Frequency analysis of random fatigue: setup for an experimental study

The frequency-domain approach to fatigue life estimation in random loading has been largely investigated due to its computational advantages, and several methods for the frequency translation of the most common time-domain methods have been proposed. Between the most known frequency methods there are the Bendat's Method, valid for narrow-band signals, and the Dirlik's formula, which is considered the best result for wide-band signals. However, the great part of the frequency methods takes the rainflow count as a reference time-domain method and uses the rainflow damage computation as the exact value to emulate. Therefore, very few experimental data for fatigue life of mechanical components subject to random loads are available in the literature. This work presents the setup for a series of experimental tests for specimens subjected to random loads, aiming at achieving experimental data to compare with the results provided by frequency methods. After a brief description of the materials used for the setup, the two-step test concept is described: firstly, the specimen will be subjected to random loads obtained by a certain PSD for an amount of time which should nominally cause a 30% of damage; then, the fatigue test will be ended on a resonance testing machine to compute the actual residual fatigue life of the specimen; this two-step testing also allows to reduce the time requested for the tests. The test bench developed for the experimental investigation is described in the paper, together with the results of some preliminary tests, aimed at verifying the feasibility of the conceived procedure.