Who Has the Best 3D Printer?
This article was published in 2010. Since then, manufacturers have developed a number of new 3D printers and additive manufacturing technology. For more information about specific models, check out the 3D printer selector tool on Make Parts Fast.
Vendors of 3D printers make many claims about the cost, speed, and capability of their systems. Todd Grimm of T.A. Grimm & Associates conducted a benchmark study of six 3D printers to examine such claims. His analysis and conclusions offer an objective depiction of acquisition expense, annual expense, hourly cost and prototype cost..
The test parts for the benchmark study included a block, a housing, and a panel. The block combined the features of thin walls, small steps, spherical surfaces and thick sections. The housing was a moderately sized, thick-walled, prismatic prototype. The security panel had features common to injection-molded products as well as some thin-walled features.
Just how fast, inexpensive, and easy to use are 3D printers? The answer varies depending on whom you ask. A recent benchmark study conducted by T.A. Grimm & Associates evaluated six commercially available 3D printers from five companies in the following four areas: Time (build time, total process time, and direct labor/automation); Cost (prototype cost, annual operating expense, and acquisition/implementation cost); Quality (dimensional accuracy, surface finish, and material characteristics); and Operations, which includes day-to-day investment (time), suitability for office use, and “greenness” (recycling/disposal)). This article will examine time and cost. A following article will examine quality.
For each of the companies, the lowest priced model in the product family was tested: Alaris30 (Objet Geometries); ProJet SD 3000 (3D Systems); SD300 Pro (Solido); uPrint (Stratasys); V-Flash (3D Systems); and ZPrinter 310 Plus (Z Corporation).
To test the printers, T. A. Grimm used three distinctly different objects: a test block, a housing, and a security panel assembly. The security panel included two pieces. Test results will naturally vary because prototype parameters such as size, volume and level of detail will influence production time, cost and quality.
The test block was a small piece that combined thin walls, small steps, spherical surfaces and thick sections. The housing was a moderately sized, thick-walled, prismatic prototype. The security panel had features common to injection-molded products as well as some thin-walled features. It included both a front and back panel, so it allowed evaluation of the fit between mating components.
Benching was not done to eliminate the variable of post processing and to evaluate the accuracy of raw prototypes. Only those secondary operations necessary to complete the parts were done; cleaning, curing, support removal, and part infiltration.
Several expenses were used in the evaluation. The acquisition expense included the cost of the system configuration used in the benchmark study, including all necessary support equipment and estimated costs for facility modifications. It did not include optional equipment that is at the user’s discretion.
For the annual operating expense, the acquisition expense was combined with ongoing expenses such as annual maintenance contracts, labor and replacement parts for routine service, consumables and material disposal. Acquisition expenses were amortized (straight line) over five years. The annual operating expense includes fixed expenses and the variable expenses associated with a single shift operation. It does not include the variable expenses of labor and material for the production of prototypes, which were captured in the cost of the individual prototypes.
The most significant factors affecting hourly cost for each system were annual hours of operation, system cost, and maintenance expense.
A standard hourly cost for machine operation was derived from the annual expense amortized over anticipated annual prototype throughput. To determine annual prototype throughput and the associated machine hours, build times were calculated for the construction of two “typical” parts. The X, Y, and Z dimensions and part volumes of the benchmarked parts were averaged to yield the “typical” part.
Construction times were calculated for the concurrent building of two “typical” parts. Using the time per run and assuming a single shift operation—nine hours per day, five days a week, and 50 weeks a year— the maximum number of runs and the daily throughput were determined. Taking into account lost time for repairs, maintenance and scheduling inefficiencies, a utilization rate of 60% was applied to the daily maximum.
The hourly rate for machine operation was calculated from the annual operating hours and annual expense. The benchmark shows a striking range of results for the hourly rates. The SD300 Pro had the lowest rates at $0.99. The ProJet SD 3000 had the highest rate at $9.39. The uPrint and V-Flash were $3.62 and $3.03, respectively. The Alaris30 and ZPrinter 310 Plus were $5.67 and $6.28.
Noted Grimm, while it is tempting to use hourly rate as a basis for system selection, “it is not a useful measure of performance or operational cost. It is determined solely for calculating the production costs of the benchmark parts.”
In addition to hourly rates, vendors also use other factors to differentiate their systems. One such factor is continuous run time–the length of time a unit can run without material replenishment. It is calculated by determining the number of runs and their total duration that can be completed before the material supply is exhausted. In the benchmark study, the consumption and build times were calculated from the “typical” part used for throughput determination.
Among the units evaluated, even the shortest continuous run time exceeded 40 hours, enough for overnight and weekend operation.
Another factor is post processing. Noted Grimm in the study, the post processing effort varied widely, with variances found in the types of processes, degree of automation, and amount of time. Your actual times will depend on part size or configuration.
The report summarizes the post processing cycles of the systems as follows:
• Alaris30: A sodium hydroxide soak followed by a manual water-jet support removal process, which takes about 1.0 to 1.3 hours (mostly unattended).
• ProJet SD 3000: Parts are heated in a convection oven to melt away supports and are then washed in hot water; about 1.0 to 1.5 hours (mostly unattended).
• SD300 Pro: Surrounding material is peeled from part and what remains is pulled out with tweezers and picks. About 0.5 hour (manual).
• uPrint: Parts are placed in the support removal system where support material is washed/dissolved from the part. About 1.0 to 1.5 hours (mostly unattended).
• V-Flash: Parts are placed in a cleaning and curing station that: 1) washes parts with a propylene carbonate solution 2) rinses parts with water 3) cures parts with UV light. Supports are then clipped from the part and sanded smooth. Time: 1.2 to 1.5 hours (mostly unattended).
• ZPrinter 310 Plus: Parts dry in the machine before being excavated from surrounding powder. They are then depowdered with a jet of air and infiltrated with an adhesive (or similar) material. Time: 0.5 to 1.4 hours (partially unattended).
The cost of 3D printers should include the cost of materials, which includes binder, inflitrant, and support material costs. The effective material cost is the total cost of consumables divided by a part’s volume. Most of the consumables expense is in model and support materials.
The ability to combine multiple parts in a build can save time, partly by reducing process overhead, but the savings vary significantly. Much (or all) of the build time comes from a fixed amount of time per layer. The ProJet SD 3000, ZPrinter310, and V-Flash demonstrated a reduction in build times of 55%, 52% and 48% respectively. The Alaris30 (22% reduction) has a smaller ratio of fixed to variable time, which yields a smaller reduction.
The SD300 Pro had an even smaller reduction of 12% . While it can benefit from part consolidation, the size of its build envelope forced the four parts to be made in three builds. The smallest reduction was with the uPrint (5%) since its build times are largely unaffected by consolidation.
When it comes to time, the expected mode of operation will play as big a role in the decision as the types of parts to be produced. For the six technologies, the housing took the longest to construct because it has many of the characteristics that increase build and processing time. The ZPrinter 310 took 3.7 hours to build it, while the ProJet SD 3000 took 20.7 hours.
For all of the technologies, build time is a function of part height, and at 3.48 in., the housing is the tallest. For the Alaris30 and ZPrinter 310, the X-Y footprint (4.87 in. x 3.88 in.) is large enough to add additional print head passes, which added time. The housing’s volume (6.95 in3) was the largest of all the parts, and this increased time for the uPrint.
Noted Grimm, “While the ZPrinter 310 Plus demonstrates its claimed advantages in speed, some of the systems fail to live up to their promises. Conversely, systems long thought of as slow can offer competitive process times. There simply are too many variables to make bold, sweeping claims. And speed may be trumped by a dependence on part consolidation or the amount of manual labor required.”
Another factor you will want to evaluate is prototype cost, which includes the expense of system operation, labor, and consumables. For the study, Grimm used the hourly rate and actual build time to derive the manufacturing cost component. For the labor component, he used a rate of $35.00/hour for all steps in the process requiring manual work. Consumable costs, which include model and support material, infiltrants and miscellaneous items, were calculated from the actual amount used and the vendor-supplied list price.
Individual process times (hours).
The variance in average cost was substantial. For the four benchmark parts the highest average cost was $168.99 for the ProJet SD 3000. The lowest cost, $40.51, was for ZPrinter 310 parts, closely followed by the uPrint with an average cost of $59.55. Noted Grimm, “Surprisingly, the lowest-priced systems do not have the lowest part costs.” The SD300 Pro’s average was $158.70, which is the second highest, and the V-Flash came in third with an average of $84.85.
Part consolidation painted a different picture. Across the board, average costs dropped from 9.2% to 36.9%. For the three systems with the highest average, the effect of consolidation was to make them almost the same cost, ranging from $106.57 to $124.04. The three systems with the lowest averages were ZPrinter 310 ($33.76), uPrint ($54.08), and V-Flash ($74.39).
Now these numbers reflect the costs for the study’s prototype parts. You should conduct your own review on prototypes that represent components typical of your designs.
The basic cost for materials for 3D printers can be high. But when you add in the binder, inflitrant, and support material costs, the true cost is usually higher. The effective material cost is the total cost of consumables divided by a part’s volume. Most of the consumables expense is in model and support materials.
According to the benchmark study, “the ZPrinter 310 had the lowest material cost, which includes powder, binder and infiltrant, at just $3.57/in.3 Coming in second, uPrint’s material cost was $6.87/in.3 Due to a considerable amount of support material and some loss of model material when purging and planerizing, the next highest costs were with ProJet SD 3000 ($10.13/in.3) and Alaris30 ($11.84/in.3). V-Flash’s promoted cost of $8.00/in.3 swelled to $12.80/in.3 when support material was added to the total material consumption. This made V-Flash the second highest in material cost. The highest effective material cost—$25.05/in.3 for the SD300 Pro—was more than 15 times the list price of $1.63/in.3 The big jump in cost for the SD300 Pro was due to the high percentage of wasted material. What came out of the machine was a 6.3-in. wide, solid brick with the part buried inside. Everything that surrounded the part was waste that was shipped back to the vendor for recycling.”
This chart illustrates the impact of part size and configuration on build time.
Dimensional accuracy is dependent on the size and configuration of a part. The Alaris30 had 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.
Results for the ZPrinter 310, though, show well that quality is dependent on what and how accuracy is measure. “In terms of ± 2 σ, this system was consistently in the middle of the pack. But when judged based on the number of points exceeding ± 0.005 in., it was 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.”
In general, the studied 3D printers delivered on claims of having quick turnarounds, simple operations and low operating expenses. And to varying degrees, they satisfy the requirements for in-office use. While the study’s charts, tables and images show clear leaders in specific measures of time, cost, and quality, it also shows that as the measures are adjusted, the relative positioning of systems change, sometimes dramatically.
Noted Grimm, “One conclusion from this benchmark is that general conclusions regarding the best and worst 3D printers cannot be made. Another conclusion is that bold, sweeping claims for any technology may not hold true for all parts and in all circumstances. These claims should be investigated in the context of the products you will be making and the operating conditions.”
Here’s a look at the range in average cost for the four benchmark parts when produced in individual builds. The highest average cost is $168.99 for the ProJet SD 3000. The lowest cost, $40.51, is for ZPrinter 310 parts, which is closely followed by the uPrint with an average cost of $59.55. The lowest-priced systems do not have the lowest part costs.
T. A. Grimm & Associates, Inc.