Mitigating tank level swings due to pressure spikes

I have a tank which is vent to atmosphere and reads level using a pressure transmitter at the bottom to measure hydrostatic pressure. Since the geometry of the tank (area per foot height) is relatively constant, I just do a linear scaling from PSI to %.

This works just fine when the tank is at a steady temperature. However when the temperature rises quickly, especially when being filled with hot water, the pressure will temporarily spike and give false level readings. The pressure resolves in about 5 seconds and the level readings return to normal.

Is there an elegant solution to dealing with this? A filter of some sort to dampen these spikes?

Use a differential pressure transmitter with the low port plumbed to top of tank above highest level of liquid. This way if vapor pressure is pushing down on liquid it is accounted for.

I do this by setting up an "Average" function in the plc. I then don't update my level reading until let's say every 1000 scans, or maybe every 500 scans. Maybe in your case you update the level read every 7 seconds or 10 seconds.

Just an Idea simple and NOT much cost.

Bob O said:

http://www.plctalk.net/qanda/showthread.php?t=108849&highlight=filter

Click to expand...

I use Mickey's simple filter logic by default for analogs and if I it turns out that don't need it, I just set the filter constant to 1 to defeat it. It works quite well for me with water tank levels in cases where there are short term disturbances. In my cases (municipal water storage tanks measured with pressure sensors like yours) the realistic rate of change of the tank levels is quite slow, so I can apply heavy filtering and make noise or erratic data disappear and still not harm normal operation.

Is the problem that you're trying to fill the tank to a specific level and you can't trust the level transmitter while you're filling?
I ran into a situation like that once. The fill pipe was vertical into the top of the tank and the force of the incoming water upset the signal from the load cells. Fortunately in that case, we had coarse and fine fill valves and the fine fill rate was slow enough so it didn't cause a problem.
At any rate try to find out what's causing the upset in the level transmitter signal while you're filling and see if you can't mitigate that before you try to "fix it in the program".

I will try some of the suggestions posted. The filter seems promising.

To clarify, the problem is that the tank is vented via a 3 inch pipe to the atmosphere. We can add upwards of 325 BBLs an hour to the tank. This is fast enough to build pressure in the headspace faster than it can be vented, especially when hot liquids are added. This causes the pressure spike and the volume reads 20-40 BBL higher than what is actually in this 200BBL tank.

Most of the tanks I deal with have an upper and lower pressure sensor, so it's easy to use a differential pressure. For this process, I only have the lower sensor to work with. I know there are better ways to measure liquid volume, but I just need the best solution for what I have.

Can you slow down the rate of filling when you approach the target volume? Specifically, at some volume before you reach the target volume, slow down to a fill rate at which the vent can catch up and you can start believing the level transmitter.
Adding the filter may turn out to be the best solution, but be aware that it will introduce some lag time. Unless you include logic to apply the filter conditionally, it will also filter the signal when you are removing material from the tank. Maybe that's not a concern.