Metal fatigue has been a topic interesting to engineers for long, because it has had a profound impact on making the design of virtually all products more reliable. As the research in this area evolved, the understanding of fatigue has been gained so tremendously that it has been become possible to conduct more and more complicated fatigue analysis or simulation without having to undergo lengthy fatigue tests for a product under design. In this work is a numerical model to predict the fatigue life of a power train component, whose metal material essentially behaves as being inhomogeneous due to a thin and hard surface layer on the top of the base material. As a typical case in the power train of automobiles, the clutch component is subject to the inertial load arising from an angular velocity as well as the torque loads, varying cyclically with time in a form other than what is called constant-amplitude. The strain or stress responses to the loads show that they depend not only on the loading history, but also on the status in the clutch engagement. As the loads have caused the component to exceed the yield somewhere in the metal structure, strain-life approaches are employed to compute the number of cycles to failure. A commonly-used linear damage cumulative summation, known as Miner’s Rule, is applied to assess the fatigue damage fraction. Through the results from the fatigue simulation, it will be demonstrated that the case-hardened surface has much less fatigue damage than the inner core. In other words, it is the core at the interface between the material zones that initiates a crack. Also, in the end of this paper, the effect of the depth of case-hardening on the fatigue life will be discussed.