A popular saying claims that 'innovation is 1% inspiration and 99% per
spiration'. In this paper, we present a method for automating most of
the perspiration involved in innovative design. We restrict our attent
ion to innovative design processes which can be structured into three
steps: discovery of a new technique, understanding it, and generalisin
g it to fit the problem at hand. The method we developed automates the
understanding and generalisation phases which involve most of the per
spiration. We present the FAMING system which demonstrates the method
for the design of part shapes in 2D elementary mechanisms, also called
kinematic pairs. We believe that the results are generalisable to oth
er domains with similar characteristics, in particular any problem whe
re geometry plays an important role.