coupling heterogeneous computational - ibpsa

COUPLING HETEROGENEOUS COMPUTATIONAL CODES FOR HUMAN-CENTRED INDOOR THERMAL PERFORMANCE ANALYSIS

Christoph van Treeck1, Sebastian Stratbücker1, Sandeep R. Bolineni1,

Carolin Schmidt1, and Daniel Wölki1

1Fraunhofer-Institute for Building Physics IBP, Holzkirchen Branch

Fraunhoferstr. 10, 83626 Valley, Germany, [email protected]

ABSTRACT Human-centred indoor thermal quality performance analysis comprises multiple physical domains and typically requires coupling different numerical solvers in a weak manner and further linking with empirical modeling approaches for thermal comfort assessment. We present the new co-simulation middleware framework CoSimA+ (Co-simulation Adaptation Platform) for coupling such heterogene-ous computational codes in distributed environments in a scale-adaptive manner. The framework is based on open-source software components such as the Internet Communication Engine ICE, the QT library and the VTK visualization toolkit. It provides services to visualize, observe and modify ongoing co-simulations and to connect sensor hardware, as well. Interfaces between codes are well defined by means of an interface definition language (IDL). The architecture supports the use of parallel computing resources.

INTRODUCTION Modeling physiological responses of the human body to transient and non-uniform environmental thermal conditions in detail comes along with the need to combine several simulation models. Usually, dedicated codes are coupled in a weak manner instead of using a monolithic solver and taking advantage of applying well-suited solvers for each domain. As the analysis includes multiple physical domains, third party software, legacy simulation codes (research codes), or even database applications need to be accordingly combined. Due to problem complexity and model interoperability there is a lack of such integrated solutions which address these specific demands. In this contribution, we briefly outline our scale-adaptive coupling strategy between such heteroge-neous computational codes for human-centred indoor thermal performance analysis. Scale-adaptivity stands for strategies for applying engineering models to balance between physically required and computa-tionally affordable model resolution. We explain the software concept of the developed middleware framework, the Co-Simulation Adaptation Platform CoSimA+.

As indicated in Figure 1, the coupled analysis comprises convective and radiative heat transfer between a virtual human and its environment at different levels-of-detail, for example, using computational fluid dynamics (CFD) and a thermal radiation solver, or by using predefined heat transfer coefficients instead. For human thermoregulation simulation, a finite element implementation of a mathematical thermo-physiological model is adapted which may be further coupled to a clothing model. Empirical models for thermal sensation and local thermal comfort perception are as well connected. The developed application framework allows further controlling sensor hardware of a physical thermal manikin, for example, by varying the heat flux emitted by a sensor as predicted by the thermo-physiological model.

Figure 1 Integrated coupling methodology

A middleware application therefore needs to provide the runtime infrastructure for managing geometric entities, to control attached simulation codes in a distributed environment, to interpolate data, to transfer data between solvers, to account for numerical issues such as time step control and solver convergence, as well as services for visualization. The paper is organized as follows. The paper first reviews existing solutions in the respective field of application; it addresses the coupling methodology and the specific software layout of CoSimA+, and presents how we transfer results between solvers for the case of a manikin model exposed to indoor climatic conditions by using the already developed prototype coupling models and subsystems.

Proceedings of Building Simulation 2011: 12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

- 247 -

REVIEW OF EXISTING SOLUTIONS Several authors address the coupling of virtual physiological models such as the Berkeley model (Huizenga et al. 2001), the Fiala model (Fiala et al. 2001), the Tanabe model (Tanabe et al. 2002) or the Yokoyama model (Yokoyama et al. 1997) with high-resolution simulation tools in order to predict the dynamic thermoregulation responses of the human body to variations of the local thermal microclimate. For example, Cropper et al. (2009) as well as Zhang & Yang (2008) report the coupling of the IESD Fiala model (Fiala et al. 2001) with CFD. Cropper et al. thereby take advantage of an integrated coupling strategy, directly interfacing with the ANSYS CFX code in order to increase convergence performance. Kok et al. (2006) present the coupling of a thermoregulation model with a flow and thermal radiation solver to predict indoor cabin climate in aircraft. Völker et al. (2010) demonstrate the coupling of the UC Berkeley model with CFD. Other work is reported by Nilsson et al. (2007), Niu et al. (2008), Streblow et al. (2008) or Zhu et al. (2008), for example. Besides research activities, some integrated commercial solutions exist such as the FEM model FIALA-FE within the software package THESEUS (Paulke 2007), which is used in the context of this paper, or the thermophysiological model TIM within the CFD software StarCD. For code coupling, most authors use file-based data exchange, network sockets, shared memory or inter-process communication. Typically, interfaces are developed for dedicated tools. Such software interfaces are therefore hard to maintain and cannot be transferred to other applications without effort. The commercial TLK Inter Software Connector (TISC) remedies this situation as it is designed as general engineering middleware platform for tool coupling (Kossel et al. 2009). In the scope of distributed parallel and high performance computing, on the other hand, the Message Coupling Interface (MPI) standard is widely applied. MpCCI (Wolf et al. 2010) is a general multi-physics code coupling interface that provides tools for parallel and mesh-based solver coupling. The so-called Common Component Architecture (CCA) is a standardization approach for a component architecture for high performance computing (CCA Forum 2011). Following a review of the existing solutions and some of its evident drawbacks, we conclude that for our particular needs it is reasonable to design and implement a new middleware architecture for coupling heterogeneous codes within distributed environments and to make use of open source software components only, as some computational tools offer limited access to

their structure by application programming inter-faces,

some coupling tools are deemed to be not flexible enough,

codes, if accessible, are written in different programming languages, and

some coupling libraries are hampered by extensive licensing costs which may easily exceed the licensing expenses of other (own) components.

COUPLING METHODOLOGY In a previous publication, we address the development of a "Virtual Climate Chamber" (Frisch 2008, van Treeck et al. 2009). With the software, we implemented and tested an interface between the Fiala (2001) thermoregulation model and a thermal zone model in order to predict transient thermal interaction between a virtual human and a surrounding rectangular space. This prototype uses predefined convective and radiative heat transfer coefficients to calculate heat transfer between manikin surface mesh and zone air and walls. Starting from this experience, CoSimA+ has been designed as middleware framework providing the runtime infrastructure to couple such subsystems like fluid flow and thermal radiation solvers with a thermophysical model and, for example, further (empirical) models for thermal comfort assessment as shown in Figure 1 on the previous page. A typical coupling methodology is given in Figure 2 below. With CoSimA+, modules are integrated and iteratively controlled by a core process, results can be visualized in a central monitor application and data is persistently stored within a database called features. Special emphasis is placed on the geometric model, which is hierarchically organized in the feature model structure. The next section details the software architecture of the middleware.

Figure 2 Typical coupling methodology within a

distributed environment

MIDDLEWARE ARCHITECTURE The Co-Simulation Adaptation Platform CoSimA+ is designed as middleware application and program-ming framework for coupling heterogeneous


- 248 -

computational codes in a distributed environment and to integrate models to form a common desktop application (Stratbücker et al. 2010).

The Simulation Runtime Environment of CoSimA+ provides the communication, configuration, monitoring and control infrastructure.

The Programming Framework consists of a documented class library. It provides software interfaces, including a developer guideline to support adaptation and integration of arbitrary simulation codes.

Figure 3 shows the software architecture which basically consists of the core, monitor, coupling, features and blackboard components. The class architecture is based on a modified model view controller pattern. The core classes form the central control unit, while the coupling service represents interfaces to external modules. The blackboard service provides means for data distribution and process synchronization. The monitor classes serve for visualization and user interaction services. Objects, entities and simulation results are managed by the features component, which serves as arbitrary simulation data container. Scene objects are hierarchically structured and elements are stored in a mesh-based surface representation. For example, segments of a manikin model may be further decomposed into individual patches.

Figure 3 CoSimA+ software architecture with

features data base, core, coupling interface and monitor classes (Stratbücker & van Treeck 2010)

CoSimA+ supports synchronous and asynchronous communication. Similar to remote procedure calls (RPC), the client-server model enables synchronous method invocation of remote objects. As a design constraint, solely open-source software components are used. The communication is based on the Internet Communication Engine ICE (2010), for the general model and view concept, the QT library (2010) is adapted and for visualization purposes the Visuali-zation Toolkit VTK (2010).

Special emphasis is placed on the interface definition methodology, as simulation codes typically are available in different programming or script languages. Therefore, a formal interface description language (IDL) defines interfaces to particular software modules. The respective definition is converted into code in the language of choice such as C++, Java, or Python. This methodology further allows creating small script-based applications from scratch, which are directly interfacing with CoSimA+. A weather data reader application is a typical example where Python may provide the whole infrastructure and avoids using a new C++ code. From a developer's point of view, a coupling interface class template needs to be adopted and attached to the specific needs. It is not visible to the programmer, how this class is connected to the main middleware and where other codes are executed, i.e. if either local or remote via a proxy connection. CoSimA+ classes form an object oriented distributed system, which is fully transparent in terms of location, access, scaling and migration. Figure 4 sketches the UML model of the main CoSimA+ components and their interaction.

Figure 4 UML model of main CoSimA+ components

MODELS AND SUBSYSTEMS The following section details some of the models and subsystems of the integrated coupling methodology, which Figure 1 sketches.

Geometric model Starting from our experience of geometric multi-segmented manikin modeling (Yue 2008, Frisch 2008), and in order to properly reproduce significant anthropometric body details, we decided to interface with an ergonomic model. The RAMSIS model is widely applied in automotive and aircraft design (RAMSIS 2005). Its body builder provides access to a wide range of anthropometric data and to a respective geometric model, if used within the CAD software CATIA. The left hand side of Figure 5 shows the RAMSIS wireframe model of a male adult with H-point, skin points and skin lines. For the available models of male and female adults, we developed a parametric surface model in CATIA V5, taking advantage of the


- 249 -

constraint-based CAD modeling features and following the surface segmentation of the thermo-physiological model of Fiala. Figure 5 details this model in the centre picture. The shell model is linked with the RAMSIS model such that modifications of the body morphology accordingly affect the thermo-physiological sub-model in terms of its degrees of freedom including surface contact with a seat model, if applicable. The constraints of the model ensure that model transformations do not lead to undesired gaps in the thermophysiological surface model. The right hand picture of Figure 5 depicts the corresponding mesh model, in this case in a resolution suitable for the CFD surface mesh.

Figure 5 Ergonomic RAMSIS wireframe model in

CATIA V5 (left), patches mapped onto thermo-physiological shell model (centre) and resulting

(here: CFD) surface mesh model (right)

Thermo-physiological model A thermo-physiological model describes the heat exchange between body and environment and models the body thermoregulation. Typically, segments decompose the body into parts, each segment consisting of layers, which are in thermodynamic contact with each other and with the ambient climate. Models include sub-models for clothing and models for blood circulation. The active thermoregulation system integrates the thermal stimuli from the various thermoreceptors and formulates reactions of signals from core and skin with respect to regulation mechanisms concerning the peripheral vasomotion, sweating and shivering (Stolwijk 1971). In this paper, the model of Fiala et al. (2001) is used, which has been initially developed at DeMontfort University (UK). Based on our software interface (Frisch 2008) to the FIALA-FE core model (Paulke 2007) of the software package THESEUS, we developed and tested a new interface between CoSimA+ and FIALA-FE. The FIALA-FE discretization is based on a finite element approach in order to represent the concentric manikin layer structure for solving the bioheat equation with spherical and cylindrical elements.

CoSimA+ operates FIALA-FE in non-coupled mode by directly interfacing with the solver. Post-processing intervals are adjusted in terms of a weak solver coupling strategy. These features allow for creating and executing multiple instances of the model. The graphical user interface to the thermoregulation model utilizes the CoSimA+ monitor framework for visualization and interaction. As the thermoregulation model is calibrated for an average human adult with physiological properties detailed in (Fiala et al. 2001), current research activities further address the individualization of thermo-physiological models in terms of modeling different morphological types by variation of key parameters such as gender, age, weight, etc. We refer to Wölki et al. (2011).

Surface convection Surface convection is accounted for by two different strategies, either

using predefined heat transfer coefficients, which are obtained from a series of CFD simulations in advance and which are accessible by a respective database through CoSimA+, or

by directly interfacing to a CFD solver with the above described interface definition of CoSimA+.

The following examples detail the application of the various solvers in terms of mapping results to the above-defined surface patches of the segmented manikin model. The example refers to the case of an individual partially exposed to forced convection and to direct sunlight within an indoor space.

Figure 6 Convective heat transfer coefficients

[W/(m²K)] computed in Fluent (left) and mapped to area weighted segment values (right)

For the first of the above-mentioned cases, Figure 6 shows convective heat transfer coefficients obtained by ANSYS Fluent computations which are mapped to the patch model. For the second case, to date a prototype is available for linking the open source computational fluid


- 250 -

dynamics package OpenFOAM to our CoSimA+ framework (Nadeem 2011). The coupling interface is fully integrated into OpenFOAM using the (here: serial) PISO and SIMPLE solver families, i.e., data is thus not transferred in a file-based manner but using the CoSimA+ communication strategy. Mayer (2010) and Nadeem (2011) provide a comparison between Fluent and OpenFOAM results for the same case, which agree well.

Thermal radiation solver Again, starting from a previously tested thermal radiation radiosity solver implementation (van Treeck et al. 2009), which is based on the adaptive view factor integration method of Walton (2002), we developed a new radiosity solver taking advantage of the performance of a graphics processing unit (GPU) in order to speed-up the view factor calculation. Figure 7 exemplarily shows long wave radiative heat transfer coefficients computed with this GPU solver.

Figure 7 Radiative heat transfer coefficients

[W/(m²K)] computed with GPU radiosity solver (left) and radiative heat transfer [W/m²] (right)

Figure 8 Solar heat gains [W/m²]: high resolution

results (left) obtained by raytracing mapped to area weighted segment values (right)

As indicated in Figure 8, likewise a GPU-based ray-tracing method is applied for calculating solar radiation incident to the manikin surface, further taking a window model into account. The solver is integrated via CoSimA+, results are mapped onto area weighted segment values. In Figure 9, the obtained convective and radiative heat transfer coefficients, as available via the database interface, are compared for the above-mentioned sample case.

Figure 9 Segment-wise convective (hc) and radiative

(hr) heat transfer coefficients [W/(m²K)]

Thermal comfort assessment It is well understood, that methods implemented in ISO 7730 (2006) are limited applicable to non-uniform thermal environments, as it focuses on the body as a whole and on thermal conditions where the body is close to thermal neutrality (Fanger 1970). For typically steady-state conditions, the equivalent temperature method (Bohm et al. 1990) can be used to correlate a non-uniform scenario to a uniform environment and to find by experiments a statistical relationship between equivalent temperatures and local thermal sensation (ISO 14505-2). Only a few publications, however, address human responses to inhomogeneous and transient thermal conditions at the same time (Zhang et al. 2010). To date, a prototype application for comfort visualization exists in CoSimA+. Equivalent temperature values (ISO 14505-2) obtained by simulation are evaluated taking correlations known from a series of experiments. Fraunhofer IBP conducted multiple series of subject tests involving questionnaires in buildings, aircraft and vehicles, cf. (Park et al. 2009, van Treeck et al. 2009), for example. Current research therefore focuses on the enhancement of existing local and global comfort models for inhomogeneous environments and transient thermal conditions, including the develop-ment of a new CoSimA+ module.

CONCLUSIONS We have presented a developed middleware application, the Co-Simulation Adaptation Platform CoSimA+, which provides a framework and a documented class library for weak coupling of heterogeneous computational codes in distributed


- 251 -

environments. Emphasis is placed on the application within the scope of human-centred indoor thermal quality performance analysis. Besides the middleware application itself, first interfaces to a thermoregulation model and to the CFD code OpenFOAM are available. Current developments comprise further interfaces ANSYS Fluent, to a new thermal radiation solver (as shown) utilizing parallel high-performance graphics hardware, a 1D/3D interface from/to the equation-based modeling language MODELICA, as well as to a new virtual thermal comfort model including a new clothing model. In another project, an interface to the Fraunhofer thermal manikin sensor hardware DRESSMAN is under development.

ACKNOWLEDGEMENTS This work is supported by the Fraunhofer Internal Programs under Grant No. Attract 692 239.

REFERENCES Bohm M., Browen M., Holmer I., Nilsson P.V. and

Noren O. 1990. Evaluation of vehicle climate with a thermal manikin. Techn. report, Swedish Inst. of Agricultural Engineering.

CCA Forum. 2010. The DOE Common Component Architecture project. www.cca-forum.org.

Cropper P., Yang T., Cook M., Fiala D., Yousaf R. 2009. Simulating the effect of complex indoor env. cond. on human thermal comfort. Proc. Building Simulation, Glasgow, UK.

EN ISO 7730. 2006. Ergonomics of the thermal environment.

EN ISO 14505. 2004. Ergonomics of the thermal environment.

Fanger P.O. 1970. Thermal Comfort. Copenhagen: Danish Technical Press.

Fiala, D., Lomas, K., & Stohrer, M. 2001. Computer prediction of human thermoregulatory and temp. response to a wide range of env. conditions. Int J Biometeorol, 45, 143–159.

Frisch J. 2008. Diploma thesis, Technische Universität München.

Huizenga, C., Hui, Z., & Arens, E. 2001. A model of human physiol. and comfort for assesing com-plex thermal env. Bldg. and Env., 36, 691–699.

ICE. 2010. The Internet Communications Engine. www.zeroc.com/ice.html

Kok J.C., van Muijden J., Burgers S., Dol H. and Spekreijse S.P. 2006. Enh. of aircraft cabin comfort studies by coupl. of mod. for hum. thermoreg. ECCOMAS CFD 06, Egmond aan Zee, The Netherlands.

Kossel R., Correia C., Loeffler M., Bodmann M., Tegethoff W. 2009. Verteilte Systemsimulation

mit TISC. ASIM-Workshop, Dresden, und DASS´2009.

Mayer A. 2010. Diploma thesis, Fraunhofer-Institute for Building Physics and Universität Ulm.

Nadeem U.U. 2011. Masters thesis, Fraunhofer-Institute for Building Physics and Technische Universität München.

Nilsson, H., Brohus, H., & Nielsen, P. 2007. CFD Modeling of Thermal Manikin Heat Loss in a Comfort Evaluation Benchmark Test. Helsinki.

Niu J., Gao N. 2008. Personalized vent. for commercial aircraft cabins, IndoorAir, Denmark.

Park S., Hellwig R., Grün G., Holm A. 2009. The interrelation of local and overall thermal comfort of passengers in aircraft cabins, Proc. Roomvent, South Korea, pp. 1027-1034.

Paulke S. 2007. FE-based impl. of Fiala‘s thermal manikin in THESEUS-FE. In: EUROSIM.

QT. 2010. A cross-platform application and UI framework. qt.nokia.com

Ramsis. 2005. Ramsis in Catia V5 user guide, Version 3.8, Human Solutions GmbH, Kaiserslautern, Germany

Stolwijk J. 1971. A mathematical model of physiological temp. reg. in man, CR-1855, NASA.

Stratbücker S. & van Treeck C. 2010. CoSimA+ Co-simulation adaptation platform, Fraunhofer-Institut for Building Physics, IBP Report 37-2010.

Streblow, R., Müller, D., Gores, I. 2008. A coupled simulation of the therm. environment and thermal comfort with an adapted Tanabe model. Proc. IndoorAir, Denmark.

Tanabe, S.-I., K.K., Nakano, J., Ozeki, Y. 2002. Eval. of thermal comfort using comb. multi-node thermoreg. (65MN) and radiation models and CFD. Energy and Bldg 34: 637-646.

van Treeck, C., Frisch, J., Pfaffinger, M., Rank, E., Paulke, S., Schweinfurth, I., Schwab, R., Hellwig R., Holm, A. 2009. Int. thermal comfort analysis using a parametric manikin model for interactive simulation, J Building Performance Simulation, 2(4): 233-250.

van Treeck C., Frisch J., Egger M. and Rank E. 2009. Model-adaptive analysis of indoor thermal comfort, BuildingSimulation 2009, July 27-30, Glasgow, Scotland.

van Treeck C., Stratbücker S., Bolineni S.R., Wölki D., Holm A. 2011. Human-centred indoor thermal performance analysis by scale-adaptive coupling between heterogeneous computational codes, Indoor Air 2011, Austin (TX), USA


- 252 -

Völker C. and Kornadt O. 2010. Bauphysik 32(6): 365-372.

VTK. 2010. The Visualization Toolkit. www.vtk.org Walton, G. N. 2002. Calculation of obstructed view

factors by adaptive integration. Tech. Rep. NIST, USA.

Wölki D., van Treeck C., Stratbücker S., Bolineni S.R., Holm A. 2011. Individualization of virtual thermal manikin models for predicting thermo-physical responses, IndoorAir 2011, Austin (TX), USA

Wolf, K., Bayrasy P. 2010. Generic coupling of 1D system codes with 3D CFD tools. In: ERCIM News, 81: Special theme, S. 27-28

Yokoyama, S., Kakuta, N. and Ochifuji, K. 1997. Development of a New Algorithm for Heat Transfer Equation in the Human Body and its Appl., Appl. Human Sci., 16, 153-159.

Yue, Z. 2008. Master Thesis. TU München. Zhang H., Arens E., Huizenga C. & Han T. 2010.

Thermal sensation and comfort models for non-uniform and transient env., Parts I-III. J Building and Environment, 45, 380–410.

Zhang, Y., & Yang, T. 2008. Sim. of Human Ther-mal Resp. in Confined Space. Proc. IndoorAir 2008, Denmark.

Zhu, S., et al. 2008. Development of a computational thermal manikin applicable in a non-uniform thermal environment (Part 2). ASHRAE HVAC&R Research, 14 (4), 545–564.


- 253 -

coupling heterogeneous computational - ibpsa

Documents