QUESTION: How can I improve the temperature regulation of an aluminium anodizing tank (±10 m³) with a temperature controlled internal cooling coil? This loop oscillates by 6 °C to 14 °C (and there are different working loads in summer and winter) because there are four tanks/exchangers using the same cooling water. Actually, the control temperature sensor is located on the tank, and there is an important dead time. I think that the first thing to do, if possible, is to minimize this dead time. Do you think that putting the control sensor in the outlet of the exchanger would help the process go better? See the drawing below (Figure 1). This way the system should have less dead time. The regulation is made via a three-way valve, two in, and one out. The actual control valve installation is quite strange for me. The liquid of the tank may take 15 minutes to be completely displaced.
Ibon Torrealdea Zarrabe / ENI S.A. / firstname.lastname@example.org
The Proposed Solution
Zarabe's question: Should the temperature sensor be moved?
ANSWER: This is a classic case where the “human element” in your plant (your boss) makes a big difference. If he/she is over 50, just put in a CW circulating pump to fix the water flow rate, a slave TIC to eliminate upsets caused by CW temperature variation, and use the traditional cascade loop shown (Figure 2) without moving the master.
If your boss is under 40 and knowledgeable about process control software, suggest an artificial neural network control (ANNC) system. In that case, you would measure the temperatures at all points, plus any other relevant variables, and would let the ANN train itself by automatically building and continuously adopting its model to the actual dead times, time constants and nonlinearities of the process. If you want to learn about ANN, read Chapter 2.18 in the 4th edition of Volume 2 of The Instrument Engineer’s Handbook.
If you're over 50
A traditional solution to the problem is to add a CW circulating pump to fix the water flow rate, a slave TIC to eliminate upsets caused by CW temperature variation and use the traditional cascade loop shown
ANSWER: The suggestion of moving the sensor to the outlet of the exchanger to reduce dead time is a good one, though incomplete. Due to heat generated in the process, the bath temperature will be higher by a variable amount. I suggest using cascade control, where the bath temperature controller sets the exchanger outlet temperature in cascade.
Greg Shinskey/ process control consultant
ANSWER: Without knowing any numbers, it’s not possible to know the several time constants involved. And without knowing the importance of the temperature control, it’s not possible to know the monetary importance of the control.
I suggest that the temperature “from” the anodizing tank should be more representative of contents temperature than the temperature “into” the tank. If the tank temperature changes much slower than the outlet of the heat exchanger, it might be advisable to use a cascade system to make response more stable.
Use two sensors and two controllers. The “secondary loop” measures the temperature from the heat exchanger and controls the heat exchanger valve. The “primary loop” measures the temperature in the tank or exit the tank, and the output of the primary controller provides the set point to the secondary loop controller. The primary advantage of this is to isolate any changes in the temperature or pressure of the fluid controlled by the secondary loop. These may change more quickly than the primary loop temperature. The secondary loop should be considerably faster than the primary, and this is tuned for good fast response. The primary is tuned to match the slower response of the tank.
Cullen Langford, consultant / email@example.com
ANSWER: On the anodizing tank temperature control, I would not move the sensor because then the process variable to be controlled would be unknown in the future. You could get very precise control to 18 °C of the exchanger outlet, but have no idea what the tank temperature is, and that is the goal. The engineer is correct in realizing that this would remove dead time, but has made a fundamental error in not monitoring the actual process variable he is trying to control. The remedy is to put in more temperature sensors, or just manually make hourly temperature measurements in the tank and heat exchanger inlet and outlet, cooling water temp, etc., until he learns his process, and then make a decision.