Jul 2, 2005 - Building HVAC equipment routinely fails to satisfy performance expectations ... the useful service life of equipment can be extended. Also, repairs can be ... In general, these tools take the form of stand-alone software products .... Figure 3is an example of a fault found during a field study.8The cooling coil ...
© 2005, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae. org). Reprinted by permission from ASHRAE Journal, (Vol. 47, No. 7, July 2005). This article may not be copied nor distributed in either paper or digital form without ASHRAE’s permission.
Fault Detection & Diagnostics For AHUs and VAV Boxes By Jeffrey Schein, Associate Member ASHRAE, and Steven T. Bushby, Member ASHRAE
B
uilding HVAC equipment routinely fails to satisfy performance expectations because of improper installation, inadequate
maintenance, or equipment failure. These problems include mechanical failures such as stuck, broken, or leaking valves, dampers, or actuators; control problems related to failed or drifting sensors; poor feedback loop tuning or incorrect sequencing logic; degraded performance caused by heat exchanger fouling; design errors; or inappropriate operator intervention. 58
ASHRAE Journal
ashrae.org
These faults often go unnoticed for extended periods until the deterioration in performance becomes great enough to trigger comfort complaints or a gross equipment failure. The term fault detection and diagnostics (FDD) refers to mathematical techniques to detect and diagnose these types of faults. About the Authors Jeffrey Schein is a mechanical engineer, and Steven T. Bushby is the leader of the Mechanical Systems and Controls Group at the Building and Fire Research Laboratory, National Institute of Standards and Technology, Gaithersburg, Md.
July 2005
By identifying and diagnosing faults to be repaired, FDD techniques can benefit building owners by reducing energy consumption, improving operations and maintenance (O&M), and increasing effective control over environmental conditions in occupied spaces. A number of studies estimate that FDD has an energy-saving potential of 10% to 30%, depending on the age and condition of the equipment, maintenance practices, climate, and building use.1,2,3,4 Significant non-energy benefits of FDD exist. By identifying minor problems before they become major problems, the useful service life of equipment can be extended. Also, repairs can be scheduled when convenient, avoiding undesirable downtime and costly overtime work. Depending on building use, better control of the temperature, humidity, and ventilation rate of the occupied spaces can improve employee productivity, guest/customer comfort, and/or product quality control. In some cases, identifying and repairing faults may make the difference between regulatory compliance and noncompliance. A number of FDD tools are emerging from research.5,6 In general, these tools take the form of stand-alone software products where trend data files must be processed off-line or an interface to the building control system must be developed to enable on-line analysis. A different approach is to embed FDD in the local controller for each piece of equipment, so that the FDD algorithm is executed as a component of the control logic. In this case, the algorithm will have local access to sensor data and control signals, eliminating the need to communicate this information over the building control network. This approach is highly scalable, and, therefore, suitable to larger HVAC systems. Any faults that are detected can be reported to the building operator using the building automation system’s alarm or event handling capability. This article discusses two FDD methods: AHU Performance Assessment Rules (APAR) and VAV box Performance Assessment Control Charts (VPACC). APAR and VPACC can detect common faults in air-handling units (AHUs) and variable-airvolume (VAV) boxes, respectively. The tools are sufficiently simple that they can operate, alongside the control logic, within the processor and memory limitations of commercial HVAC equipment controllers. APAR and VPACC produce outputs that are meaningful to building operators and robust against false alarms.
July 2005
AHU Performance Assessment Rules (APAR)
APAR evaluates a set of rules to determine the presence of faults. These rules are based on simple mass and energy balances of various subsystems of an AHU.7 The APAR algorithm is a three-step process: in Step 1, the mode of operation is determined, based on the positions of the heating coil valve, cooling coil valve, and mixing box damper; Step 2 follows with the evaluation of the rules applicable to the current mode; and in Step 3, possible diagnoses of the fault are listed. APAR is applicable to single duct VAV and constant volume AHUs with airside economizers ((Figure 1). The operation of this type of AHU during occupied periods can be classified into a number of modes, depending on the heating/cooling load and outdoor air conditions ((Figure 2). Each mode of operation can be characterized by different ranges of values for each of three control signals: the heating coil valve, cooling coil valve, and mixing box dampers. The modes are heating (Mode 1), cooling with outdoor air (Mode 2), mechanical cooling with 100% outdoor air (Mode 3), and mechanical cooling with minimum outdoor air (Mode 4). APAR has a total of 28 rules. Each rule is expressed as a logical statement that, if true, indicates the presence of a fault. Because the mass and energy balances are different for each mode of operation, a different subset of the rules applies to each mode. Some rules are independent of the operating mode and are always evaluated. Table 1 shows the set of rules that are specific to Mode 2, cooling with outdoor air. Rule 7 from Table 1 illustrates how the APAR rules were developed. This rule is based on an energy balance of the coil subsystem. In Mode 2, both the heating and cooling coil valves are closed, so the outlet (supply air) temperature should be equal to the inlet (mixed air) temperature plus the heat addition from the supply fan. Possible diagnoses associated with this rule are supply air temperature sensor error; mixed air temperature sensor error; leaking cooling coil valve; cooling coil valve stuck open, partially open, or closed; leaking heating coil valve; or heating coil valve stuck open, partially open, or closed. It is typical of APAR rules that several different faults can cause a single rule to be violated. As a consequence, a few simple rules can be used to find many different faults. Although a list of candidate faults can be made based on the violated rule(s), further information, such as a plot of trend data, is usually needed to identify the specific cause of the fault.
ASHRAE Journal
59
Mode 2: Cooling with Outdoor Air
Control Signal, %
Mode 1: Heating
Mode 3: Mechanical Cooling With 100% Outdoor Air
Mode 4: Mechanical Cooling With Minimum Outdoor Air
Heating Coil Valve Control Signal
100
Mixing Box Dampers Control Signal
Cooling Coil Valve Control Signal
0
Increasing Cooling Load and/or Outdoor Air Temperature/Humidity
Figure 2: Air-handling unit modes of operation. Mode Cooling With Outdoor Air (Mode 2)
and the difference between supply air temperature, Tsa, and supply air temperature setpoint, Tsa,s, is greater than another threshold, εt (in this example, εt = 1.7°C [3.0°F]), Tsa – Tsa,s ≥ εt then the cooling coil valve is saturated and a persistent supply air temperature error exists. Figure 3 shows the supply air temperature varies from 9°C to 13°C (50°F to 55°F) while the supply air temperature setpoint remains fixed at 7°C (45°F)—an unreasonably low value for this application. Clearly, the supply air temperature error is greater than the threshold; therefore, APAR has detected a 60
ASHRAE Journal
Toa > Tsa,s – ∆T Tsf + εt
6
Tsa > Tra – ∆T Trf + εt
7
|T Tsa – ∆T Tsf – Tma|> εt
where Tsa Tma Tra Toa Tsa,s ∆T Tsf ∆T Trf εt
= = = = = = = =
Supply Air Temperature Mixed Air Temperature Return Air Temperature Outdoor Air Temperature Supply Air Temperature Setpoint Temperature Rise Across the Supply Fan Temperature Rise Across the Return Fan Threshold for Errors in Temperature Measurements
Table 1: A sample APAR rule set for Mode 2.
5
Toa should be less than Tsa,s – ∆T Tsf
6
Tsa should be less than Tra – ∆T Trf
X
7
Tsa and ∆T Tsf – Tma should be nearly the same
X
X
Controller Logic Error
Stuck Heating Coil Valve
Leaking Heating Coil Valve
Stuck Cooling Coil Valve
Leaking Cooling Coil Valve
Outdoor Air Temp. Sensor Error
Rule No. Rule Description
Mixed Air Temp. Sensor Error
Possible Diagnoses
Return Air Temp. Sensor Error
Table 2 illustrates the variety of faults that can be detected in Mode 2 from these relatively simple rules. The data needed to implement the APAR algorithm includes: temperatures of supply, return, mixed, and outdoor air; control signals for heating coil valve, cooling coil valve, and mixing box damper; outdoor and return air humidity (if enthalpybased economizer control is used); supply air temperature setpoint; and occupancy/run status of the AHU. Some design information also is needed: the minimum outdoor air fraction required for ventilation and the economizer changeover conditions. If the APAR algorithm is embedded in the AHU controller, all the required data and design information will be available locally. Figure 3 is an example of a fault found during a field study.8 The cooling coil valve is fully open, the heating coil valve (not shown) remains fully closed, and the mixing box dampers (not shown) remain aligned for minimum outdoor air fraction. APAR determines that the AHU is operating in Mode 4 (mechanical cooling with minimum outdoor air), then applies the rules for this mode. One of the rules for Mode 4 states that if the average cooling coil valve control signal, ucc, is fully open, subject to a threshold, εcc (in this example, εcc = 1%) |100 % – ucc| ≤ εcc
Logical Expression (if true, a fault exists)
5
Supply Air Temp. Sensor Error
Figure 1: Air-handling unit temperature control scheme.
Rule No.
X
X
X X
X X
X
X
Table 2: APAR diagnoses.
fault. Facility personnel confirmed that the fault was the result of inappropriate operator intervention. VAV Box Performance Assessment Control Charts (VPACC)
The sheer quantity of VAV boxes typically found in an HVAC system, combined with their relatively limited communication capabilities, makes a strong case for embedding the FDD algo-
ashrae.org
July 2005
where zi is the normalized error at time i, xi is the error at time i, xexp is the expected value of the error, and σexp is the expected variation of the error. The expected value and variation of each of the errors are determined by calculating the mean and standard deviation of each error from a set of known fault free data.9 Separate positive (S) SS) and negative (T T) sums are accumulated. The slack parameter, kk, is defined as the amount of variation considered normal, and therefore ignored. The cumulative positive and negative sums are calculated by: Si = max[0, zi – k + Si – 1] Ti = max[0, – zi – k + Ti – 1] The final step is to compare S and T to the alarm limit, h, to determine whether the process is out of control. Figure 4 illustrates the CUSUM technique.9 The data in Figure 4 are random numbers generated around a mean value, which is gradually increased. This change is difficult to see in the raw data, but the positive sum in the CUSUM chart begins
80 60 40
10 Supply Air Temperature Setpoint 5
0
100 200 300 400 500 600 Minutes from Beginning of Occupied Period
20
700
0
Figure 3: Air-handling unit supply air temperature fault.
to rise dramatically after data point 60, indicating that a change in the process variable occurred. VPACC10 applies the CUSUM approach to detecting faults in single duct, pressure independent VAV boxes with hydronic reheat coils ((Figure 5). Most VAV boxes operate with a pressure independent control strategy similar to the one shown in Figure 6, where the airflow rate is maintained at the airflow rate setpoint, which is in turn a function of the space temperature. To make VPACC widely applicable, a small number of generic
3
25
2
20 15
1
S CUSUM
Z
Supply Air Temperature
15
Control Signal, %
zi = (xi – xexp)/σexp
Temperature, °C
rithm in the VAV box controller. However, Process Error Definition this algorithm must be computationally = space temperature – CSP: if space temperature > CSP Space Temperature Error = 0: if HSP < space temperature < CSP efficient due to the limited processing = space temperature – HSP: if space temperature < HSP power and memory available in a typical Airflow Rate Error = measured airflow rate – airflow rate setpoint VAV box controller. Absolute Value of the Airflow Rate Error = |measured airflow rate – airflow rate setpoint| A technique used in statistical process = DAT – EAT : if reheat coil valve is fully closed Differential Temperature Error control, called cumulative sum (CUSUM) = 0 : if reheat coil valve is not fully closed where charts, is a good solution for this problem. CSP = Cooling Setpoint Temperature The key characteristic of CUSUM charts is HSP = Heating Setpoint Temperature that they can identify trends in noisy data DAT = VAV Box Discharge Air Temperature that indicate a process is out of control. EAT = VAV Box Entering Air Temperature (Obtained From the AHU via the Control Network) They are also computationally lightweight. The basic approach is to define an expected Table 3: VPACC process errors. value and expected variation for a process variable, then to accumulate only the deviation of that process 20 Cooling Coil Valve Control Signal 100 variable outside the expected variation. Mathematically this can be expressed as:
0 –1
10 Upper Chart Limit
5
T
0
–2
–5
–3
–10 0
20
40
60 Data Point
80
100
Lower Chart Limit 0
20
40
60 Data Point
80
100
Figure 4: A CUSUM chart of an out of control process. July 2005
ASHRAE Journal
61
DP
Hot Water Return
M Damper
T
H Supply Air
C Hydronic Reheat Coil
Discharge Air Temperature Sensor
Airflow Rate Setpoint
Min. Airflow Rate
Fully Closed Room o
Heating Setpoint
Thermostat
Airflow Rate Setpoint
Reheat Coil Valve
Hot Water Supply
Differential Pressure Transducer
Valve Position
M
Max. Airflow Rate
Reheat Coil Valve Position
Fully Open
VAV Box Controller
Cooling Setpoint
Room Temperature
Figure 5: VAV box schematic.
Figure 6: VAV box control sequence.
process errors were defined that are meaningful to these systems (Table 3). VPACC works by applying the CUSUM technique to each of the errors in Table 3. Figure 7 shows a plot of data collected during a laboratory test of VPACC.11 An unstable airflow fault was created by altering a parameter in the controller logic to induce oscillation in the damper position. The airflow rate error varies between positive and negative, so the fault is detected by the absolute value of the airflow rate error S (positive) cumulative sum, which increases to the alarm limit, h (set at 900) twice during one day of operation. When the sum reaches h, an alarm is recorded and the sum is reset. VPACC also can be applied to other VAV box types or to VAV boxes without discharge air temperature sensors by selecting a different combination of process errors.
leaking dampers and control valves, sensor drift, and improper control sequencing. The simplicity of APAR and VPACC is a considerable advantage over other FDD techniques since they can be implemented in existing commercial controllers. This has been tested and verified for a small sample of controllers using only the programming software usually provided with the products.11 The deployment path envisioned for APAR and VPACC is for manufacturers to include these FDD tools in their standard application program libraries. One obstacle to commercialization is the lack of a simple means to establish the fault thresholds and statistical parameters that the tools need. A fault threshold expresses the severity of a fault required to trigger an alarm, and is necessary because of uncertainty in the data and operating conditions. If the threshold is too low, normal variation in the data and operating conditions may trigger false alarms. If the threshold is too high, only the most severe faults will be detected. At this time, the fault thresholds and statistical parameters are determined on a case-by-case basis for each site. This requirement for site-specific configuration is unlikely to succeed as a general practice. A field study is underway, in cooperation with several control system manufacturers, to test APAR and VPACC in a variety
APAR and VPACC Commercialization Issues
0.3
1,000
0.2
800
0.1
600 CUSUM
Qerror (m3/s)
The ability of APAR and VPACC to detect a variety of faults with varying severity and during various weather conditions has been evaluated using simulation, building emulation involving commercial building controllers, and limited field tests.8,10,11,12 The results of those studies indicate that APAR and VPACC were successful at finding a variety of faults including stuck or
0.0
400
–0.1
200
–0.2
0
–0.3
S|Q|
SQ and TQ
–200 0
60
120
180 240 300 360 420 480 Data Point (1 min. Intervals)
540
600 660
0
60
120
180 240 300 360 420 480 Data Point (1 min. Intervals)
540
600 660
Figure 7: VAV box unstable airflow fault. 62
ASHRAE Journal
ashrae.org
July 2005
of systems with the intent to develop some simple guidelines that can be used to establish fault thresholds and statistical parameters that will eliminate the need for site-specific configuration. Summary
The increasing performance demands on building automation and control systems, combined with the growing complexity of these systems, have created a need for automated FDD tools. Tools that can be embedded in the controllers offer significant advantages over approaches that depend on collecting and analyzing large amounts of trend data. Two such tools, APAR and VPACC, have been shown to be effective in detecting common AHU and VAV box faults. Work underway in collaboration with several control system manufacturers is expected to establish guidelines to eliminate the need for site-specific customization and lead to new FDD capabilities that will be offered in those manufacturers’ standard program libraries. References 1. TIAX. 2002. “Energy Consumption Characteristics of Commercial Building HVAC Systems – Volume III: Energy Savings Potential.” Final Report to U.S. DOE, Office of Building Technologies. 2. Piette, M.A., S. Kinney and P. Haves. 2001. “Analysis of an information monitoring and diagnostic system to improve building operations.” Energy and Buildings 33(8):783 – 791. 3. Westphalen, D. and K.W. Roth. 2003. “System and component
diagnostics.” ASHRAE Journal 45(4):58 – 59. 4. Claridge, D., M. Liu and W.D. Turner. 1999. Whole Building Diagnostics Workshop. http://poet.lbl.gov/diagworkshop/proceedings/. 5. Hyvärinen, J. and S. Kärki. (Ed.). 1996. International Energy Agency Building Optimisation and Fault Diagnosis Source Book. Published by Technical Research Centre of Finland, Laboratory of Heating and Ventilation. 6. Dexter A. and J. Pakanen. (Ed.). 2001. International Energy Agency: Demonstrating Automated Fault Detection and Diagnosis in Real Buildings. Published by Technical Research Centre of Finland, Laboratory of Heating and Ventilation. 7. House, J.M., H. Vaezi-Nejad and J.M. Whitcomb. 2001. “An expert rule set for fault detection in air-handling units.” ASHRAE Transactions 107(1):858 – 871. 8. Schein, J., et al. 2003. NISTIR 6994. Results from Field Testing of Air Handling Unit and Variable Air Volume Box Diagnostic Tools. National Institute of Standards and Technology. 9. Ryan, T.P. 2000. Statistical Methods for Quality Improvement. 2nd Edition, NY: Wiley and Sons. 10. Schein, J., and J.M. House. 2003. “Application of control charts for detecting faults in variable-air-volume boxes.” ASHRAE Transactions 109(2). 11. Schein, J., S.T. Bushby, and J.M. House. 2003. NISTIR 7036. Results from Laboratory Testing of Embedded Air Handling Unit and Variable Air Volume Box Diagnostic Tools. National Institute of Standards and Technology. 12. Castro, N.S., et al. 2002. NISTIR 6964. Results from Simulation and Laboratory Testing of Air Handling Unit and Variable Air Volume Box Diagnostic Tools. National Institute of Standards and Technology.
Advertisement formerly in this space.
July 2005
ASHRAE Journal
63
