1、4842 (RP-11.37) VAV Airflow Sensor Response in Relation to “Poor” Upstream Duct Geometry Wayne Klaczek Student Member ASHRAE Pat Fleming, PE Member ASHRAE Mark Ackerman, PE Member ASHRAE Brian Fleck, PhD, PE ABSTRACT A recent ASHRAE research project, RP-1137, indicated Variable air volume (?AV) term
2、inal placement has typi- cally been considered “good enough in the HVAC industry, regardless of upstream duct geometry that clearly afects the response ofthe flow sensoy. This paper describes the loss of VAV flow sensor amplijcation, un formi, and precision following common upstream duct geometries
3、from a recom- nzissioning standpoint. Two brands of VAV terminals were tested in three different sizes: 250,200, and I50 mm diameters (referred to as 1 0, 8, and 6 in. throughout thispaper). Upstream geometry was modeled after transitions that were deemed Ipoor ” during a recommissioning procedure t
4、hat was completedat three facilities duringASHRAE Research Project I1 3 7. The poor ”upstream conditions included combinations of concentric reducers, expanders, 90” elbows, and S-shape geometries with varying straight duct lengths prior to the sensor: This paper outlines some basic recommendations
5、that HVAC designers, terminal manufacturers, and commissioning agents can use to improve the accuracy of VAVairflow sensors both before installation and as part of a recommissioning procedure. BACKGROUND INFORMATION Variable air volume (VAV) systems with direct digital control (DDC) are commonly imp
6、lemented because of added stability, zone control, and greater energy efficiency. However, the literature indicates that a thorough commissioning proce- dure is required to ensure that VAV systems with DDC operate as they were intended (Elovitz 1992). Recommissioning (or retrocommissioning) VAV syst
7、ems is generally beneficial based on energy savings alone (Piette and Nordman 1996; Kjellman et al. 1996). that the calibration of individual VAV terminals was often the most significant improvement made during recommissioning (Klaczek et al. 2004; Klaczek et al. 2006). Several VAV termi- nals were
8、identified during the course of RP-1137 located downstream of “poor” duct conditions, and questions were developed regarding the response of VAV airflow sensors. It was hypothesized that VAV airflow sensor response is highly dependent on upstream duct geometry, even after a perfect recommissioning p
9、rocedure. Designers are often compelled to use transitions prior to VAV terminals due to constricted ceiling spaces; however, often ductwork is installed with poor geometry contrary to the design intent. Sometimes this is due to space constraints, while in the case of concentric reducers, these tran
10、sitions are some- times added to allow for large ducts and lower static pressure losses prior to the VAV terminals (concentric reducers are often supplied by manufacturers of the box if they do not actu- ally manufacture the specified size, for instance, downsizing a 8 in. duct to a 4 in. box). Unfo
11、rtunately, there is a common misconception that the effects of transitions are minimized if a long straight section of ducting is located prior to the VAV terminal. Variable air volume airflow sensors are designed to provide an amplified pressure signal that can be monitored with a low-cost pressure
12、 transducer. Unfortunately, poor upstream geometry can provide a significant calibration error despite the presence of long straight ducting prior to the VAV terminal. INTRODUCTION This study sought to experimentally determine the effects of poor upstream duct geometry on the response provided by Wa
13、yne Klaczek is a research engineer at C-FER Technologies, Edmonton, Alberta, Canada. Mark Ackerman is the faculty service officer for the Mechanical Engineering Department and Brian Fleck is an associate professor at the University of Alberta, Edmonton. Pat Fleming is a mechanical engineer at Hemisp
14、here Engineering, Inc., Edmonton. 202 O2006 ASHRAE. different VAV airflow sensors from a recommissioning point of view. Several nonideal duct configurations were re-created in a laboratory based on the upstream conditions commonly found in real buildings. However, it is important to note that these
15、upstream conditions are far from the worst types of tran- sitions that are possible. The geometries that were tested included a 40D baseline case (a baseline measurement taken with 40D of straight duct prior to the test location) and combi- nations of 90“ short radius elbows, concentric reducers, ex
16、panders, and S-shape geometries. The theoretical VAV sensor response was determined from basic fluid mechanics and compared to the experimental results. The VAV airflow signal was compared to the true airflow rate to determine which geometries resulted in the greatest sensor amplification loss with
17、a dimensionless flow coefficient (C). The upstream geometries were ranked in terms of the greatest losses in amplification and precision. Flow visualization experiments were also completed for a few common geometries (concentric reducer, 90“ elbow, and the 40D baseline case) to examine internal airf
18、low behavior. The results identified common problems with the current design of VAV airflow sensors and led to recommendations for manu- facturers, system designers, and commissioning agents. THEORY VAV averaging flow sensors are simple in both design and practice; as such, they are a common and cos
19、t-effective way to monitor airflow. VAV airflow sensors utilize a differential pressure signal (U) to estimate the airflow by monitoring both the total pressure (P,) and the wake pressure (Pw), as shown by Equation 1. where QTRUE = K= A AP= P - - - T - - pw - true airflow rate, m3/s a constant, dime
20、nsionless cross-sectional area, m2 differential pressure signal, Pa air density, kg/m3 total pressure, Pa wake pressure from the VAV sensors low-pressure port, Pa All of the terms on the right-hand side of Equation 1, except the differential pressure (AP = PT-Pw), are commonly represented by some co
21、nstant within a typical DDC system, which is a generally safe assumption since the air density (p) is the only variable (and it changes only very slightly with temperature). VAV airflow sensors provide an amplified pres- sure signal because the low-pressure port is located within the wake region (fo
22、llowing the sensor); the presence of a wake region is accomplished with sensor geometry, shape, or with the addition of protrusions to ensure flow separation occurs. The theoretical amplification that can be expected from a VAV airflow sensor will be proportional to the square root of the pressure c
23、oefficient (C,) at the low-pressure port, as shown in Equation 2. Thus, the relationship between the pressure coefficient (C,) and the VAV flow signal is given by Equation 3, which is generated by substituting the Pw term from Equation 2 into Equation 1 and simplifying. where C, = the pressure coeff
24、icient, dimensionless PD = the dynamic pressure, Pa V, = air velocity far upstream, ds QVAV ,/(p-pw)(1 -cp) (3) It is clear from Equation 3 that the more negative the pressure coefficient (C,) the greater the amplification of the VAV flow sensor. The VAV flow sensors that were evaluated within this
25、study consisted of four cylinders joined in the center- the cross sections of the cylinders were square (or diamond shape) and rounded (which could be approximated with a circle). The C, following a circular cylinder is a well- studied problem within classical fluid dynamics and is often approximate
26、d as -0.5 within the turbulent flow regime. The pressure coefficients following a rectangular cylinder will vary between -0.93 and -1.62 (Norberg 1993). Likewise, the C, following a rough circular cylinder will vary between approximately -0.4 and -1.0 (Shih et al. 1993), while square to diamond cyli
27、nders of varying angles, will have a C, of approximately -1.3 to -1.7 (Taylor and Vezza, 1999). Although there is clearly some variability, it can be predicted that the pressure coefficients will vary between approxi- mately -0.25 and -1.7. A theoretical fluid dynamics model was developed to predict
28、 the response of several VAV airflow sensors as a function of the velocity profile and static pressure as well as the spacing and shape of the pressure ports. It was predicted that C, shown below in Equation 4, would vary between -2.0 and -1.2 with C, values of -1.7 and -0.25, respectively, over a r
29、ange of O to 8.0 m/s. The experimental results were later found to be in agreement with this model. The most convenient way to consider the amplification of these sensors is to define a dimensionless coefficient to relate the true flow rate (eTRuE) to the flow indicated by the sensor (QVAV) (Equatio
30、n 4). QVAV AV - i c=- TRUE KAV E where: QvAv = the airflow rate indicated by the VAV sensor, amplified, m3/s (4) ASHRAE Transactions: Research 203 V = mean air velocity, m/s C = VAV sensor amplification, a convenient flow constant, dimensionless All of the experimental results presented in this pape
31、r are in terms of the dimensionless flow coefficient (C), which can also be labeled as the amplification coefficient. The VAV I sensor is amplifying the flow signal if the coefficient (C) is higher than 1 .O (conversely, if C 1 .O then the VAV sensor is no longer amplifying the signal but is, in fac
32、t, reducing it). Thus, straight ductwork (with no transitions) will provide the maximum flow coefficient, and the relative effect of each upstream geometry will be evident from the reduction in C. It is important to note that Cdoes not refer implicitly to accuracy, only to amplification. It is possi
33、ble to account for poor upstream geometry with a detailed calibration (to a certain degree, this will be shown). The common problem occurs when a simple calibration is used (for instance a one- or two- point linear calibration equation). The combination of poor upstream conditions and a simple calib
34、ration equation will result in very large flow inaccuracies. It is also the nature of these sensors that when inaccuracies do occur, the zones tend to receive too much airflow (in relation to the indicated value from the DDC system), which was observed in each of the three buildings from Rp-1137 (Kl
35、aczek et al. 2004). EXPERIMENTAL METHODOLOGY Two brands of VAV terminals (designated A and B, which are common models obtained from local suppliers) were obtained in 10, 8, and 6 in. diameter sizes. Standard sheet metal ductwork and transitions were used to assemble all geometries. Airflow was achie
36、ved with a centrifugal variable- speed fan that was capable of providing the flow rates and static pressure recommended by both manufacturers. The true airflow measurements (erRuE) were facilitated with a combi- nation of averaging arrays and orifice plates that were cali- brated with a series of pi
37、tot tube traverses (using a mechanical holder and ANSUASHUE 4 1.2- 1987 recommendations). The true flow rates were monitored within a straight duct that was always a minimum of 20D upstream and 10D downstream from any transitions. The difference between the indicated flow rate from the VAV terminal
38、(evAv) and the true flow rate (QTRuE) was repre- sented by a dimensionless coefficient (C) (see Equation 4). The laboratory experiments focused on determining this flow amplification coefficient (C) for a number of known upstream duct geometries. A long test section (10D) was attached upstream of th
39、e VAV terminal, subsequent to the transition of interest. The variable-speed fan was used to provide a constant flow rate that was slightly lower than the manufacturers recommended minimum value. Readings were taken at five different VAV damper positions: fully open (go“), partially closed (70), hal
40、f closed (45“), partially open (20), and fully closed (Oo). Fan speed was increased and the tests were repeated until the fully open flow setting resulted in an airflow rate higher than the VAV terminals recommended maximum value. This process was repeated with 5D, 3D, and 2D (and, in some cases, OD
41、) length test sections located between the tran- sition of interest and the VAV terminal. This procedure was completed for all six VAV terminals (three sizes from two manufacturers). Several duct configurations were constructed over the course of these experiments, including an ideal duct section wi
42、th at least 40D of straight duct prior to the VAV terminal. The ideal case was used to determine the baseline response for each VAV terminal, which could be well repre- sented by a linear equation to relate the flow coefficient (C40D) to the true airflow rate (eTRuE) (Equation 5). 40, m(QTR(/E)+B (5
43、) where C, = baseline amplification coefficient, taken from the rn = slope B = y-intercept All flow and pressure measurements were made with averaging voltmeters that were set to average the transducer output over ten-second intervals. In addition, the duct static pressure, VAV-indicated static pres
44、sure (note this is usually ignored in a real-world system and only the differential is used), air temperature, and barometric pressure were also monitored. The experimental points were only considered within the airflow range recommended by the manufacturers. Although the apparent loss of amplificat
45、ion is certainly not desirable, the true concern is the error associated with the flow sensors (E) when compared to the DDC system calibration. In Figure 1 the error is based on the assumption that the VAV terminal was calibrated at only the minimudmaximum flow setpoints with a simple linear equatio
46、n and not over the entire range. Although this two-point type of calibration is certainly not desirable (Figure 1 shows a clear error is incurred), it is, unfortunately, quite a common practice when recommission- ing. It is also quite common to take a single point at the maxi- mum of the VAV range a
47、nd simply assume a zero reading, which will result in even greater calibration errors. benchmark (40D) test Test Section Geometries A variety of test sections were re-created; examples of poor duct geometries are provided in Figure 2. The laboratory experiments also considered the effects of “adequa
48、te“ duct lengths prior to the VAV airflow sensors. There is a common industry misconception that 10D of straight, unobstructed ductwork prior to a VAV terminal will correct for poor geometries upstream. Test sections of varying length (including I OD, 5D, 3D, 2D, and, in some cases, OD) were placed
49、after the duct geometry of interest and upstream of the VAV terminal to see if the results could be improved by providing larger sections of straight ductwork. Examples of the laboratory setup for the baseline (40D) case, two 90“ elbows case, and the S-shape geometry case are provided in Figure 3 with a more detailed view of the S-shape geometry (Figure 3e) as an example. 204 ASHRAE Transactions: Research I VAV Teriiruial Auflou. hqe * O O Acha1 Flow Rate, GRUE (Lh) Figure I Sample results using a dimensionless flow coeficient (C). The baseline (400) VAVwas determined, whi
