Temperature control in laboratory using strange algorithm?

Hmmm..... OK - here's a pebble in the pool:

How big is the incoming duct? If the duct is big enough then there is scope for "plug flow" in the duct, where most of the streamlines in the duct are parallel, and there would be no mixing. All it would take for the effect to happen is for the heater or cooler to be offset to one side of the duct, or even over-sized so that there is a temperature variation from one side of the duct to the other, and then some parts of the incoming air would be warmer than others (obviously the configuration of the duct and the heat exchangers cannot be seen from here).

I think the hot and cold air are not mixing well. This could explain the 0.5° variation at the benches but nothing else, as I see it.

No, I have not seen this particular problem before. But, if you do not want to go crazy in the process of guessing, you better step out of the loops. You clearly have a data logging system, good. make use of it, but nothing else for your control.

Get some diodes, feed them thru constant current diodes fed by batteries. Get as many as you want to cover all physical points you think you need to cover. The beauty is, that the voltage drop at constant current ONLY dependent on the temperature. That is basic semiconductor physics. You do not need to calibrate anything, as you are only interested in verifying temperature variations (or none of that) and tracking down the source of it. This will get you to the core of your problem quickly and cheaply, no matter where it originates. Nobody, but nobody cares to argue with the results of such basic physics. Let us know about your experiences.

It seems, I did not express myself clearly, judging by the replies coming in since. So, here I attempt to lay it down clear as I can.

When you are confronted with a feedback system of any kind with plenty of variables (some not even expressed in the spec.) and you have problem, you have 2 choices:

1,. You cut the loop somewhere. You run it the as open loop, measure at many points, track and finally correct the problems. Where to cut? How to supply everything to make it look like as the original closed loop?

2,. Or you leave the closed feedback (that you do not know well enough) alone to do whatever erratic things it does. ADDITIONALLY you set up an as simple as possible open loop data gathering network. Unimpeachable quality is important, calibration is usually not at all.

Using #2, You can track down the erratic behaving component, since you have every reason to believe the measurements of #2 over what your loop tells you. I cannot tell you here and now, what the problem turns out to be. After all, the manufacturer would have eliminated it already, would they be aware of it. But, heck, we are engineers, and this is a case of discovery.

What I KNOW from long experience: if you wait for others to take serious interest in your problem's resolution without YOU TAKING THE LEAD, you have plenty to learn. You own it, you are more motivated by far, including the manufacturer.

Everything I said in the original note stands. We would like to know about your trials and tribulations. And we might be able to give further advice along the way.

These kind of snags are extremely difficult to trace from the distance...

However, you wrote: "It is extremely unlikely that the sensors are intermittent..." but it is not impossible. This is the VERY first place I would look at. and if all ok, it looks to me as a loose contact. the character of the fault, being occasional in my opinion proves just that. Some wire that is shaken by the movement of the air stream?, two different of types wire conductors spliced together and acting as a thermocouple? it got to be one of those stupid hard to detect snags.

you wrote:.."Does the internal controller have a strange algorithm?" I seriously Doubt it. No one that builds a lab like yours will be using a "strange algorithm..."

Wangito.

The sensors are thermistors, I believe, with a metal hood around them. There is very little air flow in the pipes, so they don't shake around. The "strange algorithm" was suggested by an engineer with experience in this area. He was referring to averaging, filtering, etc. What I was getting at was when 1 zone shows a 1° change, the other 3 change the opposite direction, keeping the average almost constant.

The thermometers are maybe Digital and are working a degree at a time 20°C, 21°C etc..?

Or do they show 21.2°C, 21.3°C, 21.8°C etc etc?

If you are referring to the TAC readouts on the computer, they show 0.01° resolution (i.e. 22.74). Our quartz thermometers have 0.0001° resolution. The fellow who took the most data used a calibrated thermistor and read it with an 8 1/2 digit multimeter.

We have a company that takes care of our heating / air conditioning units. They don't want to change anything as long as the system is under warranty, so the call TAC. The frustrating thing is that we are not getting feedback as to whether they are finding or changing anything, and we don't have time to mess with this anyway.

OK its not a resolution problem as I first suspected.....

The thermistor is not getting a constant supply of current. First check all connections and then check power supplies. Make sure you have proper shielding as currents can be induced depending on the equipment starting and stopping.

This sounds like a very plausible fault condition analysis. If you can determine that the thermistors go high when the power goes low then an intermittent supply problem causing an incidental high reading combined with a fairly normal temperature averaging algorithm would explain the complete system behavior you observe. If the thermistors are reasonably low voltage then adding a battery somewhere in the circuit could bridge the intermittent fault condition without necessitating locating and correcting it.

The battery will probably not trim spikes as readily although it can't hurt. Spike trimming can be accomplished with a wide variety of circuits but the warranty issues may preclude doing much of anything.

The assumtion is the problem is with the actual controllers. Has any consideration been given to air flow, and/or CW/HW valves reaction times?

Sorry to state the obvious, but, do you allow people in these labs: they tend to emit an awful lot of localised heat.

hi there, a lot of good thoughts here already.

Just ask for commissioning data and verify sensors calibration. Then review how they set-up PID loops and how they mapped input and output points for the controller.

What type of air system you have got, is it constant volume or variable air flow?

If you concerned about actual temperature at the benches then set different zones related to actual locations and stop controlling your average room temperature.

Thanks for all you comments, everybody. For those that are interested, I will try to update at least once a week.

I am not the person who makes the repairs. We have JCI to do that. I cannot cut any loops or put batteries in anywhere (especially for a system I don't have documentation on that is under warranty). The thermistors can't be traced out - the wires are in the ceiling, then I think go into a conduit to a control panel. I don't have access to a stepladder. We cannot change the design to have several separately controlled zones. There is normally 1 to 2 people working in each room, but equipment comes in and out, so the heat load is changing from day to day and quicker. The air flow is constant, and there are no automatic valves in the air flow.

Today I tightened about 200 connections in the control panel due to your comments. One problem is that the guy who had the most time to make measurements broke his ankle, and will be out for several weeks. That means more for me to do, and less time for experiments. Also today we put a quartz thermometer in the pipe of a zone that was reading higher than the others in that room. We found that zone to be reading 0.5° high. We asked JCI to check with TAC to see if it can be calibrated from our measurements. We have not heard back yet. Getting these zones calibrated seems like a logical first step, then the readouts will make more sense. This is going to take a long time to resolve.

S

Hi,

I am not aware of your situation, are you a tenant in the building or owner, etc.

It will be difficult for you to resolve if you were not involved in engineering on design stage when they put together sequence of operations.

The only way to receive answers or get it solved is to approach technical staff of your building... especially on warranty stage.

Good luck.

Most controls of this type are not process controls. HVAC does not usally have to maintain this tight of control. Maybe a higher end control should be looked into.

Time for an update. The lab is in a building with an admin area. There have been temperature control problems there too (worse than the lab). There is a separate system for the lab and the building (admin area). People have put cardboard in registers to divert or block air flow to make it more comfortable. The thermostat in that room has been a comedy of problems. No manual was supplied. After many hours of 'research', one guy determined that there were 28 zones, 4 for each day of the 7-day week. They were not all set properly. After that discovery it worked well for a few weeks until a very cold day when the area went up to about 80 Fahrenheit. There was a thermistor in the thermostat that was touching the outer wall (where the thermostat was placed). It was bent away from the wall, and later a bracket was made to move the whole thermostat away. The hole where the wires come in was plugged with a tissue. This helped considerably, but the thermostat is still not in a good location (too close to the outer door).

On the north and east sides of the lab is about a 12 inch space to the outer building wall. A 3-4 inch space was intentionally left into the admin area for air circulation around the exterior lab wall. Even though the outer building wall has insulation, the air coming out there is cold. One guy had plugged it up about half way with foam. The foam was removed because management wanted the temperatures inside and outside the lab wall to be measured in several places to see if the weather is affecting the lab temperature balance. The guy there by the crack was so cold he was threatening to go home. Cardboard was put to extend the wall past him so he wouldn't be cold.

I was not involved with those temperature measurements, but I think the conclusion was that the weather has little effect. I thought it was a wast of time in the first place, but at least management has ruled out one thing. We still have not heard back from JCI whether the zones can be calibrated from our measurements. I'm having a hard time convincing management that they should be.

So, we still have some problems, but what do you expect for 2 million dollars?

Time for another update. We had a meeting with JCI, and told them of our measurements. The concern was of conduction through the exposed foundation. It now has 1 inch of foam insulation on it. There will be 2 inches later on in the year when the weather is warmer. The Lab rooms have been behaving well lately. We measure a sensor when we get time. Management is not ready to request a drop in the set point, but it came down a little last week. The reason is unknown.

Time for another update. The temperature has been too cold again in the admin area in the afternoons. I found the manual for the thermostat and changed some options. Now there is only 1 zone instead of 28. The thermostat is placed too close to a computer monitor and blocked from circulation by some papers and things. When the AC comes on, the thermostat has to go down 2 degrees before it turns off. The cooling has too much capacity, so it gets too cold. I had to raise the settings to warm up the room. On the other end of the room it was getting too hot especially in the mornings. Turning on the fan continuously helped this problem. Nothing has been done in the lab area yet.

Another update. A coworker discovered that the thermostat in the admin area had provisions for an external thermistor, so we ran a wire, and put one where the thermostat should have been put in the first place. After adjusting the set points, it is working better than before.

In the lab we had a calibrator fail, so it was shut off. It was in a rack near to one of the sensors. This caused a change in that sensor, and affected the temperature in the center of the room where we officially read the temp. When the calibrator was replaced, everything went back the way it was. This proves once again that we don't have enough air flow coming down to force the heat down to the intake registers. I would like to replace some ceiling panels with solid ones in the center of the room to give more air flow on the edges, but money is very tight right now, so that won't happen. At least management has given to go ahead to have the set points changed, so we can get the average temperature right.

We asked for a change in set point for temp & humidity. The humidity came down as we wanted. For temp, they put in a "bias" instead of changing set point. One room came down for a few minutes, then went back to where it was. The other room never changed as far as anyone could tell, so no progress yet!