Corrosion degree detection of ferromagnetic materials using a hybrid machine learning approach and self-magnetic flux leakage technology