Low cycle fatigue tests are performed on specimens of niobium stabilized austenitic steel AISI 347 (1.4550) at ambient temperature. During the test, the fatigue specimens are equipped with eddy current probes, and it can be seen here that the impedance phase shift changes significantly at very early stages of fatigue (i.e. before cracking). Electron backscattering diffraction investigations were carried out to better connect microstructure evolution with impedance phase shifts. Machine learning techniques are employed to relate the impedance shift to the fatigue degradation. This approach allows also the derivation of fatigue life curves with few specimens. •Eddy current sensors are sensitive to fatigue-induced microstructural changes•The fatigue degradation can be identified by artificial neural networks•Fatigue life curves can be defined based on machine learning techniques