1、471 6 Effect of Boundary Conditions on the Prediction of Temperature Distribution for Curtain Walls Hua Ge, Ph.D., P.Eng. ABSTRACT Previous studies have indicated that the application of variable local film coejcients can improve the accuracy of temperature predictions for fenestration systems using
2、 two- dimensional programs, such as FRAME and THERM. Howevel; the discrepancy between simulation and test results for metalframed windows is relatively large, e.g., a typical 5OC (9F) discrepancy may exist in theframe area. One of thepossi- ble reasons was believed to be the lack of test data that w
3、ere reported with accurate measurement locations and test condi- tions. In a comprehensive experimental study carried out by the authors on the overallperformance of metal curtain walls, extensive temperature measurements enabled more accurate comparison between simulations and tests. Several sets o
4、f boundary conditions were assigned to the sill sections of two diferent curtain wall systems in the FRAME simulations. The variation of local film coejicients along the room-side frame surface was considered. The comparison between simulation results and test data indicates that the proper applicat
5、ion of more realistic bound- ary conditions can greatly improve the accuracy of computer simulations in predicting the temperature distributions and thus the condensation resistance performance for fenestration systems. The application of localfilm coejicients to the frame surface reduces significan
6、tly the discrepancy between simu- lation and test results to within fO.5OC (kO.9“F) for the stan- dard curtain wall system and to within +2OC (+3.6“F) for the highly insulated system. These improvements confirm the importance of assigning more realistic boundary conditions to the frame area for meta
7、l-framed fenestration systems. The results also suggest that the convection motion in the glazing cavity should be considered in order to obtain more accurate results for edge-ofglass in simulations. Paul Fazio, Ph.D, P.Eng. INTRODUCTION Condensation resistance is an important factor in the eval- ua
8、tion of the thermal performance of fenestration systems. The occurrence of condensation can cause higher energy consumption due to latent loads, deterioration of materials sensitive to water, and indoor air quality problems (bacterial or mold growth associated with the appearance of water). Condensa
9、tion on fenestration products is the number one source of fenestration product complaints (Dariush et al. 2001). Computer programs such as VISION/FRAME (EEL 1995) and WINDOW/THERhUOptics (LBL 1998) are inte- gral parts of the Canadian standards (CSA 2000) and Ameri- can standards (NFRC 2001) for rat
10、ing fenestration products. FRAME and THERM are two-dimensional programs to model the heat transfer through edge-of-glass and frame areas of the fenestration systems. Programs VISION and WINDOW/Optics are used to model the one-dimensional heat transfer, solar transmittance, and visible transmittance
11、through the center-of-glass. These programs have been vali- dated as a standardized alternative ofphysical tests to calculate the overall thermal transmittance (U-factor), solar heat gain coeficient (SHGC), and visible transmittance. However, the accuracy of these programs in predicting the condensa
12、tion resistance is still being validated and the simulation procedure is still under development. Laboratory physical test remains the only reliable means to determine the condensation resis- tance performance for fenestration products. A great deal of effort has been made in understanding and impro
13、ving the accuracy in predicting temperature distribu- tions over fenestration systems by simulations. In the U-factor calculations using the programs VISION/FRAME and H. Ge is a postdoctoral fellow and P. Fazio is a professor in the Building Envelope Performance Laboratory, Center for Building Studi
14、es, Department of Building, Civil, and Environmental Engineering, Concordia University, Montreal, Quebec, Canada. 02004 ASHRAE. 249 r0.7m I1 6.0171 -1 t- 3.6m hot box Figure I The experimental setup in the environmental chamber: WINDOW/THERM, an “effective conductivity“ is assigned for the glazing c
15、avity or for the frame cavity to account for the convective and radiative heat transfer, and constant film coef- ficients are applied to both interior and exterior surfaces as boundary conditions. This procedure yields good agreement between the simulated and tested U-factors. However, a large discr
16、epancy exists in temperature predictions, especially at the edge-of-glass and frame areas. By using a detailed convection/ radiation model in the glazing cavity, the accuracy of simula- tion was improved to within 1 “C to 2C (1 3F to 3.6“F) for the edge-of-glass surface (McGowan 1995; Zhao et al. 19
17、96). The simplification of using a constant surface film coefficient is believed to be another reason for the discrepancy (Sullivan, et al. 1996; de Abreu et al. 1996). Some numerical and experi- mental studies have been conducted to quantify the local film coefficients along fenestration surfaces (
18、Curcija and Goss 1993, 1995; Schrey et al. 1998; Griffith et al. 1998). Curcija et al. (1998) reported that the simulations with variable local film coefficients on both interior and exterior surfaces gave the best results for an IGU (insulated glazing unit) and a wood-framed window. McGowan and Wri
19、ght (1998) reported similar find- ings for most windows examined except for those with metal frames. Typically, the simulations predicted a 3C (5.4“F) warmer surface temperature for the frame area, with a worst case of 5C (9F) for an aluminum frame. In their study, the edge-of-glass portion (63.5 mm
20、) was divided into five segments and a constant film coefficient was assumed over each segment. The convective portion of the film coefficient was assumed to vary linearly between zero at the sightline and the value at center-of-glass over the length of 63.5 mm (2.5 in.). The radiative portion of th
21、e film coefficient accounted for the heat exchange with the frame surface. However, no local film coefficient was applied to the frame area. The poor agreement between simulation and test results for aluminum frame windows was noted by other studies as well (McGowan 1995; Carpenter 2001). thin strip
22、 of flexible PVC I high-performance GU i/ I reinforced nylon nose revised backpan Figure2 Sill sections of the curtain wall systems simulated: (a) system A and (6) system B. In a comprehensive study on the overall performance of metal curtain walls, the authors measured the detailed temper- ature di
23、stributions throughout the test specimen and deter- mined the variable local film coefficients on the room side either by measurements or by analysis. The accurately reported locations of temperature measurements and test conditions make possible a more accurate comparison between test and simulatio
24、n results. This paper explores how the local variation in film coefficients on both the edge-of- glass and frame areas improves the accuracy of simulation for the sill sections of metal curtain walls. TEST PROCEDURE A two-story full-size test specimen including two different curtain wall designs was
25、 installed in a large-scale environmental chamber (Figure I). The overall dimension of the test specimen is3.8mwideby6.7mhigh(l2ft,6in.by22 ft).Themaindiffer- ence between these two types of curtain walls is the frame configuration. The first one is a standard system, which uses a thin strip of flex
26、ible PVC as the thermal break at the mullion nose, and is referred to as system A in this paper (Figure 2a). The second is an improved system referred to as system B and has a much larger thermal break achieved by replacing the aluminum mullion nose with reinforced nylon (Figure 2b). The other two 2
27、50 ASHRAE Transactions: Research System A Glazing Panel Frame Spandrel Panel 2% in. (63.5 mm) high by 4 in. (101.6 mm) deep aluminum mullion with a continuous strip of flexible PVC as thermal break Double IGU with % in. (6.4 mm) clear annealed glass pane, % in. (12.7 mm) air gap, and conventional al
28、uminum spacer % in. (6.4 mm) clear annealed span- drel glass, % in. (19.2 mm) air gap, 4 in. (101.6 mm) rigid fiberglass insu, lation with regular steel back-pan design features used in system B include a revised back-pan design as shown in Figure 2b, which shifts the connection of the back-pan to t
29、he interior flange of the mullion tube to eliminate the thermal bridge created by the return of the back-pan and the high-perfor- mance insulated glazing units. The details and dimensions are listed in Table 1. The test specimen was tested under a series of winter conditions. The temperature distrib
30、ution throughout the test specimen was monitored by over 700 type T (copper-constan- tan) thermocouples. The detailed sensor locations and temper- ature readings across some typical curtain wall sections were reported in Ge et al. (2001). The temperature readings measured for curtain wall sill secti
31、ons under CSA winter conditions, -18C (0F) outdoor and 70F (21C) indoor, are used for the comparison to simulations in this paper. To enable the comparison between measurements and simulations, the determination of the surface film coefficients along the test specimen under CSA winter testing condit
32、ions are necessary. The indoor convective film coefficient was established experimentally by measuring the air velocity and temperature within the inner boundary layer of the wall specimen. A customized, large-scale, computer-controlled three-dimen- sional traverse system was built to carry out the
33、measurement with high spatial resolution and precision. The measurements indicated that free natural convection prevailed along the glaz- ing surface, and the temperature gradient was approximately linear within the inner thickness of the boundary layer. Thus, Fouriers equation can be used to calcul
34、ate the heat flux and to determine the local convective film coefficients. The details on the experimental setup and results were reported in Ge and Fazio (2003). The values of the local convection film coefficients measured for the center-of-glass area were averaged to repre- sent the mean convecti
35、on film coefficient for room side. The convective film coefficients for the outdoors were calculated based on the measured average air speed in the cold box for the region of interest using the correlations recommended by ASHRAE standard 142P (ASHRAE 1996). The radiative components for both indoor a
36、nd outdoor surface film coeffi- cients were determined following the procedure outlined in ASTM Standard CI 199 (2000) using the measured average temperature for the center-of-glass area. The average surface System B exterior Double IGU with % in. clear annealed 2% in. (63.5 mm) high by 4 in. glass,
37、 95%/5% Argodair gap, low-E coat- (10 1.6 mm) deep aluminum mullion ing (E = O. 1) on the exterior surface of the with reinforced nylon as a thermal inner glass pane, and aluminum spacer break % in. (6.4 mm) clear annealed span- drel glass, % in. (19.2mm) air gap, 4 in. (1 01.6 mm) rigid fiberglass
38、insu lation with revised steel back-pan with thermal break design VI1 interior Figure 3 Locations of temperature measurements und of boundary segments (dimensions in mm). film coefficients determined for CSA winter conditions simu- lated in the environmental chamber are 7.0 W/m2.K for the indoor and
39、 20.0 W/m2.K for the outdoor. SIMULATION PROCEDURE Sill sections of the two different types of curtain walls tested in the environmental chamber are simulated. In this paper, the sill section refers to the joint section including edge- of-glass, frame, and edge-of-spandrel (Figure 2). The edge-of- s
40、pandrel is defined as the perimeter area within 63.5 mm (2.5 in.) of the frame. The reason to incorporate the edge-of- spandrel into the simulation sections in addition to the edge- of-glass is to reproduce the exact configurations as they were in the tests. Since the interior surface temperature of
41、 the glazing units is much more sensitive to the indoor surface heat transfer coef- ficients (de Abreu et al. 1996), this study focuses on the effect of local variation in film coefficients on the room side, while for the outdoor side, the surface film coefficient is assumed to be constant. Therefor
42、e, the interior boundary is divided into eight segments as shown in Figure 3. The numbers labeled on ASHRAE Transactions: Research 251 the surface are the selected locations of temperature measure- ments. The edge-of-glass area, which extends over 76.2 mm (3 in.), is divided into three segments, I,
43、II, and III, and each has a length of 25.4 mrn (1 in.). The results from measure- ments show that the mullion surface temperature is sensitive to local film coefficients. For example, the temperature on the upper horizontal mullion surface was about 0.5”C (0.9”F) warmer than the lower mullion surfac
44、e because the lower hori- zontal mullion surface was exposed to the return of the back- pan, which had much lower surface temperature than the indoor air, and the air movement was restricted as well in the narrow air gap (Ge et al. 2001). Therefore, it is reasonable to assume that the accuracy of th
45、e simulation may be improved for the frame surface by applying nonuniform film coefficients over this area. The upper horizontal surface of the frame is divided into three segments, IV, V, and VI (Figure 3). The vertical mullion surface, together with the first 32 mm (1 % in.) on the lower horizonta
46、l mullion surface, is defined as one boundary segment, VII. The vertical surface of the back-pan, together with the first 32 mrn (1 % in.) on the return ofthe back- pan, is defined as one boundary segment, VIII. An air gap of about 6.4 rnm (% in.) exists between the return of the back-pan and the lo
47、wer horizontal surface of the mullion. In accordance with standard CSA A440.2 (CSA 1998), the first 32 mm (1 % in.) of the cavity is treated as boundary and the rest of the cavity is treated as regular air cavity. The film coefficient over each boundary segment is assumed to be constant and the valu
48、e at the midpoint of the segment is used. Programs FRAME and VISION are used to carry out the simulations. Six scenarios are used for variations of the bound- ary conditions, and the corresponding film coefficients for each boundary segment are listed in Table 2. These scenarios include: 1. “Conduct
49、ion” with constant surface film coeflcients. The CSA standard procedure (CSA 1998) for calculating U-factors is followed, in which the convection in the glaz- ing cavity is not considered and the surface film coefficient is assumed to be constant. The constant surface film coef- ficients used in simulations are the average values deter- mined for the tests, which is 7.0 W/m2.K for room side and 20.0 W/m2.K for outdoor, respectively. “Conduction ” with variable convection film coeflcients for the edge-ofglass area. The local values of the indoor convection film coe
