1、NA-04-6-4 Evaluation of Two New Duct Leakage Measurement Methods in 51 Homes Paul W. Francisco Member ASHRAE Erin Kruse Associate Member ASHRAE ABSTRACT Duct leakage in forced-air distribution systems has been recognized for years as a major source of energy losses in resi- dential buildings. Unfort
2、unately, the distribution of leakage across homes is far from uniform, and measuring duct leakage under normal operating conditions has proven to be diflcult. Recently, two new methods for estimating duct leakage ut normal operating conditions have been devised. These are called the nulling test and
3、 the delta-Q test. This paper presents the results of a study on 51 homes to evaluate these new methods relative to an independent benchmark and a method that is currently used. The nulling test was found toperform well as long as wind efects were minimal. Unfortunuteh the time and diflcul Modera 19
4、89; Parker 1989; Cummings et al. 1990; Andrews and Modera 1991; Olson et al. 1993; Palmiter et al. 1119951; Jump et al. 1996; Siegel et al. 1996; Davis et al. 1997; Walker et al. 1998a; Siegel et al. 2003; and Francisco et al. Larry Palmiter Bob Davis 2002a, 2002b, 2003al for a sampling ofprevious w
5、ork on this subject). Duct leakage can have a variety of impacts. It lowers the thermal efficiency of the distribution system. For heat pumps and air conditioners, return leakage can greatly affect the conditions ofthe air flowing over the coil, thereby reducing its performance. For heat pumps in he
6、ating mode, duct leakage can cause the backup heating to be used more, reducing the efficiency benefits of having a compressor. Further, large or concentrated duct leakage can cause homes to have localized areas that are uncomfortable. It has also been found through many of the above-refer- enced st
7、udies that duct leakage is not uniform across houses. Most houses have some duct leakage, but if the duct system is installed properly, the duct leakage is likely to be a small percentage ( Y al -I h a -0.05 I ?: Figure 6 Difference between delta-Q test and benchmark estimate vs. supply leakage pi-e
8、ssure. ASHRAE Transactions: Symposia options were considered here. The first is 1 Pa, corresponding to the small pressures often measured in the furnace closet. The second is 15 Pa, which is about the pressure drop across a typical filter. The last is 10,000 Pa, which is nothing more than an attempt
9、 to make the denominator of the return term of the equation large enough to effectively eliminate the term. Table 2 shows the results of these three pressure assump- tions. Both the 1 Pa and 10,000 Pa results are poor, with large non-zero results for the return leakage and values for the supply side
10、 that disagree with both the nulling test and the benchmark estimate. The I5 Pa results are better, with return leakages that are smaller and supply leakage estimates that are nearer to the nulling test. Given that the known return leakage is O for these cases, however, the results are still less th
11、an satis- fying. Table 3 shows the results with the return excluded from the analysis. These results are nearly identical to the results from the nulling test, indicating that the delta-Q test worked well for these cases as long as the change to the equations was made. Since it is straightforward to
12、 remove the return from the equation, it is again recommended that this change be made for homes without return ducts. “False-Positive” Tendency. One of the primary concerns regarding the delta-Q test is the tendency to overpredict e., false positives of large leakage). In some programs, leakage tes
13、ts are used to rate installers of Site 41 equipment, and there is little tolerance for failing installations that should have passed. Systematically overestimating the leakage is also a concern when financial decisions are being made based on the potential savings associated with sealing the ducts.
14、If a homeowner is trying to decide whether to pay $500 for duct sealing work, the measurement needs to be reliable. On the other hand, if the program is paid for by the agency, such as the utility, then it is probably more permissible to have some fraction of homes sealed that did not really need it
15、, as long as the overall qualitative assessment of the program is correct. While the results suggest that, on average, the delta-Q test worked fairly well, when the test does not perform well the result is usually to overestimate the leakage, often by a large amount. It certainly cannot be expected
16、that any method will be perfect, but it is desirable to have errors be small enough that reasonable judgments can be made. For example, a program may have a threshold of 10% leakage as percentage of air-handler flow, and any result within 2% of air-handler flow of the threshold could also considered
17、 to meet the stan- dard. Table 4 shows how the nulling test and delta-Q test perform in the context of false positives on the supply side. The leakage level in the first column is the “threshold.” For each of the nulling and delta-Q tests, the number of cases above that threshold is shown, followed
18、by the number of false 8 Delta-Q Test Benchmark Estimate Difference cfm YO Air-Handler Flow cfm YO Air-Handler Flow cfm YO Air-Handler Flow 49.5 4.2 35.4 3.0 14.1 1.2 49 45.3 6.2 9.1 1.3 36.2 5.0 Avg. 52.4 5.9 33.6 3.6 18.8 2.3 - 1 45 1 62.5 1 7.3 I 56.4 I 6.6 I 6.1 I 0.7 I ASHRAE Transactions: Symp
19、osia 735 positives, where a false positive is defined as a case where the test estimated a leakage greater than the threshold but the benchmark estimate did not. In addition, the average, mini- mum, and maximum leakage levels of the false positives are shown. For the nulling test, site 21 is exclude
20、d because it distorts the results. At this site, the nulling test estimated about 25% leakage, while the benchmark estimate measured less than 10% leakage, so it would be an additional false positive in each row. These results show that the nulling test has few false posi- tives, and those that occu
21、r have small errors. For the delta-Q test, however, the false positive rate is high, nearly 50% of cases in the 15% and 20% leakage groups. For the 10% and 15% levels, the average errors are large. This shows that, despite the reasonable performance of the delta-Q test on aver- age, there are many c
22、ases where the prediction is unreliable, both on a threshold basis and on an absolute leakage basis. In an effort to improve the results of the delta-Q test, a number of modifications were evaluated. These included using different leakage pressure assumptions, using measured leakage exponents, expan
23、ding the envelope pressure range for the test to 50 Pa, and using nonlinear regression for the purpose of fitting for the pressures. Some of these potential modifications have been proposed by the developers of the test, none of which performed as well as the original method. One new modification di
24、d provide better results, on aver- age. This modification included using duct pressures measured during each envelope pressure measurement, rather than a single pressure measured with the blower door off and the air handler running normally. The goal behind this modi- fication was to address, at lea
25、st partially, the problem of the duct pressures not remaining constant relative to the house throughout the delta-Q test. This modification reduced the average bias of the delta-Q test by about one-third. Surprisingly, the modification that performed the worst was the one using the nonlinear pressur
26、e fitting technique. The leakage estimated with this technique is about 40% greater than the benchmark estimate, on average, compared to an average bias of about 10% of the leakage for the standard delta-Q method. Therefore, the bias has increased by about a factor of four by doing the pressure fitt
27、ing. This technique has been proposed and used as a means of addressing the fact that the actual leakage pressure is not known and assumes that performing this type of fitting will find the best duct pressure to represent the leakage. The results show that this is not the case, and further work has
28、determined that, because the pres- sures in the ducts are changing during the delta-Q test, the pressures at the leaks during the test are not the same as the pressures when the blower door is off. In addition, simulation work by Palmiter et al. (2003) has shown that, due to the math- ematical natur
29、e of the equations, the fitting technique actually tends to provide pressures that match one of the envelope pres- sure sampling targets. Potential Delta-Q Modifications. CONCLUSIONS Benchmark Estimate One of the most surprising results of the study relates to the benchmark estimate. Contrary to exp
30、ectations, the perfor- mance of this test was much worse on the retum side than on the supply side. Much of the reason for this may be the lack of cancellation of errors across return grilles in a single home, since many homes have only one return grille and rarely more than three. There is also evi
31、dence that the flow hood used for return grille flows, though calibrated, did not perform well in the field. On the supply side, however, the benchmark estimate was shown to be more accurate than expected in those houses in which validation testing was performed. When comparing the benchmark estimat
32、e results on the supply side to the added known leak, the maximum error was less than 7 cfm, and no error in these houses was greater than 1% of the air-handler flow. There is evidence in the supply-side benchmark estimate results that the correction for leakage to inside was not as good as was desi
33、red for homes with a significant amount of duct leakage to the conditioned space, primarily multi-story homes. As a result, for these types of homes, the nulling test may actually be the best estimate of the methods tested. Nulling Test The nulling test was found to work very well in most houses. Fo
34、r single-story homes, where the benchmark esti- mate is the most accurate on the supply side, the nulling test showed very little bias relative to the benchmark estimate. The mean absolute error was less than 2% of air-handler flow, and the RMS error was 2.4% of air-handler flow. Because the nulling
35、 test does not perceive leakage to inside as different from airflow through registers, there is no conceptual reason to believe that the performance should be worse for multi-story homes. Therefore, though there is larger disagreement between the nulling test and the benchmark esti- mate for multi-s
36、tory homes, it is likely that errors in the bench- mark estimate are causing the discrepancy. The benchmark estimate validation tests showed little difference in the accuracy of the nulling test estimates on the return side compared to those on the supply side. The small sample size of these validat
37、ion homes prevents a firm conclu- sion from being drawn, but these results are encouraging. As determined in previous studies, the primary source of error for the nulling test appears to be noise due to wind. The test is done at a few small pressure differentials, so significant wind can make accura
38、te pressure measurements nearly impos- sible. Using longer sampling times and larger spacing between pressure targets often provides reasonable looking data, but the uncertainty is greater. Other than wind noise, the primary drawback to the null- ing test is the time-consuming nature of the setup fo
39、r the supply-only measurement. The difficult and time-consuming . 736 ASHRAE Transactions: Symposia part of the setup is the same as is done for measuring air- handler flow using a calibrated fan, so if this test of air-handler flow is also being done, then the nulling test requires little additiona
40、l setup and time. Other than homes where this testing is being done, however, the nulling test is probably only prac- tical for use by contractors in homes without ducted return systems, since, at these homes, only the unbalanced leakage portion of the test is required. This portion of the nulling t
41、est is fast and requires little setup. Examples of houses for which this would be appropriate are many manufactured homes and homes with platform returns. Delta-Q Test Qualitatively, the results confirmed previous findings on the performance of the delta-Q test. This test showed the same tendency to
42、 overestimate the leakage as in other studies, both by the authors and by other researchers. The magnitude of the overestimation was not as great in this study as it was in the previous study by the authors. The larger bias in the previous study can at be least partially explained by the small sampl
43、e size. In this study, the delta-Q test had a bias of about 2% of air- handler flow relative to the benchmark estimate of supply leakage in the single-story homes, with an RMS error ofnearly 5% of air-handler flow, The bias was smaller for multi-story homes, even when compared to the nulling test, w
44、hich is prob- ably the best estimate available for that set of homes. The delta-Q test shows better agreement with the nulling test for multi-story homes, most likely because of the increased connection between the ducts and the house in these cases. This greater connection is via additional registe
45、rs and interior duct leaks. As seen in previous studies and in theoretical modeling of the delta-Q test, one of the major problems is the failure of the assumption that the pressure difference between the ducts and the house remains constant throughout the test. The delta-Q equations implicitly requ
46、ire this to be true, but these pressures can actually change by 20% or more in practice. The result tends to be an overprediction of the flow, which is in agree- ment with both lab and field studies. The results of the modifications to the delta-Q test show that the primary factor that needs to be a
47、ddressed is the failure of this assumption. The only modification that improved the results was one that accounted, in part, for this change. This modification requires the measurement of duct pressures rela- tive to the house during the course of the delta-Q test. While it is not practical to measu
48、re the duct pressures at each enve- lope pressure, it is reasonable to measure this pressure at the extremes of the pressure range, and interpolate, It is the authors recommendation that this modification be made to the delta-Q test protocol. The pressure-fitting technique, however, performed quite
49、poorly. The average bias increased by nearly a factor of 4. These results suggest that this technique of estimating the duct pressures is inadequate and should be dropped from further consideration. Due to the symmetry of the deita-Q test, there is no reason to believe that the results would be substantially different on the return side compared to the supply side. When compared to the nulling test, which is likely the best estimate available on the return side, this holds true. On average, the delta-Q test is 2.5% of air-handler flow higher than the nulling test
