1、There are at least two reasons to use external coupling. First, each domain application has evolved separately over the years and is well proven. Rewriting the code (to be included as part of a package in another domain) could be seen as a setback from these independent advances in separate domains.
2、 There- fore, further efforts would better be concentrated at making these different domain applications communicate with each other. Second, external coupling can immediately benefit from independent developments in each domain. The separate domain applications can expand and develop in their respe
3、c- tive directions, and the external coupling mechanism can make this development available without having to (heavily) update the source code. DIFFERENT IMPLEMENTATIONS OF COUPLING STRATEGIES The focus point of coupling between CFD and BES can be represented by the convective heat transfer equation
4、 on the internal surfaces: where 4c - h= A= - Twdl = TL?f = convective heat flux (W) convective heat transfer coefficient (CHTC) (W/m2K) wail surface area (m2) wall temperature (“C) reference temperature (“C) The energy calculation of the BES is sensitive to the value of the CHTC and the reference t
5、emperature used in the above equation. Without CFD, the best BES can do is to adaptively use empirical correlations during the simulation and use the air-point temperature as the reference temperature. The main disadvantage is that it cannot include the effect oftemperature stratification around the
6、 wall or the difference in flow charac- teristics between surfaces in the same room. CFD is intro- duced to overcome those problems. However, the parameters in the equation must be resolved iteratively, by exchanging the parameters between the two programs until the values are converged. Negrao (199
7、5) describes two handshaking mechanisms between the BES and CFD: surface coupling and integrated coupling. Negrao uses the word “conflation” for coupling. However, for consistency, coupling will be used in this paper. In integrated coupling, CFD interacts directly with the thermal matrix solver and
8、resolves the exchanged parameters until the values are converged. CFD is used to solve the zone air-point temperature and the internal surface convection, while the BES provides the CFD with the internal surface temperatures. Both iteratively exchange the data until conver- gence before moving to th
9、e next time step. In surface coupling, on the other hand, the two programs work independently and exchange information at the internal surfaces. The CFD uses the boundary conditions (wall temper- ature) from the previous time step, calculates the convective heat transfer coefficient (CHTC), and send
10、s this back to the BES. The BES will then use this information to form the matrix for the zone heat balance equations and solves the matrix for the current time step. The simulation continues with CFD simulations always using the data from the previous time step. Beausolleil-Momson (2000) argued tha
11、t surface coupling brings many advantages over the integrated approach. For external coupling, the most important feature is that the surface coupling provides more flexibility in defining the coupling mechanism. With regard to accuracy, obviously integrated coupling is more accurate because it reso
12、lves the exchanged data in many iterations until converged to a certain value. Beausolleil-Momson (2000) also argued that the accu- racy will be the same if the time step is sufficiently small, although he did not elaborate on how small is small. Zhai and Chen (2001) found that in the iteration betw
13、een CFD and BES, the solution does exist and is unique. Zhai and Chen also reported that normally convergence can be reached after 4 to 10 iterations. If we take one hour as the standard time step in most BES, we can conclude that a 6- to 15-minute time step is small enough for surface coupling to g
14、et the same accu- racy as the integrated coupling. Furthermore, the CFD-predicted value of CHTC (CHTC,) can always be rejected in surface coupling. This cannot be done in integrated coupling without interrupting the iteration process. This checking mechanism of the CHTC, value before passing it back
15、 to BES is one ofthe quality assurance measures that should be used when using surface coupling. COUPLING MECHANISM Figure 1 gives an overview of the current status of surface coupling (Beausolleil-Morrison 2000; ESRU 2000). In summary, for every time step during the calculation of the convective he
16、at transfer coefficient (CHTC) of internal surfaces, the thermal domain checks whether there is any CFD call defined for that time step. If not, it continues with another mechanism for defining the CHTC of the internal surfaces. If yes, it will invoke the coupling controller to derive the CHTC from
17、a CFD simulation. The coupling mechanism consists of a pre-CFD treatment (final), the actual CFD simulation, and a post-CFD treatment. The pre-CFD treatment involves an investigative CFD simulation (so actually there are two CFD simulations for every CFD call). This investigative CFD simulation (the
18、 so- called gopher run) is a simple CFD simulation (with coarse mesh and simple turbulence model) that will classify the flow regime near each surface. Based on this classification, the coupling controller decides which boundary conditions are applied for the final CFD simulation. ASH RAE Transactio
19、ns: Symposia 61 3 Thermal domain I I CFD I I Other procedures calculation procedure I I I CFD calculation I Post-CFD treatment I I I I o - Calculate CHTC - Decide: accept or reject CHTC from CFD I I CI I I I I i Figure 1 Internal coupling mechanism in BES. The final CFD simulation mainly uses the st
20、andard k-E turbulence model. However, apart from sending the wall temperature to CFD, the coupling controller decides (1) which wall function to use, (2) whether the CHTC derived from an empirical correlation should also be sent to CFD, and (3) what reference temperature should be used in CHTC calcu
21、lation in CFD. After the final CFD simulation, the post-CFD treatment will calculate the CHTC for each internal surface based on the CFD result, i.e., the CHTCCFD, and decides whether the predicted CHTC can be used for further calculation in thermal domain. Note that the CHTCCFD is not calculated so
22、lely by CFD. Beausoleil-Morrison (2001) noted that the use of the CFD definition of CHTC is the most desired approach. However, its own limitation to accurately predict the surface heat convec- tion makes it not a fully viable approach. For that reason, Beausoleil-Morrison argued that the “CO-operat
23、ive approach is the solution, so long as the CFD solution is regarded as not accurate enough to get reasonably accurate results on its own in all flow configurations. In the Co-operative approach, CFD will calculate the convective heat transfer on the wall based on the boundary conditions set by the
24、 coupling controller. Depending on the type of boundary conditions set, there are eight different ways of calculating the convective heat transfer on the wall (the details can be found in Beausolleil-Morrison 2000). The calculated convective heat transfer is passed back to BES where it will be conve
25、rted to CHTCCFD using the BES air- point temperature as the reference temperature. There is an inconsistency in using the reference temperature between the CFD and BES: the CFD uses local temperature e, the air temperature in the adjacent cells) while BES uses the air-point temperature. This inconsi
26、stency has been acknowledged by Beausolleil-Morrison (2000). However, the Co-operative approach seems to be able to predict the impact of stratifica- tion on the wall. The calculated CHTC, is then compared with a CHTC value that is determined from empirical correlations applica- ble to the flow fiel
27、d under investigation. If CHTCC, falls within a certain predefined range, then it will be accepted and passed to the thermal domain. If not, it will be rejected and the thermal domain continues the calculation using the CHTC derived from the empirical correlation. The initial implementation of the e
28、xternal coupling uses an external CFD program for the final CFD simulation, while keeping the rest of the mechanism unmodified. This assumes that the BES has the internal capability to do the investigative CFD simulation. The result of the external CFD program will 61 4 ASHRAE Transactions: Symposia
29、 be passed back to the BES (or more specifically the CFD module of the BES) in terms of field values. This further assumes that the CFD module of the BES also have exactly the same mesh to avoid interpolation of CFD results. As a starting point, an advanced building energy program that meets the abo
30、ve requirements was selected (ESRU 2000). A commercial CFD package was also selected (Fluent 2003). However, the results of this study would apply for any BES and CFD program with the described criteria. However, as will be shown later, this initial implementa- tion has several drawbacks. The pre-CF
31、D treatment was designed to send boundary conditions (for the final CFD simu- lation) based on the standard k-E turbulence model. This model is known to be less accurate for a coarse mesh, and, in particular, it is dependent on the size of the first grid cell adja- cent to the wall (Yuan 1995). The
32、obvious solutions for this are to refine the mesh, up to the point of practicality, or simply use another turbulence model. The use of another (simpler) turbulence model at this point is more attractive, as refining the mesh is less practical. This is due to the limitation of the available internal
33、CFD program, where the coupling controller can only read one mesh for both the investigative and the final CFD simulation. If the mesh for the CFD simulation would be refined, then the computation time will increase significantly as the investiga- tive CFD run will also use the same mesh. The zero e
34、quation turbulence model (Chen and Xu 1998) is regarded applicable for the above purpose because 1. it uses less computing resource than the standard k-E turbu- lence model; 2. it does not make use of wall functions; 3. the CHTC can be explicitly defined using the Reynolds analogy: h = PefF where h
35、= convective heat transfer coefficient (CHTC) (w/m2K), = effective viscosity (Pa s), bff Pref = cp = specific heat (Jkg K), Ax = 4. effective Prandtl number (=0.9), distance between the surface to the adjacent cell (m); it has been successfully used for the coupling of CFD and energy simulation (Che
36、n et al. 1999). As the final CFD simulation does not use the standard k- E turbulence model, then there is no reason to do the investi- gative CFD run. It should be noted that the investigative CFD simulation is still an advanced coupling mechanism. However, its implementation so far is limited to t
37、he standard k-E turbu- lence model. Since the zero equation turbulence model will be used for the (final) CFD simulation, the investigative CFD simulation becomes irrelevant. Further work on this is ongo- ing and will be reported in the near fuhire. The following mechanism reflects the latest develo
38、p- ments on external coupling between CFD and BES: 1. During the calculation of the CHTC for the internal surfaces, BES checks if a CFD simulation is specified for the current time step. If not, it will continue to calculate the CHTC based on user input. If yes, the BES will invoke the coupling cont
39、roller, which will call the external CFD program to run the CFD simula- tion (using the zero equation turbulence model). After the simulation, the CFD program calculates the CHTC for each internal surface and sends the result back to the coupling controller. If the CHTC falls within predefined crite
40、ria, the coupling controller will pass the result to the BES; otherwise, the coupling controller will send a flag to BES to use its own CHTC value. 2. 3. 4. The following assumptionsllimitation apply to the current Radiative heat exchange is not included in the CFD simu- lation. As the BES includes
41、the radiation in its calculation, this will affect the boundary conditions sent to CFD, which in turn will make CFD over predict the CHTC value. How significant is the effect of this omission is recognized as an important future work for this study. The external coupling implementation uses the surf
42、ace coupling mechanism, even though the integrated conflation is obviously more accurate. The reasons are, first, because it could have the same accuracy if the time step is small and, second, because the data exchange mechanism employed by the current implementation is using the intermediate file,
43、which is not fast enough for integrated coupling. Another data exchange mechanism is the interprocess communica- tion (IPC), which is the subject of ongoing research (Yahi- aoui et al. 2004). An integrated coupling mechanism can easily be implemented for extemal coupling once there is a faster mecha
44、nism for data exchange. The CFD calculations are at steady state for every time step of the BES simulation. However, on a higher level view, all of the CFD simulations are more like quasi-steady simula- tions that change its boundary conditions every (BES) time step. The use of transient CFD simulat
45、ion is intentionally avoided, as it is more computationally expensive, and in this case this is justified by the fact that implementation. 1. 2. 3. the simulation is stable and can reach the conver- gence without difficulty and there is no transient parameter (within the BES time step) needed for th
46、e result. 4. The current implementation depends on the capability of BES to provide an extensive range of empirical correlation ASHRAE Transactions: Symposia 61 5 J Modelled Figure 2 Test cell site (jhoto from Lomas et al. 1994). Figure 3 BES model. of CHTC, against which the CHTC prediction from CF
47、D is checked. If there is no such library available, then the only way to use external coupling is to manually speciQ a range of acceptable values of CHTC. This range would have to be hard-coded into the source code, and this will make the mechanism less flexible as this range could be different for
48、 a different problem. CASE STUDY The above discussion on internal and external coupling is supported by a case study in which the separate coupling mechanisms are compared. The case study is derived from Lomas et al. (1 994) and is described in the remaining of this paper. BES Model The test cell un
49、der investigation is located in Bedford- shire, UK. Figure 2 shows the site layout and the situated test cells. The site has eight semi-detached rooms, which have a lightweight, timber-framed, construction. These rooms have interchangeble glazing panels in the south wails, and the one used for the current study has a double-glazed panel. A detailed description of the site and the test room can be found in Lomas et al. (1994). Figure 3 shows the BES model that has been used for this study. The geometry and construction are built according to the report without modification, includ
