0 f
Determination of mechanical properties in CGI- cylinder blocks by experiment and simulation
Sebastian Edbom
Examensarbete i
Hållfasthetslära
Avancerad nivå, 30 hp
Stockholm, Sverige 2014
1
Utvärdering och simulering av mekaniska egenskaper
i CGI-cylinderblock
Sebastian Edbom
Examensarbete i Hållfasthetslära Avancerad nivå, 30 hp Stockholm, Sverige 2014
Sammanfattning
I takt med att den tekniska utvecklingen går framåt ställs allt högre krav på konstruktionsmaterial. Inte minst gäller detta cylinderblock i lastbilar. Ett intressant alternativ till traditionellt lamellärt grafitjärn är kompaktgrafitjärn (CGI), som i flera avseenden har mekaniska egenskaper som är bättre lämpade för cylinderblock än det traditionella alternativet. I detta arbete kartläggs mekaniska egenskaper hos cylinderblock gjutna i CGI. De uppmätta värdena används sedan för att utvärdera ett kommersiellt gjutsimuleringsverktyg.
Arbetet visar på att det är möjligt att erhålla låg spridning i mekaniska egenskaper mellan individuella block givet att ingångsparametrarna är likartade. Samtidigt går det att konstatera att det föreligger viss spridning i mekaniska egenskaper inom varje cylinderblock.
Simuleringsprogrammet som utvärderats överskattade aldrig någon av de undersökta hållfasthetsparametrarna för någon av de positioner som undersökts i detta arbete.
Detta är betryggande då det borgar för att simuleringsprogrammet kan komma att utgöra ett pålitligt verktyg i framtida utvecklingsarbeten.
2
Determination of mechanical properties in CGI- cylinder blocks by experiment and
simulation
Sebastian Edbom
Degree project in Solid Mechanics Second level, 30.0 HEC Stockholm, Sweden 2014
Abstract
As the frontier of technology is pushed ever farther, the need for new and improved materials is always present. This thesis investigates the mechanical properties of compacted graphite iron (CGI) as it has been suggested as an alternative to traditional lamellar iron in truck cylinder blocks. The obtained values are compared to the mechanical properties predicted by a commercially available casting simulation software.
This thesis shows that the mechanical properties vary within each CGI- cylinder block with respect to position and that the technology of casting several cylinder blocks with similar properties already exists.
The casting simulation software used in this thesis was found to give conservative predictions of the evaluated mechanical properties.
3
Förord
Ett stort tack riktas till följande personer vars bidrag och insikter har hjälpt till att höja kvaliteten på detta arbete:
Jonas Lindberg, Nulifer Ipek samt övrig personal på UTMB och gjuteriet, SCANIA.
Ett extra tack riktas till Jessica Elfsberg, Mathias König och Jonas Faleskog vars trygga handledning varit både värdefull och uppskattad.
4
1 Introduction ... 5
2 Problem formulation ... 5
3 Background ... 5
3.1 Different types of cast irons... 5
3.2 Sand mould casting ... 8
3.3 Analyzing the casting process ... 9
3.4 Casting simulation – basic concepts ... 10
3.5 Casting simulation – an example in MAGMAsoft® ... 11
3.6 Earlier similar works at SCANIA ... 16
4 Method ... 17
4.1 Temperature measurements during casting ... 17
4.2 Uniaxial tensile and compression tests ... 22
4.3 Simulation in MAGMAsoft® ... 27
4.4 Comparing results ... 28
5 Results and discussion ... 29
5.1 Cast batch data ... 29
5.2 Experimental results ... 30
5.3 Results from simulation ... 36
5.4 Comparison between experiment and simulation ... 39
6 Conclusions ... 50
6.1 Experimental results ... 50
6.2 Results from simulation ... 50
6.3 Comparison between experiment and simulation ... 50
7 Further recommendations ... 51
8 Bibliography ... 52
9 Appendix A – Solidification and cooling process of cast iron ... 55
10 Appendix B – Effect of added thermocouple tip protection ... 59
11 Appendix C – Positioning of the thermocouples ... 60
12 Appendix D – Mesh convergence study ... 69
13 Appendix E – Comparison between temperatures in CGI and LGI ... 71
5
1 Introduction
The fact that a component is cast as one piece is no guarantee that the mechanical properties will be the same throughout the structure. Gas- and sand inclusions are examples of local defects that may compromise the homogeneity of the material [1].
Depending on the local solidification and cooling conditions in different regions of a component, the microstructure will differ [2]. Furthermore, shrinkage during solidification may create inner cavities in the material [3].
The ability to predict local mechanical properties of a component already before it is produced is obviously a great advantage in the development process. One way to enhance this ability is to use simulation software in order to estimate mechanical properties of different designs. However, it is of great importance that the simulation software is thoroughly verified before it can be trusted to correctly predict reality.
This work aims to measure and compare values of mechanical properties with corresponding simulated values for a particular product cast in a particular material.
2 Problem formulation
The aim of this work is to compare the mechanical properties of a SCANIA cylinder block with the mechanical properties predicted by the casting simulation software MAGMAsoft
®
. The study will be conducted on cylinder blocks cast in compacted graphite iron (CGI).The local MAGMAsoft
®
installations at SCANIA are already used in order to simulate the casting process of the cylinder block and to predict its mechanical properties. This work also aims at adapting simulation input parameters such as pouring temperature and chemical composition of the melt to the parameters found at the local foundry.Hopefully this will ensure that the estimations done by the simulation software will be more close to reality than they are today.
3 Background
The different subsections below will give the reader insight into the mechanical and microstructural differences between three common types of cast iron, as well as describing the basic idea of sand mould casting. The main concepts of numerical casting simulation are introduced and a simple casting example is explained step by step where the reader is guided through the different stages of a MAGMAsoft
®
casting simulation. Findings of some similar works done previously at SCANIA are presented and briefly discussed.3.1 Different types of cast irons
The transition of iron from a hot liquid to a room temperature solid is a combination of different metallurgical microprocesses and include several intermediate stages. The
6
main metallurgical microprocesses are discussed in “Appendix A – Solidification and cooling process of cast iron”. Below follows an introduction to the mechanical properties of three common types of cast iron. They are lamellar graphite iron, spheroidized graphite iron and compacted graphite iron.
3.1.1 Lamellar Graphite Iron (LGI)
The oldest known cast iron is LGI. It is still used today because of desirable properties such as low price, good castability and good machinability. LGI is also known for its abilities to conduct heat and absorb sound and vibrations. [1]
3.1.2 Spheroidized Graphite Iron (SGI)
Another common cast iron is SGI. Mechanically it is more ductile and has a higher ultimate tensile strength (UTS) than LGI. It is resistant to fatigue but not as machinable as LGI. The thermal conductivity is lower than in LGI and the ability to absorb sound and vibration is lower for SGI than for LGI. [1]
3.1.3 Compacted Graphite Iron (CGI)
CGI has a microstructure that is in between that of LGI and SGI. Analogously, the mechanical properties of CGI are in between the mechanical properties of LGI and SGI.
The combination of relatively good machinability and high thermal conductivity from LGI with the high ductility, UTS, hardness and E-modulus of the SGI makes CGI an attractive choice for manufacturers of cylinder blocks. [1]
3.1.4 Comparison between LGI, SGI and CGI
Table 1 shows typical material data for the three types of cast iron described above.
Table 1. Typical material data for three types of cast iron [1],[4]. The data represents as-cast properties.
Property1 LGI CGI SGI
Ultimate tensile strength [MPa] 250 450 700-750
E-modulus [GPa] 105-125 145 160-176
Elongation A [%] <1 1,5 5-6
Heat conductivity at 300°C [W/mK] 46-48 37 28-31
Relative damping coefficient 1 0,35 0,22
Hardness 10/3000 [HB] ca 190 217-241 230
The microstructural difference between LGI and SGI is that in LGI the graphite is present in the form of flakes, while it has formed spheres in SGI. In CGI the graphite has formed shapes that have a network-like structure as in LGI. However, the ends of the graphite flakes are rounded, bearing resemblance to the graphite spheres in SGI.
The graphite precipitations in CGI are shorter and thicker than in LGI. [1]
1 There exist many types of LGI, CGI and SGI. Mechanical properties may vary in a large span for all three types of iron.
7
Figure 1 aims to visualize the difference between three types of cast iron in polished samples. The black areas are graphite while the white areas are matrix material.
Figure 1. The left most picture shows LGI while the middle one [5] shows CGI. The right most picture shows SGI.
From Figure 1 it can be observed that the graphite precipitations in SGI are more spherical than the precipitations in LGI and CGI. The roundness of a CGI or SGI graphite precipitation larger than 10 μm can be evaluated according to ISO 16112 [6] by
equation (1).
In equation (1), is the area of the precipitation in a 2D cut sample and is the area of the smallest possible circle that includes the precipitation. Thus a perfectly spherical precipitation will have a roundness of 1. Precipitations are classified as nodules if the roundness is between 0.625 and 1. A precipitation with a roundness between 0.525 and 0.625 is called an intermediate and if it has a roundness lower than 0.525 it is called compacted. Through equation (2) the percent nodularity of a sample may be determined.
The standard states that the conventional definition of CGI is that the percent nodularity is between 0 % and 20 %. Furthermore, no lamellar graphite is allowed in CGI.
3.1.5 Process control
From a theoretical point of view there exist several ways of controlling if the melt will obtain LGI-, SGI- or CGI-structure in room temperature. It is possible to affect the microstructure of cast iron by altering the cooling rate [7]. Another way of controlling the microstructure is to alter the chemical composition of the melt. This is the control mechanism used by for example the company SinterCast which provides equipment for controlling industrial casting processes so that CGI is formed [8].
According to König [5] it is well known that Mg (Magnesium) has a nodularising effect on a CGI melt and Fredriksson and Åkerlind [2] state that the growth of spheroidal graphite structures is encouraged by additions of Ce (Cerium) and Mg to a melt.
8 3.2 Sand mould casting
Today, there is a large flora of casting techniques available. This includes for example sand mould casting, gravity die casting, pressure casting, centrifugal casting and continuous casting [1]. In this chapter a variant of the casting technique sand mould casting is briefly presented. This is the technique used at SCANIA for casting cylinder blocks.
The main idea in this casting technique is that two mould parts, which are made of sand, are put together with cavities remaining between them. As the melt is poured into the connected mould parts, the cavities are filled with metal which will cool and solidify. Sand cores may be placed in the cavity between the mould parts. The system is visualized in Figure 2.
Figure 2. The mould parts (dark green) and sand core (light green) are assembled. Liquid metal (red) is poured in to fill in the remaining cavities. When the metal has solidified (gray), the sand is removed.
9 3.3 Analyzing the casting process
One way to try to understand why a particular cast iron part has certain mechanical properties is to assume that the properties can be derived from known parameters such as cast geometry, chemical composition of the melt, casting temperature etcetera. The flow chart in Figure 3 visualizes this theoretical approach.
Figure 3. A flow chart describing the origins of the mechanical properties in a theoretical model.
The concept described in Figure 3 assumes that input parameters determine the temperature curve during solidification and cooling. These curves in turn explain why the microstructure is the way it is. Finally, the mechanical properties are given by the microstructure.
10 3.4 Casting simulation – basic concepts
Regardless of mathematical approach, a numerical casting simulation software needs to address pre-processing, calculation and post-processing. In most cases the calculation part is made up of the three different modelling parts: mould filling, solidification and stress/strain analysis [9]. This scheme is visualized in Figure 4.
Figure 4. The basic steps in numerical casting simulation.
Casting simulation is conducted by the implementation of different numerical techniques. Examples of these include the use of FEM (Finite Element Method), FVM (Finite Volume Method) and FDM (Finite Difference Method). [10]
Brief explanations of how the three methods can be made of use in casting simulations are given below.
3.4.1 FDM in casting simulation
As a concept FDM is quite straight forward. A mesh is introduced and values of an entity are prescribed for all nodes. The governing partial differential equation (PDE) is linearized via Taylor expansions. For example, FDM can be used to predict heat flow if the governing equation is chosen to be the heat conduction equation (3). [9]
In the above equation is the time derivative of temperature, is the second spatial derivative of temperature and is a constant. The discretized version of the PDE is used to calculate the values at each node at the next time step. The method can be explicit, i.e. Euler forward, or implicit, for example Euler backwards. [9]
11 3.4.2 FVM in casting simulation
In FVM the geometry is divided into an arbitrary number of control volumes. Inside each control volume there is a grid point. As with FDM, the aim of using FVM is to assign new values for variables at the grid points (nodes) for the next time step. Just as in FDM, this is done by evaluation of differential equations. In FVM, the algorithms may be constructed so that resulting entity fields never violate the conservation principle. This means that conservation of for example energy is guaranteed over one, or a combination of several finite volume elements. [9]
Consider again the case of heat transfer. In FVM a differential equation that accounts for updating the temperature field can be derived from the heat balance equation for one control volume.
In equation (4), is the heat change of the control volume, are the rates at which heat is conducted through the N sides of the control volume and is the rate at which heat is generated within the volume. Note that this equation is an integration over the whole control volume, thus guaranteeing energy conservation. [9]
3.4.3 FEM in casting simulation
FEM can also be made of use in casting simulation. This was the case in the work of Cho et. al. [11] where the authors used FDM in combination with FEM to analyse the thermal stresses in castings. From an original FEM mesh, all the elements that represented molten iron were eliminated. The heat flow was modelled with FDM and the temperatures were mapped onto the FEM mesh nodes. As parts of the casting solidified, the mesh elements in these regions were recreated in the FEM model with temperature dependent material properties. Thus the FEM mesh had to be updated during the whole casting process.
Instead of constantly updating the geometry, huge reduction factors can be applied to elements that are considered molten. However the penalized elements may cause convergence problems during solving. [11]
3.5 Casting simulation – an example in MAGMAsoft®
MAGMAsoft® is a commercial software for casting simulating based on FVM [10]. The algorithms of the program are to a small extent visible for the user, who is however granted the possibility of manually inserting foundry specific parameters like casting temperatures and melt composition into the software database.
In this section the casting of a sphere will be examined as an example. The first step in the casting analysis is to create a geometry. This can be done in a third part CAD program, but for a simple geometry like this, the MAGMAsoft® CAD module proves sufficient. The geometry is modelled and shown in Figure 5.
12
Figure 5. The geometry defined in MAGMAsoft® where the casting of a sphere is to be simulated.
In Figure 5, the gray sphere is the actual casting, the brown volumes are the gating system which lead the melt into the sphere. The light green surrounding block represents the mould, which is made of sand. The red cylinder on top of the brown cone represents the inlet from where the melt is poured into the mould.
Next, the geometry is meshed, i.e. divided into small blocks as MAGMAsoft® is a FVM- based software [10]. This is shown in Figure 6. Please observe that Figure 6 does not show the meshed mould because it would block the view of everything else.
Figure 6. The geometry is meshed.
It is now time to define material parameters such as heat transfer coefficients, density, thermal conductivity, alloy composition and heat capacity. In this example all the material properties of the iron are taken from the MAGMAsoft® standard CGI called GJV450. The surrounding sand is taken to be MAGMAsoft® standard Green sand.
The first simulation step is the mould filling. Though not encouraging the end-user to access the actual algorithms used in calculating flow of the melt, MAGMAsoft®
acknowledges that the algorithms include steps where momentum equations (Navier- Stokes), an energy equation and a continuity equation are solved. Pressure and venting
13
of air inside the mould are also taken into account. The local viscosity of the melt is calculated as a function of the viscosity of liquid melt and the fraction of solid metal in each element. Turbulence is modelled as an increase in local viscosity. [12]
Figure 7. Pouring of melt into the mould is simulated during the mould filling step.
In Figure 7, the mould filling is illustrated. The different colors represent different local velocities of the melt. When the fraction solid metal in an element reaches a material specific value, the velocities in this element are set to zero [12].
The solidification step follows. During this step the final microstructure of the iron is calculated. In “Appendix A – Solidification and cooling process of cast iron” the solidification process of an iron melt is discussed. As mentioned in the Appendix, several phases will appear, permanently or temporarily, as the material cools to room temperature.
Figure 8. As the heat is transported away from the melt it solidifies. Yellow indicate a high temperature and blue indicate a low temperature.
Figure 8 is a snapshot of the temperature distribution at a given time in the solidification process. The version of MAGMAsoft® used in this work takes the local phase and chemical composition in the material into account. This means that the heat capacity of the material is evaluated with respect to the locally existing phase and that the chemical composition in the melt will vary locally because of segregation. [13]
Moreover, MAGMAsoft® states that the software continuously calculates the shrinkage in each metallurgical phase and sums it up to a total shrinkage. Elements
14
that have not yet completely solidified can transport liquid to even out pressure gradients caused by local shrinkage. [13]
In Figure 9, volumes where there is a risk of remaining porosities in the room temperature solid are indicated. In this example there appears to be a risk of remaining porosities in the center of the sphere, which might be expected by people with insight into casing mechanisms.
Figure 9. Volumes where there is a risk of remaining porosities at room temperature are indicated by dark blue areas.
When the room temperature microstructure has been calculated, the simulation software will give estimations of local mechanical properties like Young’s modulus, minimum tensile strength and minimum yield strength. The calculations of some of these properties are discussed below.
3.5.1 Young’s modulus
MAGMAsoft® states that for CGI the Young’s modulus is calculated from the nodularity [14]. Sjögren et. al. [15] suggest that at least for a while equations (5) were implemented in MAGMAsoft® to predict the local Young’s modulus in a CGI casting.
In equations (5), is the elastic modulus of the CGI, is the elastic modulus of the matrix and is the elastic modulus of the graphite. The ratio is the relation
15
between length and width of the ellipsoid graphite particles and is an orientation parameter that describes the angle of the ellipsoid axis relative to the stress direction.
For randomly oriented particles . The volume fraction of graphite inclusions is denoted . The module is taken to depend on the nodularity as presented in equation (6).
The relationship between and nodularity was established empirically by Sjögren by calculation of in equation (5) for cases were all other parameters were known. [16]
The equations (5) have been found to give satisfactory predictions of the Young’s modulus in cast iron with graphite inclusions [17], where a good agreement between experimental and calculated values was obtained.
3.5.2 Hardness
In the calculation of hardness in CGI the cooling rate at solid state transformation, melt composition and nodularity are considered. [14]
For LGI, the hardness is calculated from several parameters, including the amount of primary austenite. The distance between the layers of ferrite and cementite in the pearlite matrix is also considered, as well as the cooling rate during solid state transformation which affects the hardness of pearlite. [13]
Solution hardening is a phenomenon that occurs when alloying elements interrupt the lattice of a material. The irregularity in the lattice will block dislocation movements and thereby make the material harder. The solution hardening of ferrite in the pearlite matrix caused by alloys in the melt is also taken into account in the calculation of hardness in LGI [13].
Solution hardening is also considered for SGI. For this type of iron the hardness is calculated by equation (7):
In the above equation denotes ferrite, is the Brinell hardness and denotes the fraction of different phases. and are constants whereas and
are functions of the melt composition. [13]
16 3.6 Earlier similar works at SCANIA
Several works already exist were predictions from MAGMAsoft® have been compared with the real products cast at the SCANIA foundry. In one of these [18], Eduardo measured the temperature in a cylinder block, sand cores and mould sand during casting. The material used in Eduardo’s experiment was LGI and the measurements continued until the iron had cooled to approximately ca 400°C. Eduardo used thermocouples to measure the temperature during casting and cooling. These probes were place in the cylinder block before the casting started and remained there until the measurements ended. The correlation between reality and simulation was found to be high for all the measurement points in the iron and inside the sand cores.
In another work [19] an alternative gating system was developed for a cylinder block cast in LGI with the aid of MAGMAsoft®. It is worth noticing that the authors also constructed a simple test geometry that they cast and simulated. Despite that the geometry was simple, MAGMAsoft® predicted that the whole component should solidify as LGI, whereas studies of the microstructure in the real casting showed that it in large parts had solidified as white iron. White iron is a harder and more brittle type of iron.
In a thesis from 1994 [20] Per Erikson studied the significance of the heat transfer coefficient (HTC) between iron and surrounding sand. One of the conclusions he reached was that different HTCs should be used for different stages during casting.
During the pouring stage the HTC should be constant. For the remaining stages of the casting process a temperature dependent HTC should be used. The HTC suggested by Per is the following:
In equation (8) is an increment in time, is the metal volume corresponding to one metal-sand interface, is the density of the metal, is the latent heat of the metal per kg, is an increment in the solidification level of the metal, is the specific heat of the metal, is an increment in temperature, is the area of the interface between the metal and the sand, is the temperature of the metal surface,
is the temperature of the sand surface and is the HTC.
In a technical report from 2009 [21] Henrik Pettersson found that there was a linear correlation between tensile strength and hardness in LGI2. A somewhat weaker linear correlation was found between tensile strength and average graphite precipitation size.
2 The fact that there is a positive correlation between tensile strength and hardness also for CGI is known from reference literature, for example [30].
17
4 Method
This chapter describes how temperature measurements were conducted at the SCANIA foundry in Södertälje and how the temperature curves subsequently were analyzed. Test procedure in both tension and compression are also described.
Furthermore, the methods of simulation in MAGMAsoft® are discussed together with a note on how a goodness-of-fit value is calculated.
4.1 Temperature measurements during casting
The measurements took place in March 2014. The experimental setup, equipment specifications, measurement positions and temperature curve evaluation methods are explained in this chapter.
4.1.1 Setup
The temperature was measured using type-N thermocouples (heat sensors). The thermocouples were connected to a logger that in turn was connected to a portable computer. See Figure 10.
Figure 10. From left to right: thermocouple, logger and portable computer.
The tips of the thermocouples placed in slowly solidifying areas of the cylinder block were protected by protective tubes, i.e. semi-open hollow cylinders. The protective tubes were open at one end and closed at the other. This is shown in Figure 11.
Figure 11. From top down: Thermocouple tip, protective tube and protected thermocouple tip.
18
To ensure that the protective semi-open cylinders did not cause measurement disturbances, a reference test was conducted. This test is presented in “Appendix B – Effect of added thermocouple tip protection”.
The temperature was monitored in the first and last cylinder block cast in a batch of five blocks. The first block is denoted block A and the last block is denoted block B. The five blocks including the two blocks with thermocouples are confined in the containers shown in Figure 12.
Figure 12. The temperature was monitored in the first and last cylinder block in a batch of five blocks.
4.1.2 Equipment specifications
The thermocouples used were made by Pentronics and of type N, i.e. Nicrosil-Nisil conductors. All the thermocouples were fully mantled with Inconel600 and had a diameter of 1.5 mm. The protective semi-open cylinders were made of Inconel and had an inner/outer diameter of 2 mm/4 mm. The specified maximum measurement temperature for this type of thermocouple is 1200°C. The thermocouples are specified to have a tolerance of or less for all temperatures between -40°C and 1000°C.
Two loggers were used for block A and B respectively. One was an INTAB AAC-2 and the other was an INTAB PC-logger 3100i.
19 4.1.3 Positioning of the thermocouples
In total 24 thermocouples were placed in the two cylinder blocks, 12 in each. The positions of the thermocouples are denoted C1-C12. The positions are presented in Table 2.
Table 2. Thermocouple positions.
Thermocouple Metal or sand Position
C1 Sand Thin core, 2 cm from metal
C2 Sand Thin core, 1 cm from metal
C3 Metal Thin metal
C4 Metal Thick metal
C5 Metal Thick metal
C6 Metal Thick metal
C7 Sand Thick core, camshaft
C8 Metal Close to metal surface
C9 Metal Close to metal surface
C10 Sand Mould sand, 15 cm from metal
C11 Sand Mould sand, 10 cm from metal
C12 Sand Mould sand, 5 cm from metal
A detailed description of the position of all the thermocouples in cylinder block A and cylinder block B is given in “Appendix C – Positioning of the thermocouples”.
4.1.4 Evaluation of obtained temperature curves
From the measured temperature curves in metal thermocouple positions the liquidus temperature, solidus temperature, solidification time, cooling rate at 700°C, temperature at start of solid state transformation, temperature at end of solid state transformation and the time when all phase transformations have occurred were extracted in a master thesis by Elin Nährström [22].
Simulated temperature curves from MAGMAsoft® were used to extract simulated values of the same as well as other quantities. The liquidus temperature was taken to be the temperature before the eutectic temperature where the behavior of the temperature time derivative changed. This is characterized as a kink in the second time derivative, as shown in Figure 13. This change in temperature derivative was assumed to be a consequence of latent heat released at the start of primary austenite precipitation.
The solidus temperature was taken to be the temperature corresponding to the local minimum in temperature time derivative after the eutectic temperature. This is also visualized in Figure 13. The solidification time was taken to be the time elapsed between the time corresponding to the liquidus temperature and the time corresponding to the solidus temperature.
20
Figure 13. The temperature and it’s time derivatives. The intersection lines indicate the times tL and tS at which the liquidus and solidus temperatures were evaluated. The time between times tL and tS is the solidification time.
The temperature curve displays a similar behavior at the solid state transformation temperature. From this part of the curve the temperatures at start and end of solid state transformation can be extracted together with the time for the solid state transformation. This is shown in Figure 14.
Figure 14. The temperature at tPS is the temperature at the start of solid state transformation and tPE is the temperature at the end of the solid state transformation. The time between tPS and tPE is the solid state transformation time.
After the solid state transformation there is a local minimum in the time derivative of temperature, similar to the dip at time tS in Figure 13. For the MAGMAsoft®
temperature curves this dip was assumed to correspond to the time for which all
21
phase transformations had occurred. The time elapsed between the time corresponding to the local minimum in temperature derivative after the temperature of solid state transformation and the time corresponding to the liquidus temperature is called the time when all phase transformations have occurred. This is illustrated in Figure 15.
Figure 15. The time when all phase transformation have occurred is the time between tPE and tL.
22
4.2 Uniaxial tensile and compression tests
From the cylinder blocks with thermocouples (number one and five in the cast batch, also denoted block A and block B) and from the middle one (number three), cylinder walls were cut out. The cylinder walls are shown in Figure 16.
Figure 16. The cylinder walls (pink) are numbered from the front of the cylinder block to the rear.
Test pieces were in turn machined from some of these walls according to Figures 17- 19. The test pieces C4, C5, C6 and C9 were placed in accordance with the thermocouple positions presented in Appendix C, though some of these test pieces were machined from different cylinder walls than indicated in Appendix C. The exact positions of each individual test piece is given in Figure 17-19.
The positions of R, L and U are according to the appropriate SCANIA technical regulation [23]. Û is placed the same distance from the center line as U but on the left instead of on the right side. Because of geometric limitations test pieces for positions C3 and C8 were not taken from the thermocouple positions but from volumes with similar cooling and solidification conditions according to a MAGMAsoft® simulation.
These representative positions will be denoted C3t and C8t.
23
Figure 17. Test pieces taken from cylinder block 1. Each blue structure represents a tensile test piece.
24
Figure 18. Test pieces taken from cylinder block 3. Each orange structures represent 3 compression tests.
25
Figure 19. Test pieces taken from cylinder block 5.
26
The dimensions of the tensile and compression test pieces are given in Figure 20.
Figure 20. From left to right: dimensions of tensile and compressive test pieces.
The test pieces prepared for tensile testing followed the size regulations stated by SS- EN ISO 6892-1:2009 [24]. The ends of the tensile test pieces were threaded, size M14.
A clamping extensometer of length 25 mm was used to register the strain. The tensile tests were conducted using a constant crosshead separation rate of 0.006 mm/s, which gives a constant estimated strain rate over the parallel length of 0.00015 s-1. The Young’s modulus was evaluated between 50 and 100 MPa and the tensile strength of the material, Rm, was taken to be the maximum stress measured during the test.
The proof strength was defined as Rp0.2, i.e. the stress for which the remaining plastic strain would be 0.2 % if the specimen was unloaded elastically at that instant.
A typical stress-strain curve analysed in this work is shown in Figure 21. The values of Rm and Rp0.2 are indicated. The slope of the two blue lines originating from the strain axis is the Young’s modulus. It is worth pointing out that the value of Young’s modulus is highly dependent on the interval in which it is evaluated. This is also shown in Figure 21.
Figure 21. Left: a typical stress strain curve evaluated in this work is shown. The slope of the blue lines originating from the strain axis is the Young’s modulus. Right: the value of the Young’s modulus depends on where in the stress strain curve it is evaluated.
27
The compression test pieces were of the same diameter as the tensile test pieces. As for the tensile tests, the elastic modulus was evaluated between 50 and 100 MPa from the stress strain curves. The compressive test pieces were compressed at a rate of 0.007 mm/s. A value corresponding to the tensile test value Rp0.2 was extracted at 0.2
% plastic compression. This value is called Rc0.2. 4.3 Simulation in MAGMAsoft®
The geometry of the cylinder block was imported into MAGMAsoft®. This geometry is shown in Figure 22.
Figure 22. The cylinder block as it is cast.
The gray volume is the actual casting, the brown parts are the gating systems while the light green and dark green volumes represent sand mould and sand cores. A convergence study was carried out in order to verify that the standard mesh consisting of 34 million elements was sufficient. This study is presented in “Appendix D – Mesh convergence study”.
28
4.3.1 Simplified model of cylinder block geometry
A simpler model of the geometry was developed and used for rough, fast calculations.
This geometry is shown in Figure 23, together with the original geometry.
Figure 23. A simplified model of the cylinder block was created in order to reduce computational time.
The simplified model consisted of about 2 million elements and was mainly used to evaluate the impact of different sets of sand data. Care was taken to preserve the same volume of metal in the simplified geometry as in the original geometry. As can be seen from Figure 24, the accuracy of the simplified model was evaluated against the main geometry model consisting of 34 million elements.
Figure 24. The simplified geometry was evaluated against the original geometry at position C10.
As much of the work with the simplified model consisted of evaluation of different sand data sets, the simplified model was optimized to produce a temperature curve as close to the original model as possible at position C10. If not specified, simulated values are obtained from the standard model.
4.4 Comparing results
In the result and discussion section values are calculated in order to quantify how well data points fit a model . The R2 value is defined in equation (9).
0 2 4 6 8 10
0 100 200 300 400
C10 - Mould sand, 15 cm from metal
Time [hours]
Temperature [°C]
Original geometry 34 million cells Simplified geometry 2 million cells
29
5 Results and discussion
In this section general cast batch data is presented. After this, three main subchapters describe and discuss in order: experimental results, simulated results and comparison between experimental and simulated results.
In each subchapter results are presented in the following order: results concerning temperature curves, results concerning microstructure and results concerning mechanical properties. This order is the same as suggested by Figure 3.
5.1 Cast batch data
Cylinder block A was poured in 22 seconds while block B was poured in 23 s. The time between the end of pouring of block A and start of pouring of block B was 105 seconds. During this time the three middle blocks were cast. The pouring temperature of block A was 1434 °C. The pouring temperature of block B was not measured.
30 5.2 Experimental results
In this section experimental results are presented. Results from temperature curves, microstructure and mechanical properties are presented.
5.2.1 Temperature curves
The positions of all the thermocouples are thoroughly presented in “Appendix C – Positioning of the thermocouples”. In Figure 25 the original temperature curves for all thermocouple positions, for cylinder block A and cylinder block B, are presented.
0 2 4 6 8 10 12
0 100 200 300 400 500 600 700 800 900 1000
C1 - Thin sand core, 2 cm from melt
Time [hours]
Temperature [°C]
Block A Block B
0 2 4 6 8 10 12
0 100 200 300 400 500 600 700 800 900 1000
C2 - Thin sand core, 1 cm from melt
Time [hours]
Temperature [°C]
Block A Block B
0 2 4 6 8 10 12
0 200 400 600 800 1000 1200 1400
C3 - Thin metal
Time [hours]
Temperature [°C]
Block A Block B
0 2 4 6 8 10 12
0 200 400 600 800 1000 1200 1400
C4 - Thick metal
Time [hours]
Temperature [°C]
Block A Block B
0 2 4 6 8 10 12
0 200 400 600 800 1000 1200 1400
C5 - Thick metal
Time [hours]
Temperature [°C]
Block A Block B
0 2 4 6 8 10 12
0 200 400 600 800 1000 1200 1400
C6 - Thick metal
Time [hours]
Temperature [°C]
Block A Block B
31
Figure 25. Measured temperatures for all positions for cylinder block A and cylinder block B.
0 2 4 6 8 10 12
0 100 200 300 400 500 600 700 800 900 1000
Time [hours]
Temperature [°C]
C7 - Thick core, camshaft Block A Block B
0 2 4 6 8 10 12
0 200 400 600 800 1000 1200 1400
C8 - Close to metal surface
Time [hours]
Temperature [°C]
Block A
0 2 4 6 8 10 12
0 200 400 600 800 1000 1200 1400
C9 - Close to metal surface
Time [hours]
Temperature [°C]
Block A Block B
0 2 4 6 8 10 12
0 100 200 300 400 500
C10 - Mould sand, 15 cm from metal
Time [hours]
Temperature [°C]
Block A Block B
0 2 4 6 8 10 12
0 100 200 300 400 500
Time [hours]
Temperature [°C]
C11 - Mould sand, 10 cm from metal Block A Block B
0 2 4 6 8 10 12
0 100 200 300 400 500
Time [hours]
Temperature [°C]
C12 - Mould sand, 5 cm from metal Block A Block B
32
In general the measurements agree very well. Block A has slightly higher temperatures in some positions, possibly indicating higher heat content in the block due to the higher casting temperature.
There was a malfunction in the thermocouple at position C8 in cylinder block B. The measurements at position C10-C12 in block A were interrupted and are unusable after 6 hours.
In “Appendix E – Comparison between temperatures in CGI and LGI” the measured temperatures are compared to the temperatures measured for the same cylinder block geometry cast in LGI. The temperatures measured in LGI are obtained from a previous thesis by Eduardo [18].
The temperature curves were analyzed by Nährström [22] who found the characteristics shown in the Tables 3 and 4. These values are used in some of the following result sections.
Table 3. Temperature curve characteristics for block A.
C3 C4 C5 C6 C8 C9
Solidification time [s] 124 960 916 388 190 217
Liquidus temperature [°C] 1142 1154 1153 1147 1174 1148 Solidus temperature [°C] 1096 1069 1110 1110 1122 1127 Solid state transformation time [s] 3335 3035 4620 1766 642 1604 Temperature at start of solid state
transformation [°C] 734 743 750 740 734 734
Temperature at end of solid state
transformation [°C] 680 709 707 716 702 715
Cooling rate at 700 [°C/min] -0.98 -1.042 -1.5 -0.94 -2.22 -0.92 Time when all phase trans-
formations have occurred [s] 7120 7634 10036 5312 2912 5150
Table 4. Temperature curve characteristic for block B.
C3 C4 C5 C6 C8 C9
Solidification time [s] 128 n/a n/a n/a n/a 216
Liquidus temperature [°C] 1142 n/a n/a n/a n/a 1151
Solidus temperature [°C] 1096 n/a n/a n/a n/a 1116
Solid state transformation time [s] 3080 2726 4488 1796 n/a 1460 Temperature at start of solid state
transformation [°C] 734 743 750 740 734 730
Temperature at end of solid state
transformation [°C] 680 709 707 716 702 713
Cooling rate at 700 [°C/min] -0.92 -1.133 -1.39 -1 n/a -1.07 Time when all phase trans-
formations have occurred [s] 5960 n/a n/a n/a n/a 4814
33
Some of the values of temperature characteristics for cylinder block B could not be obtained due to temporary disturbances in the measurements.
5.2.2 Microstructure
In Table 5 values of percentage nodularity are presented. These values were obtained by Nährström [22] from the same experiment as this work is based on. These values are used in some of the following result sections.
Table 5. Values of percentage nodularity.
C3 C4 C5 C6 C8 C9
Nodularity, block A [%] n/a 12 12 21 13 12
Nodularity, block B [%] n/a 13 16 21 18 15
The values of nodularity in Table 5 are the values obtained from the actual thermocouple positions. Additional values of nodularity for some of the tensile test pieces were also obtained from Nährström. These are presented in Figure 40.
5.2.3 Mechanical properties
The tensile strength was evaluated as a function of solidification time and solid state transformation time of block A. This is shown in Figure 26.
Figure 26. Tensile strength as function of solidification time and solid state transformation time respectively.
As no tensile test specimens could be taken from the very positions where the thermocouples had been placed, the origin of each tensile test has been indicated in Figure 26. Similar position means that the test piece was taken from another cylinder wall than the temperature was measured in (C4,C5,C6 and C9) or from a representative position (C3 and C8)3. Same position means that the test piece was taken from the same position as the temperature was measured in. Same block means
3 This means that values of solidification time were taken from positions C3 and C8 while corresponding tensile test pieces were taken from positions C3t and C8t.
34
that the tensile test was taken from the same cylinder block as the temperature was measured in, i.e. block A.
It is generally accepted that faster processes will result in a less coarse microstructure which in turn will give a harder and stronger material.
Fredriksson and Åkerlind [2] state that the eutectic growth rate increases with a larger under-cooling in cast materials. The authors also state that the eutectic lamella distance decreases with higher growth rate. Thus a large under-cooling (fast process) will promote a high growth rate which will give a finer structure.
The idea that a material with a finer microstructure will be harder and stronger is manifested in the Hall-Petch equation [25], which tells that yield strength of a material increases with smaller grain size. It is also pointed out by Fredriksson and Åkerlind [2]
that a material with finer structure will have better mechanical properties.
The results presented in Figure 26 might suggest that a faster process gives a stronger material also in these experiments, especially if the deviating results for position C6 can be explained away. At the moment, no such explanation has been found.
Results from the compressive and tensile tests are presented in Table 6.
Table 6. Elastic modulus and stress corresponding to 0.2% plastic deformation for tensile and compressive tests.
Mean values and standard deviations are give.
Position UL tensile
Position UL compressive
Position L tensile
Position L compressive Elastic modulus [GPa] 142.1 ± 1.91 151.6 ± 8.39 145.0 ± 5.66 150.9 ± 10.3 Rp 0.2 / Rc 0.2 [MPa] 317.6 ± 3.73 382.2 ± 4.64 325.2 ± 3.47 394.8 ± 3.13 Tensile / compression
strength [MPa]
436.4 ± 8.64 1471 ± 50.8 441.6 ± 5.65 1429 ± 95.4
For the elastic modulus and Rp 0.2 / Rc 0.2, ten values were taken from UL tensile positions, nine values4 were taken from L tensile positions. Six values were taken from UL and L compressive positions respectively.
For the tensile / compressive strength, ten values were taken from UL tensile positions, nine values5 were taken from L tensile positions. Four values6 were taken from UL and L compression test pieces respectively. The measured values are graphically presented in Figure 27 and Figure 28.
4 A disrupted stress-strain curve caused the values from one L tensile test to be considered an outlier.
5 A disrupted stress-strain curve caused the values from one L tensile test to be considered an outlier.
6 Not all compression test pieces were compressed until destruction.
35
Figure 27. Values obtained from tensile and compressive testing.
Figure 28. Elastic modulus evaluated from tensile and compressive tests.
While the elastic modulus is about the same in tensions as in compression, it is noteworthy that the Rc0.2 is somewhat higher than the Rp0.2 value. The mean Rc0.2 values are 20% and 21% higher than the mean Rp0.2 values for the two positions UL and L respectively. The mean compressive strength is 3.4 and 3.2 times higher than the mean tensile strength for positions UL and L respectively.
UL L
0 100 200 300 400 500 600
Rp0.2 / Rc0.2
Position
Rp0.2 / Rc0.2 [MPa]
Tensile Compressive
UL L
0 500 1000 1500 2000
Tensile / compressive strength
Position
Tensile / comressive strength [MPa]
Tensile Compressive
UL L
100 120 140 160 180 200
Elastic modulus
Position
E [GPa]
Tensile Compressive
36 5.3 Results from simulation
This section reports results from attempts to empirically establish relations between output parameters in MAGMAsoft®. This is done in order to get a better understanding of the assumptions and relations used in the simulation program. Insight into these relations will give the MAGMAsoft® user a better chance of successfully integrating experimentally found relations into the software.
In Figure 29, the liquidus and solidus temperatures evaluated from the MAGMAsoft®
temperature curves are plotted against the liquidus to solidus time reported by a MAGMAsoft® simulation.
Figure 29. The liquidus and solidus temperatures calculated from MAGMAsoft® simulated temperature curves as functions of liquidus to solidus time calculated by MAGMAsoft®.
It was suspected that the liquidus and solidus temperatures were connected to the liquidus to solidus time. The low R2 values for the adapted models (red lines) suggest that a linear model of describing the liquidus or solidus temperatures as functions of liquidus to solidus time is a rather bad model. Intuitively, the red line in the left graph in Figure 29 might seem like a good fit. However it is a consequence of the definition of the R2 value (refer equation (9)) that a linear model with a derivative close to zero will give a low R2 value.
MAGMAsoft® calculates a parameter called pearlitic hardness, which is a Brinell hardness assuming that all matrix is pearlite [26]. In Figure 30, this parameter is plotted against the time when all phase transformations have occurred evaluated from MAGMAsoft® temperature curves. Alongside this graph in Figure 30, another graph is plotted that shows the hardness as a function of the time between end of solidification and start of solid state transformation obtained from simulated temperature curves.
0 200 400 600 800 1000 1 150
1175 1 200 1225 1 250
Liquidus temperature as function of liq to sol time
Time [s]
Temperature [°C]
y = - 0.000753*x + 1170 R2 = 0.0063
0 200 400 600 800 1000 1 050
1075 1 100 1125 1150
Solidus temperature as function of liq to sol time
Time [s]
Temperature [°C]
y = - 0.0109*x + 1110 R2 = 0.0415
37
Figure 30. Hardness is plotted as function of the time when all phase transformations have occurred and the time between end of solidification and start of solid state transformation.
Figure 30 shows that there is a relation between the evaluated time entities and the estimated hardness of the material. The trend is that a faster solidification/cooling process is associated with a higher predicted hardness. This might be expected since a finer microstructure created by a faster process generally gives a harder material than a more coarse microstructure created by a slower process (refer also to page 34).
In Figure 31, the hardness is evaluated against the temperatures at start and end of solid state transformation.
Figure 31. The hardness is plotted against the temperatures at start and end of solid state transformation.
There is no obvious relation between the parameters evaluated in the graphs in Figure 31. A logarithmic function might be fitted to describe the relation between hardness and temperature at end of solid state transformation if the lower left value is considered an outlier. To consider a simulated value an outlier is however most likely not a good idea as measurement uncertainty does not exist in the same sense as in experiments.
In Figure 32 the hardness is evaluated as a function of the cooling rate at start of solid state transformation. It appears that a logarithmic function can describe the relation quite well. Once again a faster process results in a higher simulated hardness.
0 5000 10000 15000
160 180 200 220 240
Hardness as function of time when all phase transformations have occurred
Time [s]
Hardness [HB]
y = - 0.00414*x + 230 R2 = 0.843
0 2000 4000 6000 8000
160 180 200 220 240
Hardness as function of time between end of solidification and start of solid state transformation
Time [s]
Hardness [HB]
y = - 0.00666*x + 229 R2 = 0.805
710 720 730 740
180 190 200 210 220 230
Hardness as function of
temperature at start of solid state transformation.
Temperature [°C]
Hardness [HB]
y = - 1.1*x + 1000 R2 = 0.517
685 690 695 700 705
180 190 200 210 220 230
Hardness as function of
temperature at end of solid state transformation
Temperature [°C]
Hardness [HB]
