I. INTRODUCTION

Good mechanical, physical and manufacturing properties as well as the relatively low production costs lead to a common interest in cast irons for various technical applications [1]. Due to the combination of a relatively high tensile strength and an adequate ductility nodular cast iron is often used for highly stressed components in the automotive and commercial vehicle industry, e.g. for crankcases, crankshafts or exhaust manifolds, as well as in the wind power industry, e.g. for rotor hubs or nacelles [2][3][4][5]. An appropriate material selection and a weight-optimised dimensioning are essential for safe and economic operation conditions of such cast iron components. Therefore, the characterisation of the fatigue behaviour is of major importance. In the literature, fracture-mechanic investigations [6][7][8] as well as fatigue tests with mechanical stress-strain hysteresis [9] and in some cases temperature measurements [10] are described for nodular cast iron [11].

In the scope of the current paper stress-controlled fatigue tests were performed with specimens of the nodular cast iron ASTM 80-55-06 (EN-GJS-600) at ambient temperature on servohydraulic testing systems with a testing frequency of 5 Hz and on resonators with a testing frequency of 150 Hz. For a detailed microstructural-based characterisation of the fatigue behaviour, high-precision mechanical stress-strain hysteresis (f = 5 Hz), temperature, electrical resistance and frequency (f = 150 Hz) measurements were applied. The plastic strain amplitude ( a,p ) as well as the change in temperature (T), electrical resistance (R) and frequency (f) were plotted versus the number of cycles in cyclic deformation, cyclic temperature, cyclic electrical resistance and cyclic frequency curves. All measured quantities are based on microstructural changes due to fatigue processes in the plastically deformed volume of the gauge length of the specimens, e.g. deformation-induced matrix debonding of the graphite. On the basis of the measured temperature data a linear relation between the T values at f = 5 Hz and f = 150 Hz was found. For both testing frequencies the Woehler (S-N) curves were determined. In addition, fracture surfaces were investigated by using scanning electron microscopy (SEM).

II. INVESTIGATED CAST IRON

The cast iron ASTM 80-55-06 (EN-GJS-600) was provided by the Daimler AG in round bars with a diameter of 36 mm and a length of 300 mm. Brinell hardness measurements yield values of 235±6 HBW30. Figure 1 shows characteristic light (a) and scanning electron (b) micrographs of the investigated material. The microstructure consists of a pearlitic matrix with a ferrite content of 14.6±2 area-% and a graphite content of 9.8±0.8 area-%. As can be seen in Figure 1, the formation of ferrite is predominantly observed in ferrite zones located in the surrounding of the graphite nodules. The mean diameter of the graphite nodules was determined to 19.1±6.8 m. Table I summarises the microstructural parameters of the cast iron. In order to evaluate the hardness of the pearlite, ferrite and graphite fraction a Martens micro-hardness pattern with a width of 200 x 200 m was measured on a polished crosssection (Figure 2). The load of 0.1 N was applied without a hold time within 10 s. Afterwards, the microstructure was etched and correlated with the results of the micro-hardness measurements. The micro-hardness values significantly decrease in the sequence pearlite (3095 HM0.1), ferrite (1757 HM0.1) and graphite (597 HM0.1). Corresponding to the phase distribution, a mean micro-hardness value of 2654 HM0.1 can be calculated by multiplying the pearlite (ferrite, graphite) content with the appropriate micro-hardness value of 3095 (1757, 597) HM0.1. It is possible to revalue the Martens hardness to Vickers hardness and a revaluation according to [12] yields about 240 HV0.1. This value shows a good accordance with the Brinell hardness determined to be 235±6 HBW30. During the fatigue tests, the plastic strain amplitude  a,p [13], the change in temperature T [14] and the change in electrical resistance R [15] were measured to characterise the microstructure-based fatigue behaviour in detail. The physical quantities  a,p , T and R are a function of the deformation induced change in microstructure in the bulk of the specimen and represent the actual fatigue state [16][17][18][19]. For the measurement of  a,p an extensometer was fixed in the middle of the gauge length. The change in temperature T was detected with one thermocouple in the middle of the gauge length (T 1 ) and two thermocouples at the elastically loaded specimen shafts (T 2 and T 3 ). For electrical resistance measurements a DC-power supply was fixed at both shafts and R was measured with two wires spot welded at the transition of the gauge length and the shafts (Figure 4). Apart from the geometry, the change in electrical resistance R strongly depends on the resistivity  * which is directly related to deformation induced changes in the microstructure, e.g. dislocation density and arrangement, vacancies, micro-pinholes, micro-shrinkage cavities or micro-cracks. In the case of cast irons the measured value R is of major importance, in particular to get detailed information about the actual fatigue state with respect to damage mechanisms like graphite-matrix debonding.

Furthermore, stress-controlled CATs were performed at ambient temperature at a testing frequency of 150 Hz on a resonator using a sinusoidal load-time function at a load ratio of R = -1 until specimen failure or to a maximum number of cycles N max of 2•10 7 . In addition to the above mentioned physical quantities, the change in frequency f of the electromagnetical resonance device can be used for the characterisation of the fatigue behaviour. Caused by the functional principle of a resonator the specimen is part of the spring-mass-system of the testing setup and the measured value f depends on the damping behaviour of the specimen which depends on the deformation induced changes in the microstructure during fatigue loading [20].

IV. RESULTS

A. Load Increase Tests

Load increase tests (LITs) allow a reliable estimation of the endurance limit with one single specimen related to a maximum number of cycles N max = 2•10 6 . In Figure 5, besides the stress amplitude  a , the plastic strain amplitude  a,p , the change in temperature T and the change in electrical resistance R are plotted versus the number of cycles N for a LIT with the cast iron ASTM 80-55-06 (EN-GJS-600). The R-N curve indicates an initial decrease, among others caused by closing micro-cracks during the compression half cycles. Then, the courses of the change in electrical resistance R are characterised by a saturation state between 6•10 4 and 1•10 5 cycles, followed by an increase indicating cumulative graphite-matrix debonding and micro-crack growth. A significant change in the slope of the  a,p -N, T-N and R-N curve of the LIT occurs at  RW, LIT = 220 MPa. This stress amplitude can be used for the estimation of the endurance limit [14]. The stress amplitude 398 MPa leads to specimen failure.

B. Constant Amplitude Tests

Constant amplitude tests (CATs) were performed with stress amplitudes in the range of 220 ≤  a ≤ 340 MPa at a testing frequency of 5 Hz. In Figure 6a, the plastic strain amplitude is plotted versus the numbers of cycles. The  a,p -N curves immediately indicate plastic strain amplitudes between 0.015•10 -3 for  a = 220 MPa and 0.12•10 -3 for  a = 340 MPa followed by cyclic hardening processes in the matrix until macroscopic crack growth. After an initial increase caused by thermal conduction, the T-N curves shown in Figure 6b The CATs performed with stress amplitudes in the range 240 ≤  a ≤ 340 MPa lead to numbers of cycles to failure between 2•10 4 and 1•10 6 whereas the CAT performed with  a = 220 MPa reaches 2•10 6 cycles without failure. The stress amplitude of 220 MPa corresponds very well to  RW, LIT = 220 MPa estimated in the LIT. This underlines the high capability of LITs for a reliable estimation of the endurance limit of cast irons with one single specimen. Thus, this test procedure yields large economic advantages.

To evaluate the influence of the testing frequency on the cyclic deformation behaviour, CATs were performed in the range of 190 ≤  a ≤ 280 MPa at a starting testing frequency of 150 Hz. In Figure 7 10. The ratio T 5 Hz /T 150 Hz determined to 0.039 corresponds very well with the ratio between the testing frequencies of 0.033. This illustrates that a higher testing frequency results in higher heat dissipation per second and unit of the plastically deformed volume in the gauge length of the specimen. The increase of the temperature is a function of decreasing values of the plastic strain amplitude [22,23] and the increased heat dissipation per second for increasing testing frequencies. Figure 11 shows the Woehler (S-N) curves for both testing frequencies. As can be seen e.g. in [22], provided that the testing conditions are identical, an increase in the testing frequency generally results in longer lifetimes. To get more information about the fracture mechanisms fracture surfaces were investigated with scanning electron microscopy (SEM). Independent of the stress amplitude and the testing frequency the initiation of fatigue cracks can be often observed at defects, like micro-pinholes or microshrinkage cavities. Figure 12a shows the fatigue crack initiation at a defect in the centre of the fracture surface of a specimen, which was loaded with  a = 300 MPa at a testing frequency of 5 Hz and reached a number of cycles to failure of 4.1•10 5 . In Figure 12b, the striation area is shown in detail. The final fracture surface, which is shown in Figure 12c V. SUMMARY The present paper is focused on the cast iron ASTM 80-55-06 (EN-GJS-600). The microstructure predominantly consists of a pearlitic matrix with nodular graphite. In load increase tests at a testing frequency of 5 Hz the endurance limit of the investigated cast iron can be estimated with one single specimen. The cyclic deformation behaviour under constant amplitude loading was characterised at the testing frequencies f = 5 Hz and f = 150 Hz on the basis of the plastic strain amplitude (f=5Hz) as well as the change in temperature, electrical resistance and frequency (f=150 Hz). At both testing frequencies the cyclic deformation behaviour is dominated by cyclic hardening of the matrix and graphite-matrix debonding until macroscopic crack initiation. In the constant amplitude tests at f = 150 Hz higher values for the change in temperature were measured caused by the higher energy dissipation per second in the plastically deformed gauge length compared to fatigue tests with f = 5 Hz. With respect to the change in temperature T, a liner relation between the T values at f = 150 Hz and f = 5 Hz can be determined. The fatigue strength at f = 5 Hz is slightly higher than at f = 150 Hz because of the different load time functions, in particular triangular for f = 5 Hz and sinusoidal for f = 150 Hz.