Mark Hydeman of Taylor Engineering recently published and article in the ASHRAE Journal titled Humidity Controls in Data Centers; Are They Necessary? One of the topics explored in the article was sensor accuracy, specifically humidity sensor accuracy.
The article included data from a recent project Mark had worked on and one of the outfalls of the lack of accuracy was that several of the CRAC (Computer Room Air Conditioning Units) were fighting each other. Some units where actively humidifying while others were actively dehumidifying. The result is that unnecessary energy (money) was being spent to provide a level of control that the article postulates may not even be necessary.
That article led to an article that was to be published in CSE this month titled Money Sensors in which Mark looked at other instances were sensor accuracy and other control system issues could cost a lot of money if they were not properly addressed by the design, commissioning, and operation of the system. I ended up doing a peer review of the article and, as the result of a couple of comments I made, Mark and Michael Ivanovich (editor-in-chief of CSE at the time, now the President of the Ivanovich Group) asked me to do a couple of side bars for the article.
Since the demise of CSE resulted in the article not being published, and since I had already created the side bar content, I thought I would just share it here, starting with a discussion of relative accuracy. Even though absolute accuracy is important, at certain points in our HVAC systems, from an operational standpoint, the relative accuracy of sensors controlling a process may be more important, as is illustrated in this example where two sensors that meet a project’s accuracy specification of ± 0.5°F produce an unintended result.
The system that I will use as the basis of discussion is illustrated below and consists of a 100% outdoor air handling unit with an intake damper, filters, a face and bypass type preheat coil, a chilled water cooling coil and a supply fan.
The system is located in an environment where temperatures are below freezing part of the year, thus the preheat coil is independently controlled to protect the chilled water coil. A freezestat provides an added measure of protection by shutting down the system if the preheat process fails.
The sensors controlling the preheat and cooling coils were required to be accurate to ±0.5°F by the project specifications. By pure coincidence, the fan discharge sensor is at the upper end of the accuracy spec while the preheat coil discharge sensor is at the lower end of the accuracy spec.
On the day in question, the outdoor air temperature is a pleasant 55°F, which happens to be the set point required by the system’s controllers. The system has been properly tuned and calibrated and PI loops are being used to eliminate proportional error.
The Starting Point
I’ve added a graph to the illustration which will allow us to trace the actual temperature of the air as it moves through the system and is processed by the various heat transfer elements, which are controlled by sensors with minor errors both in terms of absolute accuracy as well as relative to each other.
The colored dots indicate the actual air temperature (green) as well as the temperature the preheat controller (red) and discharge controller (blue) would see at that point in the system. We start with 55°F air entering the system, which the preheat controller would see as ½°F too cold and the discharge controller would see as ½°F too warm.
The Preheat Process
Since the preheat controller thinks the air is ½°F to cold, it modulates the system’s face and bypass dampers to raise the temperature of the air to its set point of 55°F. But, since the reason the controller thinks the air is too cold is that it is reading the temperature ½°F low relative to reality as the result of its accuracy tolerance, the real air temperature is increased from 55°F to 55½°F.
From the perspective of the discharge controller, the air entering the cooling coil is now 1°F warmer than it should be.
The Cooling Process
Responding to what it believes is a supply temperature that is 1°F over set point, the discharge controller modulates the chilled water valve to cool the air to what it thinks is 55°F. But, because of its accuracy tolerance, it actually delivers the air to the loads ½°F cooler than desired.
The Bottom Line
The bottom line is that sensors meeting their absolute accuracy spec have taken air that was at the required temperature and used energy to simultaneously heat and cool it and deliver it at a temperature different from what is desired.
Intellectually, I think I realized that this could happen fairly early in my career when I started to really get involved with the control design process. But the reality of it never really hit me until I was a facilities engineer at a wafer fab and encountered the problem in one of our make up systems.
When you are operating systems, frequently the temperature difference across the various heat transfer elements is more important that the absolute accuracy of the sensors. In the next post, I’ll look in detail at how we addressed this issue at the wafer fab and will hook you up with a resource that can help you do the same in your facility if you so desire.
Senior Engineer – Facility Dynamics Engineering