Detailed fatigue data for a given combination of steel grade and processing offers the best means of determining fatigue performance when designing an automotive component. However, in some instances, where only mechanical properties and hardness are available, methods have been developed for estimating fatigue life from these properties. One such method, developed by Roeselle and Fatemi, allows prediction of strain-life performance using hardness (1). By compiling data from the AISI Bar Steel Fatigue Database, along with data from other sources, Roeselle and Fatemi arrived at the following modified strain-life equation:
As can be seen, a strain-life curve can be estimated from Brinell hardness number.
Figure 1 shows a graph of strain-life data for SAE 1038 medium carbon steel, normalized to 163 BHN, taken from the AISI Fatigue Data Base. Also shown is a predicted strain-life curve based on the reported hardness for the steel.
Figure 1
Figure 2 shows a similar graph for SAE 1141 aluminum fine grained steel normalized to 223 BHN.
Figure 2
As can be seen, there is good correlation between the predicted curves and the experimental curves. It should be noted, that the methodology described above has limits. It cannot account for the effects of variations in microstructure, non-metallic inclusions, prior austenite grain size, surface condition or residual stress. Thus, while it offers a convenient means of providing a preliminary estimate of fatigue life, it should be used with care.
Reference
(1) M. Roessle and A. Fatemi, International Journal of Fatigue, Vol. 22, 2000.
Bar Fatigue Blog 
The following paper used a new criterion to evaluate different empirical relations between monotonic and cyclic properties. It was reported that Rossle Fatehmi method is better for steels. For other materials, the conclusion need not be same. Refer link (http://www.sciencedirect.com/science/article/pii/S0142112310002859) Hariharan, K., Raghu V. Prakash, and M. Sathya Prasad. “Weighted error criterion to evaluate strain–fatigue life prediction methods.” International Journal of Fatigue 33.5 (2011): 727-734.
I would agree the hardness seems low for this strength. A check of the test report shows hardness was actually taken on each of the 18 fatigue samples, and the values were 191-204 Brinell. I suspect these were taken on the 0.500” grip diameter which is significantly larger than the 0.200” gage length diameter. Since the samples were heat treated after machining, the hardness would likely be higher in the smaller gage length diameter. Unfortunately, this is the data we have to work with in this case.
Response from Greg Fett, SMDI Program Manager Bar Steel Fatigue Sub-Committee
The following paper used a new criterion to evaluate different empirical relations between monotonic and cyclic properties. It was reported that Rossle Fatehmi method is better for steels. For other materials, the conclusion need not be same. Refer link (http://www.sciencedirect.com/science/article/pii/S0142112310002859) Hariharan, K., Raghu V. Prakash, and M. Sathya Prasad. “Weighted error criterion to evaluate strain–fatigue life prediction methods.” International Journal of Fatigue 33.5 (2011): 727-734.