Introduction And Background

Precision sand molds and cores using chemical binders are the primary technology for the production of U.S. automotive, aerospace, and military components. Therefore, castings with defects produced from chemically-bonded sand cores and molds are a major issue for the casting industry. 

Quality losses from the use of chemically-bonded sand systems arise from several factors. One set of factors is the variation in materials themselves, such as grain size and shape, chemical composition, binder level, and additives. Another set of factors includes parameters for the process utilized for mold/core production. The processes include cold box, no-bake, hot-box, 3D printed, and injection transfer molding. Process parameter variations are mainly work time, strip time, pouring temperature, metallostatic pressure, etc. With an ever-increasing focus on delivering near-net-shape castings, it is becoming crucial to develop advanced digital quality control approaches such as artificial intelligence (AI) to compensate for such a wide range of variation yet increasing the speed to response.

The 50mm diameter, 8mm thick disc-shaped specimen is an AFS standard chemically-bonded sand specimen. The specimen is a simple geometry, which is believed to be a key factor in reducing the inherent variability of sand binder system quality assessment metrics. Specifically, these specimens have been shown to reduce measurement variabilities across a wide range of quality metrics for chemically-bonded sands, namely density, scratch hardness, abrasion, impact, hot-permeability, fast loss-of-ignition, and thermal distortion testing (TDT). These AFS standard tests have produced quality metrics that have proven to be useful in control of precision sand systems. 

Various casting trials that use the same disc-shaped specimen have been used to complement the standard tests to try to establish a link between sand system quality metrics and defects, such as veining/penetration, gas, erosion or distortion at a specimen/metal interface. Studies have shown a relationship between casting surface defects (veins and penetration) and thermal distortion curves (TDCs) obtained with the TDT. Such results suggest that casting trial data can be integrated into a foundry’s quality control strategy to close the needed information loop as a dependent variable. Still, the TDT remains a preeminent tool to acquire digital thermo-mechanical data with variety, velocity, and veracity (Figure 1 (a) and (b)).

Methodology

The disc-shaped specimens of a phenolic urethane amine cold box (PUCB) and a 3D printed furan (FUR3D), were studied at cast iron temperatures. It is important to point out the silica sand type, size, distribution, and binder level were not normalized. The differences between the systems are not intended to be an indictment on any system. The aim would be to show and compare thermo-mechanical differences in sand binder system.

Dimensional Time-Series Data

During the experiment, dimensional data for the specimen was collected using a non-contact measurement tool. For profile data collection, a non-contact 3D measuring macroscope was utilized. This measurement system also allows the surface of the disc-shaped specimen to be captured at ambient or elevated temperature. Through the use of the 3D macroscope and its accompanying software, high-speed, high-accuracy 3D measurement can be obtained.

Thermal Distortion Test (TDT)

The TDT is suitable for measuring thermo-mechanical behavior, specifically distortions, of chemically bonded sand systems. The test temperature on the TDT apparatus is variable and can be set to mold-metal interfacial temperatures for a specific alloy. Similarly, different loading pressures can be applied to the specimen to simulate different metallostatic pressures on the core/mold material. After exposure to the temperature, if the test specimen remains unbroken, additional information is acquired, such as the presence of cracks (which causes veining), weight loss that relates to binder pyrolysis, and the amount of loose, unbonded sand at the mold-metal interface.

Directional heating of sand composites (mold and core media) generates anisotropic thermal gradients in materials. When shaped sand composites make contact with molten metal, heat is transferred from the metal to the sand causing thermo-mechanical-chemical reactions and resulting in dimensional deformation and shape distortion. Specifically, thermally-induced reactions of the binder occur simultaneously along with sand expansion and/or plastic deformation leading to distortions in the sand core or mold. For certain chemically bonded systems such as organics, reactions generally include the release of volatile materials, possible core strengthening reactions from secondary curing, and core weakening from pyrolysis can occur. It is important to understand that when acquiring the data, distortions can also be caused by the binder or aggregate. 

The specimens for two different sand binder systems were fabricated in industry, characterized, and thermal distortion tested. Results obtained from TDT are presented in Tables 1 and 2 and in Figures 3 and 4. Some of the data collected after the performance of the TDT is presented in Table 1. The figures given in the table are the final magnitudes of distortion after the TDT, as well as an index that captures the total distortion suffered by each sand binder system axially (longitudinal) and radially. Additional inspection of the specimens resulted in the observations listed in the table as well.

The TDCs for each one of the two sets of specimens are given in Figures 2 and 3. In Figure 2 typical axial (i.e., vertical) relative displacement of the PUCB and Furan specimens are shown. Figure 3 illustrates the radial (i.e., horizontal) relative displacements. For axial or longitudinal distortion (Figure 2), both sand binder systems showed undulations that indicate thermo-mechanical and thermo-chemical changes in the binder system at elevated temperature. There is an initial expansion, however––PUCB showed vivid plastic deformation after approximately 30 seconds of testing while expansion in FUR3D continued for the duration of the test. In the case of radial distortion (Figure 3), PUCB had larger magnitudes of radial displacement all through the 90 seconds of the TDT. The differences in distortions between the two systems is a reflection of the differences in their corresponding sand binder systems.

3D Macroscope Examination

The profile scanning of the specimens before and after TDT are obtained, and typical pictorials given in Table 2. The table also shows numerical information regarding the surface roughness and the volumetric change of the specimen. The change in volume is based on the scanned data, and the volumetric difference of the FUR3D specimen was the lower than the one for PUCB. The heat-affected zone in the specimens of sand with organic binders reveals unbonded sand where the binder had been completely pyrolyzed but remains intact because the specimen is undergoing compression against the hot surface at that location. However, cracks and fractures become evident after blowing away the loose grains of sand.

3D Visualization

One particular aspect that has been an issue with the TDCs is the difficulty in understanding the actual meaning of the information provided in them. Characteristics such as increase/decrease in magnitude of distortion and change in trend (i.e., slope) of the resulting distortions are sources for most of the questions that technical people will have when looking at the TDC for a given sand binder specimen. These issues are mentioned whenever two different sand binder specimens are compared. Therefore, an aim of 3D visualization is to serve as an additional aid to casting design engineers, by providing thermal distortion information in an understandable and useful manner.

The 3D visualization was developed using a commercial graphics package. The proper 3D visualization of TDT data is important to understand the volumetric transformation that occurs during the time of the test. It is essential that such visualization is based on the actual data measurement. The 3D visualization utilizes TDC data, scanned profiles, and other measurements to display 3D representations for various sand binder systems under thermal effects. The same data used in developing Tables 1 and 2 is used for developing the 3D visualization.

The visual aids will be valuable understanding of the relationships between the sand binder system and the molten metal, at a specific temperature and head pressure. Additionally, such representations will serve as an aid in diagnostic tasks regarding the quality of the casted part. The ultimate objective is for foundry engineers to use information from 3D representations to find sustainable benefits for designing quality near-net-shape castings.

This digital approach to design is what permits the development of generative AI models to augment our understanding of the mold/metal interface.

The 3D AI visualization tool will help users to better understand the results from the TDT, thus allowing them to have better information regarding a particular sand binder system under given temperature-load conditions and make superior and agile engineering decisions. To show-case the actionable data that can be obtained from this digital approach, a shell resin coated silica sand with binder levels similar to those in the sand systems in the previous figures and tables was used for generating the graphs and images in Figure 4. Figure 4a shows the predictions of a generative adversarial network (GAN) for the time series measurements (radial distortion, longitudinal distortion, and thermal gradient) taken during a TDT test for a specimen made of a shell resin coated silica sand. These predictions show an excellent agreement with TDT data obtain an actual specimen, also shown in Figure 4a. Figure 4b shows screenshots from the corresponding 3D visualization generated from the GAN predictions of the measurements for the TDT.

Additional Digital Data

Apart from thermo-mechanical data, additional property data can be collected when performing a TDT, mainly thermo-chemical. The use of digital smart sensors technology which captures time-series data; the researchers are able to measure certain off gassing compounds. Furthermore, time series data on thermal gradient and dimensional change from the hot surface to the backside of the disc-shaped specimen is achieved. From this data cookie thermal properties such as thermal conductivity and volumetric linear thermal expansion are attained.

Conclusion

Thermal distortion is the expansion, contraction, and degradation experienced by a mold or core under extreme heat and liquid pressure of the molten metal, and such disruptions might result in casting surface defects. TDT is a safe, quick, and cost-effective technique for monitoring thermal distortion in chemically bonded sand systems. 
The current interpretation of TDT curves as 2D graphs is currently manual. The TDT generates real time data of both axial and radial deformation due to thermal distortion of a disc-shaped specimen. Though a TDC provides valuable insight to the nature of specimen deformation, it may be arduous to interpret the result by technical people not acquainted with the test. The visualization tool will provide a 3D mapping for thermo-mechanical and thermo-chemical properties for various sand binder systems. Foundry design engineers will have simple imagery compared to the more laborious numerical data, which affords a machining learning (ML) capability.

Today, precision sand castings require tighter dimensional tolerance. The metal casting industry has been placing a strong emphasis on near-net-shape and thin wall castings, while simultaneously maintaining increasingly stringent dimensional reproducibility requirements. The TDT has proven to be an effective laboratory testing methodology for measurement of distortion in chemically bonded sand-binder specimens, and to determine the presence of undesired casting features. TDT property data can be shared as enhanced boundary conditions with the more advanced casting tools such as flow and solidification simulation analysis to improve casting design and aid manufacturing prediction.