Data collection and analysis has always been an integral part of any commissioning process. But modern electronics have increased the sampling frequency, accuracy and presentation of the data by several orders of magnitude over the more traditional approaches (see Figure 14). Experienced retrocommissioning providers, facilities engineers, and operators all know that most buildings will ďtell youĒ where their problems are if you spend a little time looking at what is going on. The data handling capabilities of modern data loggers and DDC systems provide a powerful tool for taking a detailed look at how a building is operating in a variety of test situations:
∑ The trending software and data loggers can be used to create a running record of the system performance during a functional test sequence. This will allow the test team to go back and review what happened if there are questions or if things were happening so quickly they could not absorb everything that was going on. The data sets also provide a record of the test results and a baseline of comparison for future testing efforts.
∑ The trend data collected after functional testing provides an ongoing record of system performance. Setting the system up to archive the data essentially places the building in a perpetual functional test mode where Mother Nature writes a new test sequence every couple of days. The information documented can be invaluable to commissioning providers and facilities personnel as they work with the building during its first operating season to fine-tune its systems to the real world environment. Continuing the process after the first year provides a good foundation and baseline for an ongoing commissioning process and measurement and verification programs.
∑ Many times, trending can be used to streamline the functional testing process on a project with a tight budget or tight schedule or in situations where equipment-shipping delays conspire with a firm occupancy date to erode away the original window targeted for functional testing. The information gleaned from trend data analysis can be used to focus functional testing efforts on the systems and problems that will yield the most benefit for the effort expended. Issues not immediately addressed can be placed on a list of projects and optimization opportunities to be pursued by the facilities staff as part of an ongoing continuous commissioning process.
A good operator with a clipboard and a pocket full of gauges was the original data logging and trending system. This approach still provides a cost effective, reliable way to document operating parameters in some situations. But, recent technology advances take this function to a whole new level in terms of sampling rate, accuracy and data presentation.
Most current projects come with a powerful built-in data logging capability in the form of the DDC control system. But, to take full advantage of this capability, it is important to start planning for it during the design and construction phases of the project, as described below.
∑ Specify and configure the points list to provide the data necessary for commissioning, operation, and maintenance: Typically, points are required beyond those associated with simply controlling the systems.
∑ Specify and configure the controllers for trending: Typically this involves addressing memory requirements, establishing sampling rates, defining file structures and file transfer protocols and dealing with general data handling and accuracy issues.
∑ Include data logger requirements in the commissioning proposal and scope of work: Even with a DDC system, data loggers may be required to supplement the information available from the it due to point limitations or because you suspect you will need to sample some processes faster than the proposed control system is capable of performing the task.
∑ Specify and configure the control system network for trending: The DDC system architecture needs to be configured to support the data handling required for trend analysis with out adversely impacting system performance for day-to-day operations. Just because a DDC system can execute the control sequence specified for a project does not automatically mean it can handle the high sampling rates and large data sets associated with trend analysis.
∑ Specify and configure the operator work station for trending: The operator work station needs to be configured to work in harmony with the trending requirements. This includes specifying the data storage capacity and the programming necessary to properly retrieve and archive the data both for the commissioning process as well as the ongoing operation and maintenance processes.
In the context of the project scope and budget, setting the DDC system up to accommodate trending will be highly cost effective and may not even impact the budget in a measurable manner. Sensors, controllers, a work station, and some sort of network infrastructure will be required. During the design, make sure that these features can handle the trending requirements so those infrastructure dollars are spent wisely. If you have to go into an existing system that was not properly configured and make the changes necessary to support trending, it can become quite expensive because you often must abandon some of the existing investment in infrastructure and replace it with new equipment. The differences in equipment first cost are probably not that great; the expense comes from having to make the investment in material and labor twice.
These topics are all discussed in greater detail in Chapter 2 and Chapter 3 of the Control System Design Guide as well as in some of the resources included with this document. Being proactive during the design and construction phases of the project to ensure that these issues are addressed will yield significant long term benefits to you as the commissioning provider as well as the facilities engineers and operators who will deal with the building and systems long after your work is complete.
One of the most important discussions you will have as a commissioning provider with regard to trending is to define the points you want to trend and how often do you want to sample them. Since you often donít really know what you are looking for in a diagnostic and troubleshooting environment, the list of points can become quite extensive. It will also be influenced by the nature of the machinery and configuration of the systems. As a result, the specifics for any given situation will probably vary from location to location and system to system. Requirements may also change as you progress through the diagnostic process. Issues to consider as you formulate your requirements include:
Points to be trended
1 All physical points; i.e. all points that are real inputs or outputs from the controllers: The interactive nature of HVAC systems will often cause the reaction of one output to ripple through the other inputs and outputs associated with the system. If you arenít watching all of the possibilities, you may misdiagnose a problem because you misinterpret the information that is available. Usually, the real problems are the things we didnít think of.
2 All set points: Set point changes can also ripple out through a system, triggering other problems or potentially being misdiagnosed as a problem that does not exist. The discharge temperature may have taken a jump because a valve failed closed or because it was commanded to a higher value.
3 Utility consumption data such as gas usage, electricity usage, and water usage: Monitoring this information is an important step in making Owners and operators aware of the resource impacts of their buildings. In addition, energy utilization patterns can often be a clue regarding system performance problems. See Appendix B: Resources for a paper that describes how this information can be used as a commissioning and operation guide (Using Utility Bills and Average Daily Energy Consumption to Target Commissioning Efforts and Track Building Performance).
1 Analog data, including set points that are automatically reset by other parameters, need to be sampled very frequently. If you donít sample the data set faster than any potential disturbance, you may be fooled by aliasing. Once you have analyzed the system for a while and know that there are not any rapid disturbances, the sample time can be increased to minimize network traffic and the need for archiving data.
Most current technology systems have a limit on sampling speed somewhere around once per minute. If you suspect disturbances are occurring with a cycle time of 3 or 4 minutes or less, then you may want to supplement the EMCS with data loggers capable of sampling rates in second (vs. minutes) for your initial analysis work.
2 Binary data and manually adjustable set points can be sampled based on change of value as long as the data is recorded with a time/date stamp. This will allow it to be viewed simultaneously with the once per minute analog data and significantly reduces the number of samples that must be stored.
A trend request based on these parameters can be quite significant involving tens or hundreds of points and thousands of samples. When making a request like this, you need be cognizant of the fact that if the system has not been designed with trending in mind, then these requirements may be unrealistic and could degrade the performance of the system to the point of being useless. Thus, you may want to initiate your request with a discussion of the capabilities of the system that will be performing the task lest you find pipe wrenches or change orders hefted your way by annoyed operators and control technicians. Based on the discussion, you may decide to temper you requirements to avoid over-burdening the system and/or the Ownerís pocket book. This can be accomplished in a variety of ways including:
∑ Focus on one system: Focusing on the detailed performance of one system at a time may be a more manageable arrangement for the DDC system and your work schedule. But, keep the potential for system interactions in mind. You may want to look at additional points from other systems concurrently with the focus system, even if you canít trend all of them.
∑ Reduce the sampling rate subsequent to your initial analysis: If your initial data set reveals that aliasing is not a concern, then you can reduce the sampling rate to a level that is appropriate for the conditions you are observing. During training, you may want to suggest that the operators occasionally increase the sampling rate to catch problems introduced by changes in the operating cycle or process. For example, a system that is perfectly stable when using chilled water as a source of cooling may become unstable when the economizer cycle must function especially at low ambient temperatures when economizer damper sizing becomes critical.
∑ Donít archive trend data: Many of the system problems associated with high sample rates are related to the network traffic imposed on the system as it archives the trend data from the controller to the hard drive on the host. Configuring the controllers to bump the oldest data with the newest data when memory fills up rather than archiving it can eliminate most of this network traffic. Of course, this will limit the historical information available to you for analysis, especially if most of the controller memory has been used for program functions. In these situations, you may need to adjust your work plan and budget to allow you to spend more time on site so you can use the data while it exists.
One of the biggest advantages of trend data is that it allows you to use graphics to paint a picture of what is going on, and when dealing with data sets, a picture is often worth ten thousand words. Tools that allow data visualization and analysis, both manually and automatically, are constantly evolving and emerging. The resources listed in Appendix B: Resources discuss several of them. Many of the control system suppliers and data logger manufacturers can furnish software modules and packages that allow the data from their system to be viewed and manipulated. Lacking any of these options, most systems can be set up to export data into a delimited file which can then be analyzed using standard spreadsheet software.
Regardless of the analysis technique employed, trend data can be a valuable aid in the functional testing process. For example:
∑ Combining information from multiple data sets on the same graph can be a convenient way to assess performance and identify inconsistencies. For instance it may be desirable to overlay outdoor air condition data from the National Weather Service on a plot of economizer performance to allow the integrated performance of the process to be evaluated.
∑ Unstable control loops become immediately apparent in a graphical presentation of the data. Donít forget about scale factors though. An out of control duct static pressure swinging half an inch w.c. around set point may ďdisappearĒ when plotted on the same axis as discharge temperature with the axis scale set at 0-100.
∑ Plotting the same data set different ways may reveal new information or make different trends easier to see. The plots in both graphs in Figure 15 are from the same economizer data set. But the plot on the left is a time series plot, useful for verifying sequencing, looking at relationships between triggering conditions and their responses, and looking for instability. The plot on the right is a scatter plot of temperature differences in the system, which reveal general performance information about the economizer like how well it maintains minimum outdoor air flow and how well it changes over.
∑ Trending may give you some insight into conditions you observed during construction that caused you concern. For instance, you may have seen something that caused you to wonder about how well calibrated some of the sensors actually were. Looking at a trend of what happened during a warm-up cycle when the system operated with full recirculation (no outdoor air) and all warm-up was provided by the zone reheat coils may provide a clue. One would expect all of the temperature sensors in the air stream ahead of the reheat coils to read the same or nearly the same value. If they donít, then maybe the calibration effort needs to be re-evaluated.
∑ Trending may help you target additional functional testing based on the performance of the system to date, thereby averting future operating problems or energy waste. For instance, if you started up your project in the summer, you may be concerned about what will happen during extreme winter conditions when the sizing of the economizer dampers and performance of the economizer cycle becomes critical. Pulling some trend data and looking at it during the swing season may tell you what to anticipate. If the system starts with minimal instability on a cool fall morning and transitions from economizer cooling to mechanical cooling and back smoothly as the daily outdoor air temperature swings from the 40ļís F to the 70ļís F, then you probably can have some confidence in the systemís ability to handle extremely cold weather. On the other hand, if you saw a lot of instability in these transitional conditions, you may decide that some additional focused functional is in order before the weather outside becomes subfreezing and the instability you are observing leads to freezestat trips, coil failures, and operator frustration.