National Conference on Building Commissioning: May 18-20, 2004
Automating the Functional Testing of HVAC Systems Richard M. Kelso, Ph.D., P.E. University of Tennessee Jonathan A. Wright, Ph.D. Loughborough University
Synopsis This paper describes an approach to functional testing of HVAC systems that can be applied during commissioning and re-commissioning. It utilizes the building digital control system to perform and record the results of step tests designed to exercise each component to the extremes of the performance envelope. Models are used to simulate engineering design intent, and are compared against the actual performance of the installed system. The models are based on first principles and empirical results. The tool was tested using actual air handling units with introduced faults and results of the tests are presented. About the Authors Richard Kelso is a Professor in the College of Architecture and Design at the University of Tennessee. He has BS and MS degrees in Mechanical Engineering from the University of Tennessee and a PhD from Loughborough University. He is retired from Kelso-Regen Associates, Consulting Engineers, a firm he founded and directed for 26 years. Dr. Kelso is a Fellow in ASHRAE and recipient of the Distinguished Service Award. He has published a number of papers in the HVAC field. Jonathan Wright is a Senior Lecturer in the Department of Civil and Building Engineering at Loughborough University. He has BS and MS degrees and a PhD from Loughborough University. He is a member of CIBSE and ASHRAE. Dr. Wright has done extensive research in the fields of controls, optimization and genetic algorithms.
Kelso et al: Automating the Functional Testing of HVAC Systems
1
National Conference on Building Commissioning: May 18-20, 2004
Design Intent And System Commissioning Commissioning is the process of assuring that the systems conform to design intent. In the highest sense, “design intent” refers to the building owner or occupants’ expectations for the conditions that they will experience in the completed building. These conditions include the luminous, auditory, olfactory, respiratory and thermal environmental surroundings. The discussion here is limited to consideration of the thermal aspects of those expectations. A sophisticated statement of thermal design intent might refer to the CIBSE1 or ASHRAE2 standards or other publications on comfort conditions. Examples of these are ASHRAE’s Standard 55, Thermal Conditions for Comfort (ASHRAE,1992), the Air Distribution Performance Index (Miller and Nevins,1972) or the Predicted Percentage Dissatisfied (Fanger,1982). It might include requirements for air quality as well. The range of weather conditions for which the systems are to be designed may also be given. Occupancy schedules would typically be stated. Heatgenerating equipment and equipment or areas requiring special conditions should be listed. A less-sophisticated owner or occupant might express design intent by simply giving a room temperature, or a range, for the spaces in the building. At the minimum, every owner or occupant has expectations. These may not be expressed clearly, but they exist none-the-less. It is the task of the system designer to interpret the design intent given him or her, or developed with his or her assistance, or to create it if one is not given. The process starts with calculated estimations of the heat gains and losses for the spaces and the building as a whole, proceeds to selection of a system, selection of the equipment to be used and layout of air distribution and equipment. Thus the designer filters the original design intent through the design process and allocates to each component the portion of the design intent for which it is responsible. The complete set of components and subsystems functions as a coordinated unit to achieve the design intent as interpreted by the designer. The design intent for the heating and cooling coils of an air-handling unit (Figure A) can be defined in terms of the requirements of the building owner and occupants, as well as by the contract documents. The owner/occupant design intent for components of an air-handling unit (AHU) could be described as: Heating conditions: maintain the room temperature within a comfortable range (defined by a single number, a range of numbers, a standard, or a PPD) without drafts, with minimum energy consumption and acceptable noise levels. Cooling conditions: maintain the room temperature and humidity within a comfortable range (defined by a single number, a range of numbers, a standard, or a PPD) without drafts, with minimum energy consumption and acceptable noise levels. Engineering design intent for these components of an AHU, a typical example of which is shown in Figure A, is usually expressed by the contract documents in the form of a schedule on the drawings. This schedule typically contains the “design capacity” of the components as follows:
1 2
The UK Chartered Institution of Building Services Engineers. The American Society of Heating, Refrigerating, and Air-conditioning Engineers.
Kelso et al: Automating the Functional Testing of HVAC Systems
2
National Conference on Building Commissioning: May 18-20, 2004
Exhaust Intake Preheat Louver Coil
Return Fan Return
Outside Air
Supply
Supply Fan Mixing Box
Filters
Heating Water
Cooling Water
Figure A: Diagram of Air Handling Unit and System Heating coil: maximum airflow and water flow rates, entering and leaving air dry bulb temperatures, entering and leaving water temperatures, air and water pressure drops, maximum heat transfer rate. Cooling coil: maximum airflow and water flow rates, entering and leaving air dry bulb and wet bulb temperatures (or relative humidities), entering and leaving water temperatures, air and water pressure drops, maximum sensible and latent heat transfer rates. The purpose of the equipment schedules presented in the contract documents is to enable the selection of a manufacturer’s components - usually from standard product lists. The data are thus given in maximum, or design, capacities. Following the publication of the engineering design intent in the construction documents, the installing contractor provides to the designer documents describing the equipment he or she intends to purchase and install. These documents are commonly termed “submittals” in the U.S. If these describe equipment different from the construction documents, and the designer accepts the submitted equipment, the submittals become the final expression of design intent, although this is sometimes subject to debate. Apart from the equipment, the contractor may provide “shop drawings” giving details of duct, pipe and equipment installations and, again, if accepted these become the final expression of engineering design intent.
Fault Detection And Diagnosis And System Commissioning For some time researchers have been investigating the concept of automated fault detection and diagnosis (FDD) of HVAC systems (Hyvarinen and Karki, 1996). In this concept, HVAC systems that have direct digital control (DDC) systems can be programmed to search for, detect and diagnose
Kelso et al: Automating the Functional Testing of HVAC Systems
3
National Conference on Building Commissioning: May 18-20, 2004
problems in the control system itself or in the HVAC system.. In FDD, models of the process provide analytical redundancy (Patton, et al, 1989, Braun & Rossi in Hyvarinen & Karki, 1996). Analytic redundancy replaces hardware redundancy, which uses dissimilar sensors, each measuring different variables, but which are functionally related by the state of the system. The application of FDD to HVAC systems has been studied extensively under the sponsorship of the International Energy Agency (IEA) Annex 25 (Hyvarinen and Karki, 1996) and Annex 34 (Dexter and Pakanen, 2001). Model-based fault detection and diagnosis uses reference models of the system or components to provide analytic redundancy. Values of output variables read from the system are compared with reference values predicted by the models. Differences between the two are indicators for detection of faulty operation. This approach can use mathematical models derived from known physical relationships, or first principles. Parameters for these models, if identified from design information, enable the model to represent the engineering design intent as the correct operation standard. Figure B is an information flow diagram to show the fault detection process. Faults detected by the presence of these differences, termed “innovations” can be diagnosed by comparison with a set of expert rules or by optimization of the parameters to fit the output variables to those of the real system. Changes in the parameters can then be used to diagnose the faults.
In p u ts
In s ta lle d S y s te m
R e fe re n c e M o d e l
P r e d i c t io n s
-
+
O u tp u ts
In n o v a tio n s
Figure B: Detection of Faults Using First Principles Models and Design Intent Parameters (Salsbury, 1996).
A building HVAC system must be commissioned when the construction schedule indicates, not when the thermal conditions are optimal. The models, then, must be reliable and accurate over a range of operating conditions, not just at full load, and they must be able to extrapolate from the test conditions at commissioning time to design conditions, under which the parameters were developed. First principles models are suitable for such extrapolation. Simple but reasonably accurate models that incorporate parameters for control characteristics such as leakage, nonlinearity and hysteresis are required. Simplicity is desirable for ease of understanding and computer coding as well as for efficient use of computer memory.
Kelso et al: Automating the Functional Testing of HVAC Systems
4
National Conference on Building Commissioning: May 18-20, 2004
An Approach To Automated Functional Testing The approach to automated functional testing in this work is based on the proposition that a model-based scheme can be used to commission an air–handling unit. An important part of this approach is the ability to use design information to establish values of parameters that will enable the models to accurately reproduce the intended performance of the system. Associated with this is the need to develop and test models that accurately portray the performance of the components over their range of operation. Still another significant task is to develop tests that facilitate the detection of likely faults. Figure C illustrates the testing procedure developed for automated commissioning. State variable inputs (boundary conditions)
Control signals
Ahu system model
Installed system
SelectedOutput V ariable
SelectedO utput V ariable
P red ictio n
+
Output
-
U ncertain ty Deviation
Figure C: Overall plan of functional testing procedure
Haves, et al, (1991) developed an open loop step test method that operated a control component from closed to fully open in a single step. This test forces operation at both extremes and allows observation under conditions that should expose faults. Many of the faults encountered during commissioning could theoretically be detected by this test strategy. A large step like this is more severe than the system would see in ordinary operation. The large step also captures the dynamic behavior of the system during and after the control motion. Tests can be conducted off-line by collecting real system performance data, simulating the same conditions with the models, and comparing the model results with the data. Tests can also be conducted on-line by directing real input data to the model simultaneously and comparing model output with real system output in real time. This requires interface software to connect the commissioning tool with the control system.
Kelso et al: Automating the Functional Testing of HVAC Systems
5
National Conference on Building Commissioning: May 18-20, 2004
Representing Design Intent By Component Modeling Models of the heating coil and its control valve can illustrate the principles of the automated commissioning tool. A basic requirement of the component model is that its parameters represent real performance measures that can be determined from design intent documents as described above. Real HVAC systems are dynamic in nature with constantly, although usually slowly, changing variables. Mathematical models that are truly dynamic have partial derivatives that are difficult to handle mathematically. Simpler models based on steady state conditions can be used under two possibilities: (1) a detector to eliminate data that is changing too rapidly to qualify as steady can be used, (2) the uncertainty in the output variables can be increased during dynamic periods. Alternatively, quasidynamic models developed from steady state models can be used. Heating and cooling coils are cross-counterflow externally finned heat exchangers. The steady state heating coil model is developed from the familiar effectiveness-NTU method (Nusselt, 1930, Kreith, 1958).
(
Q = εC min Twin − Tain
)
(
In this equation, Q is the rate of heat transfer, ε is the effectiveness, Cmin is the lesser of the water or air capacities, Tw in is water temperature entering and Ta in is air temperature entering. The steady-state water flow control valve model (Salsbury, 1996) is:
β =0
f (s ) = λ + (1 − λ )s
(
β ≠0
1 − e βs f (s ) = λ + (1 − λ ) β 1 − e
(
where s = the valve stem position, f(s) = the fraction of design water flow rate due to the inherent characteristic, λ is leakage and β is the curvature parameter. Curvature of 0 results in a linear characteristic. The function f(s) expresses the inherent characteristic of the valve, but the installed characteristic is often quite different. A parameter called the authority, A, can be utilized to account for the difference between inherent and installed performance: f ' ( s) =
1 1 1 + A s 2 − 1
0.5
(
where f’(s) is the fractional flow under installed conditions.
Kelso et al: Automating the Functional Testing of HVAC Systems
6
National Conference on Building Commissioning: May 18-20, 2004
Results Of Testing With Experimental Data For Correct Design And Operation The concepts introduced above were tested using actual air handling units at the Iowa Energy Center’s Energy Resource Station. The measured data were recorded by the energy management system and the tests were conducted off-line. A step test comparison of the steady-state and quasi-dynamic valve and chilled water coil models is shown in Figure D. The quasi-dynamic model, labeled dynamic in the figure, is the steady state model with a first order time lag.
Figure D: Comparison of model outputs with actual performance as a step change in setpoint is made. The control valve moves from the closed position to fully open.
The quasi-dynamic model only differs from the steady state model during step changes, but it does more closely capture the actual performance during these times. The chilled water coil and valve models are similar to the heating coil valve and coil models except, of course, opening the valve causes a drop in temperature. Figure E pictures the leaving air temperature as the heating coil control valve is stepped in open loop from closed to open and return. The first panel shows leaving air temperature as modeled and as measured. The second panel shows the difference between the two and the interval within which we have 95% confidence that the difference occurs. The third panel shows a value of zero when no fault is detected and and when a fault is detected. The heating coil performance as indicated by the measured leaving air temperature exceeds the model predictions when the valve is open. A fault under the normal
Kelso et al: Automating the Functional Testing of HVAC Systems
7
National Conference on Building Commissioning: May 18-20, 2004
operating conditions, which means without deliberately introduced faults in this case, would indicate a modeling deficiency or an incorrect parameter. In this case, the fact that the measured temperature is higher than the modeled temperature is an indication that coil performance exceeds expectations. In an actual commissioning, better than expected performance would probably not be considered a fault.
Figure E: Normal operation step test of heating coil and control valve.
Results Of Testing With Experimental Data For Faulty Operation A control valve that leaks when closed is a familiar problem in new systems. Debris in the pipe, improperly assembled valves and incorrectly adjusted valve actuators are possible causes. Opening the manual balancing valve in the bypass leg around the control valve as shown in Figure F simulated the leak. The manual balancing valve in the three-way valve bypass leg was closed to force the valve to act as a two-way valve. The flow indicated was therefore flow through the coil. The expected indications that a control valve leaks as contrasted with normal operation were: 1. The measured off-coil and supply air temperatures would be higher than the modeled temperatures when the valve is closed, and 2. The hot water flow rate would not be zero when the valve is closed.
Kelso et al: Automating the Functional Testing of HVAC Systems
8
National Conference on Building Commissioning: May 18-20, 2004
C o n tro l v a lv e
From coil
3 -w a y va lve b ypa ss clo se d to sim u late 2 -w a y va lve
To coil
B y p a s s v a lv e o p e n to s im u la te le a k .
Figure F: Control and manual valve arrangement to simulate leaking control valve.
Figure G shows the results of a simulated leaking heating coil valve in a 0-100 step test. The supply fan was operated at full speed and the return fan at 90% of the supply fan speed. The VAV terminals were fully open, the cooling coil control valve closed and the mixing box set open to outside air. The leak was set at approximately 0.03 kg/s (0.5 gpm)(2% of design). Note that the normal operation tests (Figure E) had shown the heating coils used in these tests have duty (capacity) in excess of their rating. A model parameter for number of rows has been adjusted to account for this in the tests in this section.
Figure G: Heating coil step test with leaking control valve Kelso et al: Automating the Functional Testing of HVAC Systems
9
National Conference on Building Commissioning: May 18-20, 2004
Figure G shows the measured leaving air temperature was approximately 20C higher than the modeled temperature at the beginning and end of the test. A leaking hot water valve would cause the air temperature to increase slightly as it passes through the coil. In this test, the leak was present before the valve was stepped open, hence the fault indication at the beginning of the test. When the valve is open, the modeled and measured temperatures are very close and the leak is only apparent when the valve is closed, as expected. Figure H illustrates the relative performance of the steady state and quasi-dynamic models in speed of detection. The steady state model gives leaving air temperature readings at one-minute intervals as indicated by the asterisks plotted along the continuous curve of the quasi-dynamic model. The asterisks plotted along the lower axis indicate points discarded by the steady-state filter as non-steady state points. The quasi-dynamic model detected a fault at 13 minutes after the step change whereas the steady state model did not detect the fault until 16 minutes after. The three-minute advantage of the dynamic model can be significant when testing several components with multiple tests for each and when several airhandling units are to be tested.
Figure H: Time delay until a heating coil control valve leak fault can be observed by the steady state model as compared with the quasi-dynamic model.
Conclusions And Recommendations These tests indicate that functional testing using first-principles and empirical models with parameters determined by design intent information can be successful. The tool presently contains models of a mixing box and damper actuators, heating and cooling coils and valves, and supply and return fans with speed controllers. The testing tool is able to detect a number of faults and a rules-based postprocessor can diagnose the faults. Further research is need in applying the tool on-line and in developing additional models. Pressure and flow models would allow more options for commissioning of fans, Kelso et al: Automating the Functional Testing of HVAC Systems
10
National Conference on Building Commissioning: May 18-20, 2004
dampers and air duct systems than thermal models can. The pressure and flow models used in the research project are less mature than the thermal models and need additional attention.
Acknowledgements The authors would like to thank Dr. Tim Salsbury and Johnson Controls Company for making possible the testing work at the IEC ERS, and Dr. John House, Mr. Curt Classen and the staff at the Iowa Energy Center’s Energy Resource Station for their assistance and hospitality during the testing period. Dr. Kelso would like to thank Dr. Phil Haves for his assistance during the early stages of the project that lead to this paper and the University of Tennessee for their support during the project.
References ASHRAE Standard 55, Thermal Environmental Conditions for Human Occupancy, American Society of Heating, Refrigerating and Air-conditioning Engineers, Atlanta, 1992. Dexter, A.L. and Pakanen, J., Annex 34: Computer-aided Evaluation of HVAC System Performance Final Report, International Energy Agency, Oxford, 2001. Fanger, P.O., Thermal Comfort, Robert E. Kreiger Publishing Co., Malabar, Fla, 1982 Haves, P., Dexter, A.L., Jorgensen, D.R., Ling, K.V., and Geng, G., Use of a Building Emulator to Evaluate Techniques for Improving Commissioning and Control of HVAC Systems, ASHRAE Transactions, New York, 1991. Hyvarinen, J. and Karki, S., Building Optimization and Fault Diagnosis Source Book, Final Report of IEA Annex 25, IEA, Espoo, Finland, 1996. Kreith, F., Principles of Heat Transfer, International Textbook Co., Scranton, 1958. Miller, P.L., and Nevins, R.G., AN Analysis of the Performance of Room Air Distribution Systems, ASHRAE Transactions, Vol. 78, American Society of Heating, Refrigerating and Air-conditioning Engineers, Atlanta, 1972. Nusselt,W., A New Heat Transfer Formula for Cross-Flow, Technische Mechanik and Thermodynamik, Vol. 12, 1930 Patton, R., Frank, P. and Clark, R., Fault Diagnosis in Dynamic Systems, Prentice Hall, New York, 1989. Salsbury, T. A., Fault Detection and Diagnosis in HVAC Systems Using Analytical Models, PhD. Thesis, Loughborough University, Loughborough. Leics.,1996.
Kelso et al: Automating the Functional Testing of HVAC Systems
11
