You are on page 1of 17

Computers in Industry 48 (2002) 181197

Experimental study on the ice pattern fabrication for the investment casting by rapid freeze prototyping (RFP)
Qingbin Liu*, Guanghua Sui, M.C. Leu
Department of Mechanical and Aerospace Engineering and Engineering Mechanics, University of Missouri at Rolla, 1870 Miner Circle, 229 Mechanical Engineering Building, Rolla, MO 65409, USA Received 14 February 2001; accepted 6 March 2002

Abstract The process of investment casting has come to occupy a key position in the range of modern metal casting techniques. Wax pattern is widely used in the investment casting. However, there exist some problems and technical difculties in using it such as wax pattern expanding, ceramic shell cracking, etc. A new investment casting technology, named freeze cast process (FCP), has demonstrated the possibility and advantage of investment casting with ice patterns. But how to make ice patterns is really an important concern in the FCP process. RFP is a novel, environmentally conscious solid freeform fabrication process that can generate three-dimensional ice patterns by selectively depositing and rapidly freezing water layer by layer. With RFP it is possible to make ice patterns with better accuracy and more exibility directly from CAD models in a short time, without the cost and other issues of mold making. In this paper, the current status of the experimental study of the RFP process was introduced. The effect of different processing parameters on the layer thickness and layer width was studied quantitatively and thoroughly. These parameters include: frequency of pulse signal, stagnation pressure, scanning velocity, distance between nozzle and building surface, duty cycle of pulse signal, layer offset (slant angle), environmental temperature, scanning trajectory, and depositing material. And further the effect of the processing parameters on the dimension accuracy and surface roughness were analyzed qualitatively. The optimal parameters were chosen and used to fabricate complex ice patterns based on these experiments and analyses. Finally, an example was given to demonstrate that an ice pattern can be used to make the metal prototype. The fabricated metal prototype has a very good appearance, although surface roughness and dimensional accuracy have not been measured. The experiment results showed that RFP provides an enabling way to quickly make ice patterns for the investment casting with better quality and more exibility. # 2002 Elsevier Science B.V. All rights reserved.
Keywords: Solid freeform fabrication; Rapid prototyping; Rapid freeze prototyping; Ice patterns; Investment casting

1. Introduction The process of investment casting has come to occupy a key position in the range of modern metal casting techniques. Over the half-century dating from
Corresponding author. Tel.: 1-573-3416778; fax: 1-573-3416408. E-mail address: qbliu@umr.edu (Q. Liu).
*

1940, the investment casting process developed from a small and specialized sector of casting activity into a worldwide and distinctive industry, reecting the importance of the product in the intensifying search for close accuracy of shape and dimensions in materials forming [1]. The near-net-shape objective is seen, not only as a means of providing the engineer with a direct, efcient and economical route to the manufacturing of a nished component, but also as a contribution to the

0166-3615/02/$ see front matter # 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 6 - 3 6 1 5 ( 0 2 ) 0 0 0 4 2 - 8

182

Q. Liu et al. / Computers in Industry 48 (2002) 181197

conservation of costly materials and energy. As well known, wax is the most commonly employed material to make a pattern. Natural and synthetic waxes and various additives are blended to achieve minimum shrinkage and close reproducibility of pattern dimensions, together with strength for stability in handling and storage [2]. However, there still exist some problems and technical difculties in using traditional wax patterns such as wax pattern expanding, ceramic shell cracking, etc. In this case, many other materials have been studied as the pattern material substitution for the investment casting. Among them, ice pattern is a good candidate. A new investment casting technology, freeze cast process (FCP), has demonstrated the possibility and advantage of investment casting with ice patterns [3]. It starts with the building of solid master and silicone mold. Then ice pattern are made with the mold and dipped into refrigerated ethyl silicate slurry and stuccoed. After repeating the dipping and drying processes, a ceramic shell is made and then it is put in room temperature and allows the ice pattern to melt, drain, and dry. A research group in Tsinghua University, China, has also made some efforts on making metal parts with ice patterns [4]. FCP has several advantages over the competing casting processes, including low cost, high-quality, ne surface roughness, no shell cracking problem, easy process operation, and faster run cycle. Table 1 lists a comparison of several casting processes [3,5]. These strengths make it a signicant alternative for production of quality near-net shape castings at reasonable cost. One major concern in the FCP process is ice pattern fabrication, which includes compensation of multi-directional water expansion during freezing, air bubble removal, ice pattern de-molding, and part complexity limitations [6,7]. These issues can be solved by making ice patterns with the proposed Rapid Freeze Prototyping (RFP) process, a novel and environmentally conscious rapid prototyping (RP) technique [8].
Table 1 Performance comparison of some casting processes Casting process Linear dimensional tolerance (in./10 in.) Surface finish (RMS) Casting defects (scale 110, 10 being worst) Sand 0.100 300700 10

Before we proceeded to describe the RFP technology, the authors would briey introduce the rapid prototyping technology rst. The rst rapid prototyping techniquesstereolithography (SLA) was invented in 1986 [9]. After that, many other rapid prototyping processes have been developed [10]. The rapid prototyping processes are capable of converting three-dimensional CAD model directly into a solid physical model. Taking advantages of computer hardware and software technologies, the CAD data of a three-dimensional object is sliced into many two-dimensional layers. These twodimensional layers can be made and bonded to previous layers one by one sequentially. Different materials and ways of building and binding these layers constitute different RP processes. SLA, laminated object manufacturing (LOM), fused deposition modeling (FDM), selective laser sintering (SLS), and three-dimensional printing (3D-P) are some of the commercialized processes. According to the form of material preparation used in the building process, the existing RP processes can be classied into four categories: liquid, powder, sheet, and gas based. The classication was shown in Fig. 1 [11]. Here, the authors just gave a very brief introduction of the rapid prototyping technology. More detailed information about various rapid prototyping technologies can be found in [9,10]. RFP provides a good way to make ice patterns with better accuracy and more exibility. RFP is a novel, environmentally conscious solid freeform fabrication process that can generate three-dimensional ice patterns by selectively depositing and rapidly freezing water layer by layer. There are two ways to deposit water: continuous deposition and drop-on-demand deposition. In the continuous process, the water is continuously extruded from a nozzle to the substrate surface or the previously built layer; while in the drop-on-demand process, the water is ejected drop by drop only if needed. Fig. 2 illustrates the process of RFP [12,13].

Lost wax No data 125 5

Lost foam 0.077 125 6

Freeze cast 0.027 90 2

Q. Liu et al. / Computers in Industry 48 (2002) 181197

183

Fig. 1. Classication of the current rapid prototyping processes.

The build environment is kept at a low temperature which is well below water's freezing point. Pure water or colorized water is extruded or ejected from the nozzle and deposited onto the previously solidied ice surface. The newly deposited water is cooled by the low temperature environment through convention and by the ice surface of the previous layer through conduction. As a result, the deposited materials freezes rapidly and binds to the previous layer rmly through the hydrogen bond. After one layer is nished, the nozzle is elevated upwards a height of one layer thickness, waiting a predetermined period for the complete solidication of deposited water, and then depositing water again to build the next layer. This procedure continues until the

designed ice part is fabricated. The most important advantages of the RFP process include: cheaper and cleaner, the potential to build accurate ice parts with excellent surface roughness, sufcient layer binding, easiness and no residue for part removal in molding process, no shrinkage issue and easiness of material expansion compensation. Detailed description and explanation about these can be found in our other papers [1416]. With RFP, it is possible to make accurate ice patterns directly from CAD models in a short time, without the cost and other issues of mold making. Unlike other RP technologies, the RFP process has its own unique characteristics. Therefore, a fundamental study of the RFP process is necessary to give

Fig. 2. Principle of rapid freeze prototyping.

184

Q. Liu et al. / Computers in Industry 48 (2002) 181197

a deep insight into the process. Water ow during deposition process is essential for the successful fabrication of ice patterns and is mainly inuenced by both the heat transfer and the material depositing operations. The previous factor has been analyzed in detail before and can provide basic understanding by giving the freezing time and temperature distribution within the cross-section of the built part and with time after deposition [17,18]. The latter factor refers to the scanning velocity, droplet ejecting frequency, ejecting pressure, and so forth. Without the proper integration of these parameters, high-quality ice pattern even with fast heat transfer rate cannot be obtained. Unfortunately, the authors mainly focused on the demonstration of RFP, application area of RFP, heat transfer analysis before and failed to quantitatively and systematically analyze the effects of various processing parameters on the ice pattern fabrication [8,1218]. Therefore, the authors studied effect of the processing parameters on the layer thickness and width quantitatively, analyzed the effect of processing parameters on dimension accuracy and surface roughness qualitatively in this paper. On the basis of the above experimental results, the optimal processing parameters were found and used to fabricate ice patterns. Finally, the ice patterns were used to make the metal prototype. At this early stage of the research on the RFP process, the fabricated metal prototype has a good appearance although surface roughness and dimensional accuracy have not been measured. This is not a paper that thoroughly studied the RFP process, but that introduced current progress of the authors' research. On-going researches mainly focus on the systematic study of the dimensional accuracy and surface roughness. 2. Experimental setup The experimental setup used in this paper is shown in Fig. 3. A personal computer controls the whole system, including three stepping motors of the XYZ moving axes and water ow rate. And more than that, the PC also stores the software that can import STL le from other CAD packages and slice it into a CLI le which represents the contour information of each sliced layer. This software can further generate NC codes according to the contour information of the

Fig. 3. Rapid freeze prototyping experimental setup.

sliced layer and the specic process parameters of RFP such as building speed and material ejection rate. Besides the PC, the hardware of this experimental system consists of the other four major parts: (1) water extruding and ejecting subsystem, (2) three-dimensional positioning subsystem, (3) electronic control device, and (4) freezer. The aim of the freezer is to provide a low temperature environment, which is as low as 20 8C. Obviously, the three-axis moving mechanism and the substrate should be placed in the freezer to ensure the quick solidication of water. The heat generated by the three stepping motors, however, increases the temperature of the table to above 0 8C. Therefore, an aluminum plate is mounted on the XY table as the substrate and isolated from the table to keep its temperature around 20 8C. The nozzle in water extrusion subsystem is mounted on the Z elevator and can move only in Z-direction. Thus, the movement of the building substrate relative to the nozzle makes it possible to deposit material in the XY plane. The water extrusion subsystem consists of a constant pressure device, a water reservoir, a feeding pipe with temperature control capability, and a miniature solenoid valve based nozzle. Constant pressure device and water reservoir are not shown in Fig. 3. The

Q. Liu et al. / Computers in Industry 48 (2002) 181197

185

pressure device provides a constant preset pressure according to the extrusion rate requirement. The temperature of feeding pipe is controlled at a proper range, allowing the water in it to be pre-cooled but still in liquid state so that the water can ow freely. The miniature solenoid valve controls material ow on and off in continuous method, and generates water droplets in drop-on-demand method. In the latter case, the pulsed signal is produced either by a HP 3312A function generator or by the electronic control device described later. The output can be monitored by Tektronix 2236 100 MHz oscilloscope. The electronic control device of this system is used to amplify the pulsed signal and control the three stepping motors and accordingly the material ow rate, which is proportional to the resultant linear moving speed of the XY table. Motors are driven to move with a linear resolution of up to 0.25 mm. The motion control card (AT 6400, provided by Compumotor Company) can output a pulsed signal with a pulse width of 0.3 ms to the nozzle at a frequency proportional to the resultant moving speed of XY table. The speed proportion factor is treated as a process parameter before NC code generation and is congured by software. The output signal is sent to a circuit which widens the pulse to 0.2 ms, making it compatible with the response of the valve. 3. The effect of processing parameters on the layer dimension, surface roughness, and accuracy As mentioned earlier, the RFP process is a newly developed RP technology. It has its own unique characteristics which distinguish itself to the other RP technologies. Therefore, fundamental study on the effect of processing parameters on the dimensional accuracy and surface nish will help give readers deep insight into the unique process. After a droplet is ejected onto the building surface, it will deform, spread, and freeze. The shape changes ends with complete freezing. Deformation and spreading help make continuous ice lines, however, overspreading will cause failure of shape control. Water freezing and spreading on the building surface are two dominating factors, which result in different layer thickness and width. Furthermore, layer thickness and width inuence the resolution, accuracy, and

surface roughness based on our previous studies [12,13]. Therefore, it is very important to study the effects of various processing parameters on the layer dimension, accuracy, and surface roughness. These include: frequency of pulse signal, f, stagnation pressure, P, scanning velocity, v, distance between nozzle and building surface, d, duty cycle of pulse signal, m, layer offset (slant angle), w, environmental temperature, T, scanning trajectory, and depositing material. The lower the environment temperature, the faster the solidication process. So, the temperature of the freezer is always set to the lowest, 20 8C and was not considered in this paper. 3.1. The effect on the layer thickness and layer width Researchers often mention the effect of layer thickness on the surface nish for the various RP technologies. Very few literatures have been focused on the systematic study on the effect of the layer width. Therefore, the layer width, except for the layer thickness, was also studied in this paper so that we could use these values to compensate the dimension difference between the measured dimension and the nominal dimension caused by the layer width. For a brand new process, it is very useful to get the value of the layer width under certain processing parameters. To get the value of layer thickness and width, vertical or slant walls with the height of 140 layers were built. The layer width can be directly gotten through measuring the vertical wall thickness. While the layer thickness can be calculated by dividing the gross height of the vertical wall by the layer number. To guarantee the accuracy of the layer thickness and width, the values at different points were measured and the average was calculated. When studying a certain processing parameter's inuence, only this parameter changes while keeping other parameters constant. The distilled water was used to build the wall to study the effect of parameters on the layer thickness and width in this section. 3.1.1. The effect of scanning velocity Fig. 4 illustrates the effect of scanning velocity on the layer thickness and width. It can be seen that the layer thickness increases with a decrease of scanning velocity. However, the layer width is not affected signicantly by the scanning velocity. This is because the

186

Q. Liu et al. / Computers in Industry 48 (2002) 181197

Fig. 4. The effect of scanning velocity on layer thickness and layer width.

effect of spreading process is much signicant than that of freezing process. Less water will be deposited on the building surface within unit length when the scanning velocity is relative fast. Normally, this phenomenon should result in small layer width. But the water has sufcient time to spread along the transverse direction before freezing, causing the almost same layer width and smaller layer thickness because of the balance between the spreading process and freezing process. From Fig. 4, the detailed values of the layer thickness and layer width under different processing parameters can be read. These values can also be used to compensate the dimension difference caused by the layer width.

3.1.2. The effect of duty cycle of pulse signal The pulse width determines the droplet size and accordingly the water ow rate. A wider pulse width leads to greater water ow rate. The pulse width can also be represented by the duty cycle of pulse signal. Fig. 5 shows the effect of duty cycle of pulse signal on the layer thickness and width. It can be seen that the layer thickness and layer width both increase with an increase of duty cycle of pulse signal. But it should be noted that the layer thickness and layer width when the duty cycle of pulse signal being 0.4 do not conform to the trend seen from the other three points. That is because the spreading process along the transverse

Fig. 5. The effect of duty cycle of pulse signal on layer thickness and layer width.

Q. Liu et al. / Computers in Industry 48 (2002) 181197

187

Fig. 6. The effect of distance between nozzle and building surface on layer thickness and layer width.

direction (perpendicular to the scanning direction) is much severe when the water ow rate exceeds certain value. So, the layer thickness is just a little more than that when duty cycle of pulse signal is 0.32 while the layer width is much greater. 3.1.3. The effect of distance between the nozzle and building surface The distance between the nozzle and building surface is not a very big issue for the other SFF processes. But it should be considered in our study because a Lee Company's INKA series miniature solenoid valve is used to deposit the water. This kind of solenoid valve is mainly used to dispense liquid so that the water coming out from the nozzle actually forms a spraying cone just like that in the process of spray forming. The

further the distance between the nozzle and the building surface, the greater spraying area can be covered. Effect of distance between nozzle and building surface on layer thickness and layer width is shown in Fig. 6. It is obvious that layer thickness decreases with an increase of distance between the nozzle and building surface while layer width increases. 3.1.4. The effect of signal frequency Just like the duty cycle of the pulse signal, the frequency of the signal determines the droplet frequency, and thus, eventually inuence the water ow rate. Consequently, the effect of signal frequency on layer thickness and width is very similar to that of duty cycle of pulse signal. Fig. 7 shows the effect of signal frequency on the layer thickness and width. From

Fig. 7. The effect of signal frequency on the layer thickness and width.

188

Q. Liu et al. / Computers in Industry 48 (2002) 181197

Fig. 7, it can be seen that both the layer thickness and width increase with an increase of the signal frequency and the spreading process is much severe when the water ow rate is more than certain value. 3.1.5. The effect of stagnation pressure Stagnation pressure is another very important factor to affect the water ow rate. More water will be deposited on the building surface when the stagnation pressure is relatively higher. The stagnation pressure also determines the droplet speed and momentum: high stagnation pressure will cause the severe deformation, spreading and splashing. It should be noted that the angle of the spraying cone is relatively greater when the valve working at a higher stagnation pressure because a miniature solenoid valve is used to deposit the water as aforementioned. Fig. 8 shows the experimental results of the effect of stagnation pressure on the layer thickness and width. It can be seen that the layer thickness with the stagnation pressure of 40 kPa is smaller than that of 30 kPa while the layer width is not signicantly more than that of 30 kPa. This is due to the fact that the high stagnation pressure results in the high droplet speed and momentum and eventually causes severe deformation, spreading and splashing. At the same time, the greater angle of the spraying cone also has a signicant impact on the layer thickness and layer width. 3.1.6. The effect of layer offset Layer offset greatly determines the achievable slant angle without the support. As the building surface is

not a horizontal plane any more, the water droplet is more easy to ow downward, facilitating the spreading process. The wall height increases with a decrease of layer offset. When the layer offset is more than 0.045 mm, a wall will crash because of the lack of the support. Fig. 9 illustrates the measured results about the effect of layer offset on the layer thickness and width. It can be seen that the layer thickness decreases with an increase of the layer offset and the layer width increases with an increase of the layer offset. 3.2. The effect on dimension accuracy As discussed earlier, various combinations of processing parameters have a great impact on the layer thickness and layer width, and thus, inuence the longitudinal and transverse dimension accuracy eventually. Layer width mainly affects the transverse dimension while layer thickness affects longitudinal dimension. Following we will briey discuss the effect of depositing material, ratio of water ow rate and scanning velocity, scanning trajectory, and layer offset on the dimension accuracy on the basis of our initial experimental results. This section mainly focused on how these factors affected the dimensional accuracy. No dimensional compensation was yet considered at this time to eliminate the inuence of layer thickness and layer width on the dimensional accuracy. 3.2.1. The effect of depositing material To make the fabricated ice pattern colorful and visible, certain dye or additive is always mixed with

Fig. 8. The effect of stagnation pressure on the layer thickness and width.

Q. Liu et al. / Computers in Industry 48 (2002) 181197

189

Fig. 9. The effect of layer offset on the layer thickness and width.

the distilled water. This will somewhat inuence the physical properties of the depositing materials and different results can be achieved to some extent. Fig. 10 shows the pictures of the built ice patterns and associated measured real dimension. The one shown in Fig. 10(a) is fabricated using the mixture of distilled water and orange water color; while the one shown in Fig. 10(b) using the mixture of distilled water and purple ink. It can be seen that the real dimensions of the two parts is pretty close. It means that the effect of these two kinds of dye on the spreading process during deposition is almost at the same level. As comparison, the nominal dimensions are also shown in the picture (in the parentheses). 3.2.2. The effect of water ow rate Only the effect of the layer width on the dimension accuracy along the transverse direction will be introduced in this section. From the discussion earlier, it can be seen that scanning velocity, stagnation pressure, duty cycle of pulse signal, and signal frequency all have a positive inuence on the layer width. And at the same time, these parameters contribute to the water ow rate. So, we could draw the following conclusion: the larger the water ow rate, the wider the layer width is. A bigger layer width can result in a larger difference between the nominal and measured built dimensions. In this case, a bigger or smaller dimension could be achieved because of the inuence of the layer width. Therefore, compensation should be considered according to the various processing parameters. For case of the vertical wall, a layer width should be added

Fig. 10. Ice patterns and measured real dimension fabricated using: (a) orange water color dye; (b) purple ink dye.

190

Q. Liu et al. / Computers in Industry 48 (2002) 181197

Fig. 11. The effect of water ow rate on the dimension accuracy: (a) small water ow rate; (b) large water ow rate.

or subtracted to the nominal dimension to compensate the dimension variation. Fig. 11(a) and (b) shows the comparison between the nominal and measured dimension when the water ow rate is small and large, respectively. The numbers in the parentheses demonstrate the nominal dimension. When the ow rate is small, the contributed parameters are like these: scanning velocity is 60 mm/s, the distance between the nozzle and building surface is 3 mm, stagnation

pressure is 20 kPa, the duty cycle of pulse signal is 0.167, the signal frequency is 400 Hz. When the water ow rate is large, the distance between the nozzle and building surface is 25 mm and stagnation pressure is 50 kPa while the others are kept the same. There is only a variation of 0.65 mm between the nominal and measured dimension when the water ow rate is small. But this variation increases to about 2.25 mm when the water ow rate become really large.

Q. Liu et al. / Computers in Industry 48 (2002) 181197

191

Fig. 12. Measured dimensions of the two built ice patterns with different rotating angle.

3.2.3. The effect of layer offset As discussed earlier, the layer width increases with an increase of layer offset while layer thickness decreases. When an ice pattern is very complex and composed of different features such as the slant walls with different angle and curve surface, it will cause the different layer thickness and width for each layer as a result. In this case, the effect of the layer offset on the dimension accuracy must be considered and some measures should be taken to compensate it. For example, we could make two ice patterns which have the same transverse sections, i.e. the same area and prole. The rst one was built up just along the vertical

direction so all its side walls are vertical. As comparison, the second one was rotated a little angle horizontally every layer based on its big end while built up vertically. So, it has two kind of side walls: vertical wall on the big end and slant wall the rest. Fig. 12 shows the measured dimensions of the two built ice patterns. The height on the two ends of the rst ice pattern is very close. Actually, the small error is resulted from the uneven expansion of solidied ice which will be discussed later in this paper. It is obvious that there exists an error of 1.1 mm between the two ends of the second ice pattern because of the different layer thickness and layer width. Fig. 13 is another

Fig. 13. Another example showing how the layer offset inuence the transverse dimension of an ice part.

192

Q. Liu et al. / Computers in Industry 48 (2002) 181197

example illustrating how the layer offset inuences the transverse dimension of a jar-like ice pattern. It can be seen that the difference between the measured dimension and the nominal dimension at the shown three points varies a little because of the different slant angle (layer offset). It can be found that even with the same ice pattern, different layer thickness and layer width should be compensated according to the ice pattern's geometry. To make an accurate ice pattern, there are two ways, in general, to solve above problem: (1) compensation by adjusting the processing parameters, such as scanning velocity and water ow rate; (2) removal of extra materials using a horizontal cutter. 3.2.4. The effect of scanning trajectory As the limitation of the performance of XY table utilized in the experiment, the scanning trajectory greatly inuence the dimension accuracy by affecting the evenness of the scanning velocity. The fabrication of a square ice pattern shell was chosen here to illustrate how different scanning trajectories inuence the dimension accuracy. Three trajectories could be used to build the square shell: (1) trajectory 1, starting from the middle point of a side wall, directly build the shell along its prole of the shell; (2) trajectory 2, starting from the center point of the square shell, scan along a straight line (perpendicular to a side wall) till the middle point of the associated side wall, and then build the prole of the shell; (3) trajectory 3, starting from the center point of square shell too, but scan along a semi-circle rather than a straight line till the tangent point between the semi-circle and the corresponding side wall, and then start printing. Trajec-

tories 1 and 2 will result in severe unevenness of the scanning velocity because of the acceleration and deceleration of the XY table. The right square part in Fig. 14 was built through the trajectory 1. It can be seen that the height of the shell is uneven and the difference between the highest point and lowest point can reach 4.5 mm. The wall thickness adjacent to the starting and ending point is also much bigger. On the contrary, the scanning velocity was kept the same adjacent to the tangent point because the semi-circle and straight line are tangent with each other when printing using trajectory 3. So, the even layer thickness and layer width can be achieved. This can be seen from the left picture in Fig. 14. 3.3. The effect on surface roughness It is well known that the layer thickness has a signicant impact on the surface nish of the RP parts. The smaller the layer thickness, the better the surface nish. But there are also other dominant factors affecting the surface nish in the unique RFP process. So, the other factors rather than the layer thickness were studied here. At this stage, the quantitative analysis of the surface roughness of the fabricated ice patterns was not performed yet because of the difculty to measure the surface roughness of ice pattern. The surface roughness is determined intuitively by the authors based on the initial experimental results. Next step, we are going to use the UV silicone to duplicate the surface of ice pattern and analyze the inuence of the processing parameters on the surface roughness quantitatively.

Fig. 14. Experimental results of the shell of a square part using different scanning trajectories.

Q. Liu et al. / Computers in Industry 48 (2002) 181197

193

Fig. 15. (a and b) The inuence of waiting time on the surface roughness of built ice patterns.

3.3.1. The effect of waiting time between two successive layers For successive part building, in order to prevent heat accumulation with layer building, there should be enough waiting time before starting to build the next layer. The previous FEM analysis shows that the time to cool down the previous layer to almost the environment temperature could be tens of seconds depending on many heat consumption factors. If the waiting time is not enough, the temperature of the built wall will rise gradually. The consequence of this is newly deposited layer cannot freeze very fast and result in uneven line width [17]. Fig. 15(a) shows the consequence of insufcient waiting time between the successive layers. The newly deposited material is kept in liquid state for longer time, and thus, the surface tension causes the material unevenly distributed, forming some sweats on the building surface. Fig. 15(b) shows a good-looking ice part. In this case, even the material ow rate is larger, the enough waiting time prevent the uneven spreading of colorized water. The longer the waiting time, the better the surface roughness. But a longer waiting time will affect the fabrication efciency. So, these two factors must be balanced. 3.3.2. The effect of depositing material The depositing material is the basis of the RFP and its performance will eventually inuence the surface quality of the built ice patterns. Different additives result in different surface tension, nozzle dragging

effect, uidity, and even melting point. Therefore, even using the same processing parameters, big difference could be produced under the certain circumstance. For comparison, two different materials were used to build a square shell separately using the same processing parameters. They are: distilled water and distilled water mixed with purple ink. The experimental results show that the one built with mixture of the purple ink with distilled water has a good surface smoothness compared with the one built with distilled water only. 3.3.3. The effect of other factors Besides the waiting time and depositing materials, there are also other factors which affect the surface roughness of the ice part very much. After coming out from the nozzle, the droplets will y a predetermined distance to the building surface. During this process, the droplets are under the inuence of gravity, aerodynamic drag, air resistance, and vibration of the XY table. Under certain circumstance, such as low stagnation pressure and long ight distance, these factors will dominate and produce a great impact on the surface. Fig. 16 shows the surface of a cylinder ice part. The ripple on the surface can be clearly seen. This is due to the fact that the inuence of aerodynamic drag, etc. is more powerful when the initial velocity and momentum of the droplet is little. Other experimental results show the effect of distance between the nozzle and building surface on the surface roughness. The longer the ight distance, the worse

194

Q. Liu et al. / Computers in Industry 48 (2002) 181197

Fig. 16. Very low stagnation pressure causing bad surface roughness.

Table 2 Processing parameters used to fabricate the Ice patterns Scanning velocity (mm/s) 60
a

Signal frequency (Hz) 400

Stagnation pressure (kPa) 30

Duty cycle of pulse signal 0.167

Distancea (mm) 10

Environment temperature (8C) 20

Nozzle temperature (8C) 5

Distance between the nozzle and building surface.

Fig. 17. Fabricated ice patterns.

Q. Liu et al. / Computers in Industry 48 (2002) 181197

195

the surface roughness. This is because the droplet will suffer more when ying a little long distance. The pictures describing the associated experimental results were not shown in this paper. 4. Ice pattern and metal prototype fabrication In other SFF processes, the solid objects are generally fabricated by scanning the boundary and interior within one layer at the same time. This is also the one method can be explored in RFP process to fabricate the solid objects. This can be called scanning mode. But the most unique fabrication characteristics of the RFP process is the lling mode. In this mode, the whole boundary or just N layers of the boundary of an ice pattern can be generated rst and then build the interior by lling the internal area with distilled or colorized water. Therefore, the building period of RFP can be signicantly faster than other SFF processes. All the ice patterns shown in this paper were fabricated using this mode. Based on the experiments and analysis mentioned earlier, the optimal parameters were chosen and listed in Table 2. A lot of ice patterns have been fabricated using these parameters. Fig. 17 shows part of the fabricated ice patterns. It can be seen from this picture that very complex ice patterns are achievable without support material. Following, an example were given to demonstrate how to generate an ice pattern using the lling mode. To generate an ice pattern like that shown in Fig. 18(c) using lling mode, three steps should be followed. Firstly, only the boundary (shell) of the ice pattern was built as shown in Fig. 18(a). Then the interior was built by lling the internal area with colorized water. To prevent the melting of the ice boundary, the lling process should be divided into several steps instead of lling the whole internal area once. This step continued till the internal area was full and the colorized water solidied completely. As well known, the water expands when it becomes ice. So, the top surface of the ice pattern was rugged as shown in Fig. 18(b). Undoubtedly, the nal step is to smooth the top surface of the ice pattern and a nice looking ice pattern was generated. To demonstrate the feasibility of applying the fabricated ice patterns to the investment casting, the ice pattern shown in Fig. 18 was utilized to produce a metal prototype. As limited by the experimental

Fig. 18. Fabrication of an ice part using lling mode: (a) building boundary; (b) lling and freezing; (c) nal ice pattern.

devices, the UV silicone was used to substitute the ceramic shell and the low-melting-point 63Sn37Pb solder alloy was used to ll the mold instead of the steel, aluminum alloy, etc. To avoid the melting of the

196

Q. Liu et al. / Computers in Industry 48 (2002) 181197

ice pattern, all the following procedures should be performed under the sub zero environment. Firstly, the UV silicone was kept in a freezer to let it cool down to the environment temperature. Secondly, the UV silicone was poured around the ice pattern which was already put inside a plastic cup in advance. Thirdly, the plastic cup with the UV silicone and the ice pattern inside was put in the freezer for a long time to let the entrapped air bubbles come up to the surface of the UV silicone. After that, a UV light was used to cure the UV silicone to form the mold. Then the cured UV silicone and the ice pattern were put to the room temperature to melt the ice pattern. Consequently, a mold was provided as shown in Fig. 19. The pouring cup and spruce which also acts as the riser in this case were added to the cured UV silicone mold. Finally, the molten 63Sn 37Pb solder alloy was poured to ll the cavity of the mold and a metal prototype could be produced as shown in Fig. 20. From Fig. 20, it can be seen that the surface prole of the fabricated metal component at its as-process status was fairly good and acceptable. It should be noted that there are some pits in certain area of the surface of the metal component. These pits can be reduced or eliminated by taking certain measures, such as pouring the molten metal in a vacuum chamber or further raising the height of the riser. The detailed value of the surface roughness has not been measured yet because a nontouch optical tester should be employed here to measure the surface roughness. From Fig. 20, we are condent that the RFP technology is an enabling technology for the investment casting.

Fig. 20. A 63Sn37Pb solder alloy component produced from cured UV silicone mold.

5. Conclusion The deposited ice layer is the basic element of a fabricated ice pattern. Its thickness and width are critical to the fabrication of high-quality ice patterns. Therefore, the effect of different processing parameters on the layer thickness and layer width was studied quantitatively and thoroughly in this paper. These parameters include: frequency of pulse signal, stagnation pressure, scanning velocity, distance between nozzle and building surface, duty cycle of pulse signal, layer offset (slant angle), environmental temperature, scanning trajectory, and depositing material. And further the effect of the processing parameters on the dimension accuracy and surface roughness were analyzed. These experiment results could be used to compensate the dimension error quantitatively and direct us how to successfully build an ice pattern qualitatively. The optimal parameters were chosen and used to fabricate many complex ice patterns based on these experiments and analyses. Finally, the mold from an ice pattern was made by curing the UV silicone and a metal component was consequently fabricated. The fabricated metal component has a very good surface nish. It can be seen that RFP provides a good way to quickly make ice patterns for the investment casting with better quality and more exibility. Future work will focus on improving building resolution and ice pattern accuracy, and on investment casting investigation and measurement of duplicated ice pattern surface. The nal goal of this research is to try to commercialize the RFP technology.

Fig. 19. A cured UV silicone mold from the ice pattern.

Q. Liu et al. / Computers in Industry 48 (2002) 181197

197

Acknowledgements The authors gratefully acknowledge the nancial support from the Department of Mechanical and Aerospace Engineering and Engineering Mechanics of University of Missouri at Rolla. References
[1] Investment Casting Institute, Investment Casting Handbook, Investment Casting Institute, Chicago, USA, 1968. [2] P.R. Beeley, R.F. Smart, Investment Casting, 1st Edition, The University Press, Cambridge, UK, 1995. [3] A. Yodice, Freeze cast process, US Patent 5,072,770 (1991). [4] X. Wu, Study on ceramic mold investment casting based on ice patterns made by rapid prototyping method, Bachelor degree thesis, Tsinghua University, Beijing, China, 1997. [5] A. Yodice, Freeze process cuts casting costs, Advanced Materials and Processes 155 (4) (1999) 3536. [6] A. Yodice, Freeze cast process ready for licensing, incast, International Magazine of the Investment Casting Institute XI (12) (1998) 1921. [7] D.M. Peters, Patterns in ice, Foundry Management & Technology 123 (8) (1995) 9096. [8] W. Zhang, M.C. Leu, C. Feng, R. Ren, R. Zhang, et al., Investment casting with ice patterns made by rapid freeze prototyping, in: Proceedings of the 11th Annual Solid Freeform Fabrication Symposium, University of Texas at Austin, Austin, TX, USA, August 2000, pp. 6672. [9] P.F. Jacobs, Rapid Prototyping and Manufacturing: Fundamentals of Stereolithography, SME, Dearborn, MI, 1992. [10] P.F. Jacobs, Stereolithography and Other RP&M Technologies: From Rapid Prototyping to Rapid Tooling, SME, Dearborn, MI, 1996. [11] M.C. Leu, W. Zhang, Research and development in rapid prototyping and tooling in the United States, in: Proceedings of the International Conference on Rapid Prototyping and Manufacturing, Tsinghua University, Beijing, China, 2123 July 1998. [12] G. Sui, W. Zhang, M.C. Leu, Study on water deposit in rapid freeze prototyping, in: Proceedings of the 11th Annual Solid Freeform Fabrication Symposium, University of Texas at Austin, Austin, TX, USA, August 2000, pp. 342349. [13] W. Zhang, G. Sui, M.C. Leu, Determination and improvement of building speed in rapid freeze prototyping, in: Proceedings of the 11th Annual Solid Freeform Fabrication Symposium, University of Texas at Austin, Austin, TX, USA, August 2000. [14] W. Zhang, M.C. Leu, Z. Ji, Y. Yan, Rapid freezing prototyping with water, Materials and Design 20 (1999) 139145.

[15] W. Zhang, M.C. Leu, Z. Ji, Y. Yan, Rapid freezing prototyping with water, in: Proceedings of the 9th Annual Solid Freeform Fabrication Symposium, University of Texas at Austin, Austin, TX, August 1998, pp. 185192. [16] W. Zhang, Y. Yan, G. Sui, M.C. Leu, Rapid freeze prototyping and its application in rapid tooling, in: Proceedings of the Presentation of the 1st International Conference on Mold and Dies, Beijing, 2000. [17] M.C. Leu, W. Zhang, G. Sui, An experimental and analytical study of ice part fabrication with rapid freeze prototyping, CIRP Annals 49 (1) (2000) 147150. [18] W. Zhang, M.C. Leu, G. Sui, Z. Ji, An experimental and analytical study of ice part fabrication with rapid freeze prototyping, in: Proceedings of the 10th Annual Solid Freeform Fabrication Symposium, University of Texas at Austin, Austin, TX, August 1999, pp. 591598. Qingbin Liu earned his BS and MS degree in 1991 and 1994, respectively from Shandong University of Technology, China. In 1996, he graduated from Northwestern Polytechnical University (NPU) in China and earned his PhD degree there majoring in materials science and engineering. After that, he went to South China University of Technology to do post-doc research on the laser materials processing. Since 1998, he did post-doc research on the rapid prototyping and rapid tooling in University of California at Irvine and University of Missouri at Rolla one after another. Right now, he is working on the dental application of the rapid prototyping technologies. Till now, he has published more than 50 papers in academic journals and conferences. Guanghua Sui earned his PhD degree in 2002 at Department of Mechanical and Aerospace Engineering and Engineering Mechanics of University of MissouriRolla. Research area includes Rapid Freeze Prototyping (RFP), Rapid Prototyping (RP).

M.C. Leu Keith and Pat Bailey Professor in Department of Mechanical and Aerospace Engineering and Engineering Mechanics of University of MissouriRolla. Main research interests include: CAD/CAM, Virtual and Augmented Reality, Rapid Prototyping & Manufacturing, Mechatronics/Robotics.

You might also like