Dimensional Tolerance Assessment of Iron Castings Production Using 3D Printed Sand Process

Jiten Shah

Foundries have adopted various additive manufacturing (AM) processes over the years––in particular, 3D printed sand using the binder jet process to make cores and molds has been embraced by over 90% of the foundries per recent surveys. The use of 3D printed sand casting is growing in the production environment, and the initial feedback is that the process results are comparable to the precise sand-casting processes. 

Currently, 3D sand printing is predominantly used in production via a hybrid approach, where the mold is made using the conventional green sand process and the complex core assembly is redesigned with a consolidated 3D-printed core. However, little has been studied and is known in the public domain about the dimensional tolerances achieved with this toolingless precision sand casting process, especially the potential of achieving truer position and better internal feature tolerances. An AFS research project aimed to identify and provide guidelines for improved dimensional tolerances with 3D printed sand iron castings for design engineers so they can design new castings or redesign current castings for precision sand casting toolingless additive process. 

Conventional casting dimensional tolerances are impacted by many factors: the solidification and cooling behavior of the alloy; rigging design (risering and gating system); wear and tear of the pattern and core boxes; type of mold and core media and binder; key casting process parameters (pour temperature, shakeout time, mold hardness, mold and core coating) and post-casting operations. Toolingless 3D printed sand is a digital manufacuring process; however, the dimensional tolerances, reproducibility and repeatability are impacted by the type of 3D printer equipment and its build parameters––type of print media and binder, post print processing, and handling. Additionally, complex highly-cored casting solidification and cooling is constrained by the core and mold collapsibility, and the orientation with respect to gravity. Complex castings typically do not show uniform shrinkage in all directions, which is typically uniformly applied as the pattern contraction allowance to the tooling. The residual stress and distortion are accurately estimated with today’s casting process simulation tools, provided the accurate thermo-physical and thermo-mechanical properties of the alloy, core, and mold materials are available. 

The metalcasting industry sees the value of core consolidation; however, the overall dimensional tolerance capabilities of 3D printed cores is unknown. This research effort was carried out using side-by-side trials and dimensional studies primarily focused on the interrelationship between cores to test and quantify the capabilities of 3D sand printed cores to eventually enable design engineers to achieve lightweight, quality parts with confidence. Ultimately, the study’s intent is to give OEM design engineers the standard process capability of this new precision sand casting process using 3D-printed sand molds and cores.

Core Consolidation Redesign

A complex-grade iron housing demonstration casting in production at the participating foundry utilized six Isocured chemically-bonded silica core assemblies (Figure 1). These assemblies were placed in a vertically-parted, two-cavity green sand mold for this research evaluation. The six-piece core assembly was redesigned with a three-piece core design as shown in Figure 2 and was validated against 3D models of the current tooling and core assembly. Figure 3 shows some of the critical dimensions, which were measured to evaluate the impact of he overall stack up and intra-core print elimination with new three-piece design. Recommended production iron sand castings tolerance grade DCTG (Dimensional Casting Tolerance Grade) is from 8 to 12 per ISO 8062 (Table 1). The housing casting specified tolerance was DCTG 9 per Table 2 per ISO 8062 standard.

3D Printing of Cores

A detail form and fitment validation of the 3D printed three-piece core assembly with pre-production trials was conducted before releasing the full production trials. Pre-trial castings were sectioned and compared, with the production castings made using six-piece core assembly using 3D scanning technique, and were found satisfactory; Figure 4 shows a typical 3D scan result comparison. The 3D files of the three-piece configuration were provided to a team of six 3D printing core manufacturers consisting of 3D printing equipment providers, academia, and leading service bureaus. Table 3 shows the summary of the various 3D printed cores with details on the substrate and binder system used. Each 3D printing participant followed their best practice and captured the following 3D print build & machine parameters:

  • Orientation and Core Layer
  • Layer thickness, build rate/time.
  • Binder type and % addition.
  • Substrate–silica or ceramic with GFN, shape.
  • Strength (transverse and shear-dog bone), LOI of 3D printed core material.
  • Build box layout.
  • Type of machine used with some machine parameters.

Each 3D printing core provider produced 25 core assemblies. Each printed core was serialized with the provider’s assigned code for traceability to maintain a one-to-one relationship to the castings poured and dimensional inspection. Each core was inspected visually for overall dimensions before shipping. Some core suppliers provided photos at every stage: during printing, de-sanding, clean up, inspection, packing, and shipping. The production foundry visually inspected all cores to ensure the serial numbers were captured, and no physical damage was seen on any of the cores. Core wash was applied using the same production process, and each core assembly was validated using the production core assembly fixture to ensure proper form and fit before placing in the mold.

Casting Production Trial Details

Once all 150 3D printed cores were available, a total of over 175 castings (25 with current 6-piece design) were poured all at once. The key process parameters were captured for each mold, and every casting was processed identically before inspection for critical dimensions using the CMM. The pour temperature range was recorded to be 2470F to 2574F (1354.4C–1412.2 C); mold hardness was in the range of 94–95 on B scale and pour time was in the range of 14.68 to 14.82 seconds.

Statistical analysis of the CMM measured dimensions was performed using Minitab software and the overall dimensional assessment with regards to repeatability, reproducibility and variability/tolerance was conducted and compared with the conventional six-piece practice.

Results and Discussions

Residual maximum principle stress and displacement estimated after shakeout and cooling to room temperature using production conditions consisted of 2548F (1398C) pour temperature; 14.7 seconds pour time, with a furan-bonded 3D printed silica core and hard stable green sand mold, which is shown in Figures 5 and 6. Notice the non-uniform cooling due to casting geometry and rigging system impacting the residual stress and distortion.

Figure 5 shows a typical build box layout on left and measured 64.9 mm dimensions for Core Supplier B with serial numbers, and there seems to be no good correlation to the location of the core in the build box (i.e., located at bottom versus top).

Table 4 shows the summary of the measurements for 64.9 mm dimension for various 3D printed cores, production cores ran with the campaign (labeled R) and historical production measurements of over 400 casting (labeled P). The 3D printed core assembly had the least amount of standard deviation and highest Cp and Cpk values.

Figure 6 shows the histogram for 64.9 mm dimensional measurements graphically.

Similarly, the summary for 104 mm dimensions is shown in Figure 7, where 3D printed ceramic sand, furan bonded shows the least variability; however, there is a shift in mean, indicating less expansion of the sand along with 3D Printed H (phenolic silica bonded).

Another observation as seen in the box plot in Figure 8 for 112.7 mm dimensional analysis indicating ceramic sand with the least variability compared to silica sand and is believed due to the fact that ceramic sand is stable at higher iron temperature range and does not go through any phase transformation or volume change unlike fresh silica sand. The current production sand cores (labeled R and P), using thermally recycled silica sand, also show the least variability relatively to 3D printed fresh silica furan bonded. These aspects need to be further investigated.

Table 5 shows further analysis of the dimensional variability of potential sources with 3D printed cores. Assuming well-controlled key casting process parameters including post casting operational variables, three different sources were quantified for the few critical dimensions listed in Column 1: 3D printing machine accuracy including reproducibility and repeatability; casting residual stress and distortion related; and the balance by 3D printed cores potentially attributed to manual post-printing cleaning, handling, and shipment. This could be a fair assumption because 3D printed cores being toolingless, the digital manufacturing process eliminated other causes of variability. 

The 3D printing machine accuracy data used the equipment manufacturer’s recommended worst case variability of 0.2% of the linear dimension as listed in Table 5. The least variable ISO 8062 DCTG Grade 8 was used for comparison, even though the housing casting was required to meet DCTG 9. Overall variation is defined as a range that is computed as a difference between minimum and maximum dimensional measured values and is reported in Column 6. Column 5 shows the mean measured value. The post printed 3D core related dimensional variability is derived by subtracting 3D printing machine-related variability estimated in Column 9 and casting solidification and cooling related distortion predicted by simulation in Column 11 from the total variability reported in Column 6 for all five dimensions. Based on the overall range seen with the 3D printed core and ISO 8062 (Table 3) DCTG grades from 5 to 8 are feasible with 3D printed core casting production of similar complex iron casting, which falls into the investment casting DCTG grades 4 to 8 as shown in Table 1.

Figure 8 is a summary of total dimensional variability contributors, by percent of the average of five critical dimensions measured in each of three categories. It appears that 3D printing machines print cores very accurately; it is the subsequent manual post printing operations (such as de-sanding, handling, shipping, and applying coatings) that induce more spread and variability as evident in Figure 9, and future research needs to focus on these aspects. Automated de-sanding, for example, coupled with controlled pressure/flow rate, contact pressure during de-sanding, and robotic handling of the printed cores can potentially reduce this variability and further tighten up the dimensional tolerances. Figure 10 is a histogram comparing 3D printed with conventional coremaking.


1. 3D printed toolingless core consolidation has the potential for tighter dimensional tolerance capabilities than conventional sand castings, as demonstrated with a few critical casting dimensions in a complex iron casting production; the expected DCTG per ISO 8062 is in the range of 5 to 8, overlapping investment casting process achievable tolerance capabilities. More such studies in iron and other alloys with additional examples are required to establish a separate DCTG for the toolingless 3D printed precision and casting process that would give engineers the ability to create new designs with thinner walls, lighter weight, and improved properties.

2. It appears that the location of the core in the build box of a 3D printer has no impact on the dimensional tolerances; however, this needs to be further researched.

3. Limited data on 3D printed ceramic sand (one batch of 25 cores) appears to show the least amount of dimensional variability compared to silica sand and is believed to be due to no phase change or volumetric expansion and contraction associated with a ceramic substrate. This needs to be further investigated with additional research including more samples and data.

4. It appears that from the limited measurement data, post-printing manual operations such as de-sanding and handling of the 3D printed cores potentially impact the dimensional variability observed and need to be further investigated.