Validating process simulation to experimentation for reinforced sand molds: simulation results confirmed the locations of stress concentration and demonstrated accuracy with the actual temperature profiles and cracking locations.
In the past few decades, simulation technology has made remarkable progress in casting design and process optimization prior to actual manufacturing, which reduces the time and cost of conventional trial-and-error methods. Although many experiments have been done to study sand molds/ cores (silica sand stiffness, sand cores expansion), simulation techniques for predicting and controlling sand molds/ cores casting process are still underdeveloped due to the complex mechanical behavior and failure mechanisms of sand materials. Thus, it is necessary to develop reliable and robust process simulation models of sand molds/cores for more efficient manufacturing. In the present study, mechanical tests of resin-bonded silica sand with 98.7% silica and 1.3% phenolic resin binder were performed. Experimental data from mechanical tests, together with some key data from literature, were used to build a material model for sand molds/cores for casting process simulation. Then, casting simulation for three different geometries of sand cups was performed, and corresponding experimental validation was carried out. The goal was to establish sand material models and apply them to process simulation and experimental validation of sand casting.
A commercial resin-bonded silica sand mixture (98.7% SiO2, 1.3% phenolic resin binder) was studied through both experiments and simulation. Mechanical tests, including three-point bending test and uniaxial tensile test, were employed to study the mechanical behaviors of the sand. A hardness test was done as a quantitative measure to confirm a uniform curing condition. Experimental data from mechanical tests were further employed to calibrate the simulation database. Cylindrical sand cup molds were designed to be filled with an aluminum alloy. The casting process was simulated using a finite element analysis (FEA) code. Laboratory gravity casting experiments were carried out to validate the simulation results. Both temperature profiles and failure positions presented good agreement between simulation and experiments. The methods and results provide valuable information on sand properties and a material model for casting process simulation.
Process Simulation and Casting Experiment Validation
To validate the material data of resin-bonded silica sand in casting simulation, sand cup molds with three different types of geometries, named as intact cup mold, flat-notch cup mold and V-notch cup mold, were designed for process simulation and experimental validation. Through the only change of geometry, the main idea is to create different stress conditions for each cup mold during heating and cooling of a casting process. Figure 1 is a schematic cross-section of an intact sand cup. It has a uniform thickness of 0.22 in. (5.7 mm), depth of 2.76 in. (70 mm), top inner diameter of 1.5 in. (38 mm) and a 1.5-degree draft for easy release from the stainless-steel mold after the sand is cured. The minimum wall thickness is the only difference among the three cup molds: 0.09 in (2.4 mm) for the flat-notch mold and 0.05 in (1.2 mm) for the V-notch mold, while the original intact mold has a uniform wall thickness of 0.22 in (5.7 mm). Additionally, both the flat notch and V notch shapes are acting as stress risers with different levels, and such changes wall cause stress variations in the casting process. All cup-shaped sand molds were made in a stainless-steel mold first by compacting using a CT-200 compaction table, followed by curing at 482F (250C) for 45 minutes in a forced air convection oven.
Default properties of aluminum alloy A356 in the simulation software were used in the simulation. Minimum mesh size was 0.04 in (1 mm) for intact cup mold, 0.035 in (0.9 mm) for flat-notch cup mold and 0.023 in (0.7 mm) for V-notch cup mold. Simulation calculations were compared with experimental results.
Tests Results and Discussions
Resin-bonded silica sand is more like a brittle material. Fracture stress of a brittle material varies in a much wider range as compared to that of ductile one. Therefore, 36 three-point bending samples were tested so that the results would be more representative and reliable. In addition to this, seven uniaxial tensile tests were completed as a supplement for mechanical properties. The reason more bending tests were planned than tensile is because normally a tension test is not preferred for brittle materials. The stress concentration near the jaws of the testing machine causes failure at the jaws, rather than in the gage section. Thus, more specimens were loaded in bending than in tension.
For the bending samples, the bonded sand behaved in a predominantly elastic manner and fractured in a brittle mode as expected. However, about one-third of the specimens exhibited a linear hardening behavior right before failure occurred. A representative curve of force vs. deflection for this bilinear behavior is shown in Figure 2. The red line was fit to both the elastic portion and linear hardening stage. The presence of inelastic behavior of resin-bonded silica sand is rarely reported. While the elastic part is expected to be the same under different loading conditions, the inelastic behavior may mainly be to three-point bending tests.
The average fracture stress of the total 36 samples was calculated to be 3.233 MPa and standard deviation was 0.585. The average elastic modulus was 2300 MPa. Sample mass was also measured and density was calculated as 1,628 kg/m3.
Based on Gaussian probability distribution function, the distribution curve in Figure 3 indicates 31 samples fall in p[+ or -]o region, 3 samples in [mu][+ or -]20 and 1 p[+ or -]30. There is only one sample that is out of p[+ or -]3o region, indicating a good consistency in bending tests.
As for tensile tests, the average ultimate tensile stress (UTS) was 1.450 MPa and standard deviation was 0.258. The average fracture stress from threepoint bending is about two times the average UTS from uniaxial tensile tests. A primary reason may be there is no axial load in a bending test as is in a tensile one. Instead, the beam type bending samples give rise to large shear deformations during tests. In the three-point bending test, the top half of the sample is in compression while the bottom half in tension, which is unlike the uniform stress in the cross-section area of tensile specimens. In this way, extra loading force, and thus higher fracture stress is required for three-point bending tests.
To ensure a consistent curing condition of tensile and bending samples for tests, a surface hardness test was employed as a quantitative measure to confirm the consistency. Hardness values were located in 60-70, with 20 being 1 mm penetration depth. The sand surface hardness value in this study may not be comparable to other systems, and it is mainly used to provide a more dependable judgement of curing degree rather than looking at the curing color with naked eyes.
Temperature Profiles Comparison
Temperature profiles of A356 from the simulation and experiment are plotted in Figure 4. Both solidus temperature (TS = 1018F [548C]) and liquidus temperature (TL = 1135F [613C]) lines are indicated on the plot. From the cooling curve measured by thermocouple, the average cooling rate between TL and TS is 0.157C/s, and the cooling rate at the initial 25 seconds of the curve is about 4C/s. The simulation result agrees well with the experimental data in the first 300 seconds. After 300 seconds, discrepancy between the experimental measurement and simulation calculation is presented in the slope and temperature values.
Temperature profiles of the sand in the original intact cup mold are drawn in Figure 5. The simulation curve shows a good agreement with the experimental measurement during the initial temperature rise region. There is little discrepancy at 100-350 seconds between the simulation and experiment, but the general trend is similar during this period. After 350 seconds, as in the case of Figure 4, more discrepancy is seen in the slope and temperature values. The reason for such discrepancy after longer periods is still not clear, but it is less critical after the solidification. Nevertheless, simulation prediction at high temperatures and initial stages of solidification and casting processes is of primary focus. For both graphs, the experimental measurements and simulation results of A356 and sand show good agreement at high temperatures in the beginning, which indicates that the improved input data and the selected material model for the sand are accurate and reliable.
Cracking Position Comparison
Process simulation and laboratory casting experiments were completed for three different geometries of sand cup molds. Figure 6a presents the simulation results of maximum first principal stress during the process. A scale bar is labeled with unit of MPa. Figure 6b and 6c show cup molds before and after pouring. From the first principal stress contour, tension is on the outer surface of cup molds and compression on the inner. The maximum first principal stress is increased from 1.332 MPa of original intact cup mold, to 2.372 MPa of flat-notch cup mold and to 4.205 MPa of V-notch cup mold. The stress concentration of the V-notch mold is on the tip of the notch, as labeled. Therefore, cracking should initiate at this location if it would occur. This also confirms it is tensile stress tearing the mold apart to cause cracking. Experimental results in Figure 6c validated the simulation results. No cracking occurred on the intact cup mold and flat-notch one. From the color difference between the flat notch and intact cup, the mold was increasingly damaged with thinner wall thickness. As for the V-notch cup mold, cracking finally initiated from the tip of notch at 5 seconds after pouring and propagated vertically along the notch. Therefore, with the only change in the wall thickness, the stress was gradually concentrated from intact mold to flat-notch mold to the V-notch mold, which eventually resulted in cracking. In this case, such phenomena could be well predicted through process simulation.
Summary and Conclusions
A resin-bonded silica sand mixture with 98.7% SiO2 and 1.3% phenolic resin binder was studied to develop a material model for casting process simulation. Thirty-six threepoint bending tests were completed with an average fracture stress of 3.233 MPa and a standard deviation of 0.585. Inelastic behavior of this sand was found and elastic modulus was measured as 2,300 MPa from the three-point bending tests. Density was also obtained from three-point bending samples as 1628 kg/m3. Seven (7) uniaxial tensile tests were done with an average UTS of 1.450 MPa and a standard deviation of 0.258. The data from mechanical tests are used to establish a material model for sand molds/cores in casting process simulation. Three types of sand cup molds were made and poured with A356. Stress was increasingly concentrated from intact mold to flat-notch mold to V-notch mold, which eventually resulted in cracking. Casting process simulation was accomplished using a FEA code.
The simulation results confirmed the locations of stress concentration, and demonstrated excellent accuracy with the actual temperature profiles and cracking locations. The established material model for the sand mold, based on the actual mechanical test data, can be used to predict casting defects related to cracking, such as veining defects.
This paper (17-002) was originally presented at the 2017 Metalcasting Congress in Milwaukee.
YAN LU, HUIMIN WANG AND ALAN LUO, THE OHIO STATE UNIVERSITY (COLUMBUS, OHIO); KEITH RIPPLINGER, HONDA ENGINEERING NORTH AMERICA (ANNA, OHIO).
Caption: Fig. 1 A schematic cross-section of intact sand cup mold is shown.
Caption: Fig. 2 This graph shows a typical bilinear force vs. deflection curve of a three-point bending sample.
Caption: Fig. 3 The Gaussian probability distribution function was applied to fracture stress.
Caption: Fig. 4 A356 temperature profiles between FEA casting simulation and thermocouple measurement are compared.
Caption: Fig. 5 Sand mold temperature profiles between FEA casting simulation and thermocouple measurements are compared.
Caption: Fig. 6 These are the simulation and experimental results of three different geometry cup molds: (a) 1st principal stress from FEA casting simulation; (b) and (c) cup molds before and after pouring, respectively.