SLS empowers customers to produce high-quality end-use parts without the overhead of manual labor and post-processing costs that many other technologies require. To achieve this, printers need to reliably produce dimensionally accurate and repeatable parts in every build. Repeatability and consistency ensure that there are as few rejected pieces as possible, and that the parts fit into any assembly without needing any rework.

There are three main areas where dimensional variation can occur:

1. Part-to-part variation within a build

2. Build-to-build variation

3. Printer-to-printer variation

This paper aims to study the first two sources of variation. This is done by having six-dimensional witness blocks spread across the top, bottom, and middle of a large build, and printing them five different times on the same printer. The blocks contain several features with dimensions from 25 mm to 138 mm, in all three major axes of the printer (X, Y, and Z). These measurements are compared and then analyzed via a statistical approach using the Process Capability Index (Cpk).

The results show an extremely good repeatability in all axes between the builds, with X and Y remaining completely stable for all prints and print locations. The Z axis, while repeatable, does show a higher degree of variation from the nominal. Below are charts of the actual measured data for the 100 mm dimension on each build.

This data shows that prints from Fuse are extremely precise, maintaining +/- 0.2 mm of the average in both the X and Y axes all the way from 25 mm to 138 mm, with features under 100 mm being repeatable within +/- 0.1 mm. In both cases, the values were slightly higher than the nominal values, so the tolerance from nominal is +/- 0.5 mm. This means that for an International Tolerance (IT, defined by ISO 286) grade of 13, the Cpk is 2.3 in X and 1.5 in Y, both of which are extremely good — a more in-depth discussion of what exactly these values mean is available in the Background section.

The Z-axis data shows that while the parts are highly repeatable between prints, there is a slight differential between the parts printed at the top and bottom of the build chamber. This is largely due to differences in the thermal conditions under which they are printed. In total, parts printed vertically have an accuracy of +/- 0.7 mm. The amount of sintered material beneath a part significantly influences its thermal behavior, which directly affects cooling rate and dimensional accuracy. Builds with varying part placements are the most difficult to control due to differences in Z-axis thermal history. However, as this whitepaper will demonstrate, even a single build can be optimized for accuracy when the part placement is locked.

However, if the top and bottom prints are taken as separate pieces, they are much more repeatable and could be easily rescaled to fit the same degree of accuracy as X and Y. For example, the measurements along the top, which can be seen below, only vary by +/- 0.4% from the average throughout all 10 measurements.

Overall, the fuse printer is able to provide highly repeatable parts that are dimensionally stable, with a Cpk of > 1.33 at IT13 in X and Y, and IT16 in Z. The basic results are summarized below.

In additive manufacturing, there are several ways to evaluate the capability of the process in relation to accuracy. As the volume of parts increases, the number of ways you can analyze this data increases, making an even more nuanced calculation that requires context and strict parameters. 

Options for evaluating variation in 3D printed serial production parts: 

1. Define the process as producing a single model in multiple locations in the same build. This is a ‘part-to-part’ basis.

2. Define the process as producing a single model in a single location within multiple builds of the same printer. This is called a ‘print-to-print’ basis. 

3. Define the process as producing a single model in the same location of a build across multiple printers. This is called a ‘printer-to-printer’ basis.

Measuring accuracy with SLS 3D printers can be especially complex; with a well-packed build chamber of production parts, there may be hundreds or even thousands of parts, each with its own potential variation. This is made more difficult by the physics of parts cooling, which can vary over the height of a build, so the extremes of the build may have different results.

However, measuring accuracy is made slightly easier by designing parts that are built especially for being measured, often called ‘witness’ or test coupons. These parts have multiple different dimensions on each given part, varying from long-distance to short-distance geometries. They are also made to be easily measured using hand tools, rather than requiring more complex measurements like 3D scanning.

In SLS 3D printing, there are two main concepts to consider when thinking about accuracy. 

The first, called precision, refers to how tightly and consistently a manufacturing process can produce parts, i.e., how well Part 1 resembles Part 2, matches Part 3. This means that for multiple parts produced, all measured dimensions are within a tight tolerance band, even if that tolerance band is not near the target dimension. In process engineering, the Process Capability (Cp) is a measure of the potential, or inherent capability, of a process to produce repeatable results, and is used to evaluate the precision of the process.

Accuracy is defined by how close a part’s measurements are to the intended dimension. A process that produces an average result that is off the target value is referred to as not being “mean-centered.” The Process Capability Index (Cpk) takes into account both the process variation, as determined by Cp (precision), as well as the error between the mean and the target value. 

Cpk can be thought of as the frequency with which the part’s dimensions can stay within the target bounds. One of the ways to set these target bounds is via International Tolerancing (IT) grade tolerances. IT grades are ISO standardized tolerance classes used in manufacturing to ensure part fitment and interchangeability. As the IT grade number increases, the tolerances increase (the acceptable range of measurements grows larger). Within each grade, the allowable tolerance increases as the nominal dimension of the part increases.

Both Cp and Cpk are measured on a six sigma scale. For applications such as medical and aerospace manufacturing that typically require the highest level of confidence, a 1.66 Cpk is generally regarded as the minimum standard. For most other industries, a 1.33 Cpk is the minimum standard. A Cpk below 1.0 is unacceptable for most applications.

We chose to evaluate Cpk on a print-to-print and a part-to-part basis, taking measurements on parts in the same overall build, with parts in several locations throughout that build. Specifically, we printed edge cases to represent variation across the entire volume of the build chamber. The build was also packed with many parts beyond just the witness blocks, to give the best possible representation of the thermal effects that can occur in the normal operation of the Fuse 1+ 30W printer. 

To test dimensions, we used a calibrated height gauge. This allows the maximum amount of accuracy in measuring the dimensions of a part without risking introducing error from either the operator or warpage that could occur. We printed the test parts in Formlabs Nylon 12 Powder because of its popularity and wide variety of uses across many different industries. 

The general summary of measurement variations can be read in the following table. 

This graph shows the variation across X and Y of the printed parts. The result is extremely repeatable, with the parts showing a deviation of no more than 0.25 mm from the average dimension for all sizes, which corresponds to +/- 0.3% for all the values. All of this results in extremely high-quality Cp measurements, with the Cp staying high for all four dimensions.

For the increasingly large parts, there is an increase in the dimension clearly visible, which does negatively affect the Cpk since it evaluates dimensional accuracy as well as the precision. Since it does follow a scaling factor, this could be modified relatively easily by decreasing the scaling factor to compensate for the oversized features.

This can be broken down into the X and Y axes for each given corner and dimension, at which point the Cpk is very good (and since Cp is always greater than or equal to the Cpk, so is the Cp).

The actual measurements in Z vary more than they did in XY, but by plotting each point out, it is easy to see the variation as a function of the particular location, for 100 mm parts. Each individual corner is highly repeatable — the table below shows the Cp values for each given location and dimension, and other than one, they are all extremely repeatable within a tolerance band of +/- 0.5%.

The accuracy of the Z dimension is slightly lower than the precision. All of the parts for 100 mm fall within +/- 0.7 mm (+/- 0.7%), and generally, parts fall within +/- 1%. This fits into an IT tolerance grade of IT16 with a high Cpk. The precise measurements can be seen on the final table below.

If values are sectioned so that only the top portion is taken, however, the tolerance can be tightened substantially. Instead of values varying by as much as +/- 1% from the average, the measurements vary only by +/- 0.4% of the average, giving a very high Cp for all measurements with a much tighter tolerance limit than IT16; the parts maintain a Cp of better than 1.67 for IT13, though the Cpk is substantially worse due to the average being significantly above the nominal.

Generally, these results will be applicable to running any Fuse Series build - printed parts on a Fuse Series printer will generally be highly accurate and repeatable. This allows for designing parts for fitment into mechanical assemblies such as drones, underwater equipment, automotive, prosthetics, etc.. In many of these cases, the innate accuracy of the Fuse Series will suffice for making parts that can be immediately used.

However, there are some scenarios where a higher degree of accuracy may be required, especially in production settings. Since, as these results have shown, the parts are extremely repeatable, they can be tuned for the specific build to get the highest level of accuracy possible, even as tight as +/- 0.1 mm. The process for doing this is going to vary slightly depending on the build, but generally will involve scaling X and Y to the perfect size (i.e. if a 100 mm part prints at 100.3 mm, it is 0.3% too large, scale that dimension to 99.7 mm and it will print perfectly), and then tracking the Z height on parts to calibrate them into spec depending on the exact position in the build.

From both this paper and Formlabs’ experience, the general rules of thumb for the Z scaling are:

• Parts are often shorter on the bottom and taller on the top

• Parts are often shorter in the center and longer by the walls

Overall, the Fuse Series is capable of producing parts with extremely precise, consistent dimensions, on par with MJF and SLS printers that cost over 10x as much, making it a good choice for the production of end-use parts, providing high-quality and accurate parts every print.

To evaluate your own parts for the Fuse Series workflow and determine the right design, scaling factors, and print orientations, contact our sales team.