OT Process question

[cjd1965]

Hi I have a hot fluid in a pipe. call it water. It is heated to 90 degress and passes through a heat exchanger which can cool over 10 sec and get the fluid to 20, and has an on/off valve for the HX cooling water.

I have been asled to try and come up with an alogithm to provide reasonably accurate batching to different temperatures. The 90C and the cooling water are fairly consistent.

I wanted to put a modulating valve onto the HX but I am not allowed due to costs(i said it will cost mor not to lol)

Any ideas? think of filling a bath with some hot and then some cold but i need to get the mixer percentage figured out

Cheers

 

[OkiePC]

It might be an application for time proportioned PID. Use a PID but then take it's analog output (CV) and convert that to pulse widths that get sent to the valves. So, both valves are opened at the start, then the water ends up to hot, so the PID reacts and lowers the pulse width time from 100% to 90% for the hot water. At some point, it continues to shorten the ON time for the hot water valve until the temperature is satisfied.

Do you have a temperature signal of the output water available in the PLC?

How long does it take for a temperature change made at one of the discrete valves to begin to affect the temperature of the water?

Any ideas? think of filling a bath with some hot and then some cold but i need to get the mixer percentage figured out

Cheers

Actually, if your valves are two position (fully opened or fully closed), it's quite a bit different than that.

 

[cjd1965]

Hi Okie
I did consider the time proportional thing. The main issue is that TT is in the pipe after the HX going towards the taget vessel.
At the start of a batch the pipe is empty so ambient temp. If a batch of say 60 deg is needed,The pid ramps up for heat due to TT being ambient, then gets hit by the 90 liquid, so the pid cools rapidly so it is swinging a lot

Also not helped by minimum batch size of 50L so i was considering some alogithm to dose x% at 90 then y% at 20 kinds of thing. I can cool the HX down on its own but dony know what its temp. is

cheeres

 

[Lancie1]

I have been asked to try and come up with an alogithm to provide reasonably accurate batching to different temperatures.

This is a thermodynamics heat transfer problem. You need equations that calculate the heat content of a mass of "water". Then you reverse-calculate to determine how long to flow hot and cold liquids (each with certain heat content per mass), to achieve a certain temperature.

 

[cjd1965]

Hi Lancie thanks for that i gonna hit up google and have a look see

 

[CX_Luigi]

Hi I have a hot fluid in a pipe. call it water. It is heated to 90 degress and passes through a heat exchanger which can cool over 10 sec and get the fluid to 20, and has an on/off valve for the HX cooling water.

I have been asled to try and come up with an alogithm to provide reasonably accurate batching to different temperatures. The 90C and the cooling water are fairly consistent.

I wanted to put a modulating valve onto the HX but I am not allowed due to costs(i said it will cost mor not to lol)

Any ideas? think of filling a bath with some hot and then some cold but i need to get the mixer percentage figured out

Cheers

you can use the formulas in the attached document to get batch temperature roughly at any desired point between hot water and cold water temperatures. I hope this will solve your issue.

 

[OkiePC]

Hi Okie
I did consider the time proportional thing. The main issue is that TT is in the pipe after the HX going towards the taget vessel.
At the start of a batch the pipe is empty so ambient temp. If a batch of say 60 deg is needed,The pid ramps up for heat due to TT being ambient, then gets hit by the 90 liquid, so the pid cools rapidly so it is swinging a lot

Also not helped by minimum batch size of 50L so i was considering some alogithm to dose x% at 90 then y% at 20 kinds of thing. I can cool the HX down on its own but dony know what its temp. is

cheeres

Straighr proportional control would work quite well if you were mixing water and could measure temperature and flow.

For your heat exchanger, you could still use a PLC based PID. You just have to manipulate the mode and controls of the algorithm to deal with the process pattern, including the empty pipe.

I would expect some sort of manual mode and biasing schedule in the PLC code. So, for example, you start a batch, put the PID in manual mode and set its CV to Stage_1_CV(recipe_n) which can be retrieved from a recipe table nad easily altered and saved.

Once you figure out the lag time of the process and the typical CV values you need for certain products, but without some form of PID or automatic means of adjustment, I don't think a straight ratio is going to be accurate and repeatable long term.

 

[Lancie1]

CX, I think this may be a very workable approach for the situation.

So CJDs algorithms would become:

Volume of Required Hot Liquid = (Tdesired - 20)/70 X Volume Total
Volume of Required Cold Liquid = (Tdesired - 90)/70 X Volume Total

Next question: What is the Total Volume of liquid to be heated?
If unknown, or measured with a flow meter, then you have to assume some units, such as 100%, or a volume of "1". Then Volume of Hot/Unit Volume = (Temp Desired-20)/70, and Volume of Cold/Unit Volume = (Temp Desired-90)/70

Test: If the Temperature Desired = 90, then:
Vhot = (90-20)/70 X Vtotal = 1 x Vtotal
Vcold = (90-90)/70 X Vtotal = 0 * Vtotal

Test: If the Temperature Desired = 80, then:
Vhot = (80-20)/70 X Vtotal = 0.857 x Vtotal
Vcold = (80-90)/70 X Vtotal = 0.143 * Vtotal

Notice that the 0.857 and 0.143 are not "% Valve Open" but can be converted to that setpoint by figuring out the ratio of Valve Opening to Volume Flow Rate.

Because you only have Open/Closed control valves, all you can do is to use "% Open Time" and figure out the ratio of the Hot Flow to the Cold Flow when valves are full open. Are they equal, or is the Hot Flow greater?

 

[OkiePC]

But Lancie, that is for mixing water and won't take into account the properties of the heat exchanger will are likely to be non-linear, and/or variable. It might make a good starting point to develop a fairly simple math based approach.

The locations of temperature and volume (flow?) sensors are going to be critical to success for short batches.

 

[Lancie1]

Yes, but he didn't give us any heat exchanger properties except to say it cooled down to 20 degrees in 10 minutes. Maybe he wants to get into the real heat transfer equations. Not me, I had to take Thermodynamics twice and still didn't like it. Calculus, differential equations, "Valve X opens for 1 minute at 5 gpm and Valve Y opens for 5 minutes at 10 gpm, now calculate how long Mrs. Z will stay in the bathwater..." I am just too old and life is too short for such shinagians!

 

[cjd1965]

Hi all
thanks a lot for the ideas and input you have given me so far. I have reviewed the system today on site. Basically the product is water based but has powders added. It is constantly circulating from a break tank at 90C to 3 vessels. Each vessel has its own heat exchanger etc so i am going to talk like there is only 1 system

Basically the heat exchanger has on/off control and the cooling water is pretty constant temperature, so the HX can cool 90 to 25 in approx 40 seconds. The weird thing is there is a modulating valve on the process line after the HX so there is a weird idea of using the flow rate to control temperature.

There is also a process on/off valve before the HX so i can cool the HX with no product, or run the product with no cooling, and can restrict the flow

I am leaning towards the data provided by CX as 1st try but open to other input. I guess its my turn to drink from the poisoned chalice.

cheers guys

 

[Peter Nachtwey]

The second equation in CX's pdf should have Thot-Tdesired.

I would use this as a feed forward or good estimate. The next batch I would modify the Tdesired using the error from the previous batch. Because the heat transfer will occur over time I doubt CX's formulat will ever reach Tdesired. I would replace Tdesired with a Ttarget that is equal to Tdesired+Tcorrection. I would modify Tcorrection using a low pass filter.
Tcorrection(n+1)=Tcorrection+k*(Terror-Tcorrection)

 

[OkiePC]

What Peter has there makes sense. His additions provide a means of adjusting to get the system to represent the actual heat transfer.

After your last description of the system I need a P&ID drawing to contribute further.

Where are these valves, vessels, and sensors schematically?

Paul

 

[cjd1965]

Hi a crude diagram attached. All equipment is plc controlled.

Cheers

 

[OkiePC]

Okay, can you tell us the size of the vessel (volume), and put some flow specifications on those valves?