3D Printed Smart Mold for Sand Casting: Monitoring Binder Curing
Additive manufacturing is now being used to generate sand molds and cores with complex and customized geometries with binder jetting, which is one of the seven subprocesses within the additive manufacturing taxonomy. A serendipitous benefit of additive manufacturing is access between the printing of layers to insert components inside the mold, and new strategies are now possible in the design, ventilation (channels to improve permeability), gating (sprues, runners to reduce the velocity of molten metal to mitigate turbulence and reduce porosity), and risers to manipulate solidification.
3D printing of sacrificial molds and cores involves binder jetting in which a binder is selectively inkjetted into a bed of sand premixed with an acid to provide the two-part furan curing reaction. After one layer of binder is jetted, the bed descends by a layer thickness (typically 250 microns) and a new layer of unbound sand is deposited uniformly across the bed and the process repeats. Interrupting the process and gaining access to the bed (either manually, in this case, but potentially using integrated robotics) is possible with an ExOne SMAX printer and the capability to integrate components (e.g., chills, thermal conduits, electronics with sensors, actuation for dynamic gating systems, etc.), and subsequently, to continue the process to construct smart molds is now possible and relatively unexplored.
Sand printing has been investigated at least preliminarily with explorations of the sand and binder materials, final casting characteristics such as porosity and surface finish, multiple-material castings, casting modeling, complex casting geometries, and even embedded sensors in molds and cores for unprecedented process monitoring.
A recent study investigated the insertion of fully embedded sensors into 3D printed sand structures during process interruptions in the binder jetting of Furan-based sand molds. The driving hypothesis is that humidity and temperature (independent input variables) could be correlated to VOC production (dependent variable) and finally to the mold strength at the time of the metal pour. Potential future experiments may include extended durations (one year or more) of data collection and reporting. Small Linux computers were employed that were not embedded in the structure (to avoid the use of batteries as required by fully isolated sensor systems) thus allowing for virtually endless data collection.
Pressure, temperature, humidity, and VOCs were measured; temperature and humidity were considered control variables that directly impacted the effectiveness of the curing. The VOCs were considered to be a product of the curing process and an indication of the extent of curing and a predictor of final mold strength. Barometric pressure was not expected to be of utility but provided an indication of the local weather, which impacted the independent variables in the uncontrolled ambient treatment.
The intent of the sensor monitoring was to ensure both that: (1) molds have had sufficient time to cure (minimum time after print before pour) and (2) to measure how long a mold can be stored in a warehouse or foundry before the binder begins to degrade and lose strength (maximum time between print and pour) for a given set of ambient conditions.
Experiment Set Up and Treatments
Test specimens were produced on an ExOne S-Max 3D sand printer (Fig. 1) utilizing an 80 AFS-GFN (grain fineness number) round grain silica sand and a furan binder system to create cubes 75 mm on a side. The recipe employed a 65% virgin/35% recycled sand mixture and an activator dosage level of 0.17% based on the batch weight.
In the center of the cube was a cavity with a small channel to the outside to allow for the exit of a 300 mm long connector cable. The sensor location within the block was centered to maintain at least 75 mm of bound sound in all directions with the exception of the cable channel. In addition, with each block of twelve sand blocks, several flexural test specimens were included in both the Y axis (in direction of the powder recoating) and X axis (in direction with the inkjetting) based on AFS Procedure 3348-18-S, Transverse Bar Specimen Preparation. These specimens were treated in the same manner as the sensorized blocks for the 28-day period.
Of the 12 sand blocks, sets of four were subjected to three experimental treatments. Table 1 describes the conditions for each treatment. The hypothesis of the experiment was that temperature and humidity would impede binder curing and reduce mechanical strength. The ambient treatment was to simulate a typical foundry scenario with changes in weather. Furthermore, VOCs were hypothesized to be an indication of binder curing completeness as an output of the chemical reaction. Volatile organic compounds are released in the process and the integral of measured VOCs could correlate with final binder strength and used to ensure that molds and cores can withstand the head pressure for the specific alloy, design, and height to ensure casting quality and yield.
This effort and future work in long duration experiments is expected to provide guidelines for the storage of molds with minimum and maximum durations for a given set of humidity and temperature conditions. These sensors are inexpensive, so the utility of using them in production environments is possible and could help ensure the quality and yield of high-value castings.
Of the 12 blocks, one block had a sensor lift above the powder bed level causing an obstruction during the subsequent powder recoating. A large, gaping hole remained and the half-built block and recoater collided. The cavity in the sand bed was manually backfilled with sand to allow for printing the top half of the block separately. After harvesting, the two detached sections of the bad block were manually bound together after inserting the sensor. This manually reconstructed block was included in the oven treatment and did not appear to have significantly different VOC behavior. Figure 5 (left) depicts the harvesting with sensors embedded and the cables exiting the sand blocks. Figure 5 (right) illustrates all six PIs reading 12 sensors directly after harvesting and just prior to subjecting the sand blocks to one of three treatments.
Figure 6 illustrates the temperature, humidity, and VOC generation for the four sensors in the oven treatment (low humidity and higher temperature). As expected, the temperature was measured between 34 C and 38 C (93.2 F and 100.4 F) and the humidity was maintained below 25% for the entire duration and all four sensors provided consistent data. The VOC generation was maintained at a level of between 20,000 and 40,000 ohms, which served as a proxy for the actual VOCs and provided a relative level that was generally consistent throughout the duration with a slight increase over time. In all four sensors, two spikes are seen at approximately days 5 and 16, which correlated with unexpected power outages. The PIs were programmed to restart data collection if rebooted and the VOC sensor requires warming up with initial values that artificially spike for the first 30 seconds. The spikes provided additional information as an actual spike in VOC generation would generally last longer and the artificial spikes inform the user of a power outage or related reboots.
VOCs measurements were hypothesized to provide an indication of the extent of curing and had not been explored previously in 3D-printed sand casting in the context of sensing within the mold. This sensing capability provides an intriguing potential ability to detect when a mold reached full strength. However, the lack of absolute calibration in this experiment rendered the measurements only useful from a relative sense to demonstrate that curing was impeded with low temperature and high humidity—both of which were measured more accurately. VOC calibration with gold-standard sensors is possible and would require trivial alterations to algorithms by adding a slope and offset correction (linear error), and the corrections are the subject of future work. Alternatively, the temperature and humidity values were consistent in terms of absolute value and together can be combined to provide a “pour/no pour” metric if monitored over the duration between printing and curing. Developing this algorithm with an established correlation is the subject of next steps in this research effort.
For the two sensors monitored in the refrigerated treatment, the initial temperature is higher as the sensors were connected in the laboratory and later transported to the refrigerator causing an initial higher temperature. Only two sensors were included in this treatment as one of the PIs stopped reading the additional two sensors. Similar to the oven treatment, the conditions were consistent for the cold, low-but-slightly higher humidity than the oven treatment. The VOCs in both cases appear to have a slow declining trend in value and although the shape of the graphs are very similar and show the power outage that was seen in the oven treatment data, the overall magnitudes of the graphs differed by a factor of three by the end of the experiment, further illustrating how the absolute values for this sensor were not valid, but rather that the general trend for any given sensor may provide insight. In this case, different from the oven treatment, the VOCs appear to slowly decline over the 28-day period.
The final ambient treatment includes two sensors for likely the same problem that affected the non-functional PI in the refrigerated case. The temperature and humidity vary more dramatically than the oven or refrigerated cases as expected in an uncontrolled environment and the VOCs in both sensors appear to remain relatively consistent throughout the duration. However, the absolute values are not consistent due to the lack of calibration.
In all eight successfully-read sensors, the values during the 28-day period of humidity and temperatures were averaged for determining the extent of the binder curing, which is hypothesized to be correlated with the final mold strength. Figure 10 shows the X and Y flexural strengths where X is the direction of inkjetting and Y is the recoating direction. For both axes, the hot/low humidity case provided improved strength and had strong correlation. This correlation could be used to provide a threshold to identify the minimum time for curing to provide sufficient strength prior to pouring metal. Future experiments will include measuring conditions for months to establish a potential maximum duration for which the molds could be stored.
This technique will further explore establishing guidelines for a window in time for which the molds could be used for casting depending on humidity and temperature conditions. The important conclusion points include:
• Sensors can be embedded successfully within molds during a printer interruption of sand molds to inform the extent of curing.
• For the specific BME680 sensor, the Volatile Organic Compound sensor mode would require calibration to be used in determining whether the sand binder had sufficiently cured to provide the mold strength required for metal pouring.
• A combined metric of temperature and humidity has been preliminarily shown to provide a higher correlation between the calculated metric and the final mold flexural strength.
• With further experiments, guidelines could be established to allow for the monitoring of temperature and humidity in a mold storage facility to calculate a minimum curing time for any specific conditions.
• Long-duration experiments will be the focus of future work to understand how long 3D printed sand molds remain sufficiently strong and whether these molds should have an expiration date with the specific consideration of the environmental conditions that the molds were subjected to in the storage facility over time.
Funding for the project came in part from the Murchison Endowment at the University of Texas at El Paso and a research grant by the American Foundry Society, Research Project Number 19-20 #12.