Although significant advances have been made in mechanism synthesis to find a mechanism generating a desired motion at its end-effector, no available synthesis methods can determine the topology of a mechanism and its end-effector location simultaneously. It is generally difficult to pre-determine the location of the end-effector relative to the input drive link because the mechanism synthesis may fail if the end-effector location is incorrectly selected. Therefore, the simultaneous determination of the mechanism topology and its end-effector location can be critically useful to advance automated mechanism synthesis. Motivated by this, we propose a neural network-based big data approach to achieve the simultaneous determination. To implement a big data approach which requires a training dataset consisting of diverse mechanisms of different topologies, we propose the use of a spring-connected rigid block model as a unified means to be able to represent diverse mechanisms. The big data approach is followed by gradient-based shape optimization to determine the detailed dimensions of the mechanism synthesized by the approach. The effectiveness and validity of the proposed method are checked with various mechanism synthesis problems.
Read full abstract