COMPARISONS BETWEEN PROTOTYPING PROCESSES .1 Materials That Can Se Formed with the Various

Processe.

The SLA process uses photocured polymers that do not exhibit great strength or toughness. Nevertheless, in the SFF family, the SLA process is the most accurate and has emerged as the industry standard for creating amaster pattern that might then be used as the basis for a casting or injection mold.

On the other hand, if asingle prototype needs to be tested to destruction, or car-ried around for a while, it really has to be made from metal or a structural plastic such as AB5. If only one or two prototypes are needed, the FDM process is an ideal choice. FDM can extrude ABS polymers and create prototypes that are between 50% and 80% of full ABS strength. For full strength plastic or metal prototypes, the standard machining process is the preferred choice, despite the more limited range of geometric shapes that can be made by machining. CNC machining is also the most likely prototyping process for thesmall batch manufacturing of 2 to 10 components.

If CNC machining is out of the question because of geometric complexity, SLS metal-powder parts might be the best choice. Beyond batch sizes of 10, it is worth considering the use of small batch casting methods. This decision will be influenced by desired accuracy, machining being better than casting. Some developments in shape deposition manufacturing and 3-D printing are leading to direct mold making (e.g., Weiss et al., 1990; Sachs et al., 2000).

4.3.2Accuracy

Accuracy is perhaps the next key feature that distinguishes the various prototyping processes. The list that follows gives some very general values for a variety of SFF processes and other more traditional processes that can be used to make one or two components.1

• Hot, open die forging:+1- 1,250 microns (0.05 inch)

• Laminated object modeling:+1-250 microns (0.010 inch)

• Investment (lost-wax) casting:+1-75 microns (0.003 inch)

• Selective laser sintering: +1- 75 to 125 microns (+1- 0.003 to 0.005 inch)-depends on part geometry

• Stereolithography: +1- 25 to 125 microns (+ 1- 0.001 to 0.005 inch)-depends on part geometry

• Plastic injection molding from a machined mold (prototyping version): +1- 50 lu 100 microns(+1-0.002 to 0.004 inch)

• Rough machining: +1- 50 microns (0.002 inch)

• Finish machining:+1-12.5 microns (0.0005 inch)

lThe rust entry corresponds to the age-old vtuege blacksmith's prototyping shop. See Wright and associates (1982) for the CNC controlled version

150 Solid Freeform Fabrication (SFF) and Rapid Prototyping Chap. 4

• Electrodischarge machining: +/-2.5 microns (0.0001 inch)

• Lapping and polishing:+1-0.25 microns (0.00001 inch)

When comparing the everyday prototyping methods, the most accurate remains machining, with easily achieved accuracies of +/- 25 microns (0.001 inch) and even half this with a good craftsperson. The next most accurate is prototypingbyplastic molding from a machined mold, with an accuracy of +/- 50 microns (0.002 inch).

After that the SLA and SLS processes are listed. For a typical component, selective laser sintering and stereolithography average out at+/-50 to 125 microns (0.002 to 0.005 inch). This is different fromthe accuracies of 25 microns (0.001 inch) quoted bythe suppliers of SLA equipment, and confrontational e-mails will prob-ably be a result of the obvious differences used in this text. However, these adver-tised accuracies of 25 microns (0.001) are for simple linear objects. Some users have to make complicated computer casings and medical monitors where open shell struc-tures "warp and shrink all over the place," to quote one user. In some cases this warping and shrinking worsens the SLA accuracy to as much as +/- 375 microns (0.015 inch).

For SFF, the other accuracy consideration is stair-stepping. Mentally picture the soccer ball again, but this time with perfectly smooth surfaces. Now approximate the soccer ball by representing it as a stack of thin slices. The largest diameter slice is at the equator; the smallest slice is at the poles. Figure 4.13 from Jacobs (1996) shows that the approximation to the soccer ball becomes worse as the bounding curve comes up around the object toward the poles. In addition, the loss in accuracy/fidelity is related to layer thickness. Since SLA processes are improving all the time, layers down to 25 to 50 microns (0.001 to 0.002 inch) are now possible, therefore giving better and better accuracies. As with all manufacturing processes, the process then does take longer and more cost is involved.

Investment (lost-wax) casting is listed at +/- 75 microns (0.003 inch). Thus if the casting process is used to make a short-run prototyping mold and then the part is injected in plastic, it would seem to offer about +/- 125 microns (0.005 inch), but some hand finishing and some cosmetic work on the mold will give as good a plastic part as the cast mold.

Large layer thickness Medium layer thickness Fine layer thickness F'iple 4.13 The stair-stepping approximation in SFF processes

CAD~esign

CAp

4.3 Comparisons between Prototyping Processes 151

4.3.3 Lead TIme of Prototypes

With an in-bouse dedicated FDM machine, a part can be produced within a 24-hour period. For an ongoing design activity where a design team needs a series of proto-types-for the look and fit of subcomponents and subassemblies-the FDM process is ideal.

An in-house integrated CAD/CAM system for machining can generate a simple part in a "morning's work," whereas more complex parts will take two to three days. An in-house stereolithography machine will also create the same parts in two or three days, measuring the time from receiving the" .STL" file to a fully cured product. The curing time, incidentally, is an added time factor, often overlooked when rival companies develop their advertising literature and compare their partic-ular process with others.

If an in-house machine is not available, it should be realized that the SLA service bureaus are swamped with business in today's economy. Unless a special cus-tomer relationship exists, turnaround time of one to three weeks is more probable.

Given the need for some negotiation with a client, and the need to check incoming computer files, the actual turnaround time may be longer still. Nevertheless, the rapid prototyping shops are selling "service and speed" rather than "fidelity."

For small batches (10 to 5(0) of injection molded plastic parts, customers can expect a three- to six-week turnaround time. The steps might be (a) an SDRCflDEAS or Pro-Engineer CAD file is received from the Internet, (b) files are checked, (c) an SLA master is made, (d) an aluminum mold is cast, and (e) the fin-ished batch of 10 to 100 is injection-molded in ABS plastic.

4.3.4 Batch Size

Chapter 2 describes the influence of batch size. For just one component, SFF processes-such as stereollthography.fused deposition modeling, and selective laser sintering-or machining is the obvious choice. Small-batch casting in metal, or small-batch injection molding in plastic, is used for small-batch runs between 50 and 500.

4.3.5 Cost

In general, cost increases with fidelity and accuracy needed, fur all the rapid proro-typing processes. Figure 4.14 shows why this is so. Specifically, if the designer desires more accuracy, the" .STL" files will need to be of finer resolution, the slicing will also be thinner, the laser will make more scanning paths, and the time and hence costs will increase. Also, all prototyping processes (SFF or machining) require some hand fin-ishing, sanding, and deburring. Obviously, costs increase if tbe designer prefers a smoother surface finish. In all prototyping processes there is also a relationship between complexity, surface finish, accuracy, and cost. For SFF, Figure 4.14 shows that overhanging features require explicit support especially for SLA. For the arrangement on the right of Figure 4.14, the support columns have to be broken off by hand after manufacture. This usually leaves small stubs on the surface, which must then be sanded away.

Sacrificial material

Solid Freetorrn Fabrication (SFFl and Rapid Prototvptna Chap. 4 152

~

O'~~;

•. """",,g~.

__ Forms

..•.- cavity feature

a. Complementary support

JIIpre 4.14Supporting structures for SLS and SLA (courtesy of Lee Weiss).

b.Explicit support

Rapid prototyplngmaclrine (RP~)

TABLE 4.4 Rapid Prototyping Machina Cost-Alao sea Section 4.3.6 for Installation and the Like (aa of March 2000)

MachinecOlIt

Sinterstation25()()pIWI (faster than the above) (SLS)

$210,000

The costs shown in Table 4.4 are the base cost of the machine. It should be empha-sized that there are also additional miscellaneous costs of a warranty, installations.

and so on. For example, the Helisys 2030H LOM machine has a base price (as of March 2000) of $275,500, which actually includes a first-year service warranty.

Installation is estimated at $3,000; training at $2,000. Additional options include a chamber heating module at $4,499 and an initial supply package at $5,995.Thus the total for the complete package is $292,494. This example is not meant to endorse or criticize the LOM machine; rather it shows the real cost of doing business. All the machines in the table have such setup costs, which add 10% to 20% onto the base price. Some processes such as SLS also require a supplementary room for powder preparation and venting. Further data on cost comparisons (Table 4.4), materials (Table 4.5), part size (Table 4.6), and total part cost (Table 4.7) now follow. Figure 4.15 compares accuracy.

4.3 Comparisons between Prototyping Processes

4.3.7 Commercial Comparisons of Cost and Capability

153

Rapidprototyping

TABLE4.5 Modeling Material Comparison

Liquid 3-D prtnnng followed by machining

x

TABLE4.6 Maximum PartSize Comparison (as of March2000)

Company

TABLE4.7 Rapid Prototyping Process, Speed andCostCompariscn-c-Chrvsler Benchmarking Test Reported In "Rapid Prototyping Report," Vol.1, No.6, June 1992.NoteThatThisComparison WasDone with 1992 Machines Such as the3D Modeler by Stratesvs. ManyMachines Suchas theSinterstatian 2500P1u• HaveBecome MuchFasterSinceThen.

Rapid prototyping process Machine Total part cost

Total process time

"Assumes 35 parts built simultaneously.

154 Solid Freeform Fabrication (SFF) and Rapid Prototvptns Chap. 4

ComparisoD of Approxlnule AeeurllCY of RapidPwlutypinKPIon:_

Sinterstation 2000(SLS) Sinterstation 2500(SLS) SOC4600(SGC)

0.005

0.005

0.006

13~'

~ l5. LOM-2030H (LOM)

~~

~ FDM 2000 (FOM)

0.006

0.01

0.005 0.003

SLA-350(SLA) 0.003

SLA-500(SLA) OJXl3

o 0.001 0.002 0.003 0.004 0.(lO5 0.006 0.007 0.008 OJXl9 0.01 Accuracy

(inches perinch)

Note thatthemachine suppliers quote in "thou' andthatone "thou' '" 25 microns Figure4.15 Comparison of accuracy (as of March2000).