Available online at www.sciencedirect.comsCIENCEs @oInEoT.Transactions of、Nonferrous MetalsSociety of ChinaScienceTrans. Nonferrous Met. Soc. China 16(2006)s335-s339Presswww.csu.edu.cn/ysxb/Modeling simulation and experimental validation for mold filling processHOU Hua (侯華), JU Dong-ying(巨東英)2, MAO Hong-kui(毛紅奎), D. SAITO 21. College of Materials Science and Technology, North University of China, Taiyuan 030051, China;2. Department of Mechanical Engineering, Saitama Institute of Technology, Fusaji, Saitama369-0239, JapanReceived 20 April 2006; accepted 30 June 2006Abstract: Based on the continum equation, momentum conservation and energy conservation equations, the numerical model ofturbulent flow fllig was introduced; the 3-D free surface vof method was improved. Whether or not the numerical simulation resultsare reasonable, it needs corresponding experimental validations. General experimental techniques for casting fluid flow processinclude: thermocouple tracking location method, hydraulic simulating method, heat-resistant glass window method and X-rayobservation etc. The hydraulic analogue experiment with DPIV technique is arranged to validate the fluent flow program forlow-pressure casting with 0.1X 10 Pa and 0.6X 10 Pa pressure visually. By comparing the flow head, liquid surface, flow velocity,it is found that the flling pressure value influences the flow state strongly. With the increase of the flling pressure, the fluid flowstate becomes unstable, the flow head becomes higher, and the flling time is reduced. The simulated results are accordant with theobserved results approximately, which can prove the reasonability of our numerical program for flling process further.Key words: turbulent flow flling; VOF method; numerical simulation; experimental validation; DPIVseveral thousands even millions, hence the computation1 Introductionefficiency is considerably low, so it is important todevelop more efficient and precise arithmetic. In addition,Numerical modeling provides a powerful means ofcasting flling experiments are usually difcult to carryanalyzing various complex phenomena, such as thermalout and simulation results are often unsatisfied etc.flow, heat transfer and solidification as well as theircoupling problems occurring during flling process of2 SOLA method in numerical analysis of fluidcasting. Simulation allows researchers to observe andmechanicsquantity what is not usually visible or measurable duringMass conservation and momentum conservation arereal casting processes. The results of such simulations isto help shorten the design process and optimizerepresented by the continuum equation and N-S equationconditions of casting process to reduce scrap, use lessrespectively as follows:energy and, of course, make better filling process ofaU. aV. oW-= ((1)casting[1]. Recently, numerical analysis method anDx oy 0:modeling of the fluid flow for casting process has beenUstudied in previous work[2- -8]. Whether or not the+U_+Wnumerical simulation results are reasonable, it needsaDx0corresponding experimental validations..1p(2u, 2U, 2∪experimental techniques for casting fluid flow processρ 0x0x2 oy2 0z2 .include: thermocouple tracking location method,..aV..0.oVhydraulic simulating method, heat-resistant glasswindow method and X-ray observation etc .In this paper, the hydraulic flling process is carriedout to validate the thermal flling numerical simulation.1 op.。 (or. 2V 2v)In flling algorithm, the iterative times often extend toρOx中國煤化工TYHCNMH GCorresponding author: HOU Hua; Tel: +86- 13934153099; E-mail: houhua@263.net.HOU Hua, et alTrans. Nonferrous Met. Soc. China 162006)s336L+v+ w.0Wof free suface units (0 F(ij, k-1), consider that the fluidSome methods have been explored to improve theexists above the free surface, otherwise fluid existscalculation efciency with decreasing iteration frequencybelow the free surface.and quickening convergence speed.\nd now there areSince the "Donor-Acceptor" method is originallyuch methods as the conjugate gradient method,adopted in 1-D or 2-D simulations, in 3-D case, fluid canpre-treatment中國煤化工d，dyamieflow into or out of the six faces of a mesh unitsimultaneously, so there will appear some“false”relaxation fact(HH出2so proposedCNMHGelandreducediffusions, and this method can bring some errors, evenmixture modelresult in interface vague phenomenon. There have plentycalculation..s337HOU Hua, et al/Trans. Nonferrous Met. Soc. China 16(200)The internal and extemal area separationsimplification algorithm is adopted in this paper.LaserThe liquid metal fling area is divided into twoGlass mouldparts: the intenal unit and fre surface unit. At everytime step dt, free surface units move continuouslyLight sheetforward, and form new free surfaces units to form newintermal and external area as shown in Fig.2.↑Liquid flow :Image analysismeraFig.3 Diagram of DPIV setOwing to the successive iluminations, discrete .8十images of the particles are recorded. At high seeding19densities, however, “tracks" of individual particlescannot be ientied, much less analyzed[9]. Hence, themean displacement between successive iluminations ofseveral particles in a so. called interrogation area has to/1 / 1+drFig.2 Chart of internal and extermal zone separationbe estimated. To this end, the image is digitized and thespatial correlation of the light distribution is calculated.With this scheme, the velocity, pressure contained inWith an iterogation area, the auto-covariance functiona unit docesn't change again after the unit becomes anis estimated by means of Fast Fourier Transforming (FFT)internal unit. All changes occur in the free surface unitstechnology[10]. The mean displacement of the particlesand their nearest units. The iteration calculation of speedin theinterogation arca between two sccessiveand pessure is crried out in this area. As shown in Fig.iluminations is obtained from the shift in the peaks.2, the iteration calculation at t+dt will be crried out onlyIn this study, gas bubbles taken as particles arein the 1-10 units. So the grids number of participating inadded to the water, a laser beam from 1.3 w argon-ionthe iteration decreases grealy to raise calculationlaser is transmitted onto the glass mold (as shown ineficiency. Yet the absence of pesure and speed in theFig.3).intemal units can afect the calculation precision andThe size of the glass mould is 320 mmx 280flow state.mmx200 mm, its bottom ingate diameter is 30 mm. Incalculation the mesh size is 5 mmX5 mmX5 mm, and4 Principle of_ digital paraticle-imagethe total meshes are 143 360. The water fling is criedvelocimetry (DPIV)out under 0.1X 10' Pa and 0.6X 10* Pa pressure Here,the findings from the visualization (experiments, from theFig.3 presents the chart of the hydraulic analoguesimulation study, and from the DPIV measurements are ;experiment with DPIV intallation. This technologymutually compared.exploits the behavior of tracer particles added to a fluidFig.4 presents the observed (a1-h1) and simulatedflow, as the particles are very small and have nearly the(a2- th2) surface waves under the pressure of0.1X10* Pa.same density as water, they act as flow followers in theFigs.2 (c3- -g3) ilstrate the observed streamlines. Figs.4local transient velocity field without disturbing the flow(a4-h4) show the simulated flow velocity evolution.tself. Gerally, a laser beam transformed into a thinWe can see from Figs.4(al- -h1) that, at beginning,plane of light iluminates a section of the flow containingthe liquid surface is lower, and the flow head is bigger.a certain number of strongly refctie tracer particles.With the increase of flling volume faceion, the liquidThe iluminated particles are mapped on an image plane,surface ascends slowly and steadily, the flow headand their positions are recorded by means ofbecome more and more plain and is submerged byphotographic plate, film, video, or a charge coupledsurrounded wal中國煤化工fling volumedevice camera. From subsequent recordings with afraction arrivesall flow head,known separation time O, a two-dimensional velocityand the flow hMHc N M H Gling volumepattern of the flow in the light sheet is determined.fraction 60% to the end.HOU Hua, et al/Trans. Nonferrous Met. Soc. China 16(2006)s338In Figs.4(a2 -h2)， with the increase of flling(al)(a2)(a4)volume fraction, the liquid surface ascends slowly andsteadily, the flow head becomes more and more plain,which is approximately consistent with the observationfrom Figs.4(al -h1), yet, the simulated flow head has notdisappear until end.Flling 5%In Figs.4(c3- -g3), from the state of streamlines, we(61)(b2)](b4)can see the flow state and the distribution characteristicof velocities. From Figs.4(c3- -g3) and (c4- g4), we cansee that, the simulated flow heads in Figs.(c4- g4) aremore sticking out than that in Figs.(c3- g3), the simulatededdy currents at both sides of the Core vertical upwardFilling 10%stream are relatively weaker than that in Figs.(c3- -g3).That is because in the experiment, there are two small(cL)(c2)(c3)(C4)metal columns at the bottom of the rectangle vessel,which can hinder the flow of the two sides' liquidstreams and cause the decomposition of the verticalupward speed to sides' directions, so the flow head isFilling 20%lower and the eddy current is stronger. In calculation, thedD)(d2)1 (d3(d4)small cylinders are neglected for simplicity; so thesimulated results are somewhat different with theexperimental results.In Figs.4(a4- h4), the direction and length of thearrow respectively represent the direction and magnitudeFilling 30%of the velocity. It can be seen that the maximum velocity| (eD)(e2)(e3)(e4)is in the center plane of the flow head and verticallyupward. There have been velocity side componentsamong the velocity of other positions. From the velocityvector distribution, it also can be found that, there existFilling 40%ome circuit flows around the flow head during the(f2)(3(f4)increase of the liquid surface. These are also accord withthe DPIV results.Although the simulated results and the DPIV明國findings do not agree completely in terms of the state offlow head, the trends are similar. Fig.5 presents theFilling 50%observed (a1-h1) and simulated (a2- -h2) surface waves. (g3)(g4)under 0.6X 10 Pa. Figs.3(a3 -h3) show the simulated(g2)flow velocity evolution.We can see from Figs.5(a1-h1) that, at beginning,the liquid surface is lower, and the flow head is quitebigger. With the increase of flling volume fraction, theFilling 60%liquid surface ascends with some lttle disturbance, the6D(h2)(h4)19flow head is always sticking out obviously until the end.The simulated results in Fig.5(a2- -h2) are appro-ximately accordant with above experimental findings. Itis different that the simulated liquid surface ascendsquietly, yet, a中國煤化工t mathematicmodel suitablecon, and thisFilling 90%Fig.4 Comparison with observed and simulated flling processnumerical simMYHC N M H G explorationunder pressure of0.1X10' Pastage, so the comparison of simulation and experiment is.s339HOU Hua, et al/Trans. Nonferrous Met. Soc. China 16(2006)In Figs.5(a3- -h3), it can be seen that the maximumI (alD)(a2)(a3)velocity is in the center plane of the flow head andvertically upward. There have been velocity sidecomponents among the velocity of other positions. Fromthe velocity vector distribution, it also can be found that,there exist some circuit flows around the flow headFilling 5%during the increase of the liquid surface. These are alsoaccordant with the DPIV results.b1)(b2)(b3)Although the simulated results and the DPIVfindings do not agree completely in terms of the liquidsurface, the trends are also similar.Compared to the pressure of 1.0X 104 Pa, theFilling 10%differences of the center and surrounded velocities areobvious with that of 6.0X 10* Pa pressure.C1)(e2)(c3)Comparing Fig.4 with Fig.5, we can find that, theflling pressure value influences the flow state strongly.With the increase of the flling pressure, the fluid flowstate becomes unstable, the flow head becomes higher,and the flling time is reduced.Filling 15%Based on above discussions, it can be concluded thatd)(d2)(d3)the simulated results are accordant with the observedresults approximately, which can prove the reasonabilityof our numerical program.Filling 25%References| (e1)(e3)[1] HOU H. Studies on Numerical Simulation for Liquid metal Fllingnd Solidification During Casting Process[D]. Japan: SaitamaInstitute of Technology, 2005HOU H, JU D Y, ZHAO Y H, CHENG J. Numerical simulationfordendrite growth of binary alloy with phase-field method[J]. J MaterSci Technology, 2004, 20(12): 45- -48.Filling 40%] HOU H, ZHAO Y H, CHU z, XU H. Numerial simulation of 3-D(f3)displacement fields of steel casting during solidifcation process[J].Foundry Technology, 2002, (3): 145-149. (in Chinese)4] YOU B K. Temperature Measurement and Instrument: Couple andThermal Resistance [M]. Document Press of Science and Technology,1990.COMPBELL J. Invisible macrodefects on castings[J]. Journal DeFilling 50%Physique iv, 1993(3): 861-872.| (gD(g2)JUD Y , INOUE T. Simulation of soidifcation and heat flow in stripcasting process by twin roll method[A]. Proceeding of Conference onComputer- asisted Design and Process Simulation[C]. Tokyo, 1993.84-98[7] JU D Y, INOUE T Analysis of heat fluid flow incoporatingsolidification and its application to thin slab continuous castingFilling 70%process([]. Transaction of the Japan Society of MechanicalEngineering (JSME), 1993, 59(565): 2189- -2195.，(h2)業(yè)8(h3)[8] KAPRANOS P, BARKHUDAROV M R, KIRKWOOD D H.Modeling of structural breakdown during rapid compression ofsemi-solid alloy slugs[A]. 5th International Conference onSemi-solid Processing of Alloys and Composites[C]. Golden, 1998.23- -25.HOU HUA. Studies on Numerical Simuation for Liquid metalFilling 80%Filling and Solidification During Casting Process[D]. Japan: SaitamaInstitute of Technology, 2005.Fig.5 Comparison between simulated and observed flling10] WESTER J. Delft University of Technology[D]. Delft, Netherlands:process under 0.6X 10 PaDelft University ofTechnology, 1993.(Edited by CHEN Ai-hua)中國煤化工main focused on the flow head and the eddy currents.Hence，the numerical method needs to be improvedMYHCNMHGfurther..