CASE CAS E STU STUDY DY 203
Computer simulation of casting process of aluminium wheels – a case study Yeh-Liang Hsu* and Chia-Chieh Yu Department of Mechanical Engineering, Yuan Ze University, Taiwan, Republic of China The manuscript was received on 9 May 2005 and was accepted after revision for publication on 20 October 2005. DOI: 10.1243/09544054JEM381
Abstract: Aluminium disc wheels intended for normal use in passenger cars are commonly produced by gravity casting. If the cooling process and the initial temperature of the mould are not wel welll con control trolled led,, shr shrink inkage age cav cavity ity wil willl occ occur ur aft after er sol solidi idific ficatio ation, n, cau causin sing g lea leakage kage in the th e di disc sc wh whee eel. l. In th this is re rese searc arch, h, a ca cast stin ing g si simu mulat lation ion sof softwa tware re is us used ed to si simu mulat late e th the e castin cas ting g pro proces cesss of alu alumin minium ium whe wheels els.. The cas castin ting g sim simula ulation tion is don done e ite iterati rativel vely y unt until il the mould temperature converges to a stable temperature. A ‘shrinkage index’ (SI) is defined to prov pr ovid ide e a qu quan anti tifi fied ed in inde dexx of ca cast stin ing g qu qual alit ity y of al alum umin iniu ium m wh whee eels ls,, ba base sed d on th the e phen ph enome omeno non n of li liqu quid id en entra trapp pped ed at th the e jo join ints ts of ri rim m an and d sp spoke okess of th the e wh whee eell wh wher ere e shrinka shr inkage ge cav cavity ity usu usuall ally y hap happen pens. s. Thi Thiss shr shrink inkage age ind index ex sho shows ws good cor correl relati ation on wit with h the alumin alu minium ium whe wheel el lea leakage kage tes testt res result ults. s. Thi Thiss pap paper er also dis discus cusses ses the inf influe luence nce of coo coolin ling g process parameters on SI, including initial mould temperature, and geometry of the wheel, which verifies engineers’ empirical data. This iterative simulation process and SI can be used to predict the casting quality of aluminium wheels and to find the optimal parameters of the casting process. Keywords: aluminium disc wheels, casting, shrinkage cavity, liquid entrapped
1 INT INTROD RODUCT UCTION ION Aluminium disc wheels intended for normal use on passenger cars are commonly produced by gravity casting. Figure 1 shows the four casting moulds – top mould, side mould, bottom mould, and support mould – for an aluminium disc wheel. The cooling conditions are applied to the moulds. If the cooling process and the initial temperatures of moulds are not well controlled, shrinkage cavity can occur after solidif soli difica ication tion,, cau causin sing g lea leakage kage in the dis disc c whe wheel. el. Several Sev eral pra practi ctical cal str strateg ategies ies are ofte often n emp employ loyed ed to preven pre ventt thi this. s. The These se inc includ lude e dri drilli lling ng air ves vessel selss to increase the rate of heat transferred from the joints of rim and spokes of the wheel, and spraying vapour on the bottom mould to increase the cooling rate. In aluminium wheel manufacturing, these strategies are applied currently on a ‘trial-and-error’ basis, and depend heavily on the experience of engineers.
*Corresponding author: Department of Mechanical Engineering,, Yua ing Yuan n Ze Uni Univer versit sity, y, 13 135 5 Yu Yuan an Tun Tung g Roa Road, d, Chu Chung ngli, li, Taiwan, Republic of China. email:
[email protected]
To im impr prov ove e th the e qu qual alit ity y of fo foun undr dry y pr prod oduc ucts ts has lon long g bee been n a res researc earch h iss issue ue in man manufac ufactur turing ing industry. Numerical models are developed to predict the mec mechan hanica icall cha charact racteri eristi stics, cs, shr shrinka inkages ges,, and porosi por osities ties.. The cas castin ting g pro proces cesss and the effe effecti ctive ve param pa ramet eter erss ar are e ca care refu fully lly st stud udie ied d to ad addr dres esss th the e improvement schemes. Tiwari and Roy [1 [1] used neural networks to build an in inte telli llige gent nt sh shri rink nkage age mi mini nimi mizat zatio ion n mo modu dule le,, which learns the real behaviour of the solidification process so that it can perform the task of casting desi de sign gn fe feat atur ure e mo modi difi fica cati tion on in re real al ti time me an and d intens int ensify ify the pro proces cesss of dir direct ection ional al sol solidif idifica icatio tion. n. Seetharamu et al . [2 [ 2] used the finite element method to simulate the heat transfer process accompanying the soli solidif difica ication tion proc process ess.. The res results ults of res residu idual al stresses, shrinkage, and thermal stresses were compared par ed wit with h avai availab lable le exp experi erimen mental tal dat data. a. Bou Bounds nds et al . [3] modelled the formation of macro defects, macro porosity, mis-runs, and pipe shrinkage, explicitly as a function of the interaction among freesurface fluid flow, heat transfer, and solidif solidificatio ication n in arb arbit itrar raril ily y co comp mple lexx th thre reee-di dime mens nsion ional al ge geoometries. metrie s. Midea et al . [4] illust illustrated rated four exampl examples es
204
Yeh-Liang Hsu and Chia-Chieh Yu
in which casting process modelling is combined with other computer modelling to optimize cast component manufacturability. Shenefelt et al . [5] used ‘criteria functions (CFs)’ based on thermal environment to provide a means for estimating shrinkage porosity within a casting. In particular, the use of mould filling and solidification commercial simulation software to investigate the filling patterns, velocities, and temperature distributions, has become increasingly popular. Spittle et al . [6] used MAVIS, a heat transfer/ solidification simulation package, to predict the temperature distributions in a permanent mould. Drezet et al . [7] analysed a nominal ingot using the finite element software ABAQUS to compare two different casting speeds and free mould designs and obtained more uniform thickness. Kreziak et al . [8] utilized SIMULOR, a filling and solidification simulation software, to simulate a quarter of an automobile wheel cast. The results were validated
Top mould
by experimental data that temperatures were measured from different positions. In the meantime, they used a simple sample to study the effects of cycle time, preheat temperature, and die coatings. The current paper presents a case study of using computer simulation for the casting process of aluminium wheels of a local manufacturer in order to establish a process to find out the optimal cooling conditions to avoid shrinkage cavity. The ‘liquid entrap’ phenomenon at the joints of rim and spokes of the wheel during the casting process causes shrinkage cavity of the final aluminium wheel. In this research, a ‘shrinkage index’ (SI) is defined to describe the amount of entrapment of liquid. It provides a quantified index of casting quality of aluminium wheels. The casting simulation is done iteratively until the mould temperature converges to a stable temperature. This paper starts by describing the simulation model, the simulation process, and defining SI. Correlation of the SI with the aluminium wheel leakage test results of the local manufacturer is then investigated using the iterative simulation. The influences of cooling process parameters on SI are then discussed, including initial mould temperature and geometry of the wheel. 2
Side mould
Button mould
Support mold
Fig. 1 CAD models of casting moulds for an aluminium disc wheel
Air cooling
SIMULATION OF CASTING PROCESS OF ALUMINIUM WHEELS USING ProCAST
Many major foundries use commercial software such as ProCAST and MagmaSOFT, to simulate filling and solidification of castings. In this paper, ProCAST is used to simulate the casting process of aluminium wheels. ProCAST uses the finite element method and can be employed to analyse a wide variety of fully coupled thermal, fluid, stress, and microstructure prediction problems in the casting process [9]. Figure 2 shows the finite element model of a 15 in aluminium wheel and its moulds. The interfaces of each part are coincident in the model.
Water cooling
Fig. 2 The finite element model of an aluminium wheel and its moulds
Computer simulation of casting process of aluminium wheels
Table 1
205
The material properties of the alloy
Fraction solid
Density
Thermal conductivity
Temperature (C)
%
Temperature (C)
kg/m 3
Temperature (K)
W/(mK)
557.98 558.00 562.00 564.00 566.00 568.00 570.00 570.50 574.00 578.00 582.00 586.00 590.00 594.00 598.00 602.00 605.00
100 92.10 90.16 88.20 82.81 71.54 53.90 45.10 44.10 41.65 38.70 34.40 28.42 22.54 15.20 7.06 0.00
25.00 656.00 664.00 700.00
2702.00 2540.00 2380.00 2369.00
406.67 420.16 432.26 477.02 589.52 666.13
146.50 153.26 154.94 166.76 167.43 166.08
Four-noded tetrahedral elements are used. The material used is AlSi7Mg (ASTM A356, JIS AC4C) casting aluminium alloy, which can be found in ProCAST’s material database. The heat capacity of the alloy is 963 J/(kg K), and the latent heat is 3.98 · 105 J/kg. Table 1 lists other thermal characteristics of the alloy, which are functions of temperature. The boundary conditions of the simulation are based on the wheel manufacturer’s real parameters. In casting, one or more of the following remedies are often used to prevent unfilled cavity: preheating the die, insulating some/all of the cavity surface with die coating, and increasing the filling velocity. In the current simulation, the temperature of the melted alloy is 720 C. The cavity volume of the model is 7.474eÀ3 m3, and the fill duration is 16 s. The filled velocity is 0.673 m/s through the sectional area of 0.693 m3. The die is preheated to 360 C, while the ambient temperature is 30 C. For a coated die with a metal–mould heat transfer coefficient of 300 W/m2 C, the whole cavity can be filled in the simulation. Figure 2 also shows the locations of air cooling (by blowing cold air to the side mould) and water cooling (by spraying water to the bottom mould). In the simulation, cooling parameters are also based on the wheel manufacturer’s real parameters. The heat transfer coefficient of air cooling is 700 W/m2 C and its affected area is 4800 mm 2. The heat transfer coefficient of water cooling is 2200 W/m2 C and its affected area is 88 822 mm2. The filling of melted metal completes after 16 s. Air cooling starts at 66 s and continues until the end of casting (240 s). Water cooling starts at 126 s and lasts for 40 s. Finally, the casting wheel is picked out of the cavity at the end of casting. The
mould is cooled down for 30 s, and the next casting cycle starts. Figure 3 shows the solidification process of a typical aluminum wheel (referred to as ‘Type A’ in this paper) from ProCAST. After approximately 150 s into the casting process, the liquid starts to be entrapped at the intersection between the spokes and rim. At the positions indicated by the circles, the surrounding regions become solidified. Both the central riser and the rim riser cannot provide liquid metal. The position of entrapped liquid is coincident with a volume where the aluminium wheel actually fails, as shown in Fig. 4. The simulation shown in Fig. 3 assumes constant initial mould temperature everywhere. However, this is not true in reality. Figure 5 shows the temperature distribution of the casting process of an aluminium wheel simulated by ProCAST. The temperature scale range is 250–466 C. During the casting process, the temperature distribution changes after filling, air cooling, water cooling, and casting out of cavity. In a continuous casting process, this final temperature distribution of the mould becomes the initial temperature distribution of the mould for the next casting. As mentioned earlier, Spittle et al . [6] used MAVIS to predict the temperature distributions in a permanent mould. The mould for the production of AlSi7Mg alloy castings had been used to assess the influence of mould design modifications and water cooling on the steady state temperature distribution in the mould and the freezing characteristics of the casting. Results matched very well between experiment and simulation, where a steady state was assumed to be achieved in any batch run without water cooling after 20 castings.
206
Yeh-Liang Hsu and Chia-Chieh Yu
131 s
152 s
173 s
Fig. 3 The solidification process of a typical aluminium wheel from ProCAST
Fig. 4 The position where the aluminium wheel actually fails
In the current research, an attempt is made to evaluate the cooling parameters only after a steady state mould temperature is reached. ProCAST simulation is performed continuously for ten cycles. In the first simulation, the mould temperature is assumed to be constant everywhere at 360 C. The final temperature distribution of mould of the simulation is then used as the initial temperature distribution of mould in the next simulation. Figure 6 shows the results of ten simulations and shows the maximum temperature (D), the minimum temperature (r),
Fig. 5 Temperature distributions of mould during the casting process
Computer simulation of casting process of aluminium wheels
207
Fig. 6 The maximum, minimum, and mean temperature of the casting mould in each simulation
Fig. 7 Liquid entrapment at the joint of rim and spokes
and the mean temperature (o) of the casting mould in each simulation. The temperature distribution of the casting mould reaches a steady state after ten simulations – the relative changes in maximum temperature, minimum temperature, and mean temperature are all less than one per cent. Therefore, in this research the casting process simulation results are observed after ten cyclic simulations. 3 DEFINITION OF A SHRINKAGE INDEX Figure 7 shows the solidification of the vertical section of wheel Type A. The solidification scale of
Fig. 7 is from 0.3 to 0.9 (0 for pure liquid and 1 for solid). Critical fraction solid, the point at which the alloy is solid enough that liquid feed metal can no longer flow, is assumed to be 0.7. As shown in the left of Fig. 7, the solidification scale reaches 0.7 in the rim of the wheel indicated by the dashed circle after approximately 131 s. The riser on the top of the rim (rim riser) cannot provide melted alloy to the joint of rim and spokes. As the cooling process continues, the solidification scale reaches 0.7 in the spoke of the wheel after about 173 s. This portion is thicker and is located further from the risers, so longer solidification time is required. Now the central riser cannot provide melted alloy to the joint of rim and spokes. Therefore, liquid entrapment occurs at the joint of rim and spokes. Kreziak et al . [8] showed that there is no risk of shrinkage under where the solidification presents an orientated temperature gradient from the top to the running system, and the critical solid fraction isochronal chart does not show liquid entrapped areas. When a part of the casting is locally being solidified without feeding from the system, this is a high-risk area. Therefore, the volume of the liquid-entrapped portion in the wheel can be used to indicate the level of shrinkage in the wheel. An SI is employed in order to define quantitatively the level of shrinkage from the simulation results by ProCast. Figure 8 shows the portion of the wheel where shrinkage cavity usually happens. ProCAST can output the solid fraction at each node at a certain instant. It is difficult to output the volume of a portion of the casting in ProCAST, and the sizes of the finite elements are almost equal. Therefore, the number
208
Yeh-Liang Hsu and Chia-Chieh Yu
Fig. 8 The portion of the wheel where shrinkage cavity usually happens
Fig. 9 SI with ten cyclic simulations
of nodes with solid fractions less than 0.7 at the instant when both risers become invalid are employed as the SI. As discussed in the previous section, in this research, the wheel casting process was simulated continuously for ten cycles in order to reach a steady state mould temperature. Figure 9 shows the SI of these ten cycles for wheel Type A. SI is high for the first several simulations, but soon converges to approximately 20 at the eighth to tenth simulation. This also shows that a reasonable simulation result can be obtained after a steady state mould temperature is reached. In aluminium wheel manufacturing, every casting wheel must pass a ‘leakage test’ to guarantee that the air will not leak through shrinkage cavities. The ‘leakage ratio’ is the ratio of the number of wheels with leakage and the total number of wheels tested. Table 2 shows the test data for three types of wheels (Types A, B, and C) from the local aluminium wheel manufacturer and their corresponding SI from our
Table 2 Type
Comparison of leakage test data and SI A
B
C-3
C-2 C-1
Total number of wheels tested 917 1135 1139 20 20 Number of wheels with leakage 86 203 316 8 12 Leakage ratio % 9.4 17.9 27.7 40 60 SI 20 73 111 122 132
simulation. Types C-1, C-2, and C-3 are almost identical wheels with slightly different geometries, which will be discussed in later sections. The leakage ratios of Type C-1 and C-2 were very high and were not approved for mass production. Therefore, there were only 20 test samples of Types C-1 and C-2 available for testing. Types A, B, and C-3 are mass-produced wheels. Figure 10 shows the relation between leakage ratio and SI. When SI is high, the leakage ratio will be high. If more data are accumulated, SI can be a good index for predicting the leakage ratio in the leakage test.
Computer simulation of casting process of aluminium wheels
Fig. 10
Table 3
SI versus leakage ratio Table 5
Effect of starting time of air cooling
Starting time 16–240 s 26–240 s 66–240 s 106–240 s 146–240 s SI 21 20 20 26 35 Liquid 163 s 164 s 164 s 164 s 164 s entrapped
Table 4 Duration SI Liquid entrapped
209
Effect of initial mould temperature
Initial mould 300 C 330 C 360 C 390 C 420 C 450 C temperature SI 20 21 20 30 39 41 Liquid 155 s 159 s 168 s 168 s 172 s 176 s entrapped
Effect of duration of water cooling 20 s X X
30 s 20 169 s
40 s 20 168 s
50 s 26 160 s
60 s 37 153 s
70 s 45 151 s
4 EFFECT OF COOLING PARAMETERS Using the cyclic simulation process and the SI described in the previous sections, this section discusses the influences of the timing of air cooling and water cooling on SI. Table 3 shows the SI and the time when liquid is entrapped with variation in starting time of air cooling from 16 s to 146 s. The time when liquid is entrapped is not influenced by air cooling because air cooling is applied on the side mould. Only the rim riser is affected. The central riser is not affected. As expected, SI increases if air cooling is applied late and thus solidification at the joint of rim and spokes is slower. However, if air cooling is applied earlier than at 66 s (the starting time of the current casting process), no decrease in SI is observed. If air cooling is applied at 16 s (immediately after filling is completed), SI increases slightly because the rim riser cools down too fast and soon becomes inactive.
Table 4 shows the SI and the time when liquid is entrapped with variation in duration of water cooling from 20 to 70 s (water cooling still starts at 66 s). Water cooling is applied to the bottom and has a significant effect on the central riser. Liquid entrapment occurred earlier when the duration of water cooling is long, and SI increases. Water cooling for less than 40 s (the water cooling time of the current casting process) will not further decrease SI. However, when water cooling is less than 20 s, the casting will not solidify at the end of the casting (240 s) because there is not enough cooling. The initial mould temperature is also considered in the casting process. Table 5 shows the SI and the time when liquid is entrapped with variation in the initial temperature of the mould in the range of 300 to 450 C. When the initial temperature of the mould is high, the solidification is slow and SI is high, although the time when liquid is entrapped is long. Decreasing the initial mould temperature to less than 360 C (the mould temperature of the current casting process) will not further reduce SI because liquid is entrapped early. From the analysis of results given in Tables 3 to 5, it follows that the timing for air cooling and water cooling, as well as the initial mould temperature in the current casting
210
Yeh-Liang Hsu and Chia-Chieh Yu
Thickening
Fig. 12 Thickened portion of the rim cavity
Fig. 11 The geometry and cross-section of wheel Type A and Type C
Table 8 Table 6
Cooling conditions of simulation Casting parameters
Alloy filling time Air cooling time Water cooling time Initial mould temp. Wheel solidification time
Table 7
Rim riser inactive Central riser inactive SI
24 s 24–210 s 74–134 s 360 C 210 s
Simulation results Type A
Type C
124 s 161 s 31
116 s 116 s 132
process of the local manufacturer, obtained from the engineers’ empirical adjustment, seems already to be close to the optimum. 5 EFFECT OF WHEEL GEOMETRY Figure 11 shows the different geometries of wheel Type A and Type C. Wheel Type A has five spokes and the cross-sectional area of a spoke is 1060.7 mm2. Wheel Type C has ten spokes and the cross-sectional area of a spoke is 405.3 mm 2. Table 6 shows the current cooling parameters for wheel Type C. Comparing with wheel Type A, the filling of wheel Type C takes longer because the crosssectional area of the spoke is smaller. Air cooling is applied right after filling is completed and the duration of water cooling is extended. For comparison purposes, both wheel Type A and wheel Type C were simulated using the cooling conditions given in Table 6. As shown in Table 7, the SI for wheel Type A is much better than that of wheel
Different thickening of rim
Wheel type
C-1
C-2
Thickening of rim SI Liquid entrapped
0 132 116 s
þ0.75
122 127 s
C-3 mm
þ1.5
mm
111 130 s
Type C, which indicates that the geometry of the wheel makes a great difference in the casting process. For wheel Type A, the rim riser becomes inactive much earlier than the central riser. For wheel Type C, both risers become inactive at about the same time because the cross-sectional areas of the spokes are small. In addition to adjusting cooling parameters, engineers also often modify the geometry of the casting cavity to improve the quality of the casting. For example, in wheel Type C, engineers decided to thicken the portion of the rim cavity (Fig. 12), so that a melted alloy would not solidify too fast in this portion and the rim riser can provide enough melted alloy into the joint of the rim and spokes. Table 8 shows the SI and the time when liquid is entrapped with variation in thickness of the rim of wheel Type C. When the thickness of the rim increases, liquid is entrapped late, and SI decreases. Type C-3 has the best SI; however, it is also the heaviest among the three. As shown in Table 2, the leakage ratio in the leakage test of Type C-1 is 60 per cent, and that of Type C-2 is 40 per cent, while the leakage ratio of Type C-3 drops to 27.7 per cent.
6
CONCLUSION
Numerical simulation is a powerful tool for many industrial applications. While efforts are made to produce accurate simulation results, it is often difficult to model accurately the boundary and loading conditions in many real industrial applications,
Computer simulation of casting process of aluminium wheels
such as the casting of aluminium wheels. In many applications, it is also difficult to validate accurately the simulation results with physical measurement data. However, numerical simulation still provides the correct ‘trend’, if not 100 per cent numerically accurate, of the performance and quality of the final product. On the other hand, to implement numerical simulation as part of the everyday manufacturing process, a standardized simulation process needs to be established, and simple indices, which correctly describe the trend of the performance and quality of the final product, should be obtained from the simulation results. In this paper, casting simulation software ProCAST is used to simulate the casting process of aluminium wheels of a local manufacturer. A cyclic simulation process is established to simulate properly the temperature distribution of the mould in real casting process. SI is defined to describe quantitatively the level of casting shrinkage from casting simulation. Matching with the leakage test results of the five different wheels, SI shows good correlation with the aluminium wheel leakage test results. The effects of cooling parameters and geometry of the mould cavity on SI are also discussed. Engineers’ empirical data concerning modification of cooling parameters can be verified. If enough leakage test data for a given aluminium wheel manufacturer are accumulated, the relation between SI and leakage ratio of final casting wheels can be established. Leakage ratio of a new wheel can be predicted using SI. For future research, the current authors are also investigating optimization of casting parameters to obtain the best casting quality. The objective can be to minimize SI (therefore the leakage ratio of final casting wheels). Casting simulation described in the present paper can be used as the function generator of SI.
211
REFERENCES 1 Tiwari, M. K. and Roy, D. Minimization of internal shrinkage in castings using synthesis of neural networks. Int. J. Smart Engng System Des. 2002, 4(3), 205–214. 2 Seetharamu, K. N., Paragasam, R., Quadir, G. A., Zainal, Z. A., Prasad, B. S., and Sundararajan, T. Finite element modeling of solidification phenomena. Sadhana Acad. Proc. Engng Sci., 2001, 26, (1–2), 103–120. 3 Bounds, S., Moran, G., Pericleous, K., Cross, M., and Croft, T. N. Computational model for defect prediction in shape castings based on the interaction of free surface flow, heat transfer, and solidification phenomena. Metall. Mater. Trans. B: Process Metall. Mater. Processing Sci., 31, (3), 515–527. 4 Midea, A., Nariman, R., Yancey, B., and Faivre, T. Using computer modeling to optimize casting processes. Modern Casting , 2000, 90 (5), 4. 5 Shenefelt, J. R., Luck, R., Berry, J. T., and Taylor, R. P. Solidification modeling and porosity control in aluminum alloy castings. Manufacturing science and engineering , 14–19 November 1999, Nashville, Tennessee, USA, pp. 507–511. 6 Spittle, J. A., Brown, S. G. R., and Wishart, H. Experimental and computational evaluation of the influence of permanent mould design on solidification of A17SiMg casting. Light Metals: Proceedings of Sessions, TMS Annual Meeting. Proceedings of the 1999 128th TMS Annual Meeting ‘Light Metals 1999’, 28 February–4 March 1999, San Diego, California, USA, pp. 951–958. 7 Drezet, J.-M, Rappaz, M., and Krahenbuhl, Y. Modelling of thermomechanical effects during direct chill casting of AA1201 aluminum alloy. Mater. Sci. Forum, 1996, 217, Part I, 305–310 . 8 Kreziak, G., Rigaut, C., and Santarini, M. Low pressure permanent mould process simulation of a thin wall aluminum casting. Mater. Sci. Engng A: Struct. Mater.: Properties, Microstruct. and Processing, 1993, A173 (1–2), 10. 9 ProCAST. User’s manual and technical reference. Based on ProCAST version 3.1.0, C1–C57.
