For the monitoring of components subjected to cyclic loading, non-destructive examination techniques are promising that are sensitive to microstructural changes induced by the fatigue degradation. If these techniques can be applied as online monitoring, this approach would enable to predict the end of service life more precisely than by design consideration. Also, fatigue degradation could be identified much earlier than possibly by crack-sensitive inspection techniques.
In this study, four sensors are applied to strain-controlled constant amplitude fatigue tests for 18 AISI 347 specimens: The mechanical force is measured by the testing machine’s load cell, while resistivity measurement, eddy current probe and thermal camera system record data during the test. The obtained sensor data is combined (sensor fusion) with a Palmgren-Miner fatigue degradation parameter and used to train an artificial neural network. This allows to evaluate the potential of such an online monitoring approach under idealized conditions. It can be shown that the combination of the different sensors is able to reach a better accuracy than single sensors, since the sensitivity to microstructural changes in different stages of the overall fatigue life is specific for the different sensors. It is proposed how this approach and the obtained models allow to estimate the total fatigue testing time per sample, updated after each new test.