Due to the large individual differences between rock samples

the upper stress ratio is unable to be determined and the quantitative relationship between Sand lnN can not be established in rock fatigue analysis.Therefore

it is important to estimate the unconfined compressive strength（UCS）of rock sample.Firstly

the error propagation law and the influence of the individual differences of rock samples on fatigue life are studied on the basis of a large number of test data under uniaxial compression and S-N equation.The results show that the error increases dramatically during spreading from UCS to upper stress ratio

then to fatigue life.A very small standard deviation in UCS may lead to great dispersion in fatigue life.So

the UCS of rock sample instead of the statistical average UCS of rock should be used in fatigue analysis.Secondly

an idea applying curve fitting method to evaluate the UCS of rock sample based on a modified regression peak constitutive model is put forward.And the test data under static loading is fitted with this model.The fitting result show that the evaluated UCS is very close to the true value.Finally

based on the tested data of the static loading segment in cyclic loading test

this curve fitting method is used to analyze the experimental data of four samples under the same test condition.The results show that the UCS of each sample is different and the greater the UCS

the smaller the upper stress ratio and the loger the fatigue life will be.Meanwhile

there is existing obvious linear law between Sand lnN.Therefore

the method revising the upper stress ratio with the sample UCS evaluated by the modified regression peak constitutive model can overcome the inherent defect of the conventional fatigue analysis method.