A 2.25Cr-1Mo steel was subjected to variable amplitude block straining at test temperatures of 22, 350, 450, and 550°C. A loading block having 500 strain peaks with an exponential probability density distribution and a constant diametral strain rate was generated. The loading block was analyzed by the rain flow procedure, closed hysteresis loops were identified, and peaks forming a cycle with stress amplitude smaller than half of the fatigue limit were deleted. The strain block characterized by the maximum strain amplitude was applied to a set of cyclindrical specimens in a computer-controlled electrohydraulic testing system.

Stress amplitudes of all closed hysteresis loops versus the number of blocks (i.e., the variable amplitude cyclic hardening-softening curves) were plotted. Similar hardening-softening behavior as in constant amplitude straining was observed. A plot of stress amplitudes versus plastic strain amplitudes in a block at the half-life (i.e., the service stress-strain curve) was obtained. The parameters of the service stress-strain curves were evaluated in dependence on the maximum strain amplitude and on the temperature. They were compared with the basic cyclic stress-strain curves.

The fatigue life in variable amplitude block loading (i.e., the number of blocks to failure) was plotted versus maximum strain amplitude. The experimental results were compared with the predictions using a linear cumulative rule combined with a rain flow counting procedure and the Manson-Coffin fatigue life curve. The actual to predicted life ratio was 0.4. The observed discrepancy was discussed in terms of the mechanism of fatigue damage under spectrum loading.