In spite of the large number of wear models found in the literature, n
o model can predict metal wear a priori based only on materials proper
ty data and contact information. The complexity of wear and the large
number of parameters affecting the outcome are the primary reasons for
this situation. This paper summarizes the current understanding of we
ar modelling for metals. Several recent approaches such as wear mappin
g and wear transition diagrams have suggested some future possible dir
ections for improvement. Some success has been achieved in describing
severe wear of steels under unlubricated conditions using thermomechan
ical approaches. However, modelling of mild wear remains problematic,
especially under lubricated conditions. In mild wear, asperity contact
events dominate the wear processes. A single asperity collision simul
ation apparatus has been used to study asperity-asperity contact pheno
mena. Shear strain and strain accumulation were found to be the domina
nt underlying causes for wear. It is proposed that future research in
wear prediction for metals incorporate the following aspects: wear map
ping, temperature, shear strain response, boundary lubricating film st
rength, and surface roughness. (C) 1997 Elsevier Science Ltd.