition (100%) flow, run a step change (100 to 50% flow) in Stream 1, and keep all controllers at manual and monitor your streams until it stabilizes.

 

Having that information will help you see how the the system behaves dynamically without tuning effects and you can use it as a starting point.

 

Another issue I noticed the pressure of Stream 5 decreased following the reduction in Stream 1 flow rate.

 

 

Does V-1 have any pressure control or does it float with MeoH Column?

 

At face value, you should not lose pressure in Stream 5 as you are reducing your flow -> less hydraulic resistance. Of course you also need to factor in the two-phase behavior of HEx E-1 and the piping in Segment 5. 

 

 With a lower pressure entering the flash drum, wouldn’t more volatile components be expected to vaporize, possibly reducing the liquid level instead of increasing it?

 

Not unless you have more cooling at your exchangers, which pushes Stream 5 to bubble point.

 

Also, the level controller output reaches 100%, but the PV (level fraction) is still only around 0.6 — well below the high limit of the controller range. Why does the controller saturate even though the level hasn’t hit its upper range?

 

 

Aggressive tuning perhaps. Improper range?

 

For Stream 9 (the syngas stream), steady state was reached approximately 20 minutes after the disturbance, but at a higher temperature. I think this is because the lower flow rate and pressure of Stream 1 lead to a longer residence time in the heat exchanger, allowing more heat recovery. 

 

 

Higher residence time is a consequence of the HEx operated below its design, which effectively have surplus area (since feed is at turndown), leading to more effective heat transfer.

 

The heat duty of the reactor also decreased due to the reduced feed. Since the shell side of the reactor is simulated as a flash drum (V3), and we set its duty to zero, less energy is available to generate steam.

 

 

If you keep Stream 10 temperature constant. 

But the way the model setup only holds if reactor tubes are flooded with BFW. If the tubes are not flooded (which I highly doubt, seeing the insane syngas temperature), then the model will not account for varying surface area.

 

The amount of steam decreases rather if we set the duty of the drum to 0 excluding the heat from the reactor then heat is utilized to increase the temperature of bfw and generate less generate because its supplied at conditions of 240 C and 48 bar. 

 

I can't I understand it fully, but if my understanding is correct, the main effect of lower R-2 flow is having a built up of  V-3 level because you can't fully vaporize the BFW feed rate (still set at 100% flow), until such time the controllers act on the change.

I am not sure why the bfw flow rates is oscillating wildly in response to very small changes in the level fraction.

 

 

If you have variation in V-3 pressure, then you need to account for liquid flashing inside the vessel to vapor (which is saturated) without needing heat input, whenever V-3 pressure drops, which leads to more HPS flow, which may lead to pressure backup and increase in vessel pressure, and the cycle continues.

 

Another thing is you are controlling BFW flow (instead of blowdown, which is fixed in your model), so you'd expect this to change.

[Kakashi-01]

I'm currently tuning each controller in my system individually. I started with the flow controller of stream 1 by performing a bump test. I set the controller to manual mode and adjusted the controller output from 0.50 to 0.55.

A couple of questions came up during the process:

 

Step Change Behavior: When introducing a step change in manual mode, shouldn't the change in controller output be instantaneous? 

 

Model Selection for Tuning: After generating the process response data, how do I choose an appropriate process model to fit the data (e.g., first-order plus dead time)? Any tips for identifying the right model and extracting the parameters needed for tuning?

 

Validating the New Tuning Parameters: Once the controller is back in auto mode with the new tuning parameters, should I evaluate its performance by applying a setpoint change and observing how the process variable (PV) responds? 

 

Attached Files

•  Scenario 1.JPG   32.54KB   0 downloads
•  XC1.JPG   60.12KB   0 downloads

[Kakashi-01]

Does V-1 have any pressure control or does it float with MeoH Column? At face value, you should not lose pressure in Stream 5 as you are reducing your flow -> less hydraulic resistance. Of course you also need to factor in the two-phase behavior of HEx E-1 and the piping in Segment 5. 

 

V1 has no pressure controller. Only the level of V-1 is controlled. so how can stream 5 not lose pressure when the flow rate of stream 1 has been reduced as a result of valve throttling?

 

I plotted the the temperature profile of stream 5. Initially the temperature dropped to 36 C which is expected since we are dealing with feedback loops. The temperature must measure the lower temperature of stream 5 and then the temperature control sends a signal to the actuator to reduce the valve opening because the cooling water is initially at the design rate and it would take time to bring down the flow rate smoothly. However, why does the cooling water return temperature increase to 52 C compared to 40 C (before the flow cut)? 

 

Not unless you have more cooling at your exchangers, which pushes Stream 5 to bubble point. Aggressive tuning perhaps. Improper range?

 

Yes, Stream 5 was definitely cooled to temperatures below the setpoint for about 2 hours, which I believe explains why the level increased above the setpoint. This triggered the level control valve to open in order to drain liquid from the vessel. However, I cannot fully explain why the valve output reached 100%.

The vessel has a volume of 1.8 m³, with an estimated residence time of about 5 minutes. The level controller operates within the correct range, between 0 and 1. I suspect the vessel is relatively small compared to the inlet flow, and the inlet flow rate is balanced by the liquid accumulating in the vessel, along with the vapor and liquid leaving it.

When the inlet flow was suddenly reduced, and considering that the duty is set to 0  (i.e., the liquid is not flashing), the liquid couldn't be vaporized to help relieve the excess volume. Therefore, it had to be drained more aggressively. This might explain why the valve reached 100% output to quickly reduce the level and compensate for the sudden drop in feed.

Attached Files

•  V1.JPG   55.42KB   0 downloads

[Dacs]

Step Change Behavior: When introducing a step change in manual mode, shouldn't the change in controller output be instantaneous? 

 

 

For the controller output, yes. For the OP, you need to consider the final element response time (such as valve stroke rate).

 

For due diligence, check if you have ramping enabled in your controller.

 

Model Selection for Tuning: After generating the process response data, how do I choose an appropriate process model to fit the data (e.g., first-order plus dead time)? Any tips for identifying the right model and extracting the parameters needed for tuning?

 

 

I don't do tuning myself (I'm not a control engineer), but my limited knowledge with process control is (a) first order cannot oscillate while second order can and ( dead time is inherent for processes where there's considerable distance between PV and OV sensing points, such as systems that involve interconnecting pipes with considerable distance in between.

 

Validating the New Tuning Parameters: Once the controller is back in auto mode with the new tuning parameters, should I evaluate its performance by applying a setpoint change and observing how the process variable (PV) responds? 

 

 

From the process point of view, good controls will try to achieve the least disruption in PV. I'm sure there are metrics on how to quantitatively measure it, but I'm not in the position to say in more detail.

 

V1 has no pressure controller. Only the level of V-1 is controlled. so how can stream 5 not lose pressure when the flow rate of stream 1 has been reduced as a result of valve throttling?

 

 

I got myself mixed up. I was thinking of a lower backpressure due to lower flow, but you're right. If you have another PC downstream of V-1, it should result in lower V-1 pressure due to decreased backpressure when you suddenly operate at turndown.

 

I plotted the the temperature profile of stream 5. Initially the temperature dropped to 36 C which is expected since we are dealing with feedback loops. The temperature must measure the lower temperature of stream 5 and then the temperature control sends a signal to the actuator to reduce the valve opening because the cooling water is initially at the design rate and it would take time to bring down the flow rate smoothly. However, why does the cooling water return temperature increase to 52 C compared to 40 C (before the flow cut)? 

 

 

 

 

However, I cannot fully explain why the valve output reached 100%.

 

Do you have the plot of controller outputs of both loops? 

 

The vessel has a volume of 1.8 m³, with an estimated residence time of about 5 minutes. The level controller operates within the correct range, between 0 and 1. I suspect the vessel is relatively small compared to the inlet flow, and the inlet flow rate is balanced by the liquid accumulating in the vessel, along with the vapor and liquid leaving it.

When the inlet flow was suddenly reduced, and considering that the duty is set to 0  (i.e., the liquid is not flashing), the liquid couldn't be vaporized to help relieve the excess volume. Therefore, it had to be drained more aggressively. This might explain why the valve reached 100% output to quickly reduce the level and compensate for the sudden drop in feed.

 

Keep in mind V-1 operates at saturation. In principle, if you lose V-1 pressure, the liquid inside will boil off.