By Leslie Langnau, Managing Editor
Quality is composed of several parts; the accuracy of dimensions or measurements, and the more subjective “is it what I want?” In a recent study, Todd Grimm examined several factors associated with quality for six 3D printers to address this question.
In the last issue of DMR, we covered the factors involved in calculating the true cost of six 3D printers. This article looks at quality and what Todd Grimm’s report uncovered for material properties, surface finish, and dimensional accuracy.
Material properties affect so many aspects of parts made from 3D printers. At a minimum, parts must withstand the wear and tear of routine handling as they are passed from one designer to the next. They must be durable, which depends on a number of mechanical and thermal properties, and in some cases material exposure to UV light or moisture.
In Grimm’s report, each part from the 3D printers was reviewed in the context of a form or fit prototype. Routine handling was one test. Another involved slightly oversized screws that were driven into two freestanding bosses that have 0.100 in. walls. This test simply measured the number of full turns before the bosses failed.
Here is a summary of the observations, listing the 3D printers from best to worst in terms of durability.
The uPrint builds parts out of ABS plastic. Thus, these parts were the strongest among the group. There was no damage during routine handling, and when intentionally attempting to break the parts, nearly all features stood up to the pressure. The only damage inflicted was on a small post (0.110 in. dia.), a gusset (0.020 in. thick) and an access cover (0.070 in. thick). To do this damage took quite a bit of force. For example, the access cover did not break. Instead it deflected nearly 45° before the sidewall split along a layer.
The uPrint also did well with the screw test. Both bosses held up while the screws were driven to the bottoms.
All parts survived routine handling without any damage, and thick-walled features would not break under a good deal of pressure.
The thin-walled features, such as the 0.010 in. and 0.020 in. ribs on the test block and 0.030 in. half-lap joint on the security panel, did break under moderate pressure. Yet, the sidewalls (0.070 in. thick and 0.920 in. high) of the security panel-back would not break without an excessive amount of force. The access cover on the security panel-front broke at its base with less force that that for the uPrint part, but the amount of pressure needed was still impressive.
One concern arose when the security panel-back was lightly twisted. A large diagonal crack appeared across the bottom face. This “brittleness” is contrary to what was observed with other features.
Alaris30 performed well with the screw test. One screw was driven to the depth of the boss without breaking; the other was driven half way before the wall failed.
The next three systems, while ranked in order, were very close in durability performance.
3—ZPrinter 310 Plus
The material this system uses is based on powders, binders, and adhesive infiltrants. Parts were reasonably durable even though the material has been characterized as weak and fragile. Before infiltration, however, parts are much more fragile than any others in the benchmark.
There was only one small break during routine handling. Somewhere between inspection and return of the parts, a small piece of the thin half-lap on the security panel back broke off. Another unexpected result was that features with thicknesses greater than 0.080 in., such as the corner posts on the security panel, did not fail.
Lack of durability was apparent for thin-walled features. All broke with low to moderate pressure. In some cases, the break had a brittle feel. In others, the features had a soft, yielding feel when they failed. For example, when the screws were driven into the housing, they slowly gave way (instead of sharply cracking) after four to six turns. (Note: The Grimm report was published during the fourth quarter of 2010. ZCorporation no longer sells the ZPrinter 310.)
4—ProJet SD 3000
Parts made from this printer showed limited durability during routine handling. The first break, which was on a thin, 0.010 in. rib, happened when the test block was repositioned during photography. Light thumb pressure in the wrong spot broke the rib. In inspection and transit, several other features were damaged. This included a large section of the half-lap (0.030 in.) on the security panel breaking off.
A slight twist of the security panel back produced a diagonal crack on the bottom face.
All small or thin-walled features failed with low to moderate force. Yet, this system did perform better on the screw test than did the ZPrinter 310. In both cases, it took more than six turns to break out the sidewalls of the bosses.
Parts from this printer did quite well in routine handling—there was only one small chip on a thin wall of the test block. But the sidewalls of the security panel were easy to break.
All small or thin-walled features failed with low to moderate force. Yet, the half-lap of the security panel was slightly stronger than that on the ProJet SD 3000.
V-Flash did poorly on the screw test. One of the bosses broke out before completing a single turn of the screw.
With PVC as its modeling material, though tough, it cannot hide the weakness of the adhesive bond between layers. All features with wall thicknesses less than 0.080 in. were susceptible to breaking with low to moderate pressure. However, thicker features were quite strong and durable.
When light force was applied to the access cover and large corner bosses, the features broke cleanly and easily along the plane of a layer. For the sidewalls of the security panel, the layers would break free of one another rather than breaking outright.
If built such that all tensile and flexural loads would be perpendicular to the layers, the strength could be impressive, but this is not a practical expectation.
While the SD300 Pro did better that the V-Flash when the screws were driven into the bosses, the weakness of the layer bonds once again became apparent. Rather than splitting out the sidewalls, the screws lifted and separated layers of the bosses in multiple locations.
The surface finish of parts produced with 3D printers cannot be accurately depicted with measurement tools such as surface profilometers. There is too much variance and inconsistency to describe these prototypes with a simple Ra value. The surface finish is different for flat planes, vertical walls and spherical shapes. Additionally, on many surfaces, the finish is often non-uniform. For these reasons, the surface finish is characterized through part observations. Also, since the results vary so greatly from feature to feature, the systems were not ranked.
Alaris30: Overall, this system produces smooth surfaces with little stair stepping. There is evidence of “streaking” from the print head, and vertical surfaces are smooth, but they have some “chatter” where a surface juts in and out.
ProJet SD 3000: Surface finish is very good, but some texture and roughness are seen. Small radii, angled surfaces and spherical shapes are rather smooth with little evidence of stair stepping.
SD300 Pro: Flat, horizontal surfaces are perfectly smooth and glossy. However, sidewalls have varying degrees of chatter. Stair stepping is moderate.
uPrint: The top and bottom faces of all parts have a distinct pattern. The vertical surfaces have obvious layer lines, and angled surfaces are clearly stair-stepped. While the surfaces are not smooth, they are extremely uniform.
V-Flash: While there are some textures and layer lines on sidewalls, the surfaces are quite acceptable and reasonably smooth. On the other hand, the supported surfaces, which were lightly sanded to remove the support structure, are marked with a series of small pocks and raised bumps.
ZPrinter 310 Plus: All surfaces have similar finishes. All features have a slightly “fuzzy” feel and appearance that can be likened to 220-grit sandpaper. This texture has the advantage of hiding stair stepping on angled and spherical surfaces.
As with surface finish, there is simply too much variance in 3D printed parts to rely on a few point-to-point measurements to characterize these prototypes. White light 3D scanning technology was used to describe up to 2.8 million points for each benchmark part.
The following information cannot be matched to a traditional ± X in. statement. Instead of the tolerance between two points, such as the center-to-center position of two holes, the 3D scanning data present the 3D distance of a point on a test part to the corresponding point on the CAD data. So, what is shown is a profile tolerance. While Grimm’s report goes through each of the test parts, for brevity, we will focus on the test block. Information on the accuracy of the other parts is available in Grimm’s full report.
The graphic shown above presents a plot of the standard deviation of all measurement points from their nominal locations on the CAD file. In this chart ± 2 sigma (two standard deviations) is plotted for each technology. For this data, a smaller bar (value) centered on 0.00 shows a higher accuracy. For example, Alaris30 has the best dimensional accuracy with a 2σ of 0.0054 in. centered on -0.0004 in. (the mean). V-Flash has the poorest accuracy with 2σ of 0.0176 in centered on -0.0027 in.
The data are shown in table, which also lists the percentage of all points that fall within ±2σ. Two other items presented in the table are the percentage of points that exceed ± 0.020 in. and ± 0.005 in.
To visually interpret the dimensional quality of the test block, the error map on the previous page presents error data for five of the benchmark systems. These error maps color code the deviation of each measurement point using the legend to the right. In this error map, and all that follow, green shows areas between +0.005 in. and -0.005 in. Yellows and oranges show points that are higher than the design intent (above the surface) while cyan and blue show those that are lower (below the surface). To highlight regions that are grossly inaccurate, bright red shows areas that are higher than 0.020 in. and deep purples show those lower than -0.020 in.
Through color, a clearer picture emerges. For example, the ProJet SD 3000 shows greater variance than one would expect after reviewing the chart. This is true because a high percentage (32.4%) of the points exceed ±0.005 in. while 99.8% are within ±0.020 in. Conversely, the uPrint looks better with only a slight improvement of the ± 2σ. This is because a low percentage (10.02%) exceeds ±0.005 in., but 1.18% exceed ±0.020 in.
While the uPrint part is mostly green, light cyan and light yellow (good accuracy), the red faces on the ribs mark a problem area. These ribs are 0.010 in. and 0.020 in. thick, which is below the minimum that uPrint can replicate. So, the system creates walls that are too thick, which throws off the 2σ value.
The inaccuracy of V-Flash is depicted well in this figure on the right-hand wall. On one surface, the part goes from very low to moderately high. Conversely, the accuracy of the Alaris30 is clear with the abundance of green.
As revealed by the standard deviation plots, inspection tables and error maps, dimensional accuracy is dependent on the size and configuration of a part. The Alaris30 has the best accuracy overall with an average of 11.9% of the measurements exceeding 0.005 in. It was also the best for three of the four parts. The second best performer is the uPrint (27.4%). This is in spite of the significant number of features below 0.030 in. that the machine made oversized.
Although this chart shows the SD300 Pro to be in the third position, this is an inaccurate reflection of its accuracy. For this system, the data do not include the missing features that it could not build. It also excludes the test block, which was not made because much of the geometry was too small for the process.
With an overall value of 35.8%, the ProJet placed fourth in accuracy. However, this value would likely decrease if the housing were re-built such that its accuracy is more consistent with the other three parts. The fifth place system, ZPrinter 310 confirms that statements of quality are dependent on what and how accuracy is measured. In terms of ± 2σ, this system was consistently in the middle of the pack. But when judged based on the number of points exceeding ± 0.005 in., it is second to last with an aggregate value of 39.3%. The last place system, V-Flash had the highest percentage by all measures but one.
T. A. Grimm & Associates, Inc.