To clarify frequency-dependent attenuation effects of fluid temperature fluctuation on fatigue strength, thermal fatigue strength tests subjected to superposed sinusoidal temperature fluctuations were performed by the SPECTRA test facility. Fluid temperature waves were generated by superposition of sinusoidal waves, where frequencies were 0.05, 0.2, and 0.5Hz. Two types of superposed waves were selected for the tests, dual and triple ones. The dual one was obtained by superposing two sinusoidal waves whose temperature ranges and frequencies are respectively 200 centigrade and 0.05Hz and 60 centigrade and 0.5Hz at the inlet of test piece. The triple one was the superposition of three sinusoidal waves whose temperature ranges and frequencies are respectively, 150 centigrade and 0.2Hz, 75 centigrade and 0.05Hz and 50 centigrade and 0.5Hz at the inlet of test piece. The longest periods were 20 seconds for both types of waves and it is the fundamental cycle for the thermal fatigue tests. For the dual case, 73,810 cycles fatigue test was performed while for the triple one 116,640 cycles were performed. After these fatigue tests, cylindrical test pieces were cut away from the test loop, and cracks were observed on the inner surface of the test pieces. For the dual wave case, crack initiation occurred from 400 to 600mm position from the inlet of test piece. For the triple wave case, crack initiation occurred from 400 to 600mm position from the inlet of test piece. The corresponded fluid temperature range to crack initiation is from 205 to 220 centigrade for the dual one and from 195 to 215 centigrade for the triple one. Fatigue lives at crack initiation positions were evaluated based on the test conditions. Adopting power spectrum density functions and frequency transfer functions, fatigue lives were predicted within a factor 3 as predicted for single sinusoidal temperature waves in the other tests. To confirm advantages of these functions, fatigue life estimations were compared with those obtained without using these functions. Based on the compared results, these functions are necessary to predict accurate fatigue lives.

This content is only available via PDF.
You do not currently have access to this content.