Temp control with large dead time

I have an application in which we are trying to contol water temperature to 230 DegF using a 4KW heater coil wrapped around a Pyrex tube. The water passing through the tube is heated indirectly by the Pyrex tube. A control signal change results in a temperature change 6 to 7 minutes after the change to the heater control signal. Any advice to ensure that the the PV doesn't exceed SP by more than 1.5 DegF?

You need to indicate what is the velocity of the water inside your pyrex tube, and more importantly where is temperature detector is placed. Is the velocity of the water constant or varying? Having said that, pyrex is not the best of conductors of heat, so a delay is to be expected.

In such a situation, the use of the derivative element in your PID controller is highly recommended, but tuning up is required. You would have to put more weight on the derivative element rather than the integral element. If you have varying water velocities keeping within the 1.5F overshot limit may be tricky.

Joe -

As you probably know, it is the large dead time is what makes the process hard to control. So the first step is to ask if there is anything you can do to reduce the dead time. Why do you have so much dead time? Sometimes you can relocate the temperature measurement and get some improvement.

Now, let's look at how the controls will respond to an upset. With purely feedback (i.e. PID) control, the best performance you can reasonably expect is a return to setpoint in approximately 10 x the dead time. Given your 6 minute dead time, this means approximately 1 hour to return to setpoint. The degree of the excursion from setpoint depends on the characteristics of the upset to the process. I am going to assume that 1 hour response is probably not acceptable...

So we need to look at a few other options.

Option 1: Feedforward control. If you can measure the inlet temperature, then you can start heating up the tube as soon as the INLET temperature drops. You don't have to wait until the water temperature at outlet starts showing problems. This is like driving a car up a hill - you step on the gas BEFORE you start up the hill. Click on this link to see a recorded webinar on Feedforward control. http://www.expertune.com/r2.asp?f=AList&l=pastwebinar.asp?name=FeedforwardAn dRatio200703Mar20&nameDesc=Feedforward%20&%20Ratio%20Control

Option 2: Dead-Time Compensation You can apply a dead-time compensator, which will use your knowledge of the process to more aggressively control the temperature. Keep in mind that dead-time compensation techniques are sensitive to the accuracy of the model. So be careful if your process dead time actually changes over time. For more info, do a web search on "Smith Predictor".

Option 3: Process Re-Design This is usually the most costly option. But consider, for example, a small feed preheater, with fast response, to deal with feed temperature upsets. Or, even better, re-design the heater to get rid of the large dead time.

Good Luck!

George Buckbee
ExperTune, Inc.

Joe

You may try several approaches, including MPC, feedforward, etc. But you will get much better results reducing the time lag. 4kW seem to be a lab experiment. If you are heating only water, can't you put the heater coil inside the tube?

My 2 cents.

Marcos

I would use a cascade loop with the secondary loop controlling the heater element temperature. I am assuming the heater element has quite a lot of thermal inertia and this is the reason for your 6 minute time lag. As your water approaches set-point the element temperature reduces to avoid overshoot.

Roy

if you can not relocate the temp. sensing element, then attaching another element at the entrance will allow the system to know what it is dealing with and get a head start, and type of control would not be with pid this will never be able to compensate, there is procedure known as sampling control that the output controler will hold steady for the 6 minutes and then adjust after sampling the reading and along with the feed forward you controller can start to adjust for what is comming in.