Issue34

P. Gallo et alii, Frattura ed Integrità Strutturale, 34 (2015) 180-189; DOI: 10.3221/IGF-ESIS.34.19 184 Fatigue equipment and testing procedures The fatigue tests are conducted on a servo-hydraulic MTS 810 test system with a load cell capacity of 250 kN. The system is provided with a MTS Model 653 High Temperature Furnace. It is ideal for a wide variety of high-temperature tests, including tension, compression, bending and fatigue testing of different materials, metallic and not. The furnace includes the MTS digital PID Temperature Control System and is configured for two heating zones which can be independently temperature-controlled through high precision thermocouples. The nominal temperature ranges from 100°C to 1400°C and the control point stability is about ± 1°C. Since the wedge grips are affected by the heat of the furnace, they are equipped with a cooling system that keeps the temperature low enough in order to not provoke any damage to the test- instruments. In order to completely characterize the high temperature behavior of the considered alloy, firstly load-controlled fatigue tests were carried out at different temperatures. More precisely, the hour-glass shaped specimens were tested at room temperature, 360°C and 650°C; the V-notched specimens were tested at room temperature, 360°C, 500°C and 650°C. The specimen was heated to reach the desired temperature and after a short waiting period (10 minutes) necessary to assure a uniform temperature, the test was started. The temperature was maintained constant until specimen failures thank to the PID temperature control system. The uniaxial tensile fatigue tests were carried out over a range of cyclic stresses at the constant frequency of 5 Hz; the nominal load ratio R was kept constant and equal to 0. Regarding, instead, the investigation on the influence of the surface roughness on the crack initiation, the plates with central hole were tested only at the service temperature of 650°C. Because of the available equipment, it was not possible to monitor continuously, in real time, the specimen hole. For this reason, an alternative procedure has been adopted for the cracks detection: once reached a specific number of cycles, the test was temporarily stopped and the specimen checked through an optical microscope with the aim to detect any sign of cracks initiation. This operation was repeated until a crack was detected. The intervals, at which the tests were paused, were smaller as the number of cycles increasing (e.g. defining intervals 2500 N length, where necessary). After some calibration trials, a good reliable procedure was defined, especially for high number of cycles (i.e. for low loads) that are the most interesting for the final application. The values of the stiffness, registered in real-time by the machine, also helped to define the procedure and the crack detection: the experimental evidences shown a significant drop of the stiffness as one or more cracks initiated. For this reason, that variable was very useful as a kind of warning that something was happened. Once detected stiffness variation, in fact, systematically after a few number of cycles (about 10000 to 30000 cycles) a crack appeared visible at the optical microscope. So the visual detection helped to define a good number of cycles range at which the crack initiated, while the stiffness variation defined a more accurate number of cycles within that range, a posteriori, analyzing the stored data. The uniaxial tensile fatigue tests were carried out over a range of cyclic stresses at the constant frequency of 5 Hz while the nominal load ratio R was kept constant and equal to 0. The following values of the surface roughness, as the arithmetic mean deviation of the roughness profile (ISO 4287:1998), were considered for the plates with central hole: 2μm, 1μm, 0.4μm, 0.15μm. These values have been checked trough a “MarSurf PS1” that is an on- site surface roughness measurement instrument. R ESULTS Fatigue curves he fatigue data were statistically re-analysed by using a log-normal distribution and are plotted in terms of nominal stress ranges (referred to the net area) in Fig. 5. More specifically, Fig. 5-a shows the fatigue data of the hourglass specimens, the Wöhler curve (mean curve, P s = 50%), the Haibach scatter band referred to 10% and 90% probabilities of survival (for a confidence level equal to 95%) and the inverse slope k of the curves. Data from specimens tested at room temperature and at T=360°C are found to belong to the same scatterband, with a value of the scatter index quite low, T σ =1.29. The scatter of the specimens tested at T=650°C, instead, is higher being T σ =2.00, which show also a strong decrease of the fatigue strength combined with a strong variation of the slope. A vertical line is drawn in correspondence of one million cycles where the mean values of the stress range are given to make the comparison easier. At 10 6 cycles the stress range is equal to 675.14 MPa when T  360°C, while it is equal to 95.23 MPa at 650°C. Fatigue data of the specimens weakened by lateral V-notches are shown in Fig. 5-b at different temperatures. The run-out specimens (marked by tilted arrow) were excluded from the statistical analysis. It is evident that up to 500°C there are no differences with respect to the room temperature, whereas a substantial decrease of fatigue strength can be observed at 650°C. The scatter-band related to the specimens tested at T=650°C is compared with that summarising data obtained for T  500°C. At one million cycles, the value of the stress referred to a probability of survival of 50% decreases from 213.12 T