Introduction

The casting process has an essential impact on the creation of the local microstructure of a casting. These local variations in microstructure are leading to locally varying mechanical properties. The properties of cast iron castings are not only depending on their geometry, but are mainly driven by the graphite morphology, the microstructure, and defects. Therefore, the chosen metallurgy and process control are essential parameters for the performance of a casting.

The conventional casting design process does not consider the impact of microstructure variations on the fatigue/lifetime performance of castings. The designer lays-out castings based on established standards. These standards assume homogenous properties throughout the entire casting. As a result, only one material dataset is considered in fatigue/lifetime prediction analysis tools.

Local residual stresses and microstructure variations are rarely recognized in lifetime prediction simulations. On one hand, this leads to underestimating failure risks, while on the other hand to not utilizing the full performance potential of the material. Designers are uncertain how close the expected mechanical properties match the ones found in the real castings. A conservative design approach requires them to apply safety factors, which lead to unnecessarily high weight and resulting costs.

Foundry engineers are suffering from undesired consequences created by this approach as well: Safety factors result in larger wall thicknesses, which increase solidification times, which usually lead to decreased mechanical properties. Thicker walls are also tougher to feed and are increasing residual stresses within the casting during its cooling process.

The necessity to couple casting process simulation with fatigue/lifetime prediction analysis is obvious. The full potential of cast materials can only be unlocked when as-cast properties, meaning locally varying material properties resulting from the casting process, are transferred into these tools. Casting process simulation tools need to provide answers to questions foundry engineers and casting designers have. It is essential to not only qualitatively but quantitatively describe material and mechanical properties through simulation.

Figure 1 shows the conventional computer-aided engineering (CAE) design process of castings on the right side (surrounded by a blue dashed line). Only the expected extreme external load cases and temperatures experienced by the component in its application are considered.

Figure 1
figure 1

Potential integration of casting process simulation in the casting design process.

Full size image

The integration of local properties provided by casting process simulation allows the designer to customize the casting design for the specific requirements of an application. The transfer of residual stress distributions caused by the casting process or heat treatment and their consideration as additional load in stress analysis runs is easily accomplished and already standard operating procedure during the development of, for example, cylinder heads. Unfortunately, the impact of local property variations on the durability of a cast component is only recognized rudimentarily. The correlation of local properties and fatigue/lifetime prediction has only been done experimentally for specific components, which makes it difficult to transfer the findings to other parts.

It is necessary to close the gap between already available local property results from casting process simulation and local cyclical material properties to derive a geometry independent coupling of the two. This was realized for the first time for several cast iron materials and applications in the framework of German research project “MABIFF”.1

The concept of this research project was to couple the varying microstructures of cast irons (GJS-400 and GJV-450), predicted by casting process simulation, with the lifetime prediction for castings. Experiments were used to derive S–N curves (Woehler curves), which resulted in the development of a closed chain between casting process and the prediction of the final lifetime of a cast component. Local properties driven by the production process of a casting can now be transferred into and considered by lifetime prediction tools (Figure 2).

Figure 2
figure 2

Process of transferring property distributions from casting to lifetime prediction analysis.

Full size image

Microstructure Prediction for Cast Iron

Metallurgy and alloying components have an essential influence on the final microstructure and resulting mechanical properties of a casting. The chemical composition and inclusions, the melt treatment (charge materials, melting method, treatment, and inoculation) as well as the local cooling conditions are of utmost importance. Foundry engineers use these process variables to dial in the desired microstructure (graphite form, ferrite/pearlite ratio) and to avoid undesired defects (i.e., porosity or dross) and microstructures (i.e., graphite deformations or chill).

Simulation programs need to be able to predict the kinetics of the creation of the different phases locally during the entire solidification and cooling process.2 This requires, besides the consideration of alloying elements, the consideration of the inoculation and the melt treatment process. The impact of these is usually overlapped with the local cooling conditions within the casting. The calculation of the plain macroscopic solidification and cooling behavior cannot consider these parameters. Microstructure simulation is required to calculate at any time in any location inside the casting, which amount of which phases is created based on the parameters mentioned above. Figure 3 provides an overview of all parameters considered, which models are used, and what results can be derived.

Figure 3
figure 3

Simulation of cast iron microstructure under consideration of input parameters, models utilized, and derived results.

Full size image

Besides the gating and riser system and the geometry of the casting, casting process simulation considers the chemical composition, the melt treatment, and the inoculation, as well as other relevant process parameters. The program utilizes these input parameters and local cooling conditions to calculate the locally available inoculation sites, the growth of all phases, the impact of segregation to calculate the solidification process and resulting local microstructure, and its properties.

The calculation of all phases during the solidification process allows for the prediction of the final microstructure when the casting is completely solidified (number of nodules/eutectic cells, gray/white microstructure, austenite/eutectic ratio for lamellar/ductile iron and, additionally, nodularity for CGI).3 During the following cooling process, the diffusion of alloying elements within the austenite is considered to predict the amount of graphite. The additional consideration of segregation effects of alloying elements allows for the accurate prediction of ferrite and pearlite growth during the eutectoid phase transfer. The calculation of cooling conditions below the eutectoid phase transformation leads to the prediction of phase ratios of the matrix (ferrite/pearlite ratio, pearlite packaging). With this, it is possible to predict the static local mechanical properties over the entire casting (UTS, YTS, hardness, elongation, E-modulus).

Experimental Research to Locally Couple Microstructure and Expected Lifetime

The framework of this research project included the evaluation of major microstructure characteristics of a ductile iron (GJS-400 or….) and a compacted graphite iron (GJV-450 or…) and their impact on component lifetime. For GJS-400 the impact of its ferrite/pearlite ratio and for GJV-450 its nodularity and pearlite content was evaluated. Both characteristics can be locally predicted with casting process simulation. The consideration of defects like shrinkage, porosity, or dross was excluded on purpose in this research project.

To realize a complete chain of information, both materials underwent comprehensive cyclic load testing using tension–compression and cyclical bending tests for various microstructures.4

It was essential for the development of a representative correlation between microstructure and component lifetime to use typical microstructures found in the relevant castings in the test pieces. Due to the required sizes of the test pieces, there was only a limited amount of pieces that could be derived from the castings (windmill bedplate/frame, bearing support, and crankcase). Therefore, additional test castings needed to be poured (Figure 4).

Figure 4
figure 4

Test piece location for ductile iron bedplate (a), test casting (b), CGI crankcase (c), and step casting (d).

Full size image

The different melts were dialed in by the research partners in a manner to achieve typical variations in nodularity for CGI found in crankcases. The different wall thicknesses in castings and test castings lead to sufficient variation in ferrite/pearlite ratios for the gray iron.

The samples were machined and polished to exclusively evaluate the influence of the microstructure on the lifetime performance. The impact of the casting surface was not part of this research.

The fatigue test runs were performed under tension–compression and cyclical bending conditions. Tension as well as elongation monitored test runs was performed, as the lifetime prediction analysis uses both, tension and elongation concepts.

After failure, the fracture surfaces of the samples and the local microstructure were evaluated using automatic image analysis according to DIN EN ISO 945. Twenty-two characteristics were evaluated, specifically the number, shape, and size of the graphite particles, as well as the ferrite/pearlite ratio. The chemical composition was considered as well. Figure 5 shows an example of how the nodularity of a sample from the step casting was evaluated.

Figure 5
figure 5

Image analysis for the determination of nodularity. Nodularity 13.6 % (magenta: spherulites, turquoise: vermicular graphite, red: particle not considered).

Full size image

The microstructure characteristics found were also used to validate the accuracy of prediction of the casting process simulation tool used in this project. Figure 6 shows an example of a comparison between nodularity prediction and measurement in the step casting and the crankcase. All values provided by the simulation are within the measurement tolerance of the measured samples.

Figure 6
figure 6

Comparison of calculated and measured nodularity for step casting and crankcase.

Full size image

Comprehensive variance and regression analysis were used by the project partners to derive S–N curves (Woehler curves) from the correlation between the microstructure distributions found in the samples and castings and the durability test runs. Two of the S–N curves developed are shown as an example in Figure 7.

Figure 7
figure 7

Example of measured S–N curves for ductile iron with two different pearlite contents.

Full size image

Figure 8 shows the mathematical correlation between predicted and measured durability values for samples in tension–compression runs.6 It shows a correlation of 87 %. These findings were implemented into the MAGMA5 simulation program. Now it is possible to predict durability values with the consideration of process conditions at any location in a casting. The local durability values can easily be transferred into lifetime analysis programs (Figure 9).

Figure 8
figure 8

Predicted versus measured tension–compression durability values for ductile iron. Displayed is the deviation from the median durability. The left and right limit depicts the range of the measured deviation.

Full size image
Figure 9
figure 9

Calculated durability values as function of the local microstructure prediction.

Full size image

The durability values distributions developed in this project provide a significant qualitative and quantitative improvement compared to the currently used conventional method. This integrated virtual process chain was validated on one ductile and one compacted graphite iron alloy.

Application Examples

Local Durability Prediction in a Crankcase

Two major microstructure characteristics drive the strength of compacted graphite iron: the shape and size of the graphite particles and the ferrite/pearlite ratio (assuming no defects or gray/white microstructure is present). Predicted nodularity and pearlite content of test castings and of an Audi 3,0l V6 TDI CGI crankcase were used for the coupling of local microstructures with durability values and validated by measurements. The most stressed areas of the crankcase are shown in Figure 10.5

Figure 10
figure 10

Most stressed areas of the crankcase (colored).

Full size image

The MAGMA5 simulation tool was used to predict the nodularity and pearlite content in the critical areas. The program predicted a nodularity of about 20 % and a pearlite fraction of more than 90 % in the critical areas (Figure 11). Based on these values, MAGMA5 calculated local durability values (see also Figure 9), which were used in the lifetime analysis program.

Figure 11
figure 11

Calculated distribution of nodularity (left) and pearlite fraction (right) in the most stressed areas of the CGI crankcase.

Full size image

The consideration of local S–N curves leads to double the number of cycles until failure vs. using the conventional method. This explains why no castings failed on the test stand (Figure 12).

Figure 12
figure 12

Predicted durability in highly loaded areas considering the pearlite fraction leads to higher cycle numbers to failure values. The critical red indication is eliminated in the new lifetime prediction.

Full size image

Durability Prediction of Gray Iron Bearing Support

A durability strength value of 190 MPa was measured for a gray iron bearing support. Conventional methods, which are based on established standards, showed only a value of 140 MPa for this material.7 Results of fatigue tests for the bearing support are shown in Figure 13 for two different melts.

Figure 13
figure 13

Fatigue test results for bearing support made with two different melts (charges).

Full size image

Besides showing these quantitative differences in values, it is also possible to clearly define potentially weak areas. Figure 14 compares the conventional (microstructure independent) and the improved prediction of crack initiation areas in the bearing support. Only the microstructure distribution-based durability calculation predicts the correct location of weak spots in the casting.

Figure 14
figure 14

Comparison of weak spots in a bearing support. Only the new method shows the correct crack initiation locations.

Full size image

Outlook and Summary

The results of this research project are only the first step toward the integration of process and application simulation with the goal to predict correct and robust fatigue/lifetime values. One essential goal of the work presented here was to realize a new methodology and to show its potential. The local material behavior of many cast alloys is not only defined by their microstructure distribution, but also by local defects. This is especially important for ductile iron materials. Their performance can be significantly reduced by inclusions and dross. The calculation of these effects, so they can be considered in predictive tools, is currently being worked on.

The proposed concept of local durability values offers great potentials in the design of safety critical components, like in wind energy components. Figure 15 shows the local durability values for a ductile iron (GJS-400) windmill base frame.

Figure 15
figure 15

Prediction of locally varying durability values compared to measure values in a windmill base frame.

Full size image

The red dots in Figure 16 on the left side show the local fatigue strengths (351–455 MPa) based on conventional standards. That approach leads to a very conservative design compared to when locally predicted durability values would be used. The conventional approach puts ductile iron to a disadvantage.

Figure 16
figure 16

Calculated (blue) tension–compression durability for ductile iron (GJS-400) compared to predicted standard values (red) in the same locations.

Full size image

Casting buyers and designers have learned over the last years that early information regarding expected property distributions in castings is not only contributing to quality assurance and risk minimization. They increasingly utilize this information as a chance to reduce weight and optimize the performance of castings. An optimal, meaning application load and weight optimized, casting design is only achievable if the designer can fully unlock the potential of the material. The integration of process and application simulation offers the development of more realistic design rules for castings.