Extensive field experiments (the 'Silinge wheel flat experiments') with a m
oving train have been designed, performed (at Silinge) and evaluated. More
than 200 wheel flats were formed under controlled conditions involving diff
erent wheel loads, train speeds and sliding durations, and the friction coe
fficient between the wheel and the rail was also varied (and indirectly mea
sured). Samples extracted from flats of the tested wheels have been metallo
graphically examined with respect to phase transformations and cracks. A nu
merical model for wheel flat prediction has been qualitatively verified and
quantitatively calibrated. In the experiments, martensite was found beneat
h all flats and cracks were observed in most cases. It is concluded that th
e risk for future spalling should be considered for all wheelsets with flat
s. A damaged wheelset should be taken out of service as quickly as possible
. When reprofiling the wheels, all martensite and an additional layer of se
veral millimetres should be machined off.