Out of my depth - Feed forward, Cascade?

[rupej]

I would do a combo of what is suggested above. Do the proportional level control with filtering, plus have a high level setpoint to force the pump to max speed, and a low level to shut it off. That gives you a smooth control most of the time, but also a backup plan in case the filtered response doesn't keep up.

 

[osmanjdt]

use comparator logic

 

[Paullys50]

Remember my goal is to keep the output as constant as is reasonable. Not exact, that's impossible...

What is your definition of "constant" and "reasonable"?

 

[osmanjdt]

i remember this word ALARA.

 

[Tom Jenkins]

A primer on flow: Level control is generally used to make flow out of a wet well match the flow in. This is common in sewage pump stations, for example. Flow control is generally used when a process requires constant water flow. Supplying water to an irrigation system would be an example.

You haven't identified the process, so it is hard to pin down your objective. I am going to assume centrifugal pumps. Let's also assume a flow meter on the effluent pipe and a level transmitter on the reservoir.

You indicate that you want constant flow out of the reservoir, which indicates that level fluctuations are acceptable (and inevitable). Use any feedback control, PID or floating control or whatever, to modulate pump speed and compensate for head changes across the pump from discharge restrictions and reservoir level. Flow is the setpoint and controlled variable, speed is the manipulated variable.

If the influent flow rate is higher than the discharge flow rate the level will rise. If influent flow rate is lower than discharge flow rate the level will drop. Filtering the flow and level signals just takes the "bounce" out of the process measurement - it doesn't change the physics.

I suggest that you have high and low level setpoints with a deadband on each. If the level rises above the high level or falls below the low level you go to a cascade control. That would incorporate a feedback control on the tank level. The output of the feedback control would be a flow setpoint, which is a common level control loop technique. Have the flow setpoint to the flow control loop increase and decrease as required to maintain the variable flow setpoint.

When the flow rate setpoint from the level control loop is less than or equal to the original flow setpoint, or the level gets below the high level setting minus the deadband, go back to flow control based on the original setpoint. Invert the logic on the low level.

 

[baldone]

What technique would I employ when I have a tank that I am pumping out of at the same time water is flowing into it and I want to keep the flow as consistent as possible while making sure the tanks level doesn't get out of hand?

Currently I focus on the tanks level and a PID controls that quite well. A side effect due to the inconsistent inflow is that the VFD driven pump changes speed to maintain the level and the effluent flow obviously varies along with it. Sometimes quite a bit.

I would like to change this configuration so that the effluent flow remains much more consistent and the tank acts as the influent accumulator and changes to the level are permitted within a certain boundary with only a slight increase/decrease being felt as effluent flow.

I of course will need to sacrifice consistent effluent flow should the Tank level change dramatically or get to a dangerous level, but otherwise a couple feet of variance is OK if it would give me more consistent effluent flow.

What technique would I use to create this 2 input type of control? I have never created a cascade of PID's or "Feed Forward" (but have heard the words before ) so I would need to do my homework to figure out how to construct one, should one of those options, or some other, be the proper technique to use in this instance

I would greatly appreciate opinions on the correct control strategy I should follow.

I make a four point chart that would either generate an inflow target based on tank level or an outflow target based on a tank level if you are more concerned about a consistant inflow rate. the four points make up 3 seperate linear scales.

Essentially the points on the chart are flow level vs flow target with the outer points being min and max allowable flow.

 

[kamenges]

Personally, I would treat this less like a classical setpoint control problem and more like a process logic problem. Also, as you think about this, separate the analysis of the functional requirements of the system from the programming techniques you will use to accomplish that. Said another way, figure out what you want to do before you try to figure out how you are going to do it. Trying to do both at the same time is pretty difficult.

The easiest way to come up with the "what you want to do" part is to imagine you are the pump speed command generator. You have only the information available to the existing controller, which I believe is simply tank level. You also have a dial that sets pump speed. Given your desired process constraints and the available information, how would YOU decide what the speed should be. As for myself I wold think I would let the level wander around all it wants in the middle 50% of the tank but I would be VERY interested in how quickly the level is changing.

Keith

 

[Peter Nachtwey]

I would be VERY interested in how quickly the level is changing.
Keith

Yes, I still think the low pass filter idea with the proportional band is the best idea but I would augment that with a new twist. I would make the time constant for the low pass filter variable as a function of the rate of change and the level. I would make the time constant very long for when the rate of change is small but shorter when the rate of change AND level deviate from a quiescent state. This way the time "constant" would be very short if the level and rate of change is high so that the tank doesn't over flow and yet the time "constant" can be much longer for when not much is happening.
Changing the time "constant" is easy.