S-N curves of structural materials are obtained through fatigue tests. These tests are often performed using five different stress levels, with fifteen test specimens for each stress level. This kind of test plan provides estimates that are less precise compared to other experimental plans, for example, the ones called optimum plan or compromise plan. The main reason for this drawback is the use of the same number of specimens for all stress levels. It has been observed that less precise results are obtained for lower stress levels because failure occurs less frequently. That is why more specimens should be used for lower stress levels as compared to higher stress levels. As long as the number of specimens to be tested at low stress levels is increased, the total number of failures will also increase, which allows one to develop a more precise data analysis. The objective of this work is to present an alternative experimental plan to obtain S-N curves, which intends to provide accurate estimators. A practical application is done for planning a fatigue test, in a flex-rotating machine, to obtain the S-N curve of SAE 8620 steel.Copyright © 2002 by ASME