O

Tp

Fa

b

ARAA

KAAIS

1

trriHaoa

wtoeatd

e(

1

Journal of Cultural Heritage 16 (2015) 560–566

Available online at

ScienceDirectwww.sciencedirect.com

riginal article

he architecture of warehouses: A multidisciplinary study on thermalerformances of Portus’ roman store buildings

rancesca Pagliaroa, Evelyne Bukowieckib, Franco Gugliermetti a, Fabio Bisegnaa,∗

SAPIENZA University of Rome, Department of Astronautical, Electrical and Energetic Engineering, Via Eudossiana 18, 00184 Rome, ItalyInstitut de recherche sur l’architecture antique (IRAA) – CNRS, USR 3155, France

a r t i c l e i n f o

rticle history:eceived 5 February 2014ccepted 4 September 2014vailable online 5 October 2014

eywords:rchaeology

a b s t r a c t

In the framework of the French Research Agency program “Entrepôts et lieux de stockage dans le mondegréco-romain antique”, a multidisciplinary study on the archaeological, historical, constructive and archi-tectural characteristics of Ostia and Portus’ warehouses is ongoing. The final aim of the project is tounderstand the role these warehouses played in the process of storage and distribution of food sup-plies. The starting point is the study of the degree of knowledge the ancient people had in storagetechniques. Numerical predictive computational models, supported by in situ measurements to char-

ncient warehousesn situ measurementsoftware simulations

acterize dimension, geometry and materials, are in this paper considered as essential “tools” for thehistorical, archaeological and functional interpretation of the roman store buildings. In this paper, differ-ent hypotheses of the architecture of Portus’ store buildings are analysed and compared using computerfluid-dynamic models in order to support archaeologists in their effort to better understand how ancientwarehouses were built, managed and used.

© 2014 Elsevier Masson SAS. All rights reserved.

. Research aims

In the field of archaeology, it is easy to run into studies and caseshat cannot always be interpreted. Some examples might be rep-esented by destroyed sites where remains do not allow for theeconstruction of the original building or by the lack of consistentnformation about systems and methods used in ancient times.istorical and architectural evidences sometimes allow makingssumptions that may not be fully confirmed. A good way to supplyther information and evidences could be coupling archaeologicalnd numerical analysis.

A good example in this way is represented by Ostia’s and Portus’arehouses, focused in this paper, in which a scientific collabora-

ion between IRAA (Institut de recherche sur l’architecture antiquef CNRS) and DIAEE (Department of astronautical, electrical andnergetic engineering of “Sapienza” University of Rome) developedn approach based on CFD simulations to overcome some limits in

he archaeological reconstruction of the volumetric architecturalesign of destroyed structures in Portus.

∗ Corresponding author. Tel.: +39 06 4814332; fax: +39 06 4880120.E-mail addresses: francesca.pagliaro@uniroma1.it (F. Pagliaro),

velyne.bukowiecki@gmail.com (E. Bukowiecki), franco.gugliermetti@uniroma1.itF. Gugliermetti), fabio.bisegna@uniroma1.it (F. Bisegna).

http://dx.doi.org/10.1016/j.culher.2014.09.004296-2074/© 2014 Elsevier Masson SAS. All rights reserved.

2. Introduction

Ostia’s and Portus’ warehouses were important elements in thehub for trade between Rome and its provinces [1,2]. Archaeologicalstudies have often left unsolved many questions related to storagesystems, because what remains does not always allow the entirereconstruction of the original structure [3,4].

Predictive computational models are one of the latest applica-tions in the world of cultural heritage. They analyse and representthe physical environment through the numerical/geometrical res-olution of complex mathematical equations, which govern thethermo fluid dynamics and the visual and the acoustic phenomena.

At today, predictive numerical methods are not only intendedand used for architectural/engineering solutions related to theconservation of cultural heritage (covering, microclimate, lighting,ventilation), they but can also represent an essential “tool” for thehistorical and archaeological interpretation of ancient structures[5–8].

Nowadays, the archaeological data and evidences can be usedin a hermeneutic approach, in which different hypotheses canbe analysed to reconstruct the past, using a virtual model insidean iterative process driven by archaeologists. This hermeneutic

approach partially erases the sense of extraneousness in computersimulations that had characterized the early stage of numericalapplications. This last approach, as example, can be found in Chour-mouziadou and Kang [9], Gugliermetti et al. [10] for the acoustic

F. Pagliaro et al. / Journal of Cultural Heritage 16 (2015) 560–566 561

house

ef

lanwcAisFt[Ge

bsb(bartaehth

3

iis

Fig. 1. Topographic inclusion of Traian ware

nvironment referred to ancient theatres, or in Navvab et al. [11]or the visual and luminous aspects of museum.

Most of the authors used CFD simulation to study and ana-yse indoor environments, mainly to define the best conservationnd preservation conditions for cultural objects, while there areot many studies devoted to support directly the archaeologists’ork. For example, Albero et al. [12] simulated the virtual micro-

limate conditions in the Hall with the gilded vault in the “Domusurea”, while Balocco [13] considered the interaction between

ndoor and outdoor microclimatic conditions and the thermophy-ical behaviour of the Hall of the two hundred in the Old Palace oflorence. Besides Oetelaar et al. studied [14] temperature distribu-ions in a Roman bath, and Hussein [15] and Hussein and Shishiny16] investigated the low-speed wind environment around thereat Sphinx. Finally, Gan [17] considered interaction betweenxternal and internal airflow through a complex historical building.

This paper presents a reconstruction of Portus’ storage buildingsased on a multidisciplinary approach in which different hypothe-is of warehouses’, architectural aspects and materials, developedy archaeologists, are analysed by computational fluid dynamicsCFD) models in order to determine the resulting compatibilityetween indoor microclimatic conditions, linked to the consideredrchitectural reconstruction, and storage requirements. The paperepresents a first step to the knowledge of the ancient conserva-ion techniques. It is oriented to solve one of the main concernsrchaeologists have in this study, that is to demonstrate the exist-nce of external openings. In this way, it has been demonstratedow the numerical analysis can give a fundamental contribution inhe validation and the proposition of archaeological reconstructionypothesis.

. Topography of the site

The Traian warehouses in Portus are characterized by theirnterconnection with the harbour system [18]. The entire histor-cal structure is enclosed by open sea on the West side and by aystem of canals and docks on North and East sides (Fig. 1). This

s of Portus in the port system [18] (Fig. 91).

topographic position influenced the transport and the storage ofgoods.

Maintaining a certain symmetry, the Traian warehouses presenta “U” shape around the dock opening at East, in the direction ofthe hexagonal basin of the port of Traian. The rectangular block,consisting of three buildings and a central basin, is about 315 mlong by 175 m wide. The first archaeological investigation put inevidence the existence of about 150 storage cells on the groundfloor. The presence of many traces of stairs definitely testifies theexistence of another level of storage that would double this numberto obtain a total of 300 independent warehouses.

Considering only the ground floor, the surface area useful forthe storage arrived at about 12.150 m2, more than one hectare. Theentrance of the cells was always oriented to the external street forgoods transportation [19] (Fig. 2).

4. Archaeological and historical approach: geometry andstructure of the cells

The geometry of the cells (Fig. 3) is generally the same, butoften some details change from one cell to another, such as thearrangement of the air gap. Virtual reconstruction of Cell F.47(Fig. 4), realized basing on in situ measures and archaeologicalconsiderations, has been taken as an example for the architecturalreconstruction of the model. It is located on the West side of thewarehouses and looks out into the Portico of Claudio.

The Cell F.47 measures approximately 14 m (length) by 6.30 m(depth). The covering is composed by a stone barrel vault and themaximum high of the cell is about 8.5 m (from the floor up to thekeystone). In analogy with other Roman storage cells, it is possibleto suppose that this lengthened structure is typical of all the storagecells of Ostia and Portus’ warehouses.

The travertine threshold was raised by about 30 cm from the

external ground-level and the pins were arranged to allow thedoors opening towards the outside, thus, facilitating the mainte-nance of goods. At the end, a device consisting of two pillars of about60 cm width completed this locking system. The pillars extended

562 F. Pagliaro et al. / Journal of Cultural Heritage 16 (2015) 560–566

Fig. 2. Current planimetry of Portu

Fig. 3. Axonometric reconstruction of Cell F.47 [19] (Fig. 143).

Fig. 4. General view of cell F.47 with raised floor systems (suspensurae) (D. Gauss).

s’ warehouses [18] (Fig. 86).

the opening walls in order to prevent the reversal of unpackedgoods in the outside area.

This device was not provided in all warehouses. Therefore, itcould be assumed that this choice was related to the storage ofunpacked goods, like cereals, especially wheat [20]. Despite that, itis not possible to conclude that all warehouses equipped with thosedevices were destined to cereals stock considering the enormousquantity and variety of other goods arriving daily from all parts ofthe Empire.

Portus’ warehouses were protected from moisture and suddentemperature changes through the considerable thickness of thewalls (90 cm) and through an elevated floor system called “sus-pensurae”. This system was detected in the majority of the cellsinvestigated (Fig. 5). In particular, in Cell F.47 the system includes

10 channels of approximately 30 cm width and 50–60 cm height,delimited by brick walls about 30 cm thick. This structure wastopped by a double row of large square bricks called “bipedales”(60 × 60 cm) which ensured a good stability. Above this thick floor,

Fig. 5. Example of access of one of the storage cells – cell A.7 [18] (Fig. 89).

F. Pagliaro et al. / Journal of Cultural Heritage 16 (2015) 560–566 563

ed m

atelcc1

titctaCbptata

tpyrvtIpe[

5a

5

odmbtsaHt

stored in Portus, as wheat harvesting takes place mainly betweenthe end of June and the beginning of July. Then, to investigate themaximum possible variations of microclimatic indoor conditions

Table 1Duration of wheat warehousing (in days) [21].

Temperature

Moisture (%) 5 ◦C 10 ◦C 15 ◦C 20 ◦C 25 ◦C 30 ◦C

13 180 115 9014 160 100 50 3015 100 50 30 1516 130 50 30 20 817 65 35 22 12 518 130 40 25 17 8 219 70 30 17 12 5 020 45 22 15 821 30 17 11 7

Fig. 6. Representation of the discretiz

n additional layer of “cocciopesto” of 15 cm was arranged, pro-ecting hermetically the entire floor. It is reasonable to assume thexistence of a protection of the walls internal facade composed by aayer of hydraulic plaster. Furthermore, studies carried out in otherompound areas showed that the walls were covered with a “coc-iopesto” layer of about 4–6 cm up to an amount of approximately.80 m.

Among the “cellae” ruins, no traces of external openings, withhe exception of the door, or communication channels between thenternal and the external environments do exist. The remains ofhe “suspensurae” channels of Cell F.47 show no communication oronnection with the exterior environment, but they are connectedogether through two series of transverse openings, which wererranged approximately between 1/3 and 2/3 of the cell length.hannel floor was not continuous and consisted in a structure ofrick fragments that allowed a controlled ascent of moisture andrevented a big part of it through capillary structures. Nevertheless,he absence of communication channels with the outside seems toddress their main function as insulating air gaps that protectedhe interior of the cell from too tough thermal changes, creating anir cushion to stabilize the temperatures.

If the hypothesis that these structures were used mainly forhe storage of wheat can be considered as realistic, then the higherishability of this food material becomes relevant in the anal-sis, interpretation and reconstruction of the environment. In aoom completely closed to the outside, the wheat wastes awayery quickly. For the wheat storage, the microclimatic conditions ofhe interior environment should remain into an acceptable range.n particular, temperature and relative humidity are two criticalarameters for a good preservation of food and have a direct influ-nce on the development of microorganisms and bacteria (Table 1)21].

. Numerical approach and interpretation ofrchaeological evidence

.1. Mesh and modelling

A numerical model taking into account the thermal behaviourf the warehouse is very complex as it involves three calculationomains. Two fluid domains, represented by the external environ-ent and by the internal storage air space, were modelled and they

oth interact with a third one represented by the solid envelope ofhe warehouse. This paper only represents a first approach to the

tudy of the ancient “horrea”: consequently, a two-dimensionalnd steady state modelling was used to simplify the calculation.umidity effects were not considered to investigate the maximum

emperature difference inside the warehouse.

odel and of the boundary conditions.

The geometric module assumed for the calculation, 9.6 min height and 18.7 m in length, is made of four typical cells(2 downstairs and 2 upstairs), with the same rectangular shape anddimensions, with walls and roof thickness respectively of 0.9 m and0.5 m.

Dimensions and geometry of the external domain is modelledin accordance with the guidelines proposed by Franke et al. [22].Consequently, if H is the height of the considered model, the exter-nal fluid domain should be extended 10 H above, 10 H before, and20 H after the same model.

After a mesh sensitivity analysis, the model was discretized in astructured mesh composed by around 550,000 cells (Fig. 6).

The more relevant phenomena are inside the cell. For this rea-son a finer interval size of 0.2 m was chosen for internal and soliddomains and for the closest area around the building. In the remain-der of the external domain, a coarse mesh was chosen (Fig. 7).

The thermal behaviour of the cells without and with ventilationopenings towards the outdoor has been analysed basing on the dif-ferent available archaeological reconstructions. For the dimensionof the openings, the studies about the Portus Severianus and thelittle market of ancient Ostia [op. cit. 10, 11] suggest to assume, byanalogy, a model with 4 external and 2 internal rectangular open-ings, with the same size of 0.12 × 0.72 m (Fig. 8).

5.2. Simulation settings

The summer season was the period in which the wheat was

22 23 3 8 623 17 10 7 524 13 8 4 425 10 8 6 3

564 F. Pagliaro et al. / Journal of Cultural Heritage 16 (2015) 560–566

t(

amtWntot

pBt

Table 2Simulations summary.

Case Date Hour Openings

A.1 21th July 12.00 PM No

Fig. 7. Mesh detail.

he simulations are referred to the mean day of July at 12.00 at daycase A) and night (case B).

Outdoor air temperature, solar irradiance and air velocity datare derived from [23], with reference to “Fiumicino”, that is theunicipality in which is located Portus nowadays. With the excep-

ion of the direction of the wind, considered incoming from theest, the effects of the exposition to the sea of the West wall were

ot taken into account. Both the wind flow speed at the entrance ofhe external domain and the temperature of the soil under the floorf the warehouse’s structure are assumed to be constant, respec-ively at of 0,5 m/s and at 288 K.

Table 2 reports a summary of the simulations that have beenerformed by Ansys Fluent 14 [24] calculation package, with the

oussinesque approach, in which the k-� method is used to studyhe fluid turbulence.

Fig. 8. Schematic representation of the close case (a) and the open case (b).

A.2 21th July 12.00 PM YesB.1 21th July 12.00 AM NoB.2 21th July 12.00 AM Yes

In the case A an air temperature of 300.25 K and a pressure of101.325 Pa were imposed. The external ground temperature wasset at 302 K while the temperature of the soil under the floor of thebuilding was set at 288 K.

The radiation was not used in the calculation model. A sol-airtemperature was imposed to the outside layer of the cover to emu-late the effects of radiative exchange between the sun and thestructure. It has been calculated analytically through the equation:

Tsol−air = Ta +(

�inc• ˛

he

)(1)

where� absorption coefficient= a 0.72he abduction coefficient = 25 W/m2KTa air temperature= 300,25 K�inc incident radiation perpendicular to the surface = 829 W/m2

From (1), we can assume roof Tsol-air is 324.3 K for the case A.In the case B, an air temperature of 283.95 K and a pressure of

101.325 Pa were imposed. The external ground temperature wasset at 283.95 K while the temperature of the soil under the floor ofthe building was set at 288 K.

In this case, Tsol-air is not included since the incident radiation isequal to 0.

6. Results and discussion

The calculated external velocity field for the case B is reported inFig. 9, while Figs. 10 and 11 show the contours of static temperaturein the internal domain and in the solid domain in the daytime (casesA.1 and A.2) and in the nighttime (B.1 and B.2).

In the case A.1, the temperature of the upper “cellae” 1 and 2,which are in direct contact with the roof exposed to solar radia-tion, increases compared to the external temperature, reaching anaverage temperature of around 314 K. The lower “cellae” temper-ature is influenced by the soil. Therefore, there is a temperature

stratification from the lower temperature of the soil to the highestof the roof (average temperature around 300 K). In the case A.2, thepresence of openings allows a decrease of the high temperature ofthe covering due to the flow entrance from the outside domain.

Fig. 9. External velocity field: contours (a) and vectors (b) in m/s.

F. Pagliaro et al. / Journal of Cultural Heritage 16 (2015) 560–566 565

Itpcwpm

n

cte

pu

Fig. 10. Temperature distribution for cases A.1 (a) and A.2 (b) in Kelvin.

n the nighttime, with the absence of solar radiation, there is noemperature increase in the indoor environment. Nighttime tem-eratures are inverted compared to daytime temperatures. In thease B.2, the presence of openings and the direct contact with thearmer external temperature allow the reaching of higher tem-eratures compared to the case B.1, were the presence of the solidass allows a slower heating of the internal air.Figs. 12 and 13 show the vectors of air velocity within the inter-

al air domain; results are very similar in daytime and in nighttime.In A.1 and B.1 cases, as there are no openings, air motion is

aused by temperature differences between the ground, the air andhe roof. These movements are very small and the internal air isssentially stagnant.

In A.2 and B.2 cases, the presence of openings creates airflowaths between the indoor and the outdoor environment. In thepper cells of case B.2, the incoming air speed is lower than in case

Fig. 11. Temperature distribution for cases B.1 (a) and B.2 (b) in Kelvin.

Fig. 12. Velocity vectors for cases A.1 (a) and A.2 (b) in m/s.

A.2. This phenomenon is due to the presence of a cold roof that alsocauses a change in the flow pattern. In particular, in the upper cell2 it creates two zones of recirculation instead of a large one as incase B.2.

Results show an indoor temperature field that is greatly influ-enced by solar irradiance in its value and in the direction of verticalthermal gradient, that appeared to be opposite during the dayand the night. During the daytime, the presence of high temper-atures due to the solar radiation on the roof does not allow tomaintain acceptable internal temperatures for wheat storage. Com-paring the temperature values with the Table 1 in the case A.1, thetemperature will be always too high even if the humidity is verylow. Therefore, if we accept this hypothesis, it will mean that thewheat was stored only for short periods of time or the warehouses

were used to conserve other kinds of goods. Otherwise, we shouldassume that the warehouses had openings to the outside whichallowed air changes and a faster dissipation of heat from the inside.

Fig. 13. Velocity vectors for cases B.1 (a) and B.2 (b) in m/s.

5 ltural

Tr

7

otaittottibcc

acutwbowhstrao

cpbsi

R

[

[

[

[

[

[

[

[

[

[

[

[

[

66 F. Pagliaro et al. / Journal of Cu

hrough of all these numerical results and considerations, we haveeached the goal proposed with this approach.

. Conclusion and future developments

In this study, archaeologists developed different reconstructionsf Portus’ warehouses that are modelled with a simplified geome-ry in order to evaluate the indoor thermal conditions, by means of

CFD computational package. The goal is to compare the compat-bility between thermal requirements for food conservation andhe thermal fields calculated for the different models to identifyhe more realistic reconstruction. Then, calculations are performednly for a day of July, for 12 o’clock PM and AM, as summer washe period in which foods were stored in Portus. In this approach,he influence of mass inertia that reduces indoor thermal gradients not considered, a steady-state condition is considered, and onlyi-dimensional models are developed. The results show the physi-al need and consequently the existence of openings in the Portus’ells as mandatory to guarantee good food storage.

By means of CFD tools, a simple 2D thermal and fluid dynamicnalysis was performed in a virtual reconstruction of a set of Portusells. A very interesting result was achieved, as numerical sim-lation showed that the archaeological reconstruction based onhe absence of openings does not match with the conservation ofheat in Portus’ warehouses. Consequently, one possibility would

e to follow the archaeological evidences: accepting the absencef openings, it is easy to demonstrate the impossibility to store theheat for long periods. This consideration allows to formulate otherypotheses, like the use of those warehouses only for unloading,hort storage and distribution of goods, or at least the use of themo store goods different from wheat. On the other side, this researchepresents the starting point for the demonstration and validation,nd then the investigation on the shape and the localization of thepenings within the warehouses.

As a conclusive remark, we want to state the importance of theonnection between archaeological and numerical analysis to pro-ose and validate reconstructive hypotheses of no more existinguildings as a useful tool for archaeologists in their effort to repre-ent the ancient buildings as they were and to interpret their rolen the context of a city and society.

eferences

[1] E. Bukowiecki, et al., Chroniques des activités archéologiques de l’écolefranc aise de Rome. Italie centrale : Portus : les entrepôts de Trajan, MEFRA125-1 (2013) [Rome].

[

[

Heritage 16 (2015) 560–566

[2] http://www.entrepots-anr.fr/p-ostie en.htm[3] S. Keay, M. Millett, L. Paroli, K. Strutt, Portus. An archaeological survey of the

Port of Imperial Rome Archaeological Monographs of the British school, 15,2005.

[4] E. Bukowiecki, C. Rousse, Chronique – activités archéologiques de l’écolefranc aise de Rome : Ostia antica : étude sur la construction des Grandi Horread’Ostie, 2007, pp. 283–286 [MEFRA 119-1].

[5] G. Lock, Using computers in archaeology: towards virtual pasts, Routledge,London and New York, 2003.

[6] T.L. Evans, P. Daly, Digital archaeology: bridging method and theory, Routledge,London and New York, 2006.

[7] F. Gugliermetti, Can computational models improve investigations and diag-nostics in cultural heritage, in: Proc. of Int. study meeting on Science andcultural heritage, Palermo 18–21 March 2007, Quaderni di Palazzo Montalbo,Priulla, Palermo, 2008, pp. 201–220 [ISBN: 9788861640863].

[8] F. Gugliermetti, From the virtual model to the archaeological world: the chal-lenge of the predictive computational method, in: Proc. on Int. conferenceexchange for a challenge, Kocaeli, Turkey, 2010 [exchange.kumid.net].

[9] K. Chourmouziadou, J. Kang, Acoustic evolution of ancient Greek and Romantheatres, Appl. Acoustics 69 (2008) 514–529.

10] F. Gugliermetti, F. Bisegna, A. Monaco, Acoustical evolution of the Roman the-atre of Ostia, Building Acoustics 15 (2008) 153–168.

11] M. Navvab, F. Bisegna, F. Gugliermetti, Evaluation of historical museum inte-rior lighting system using fully immersive virtual luminous environment, SPIEOptic. Metrol. (2013) 13–16.

12] S. Albero, C. Giavarini, M.L. Santarelli, A. Vodret, CFD modelling for the con-servation of the gilded vault hall in the Domus Aurea, J. Cultur. Herit. 5 (2004)197–203.

13] C. Balocco, Daily natural heat convection in a historical hall, J. Cultur. Herit. 8(2007) 370–376.

14] T. Oetelaar, C. Johnston, D. Wood, L. Hughes, J. Humphrey, A computationalinvestigation of a room heated by subcutaneous convection – a case study of areplica Roman bath, Energy and Buildings 63 (2013) 59–66.

15] A.S. Hussein, Wind flow modeling and simulation over the Giza Plateau culturalheritage site in Egypt, J. Comput. Cultur. Herit. 2 (2009) [art. No. 6].

16] A.S. Hussein, H. El-Shishiny, Influences of wind flow over heritage sites: a casestudy of the wind environment over the Giza Plateau in Egypt, EnvironmentModel Software 24 (2009) 389–410.

17] G. Gan, Interaction between wind and buoyancy effects in natural ventilationof buildings, Open Construct. Building Technol. J. 4 (2010) 134–145.

18] G. Boetto, E. Bukowiecki, N. Monteix, C. Rousse, Portus. Les entrepôts de Trajan,Chronique – autres activités archéologiques franc aises en Italie, MEFRA 122-1(2010) 303–310.

19] E. Bukowiecki, C. Panzieri, S. Zugmeyer, Portus. Les entrepôts de Trajan,Chronique – autres activités archéologiques franc aises en Italie, MEFRA 123-1(2011) 349–357 [Rome].

20] J. Salido Dominguez, Horrea Militaria. El aprovisionamiento de grano al ejércitoen el occidente del imperio romano, 14, Anejos de Gladius 14, Madrid,2011.

21] M. De Lucia, M. Assennato, Agricultural engineering in development – post-harvest operations and management of food grains, FAO Agricultur. Serv. Bull.93 (1994) http://www.fao.org/docrep/T0522E/T0522E09.htm#Storage

22] J. Franke, A. Hellsten, H. Schlunzen, B. Carissimo, Best practice guideline for theCFD simulation of flows in the urban environment, in: COST action 732, quality

assurance and improvement of meteorological models, 2007.

23] L. Mazzarella, Dati climatici G. De Giorgio, Proc. of Giornata di Stu-dio Giovanni De Giorgio, Politecnico di Milano, 18th November, Milano,1997.

24] Ansys Inc., ANSYS fluent software package, User’s manual, version 14, 2011.