The configuration design of a mechanical product can be performed efficiently when it is based on a functional model. There are some function-based design methodologies that decompose functions from the abstract level to the concrete level by matching the functions to physical parts. However, executing innovative design is difficult when the function is matched only to pre-defined parts. This paper describes the configuration design process of a train bogie with a design expert system that uses function taxonomy and TRIZ. The design expert system can propose a functional model of a new part that is not on the existing parts list of train bogies. Levels of design knowledge which describe the operations of the mechanical product at different levels of abstraction are introduced. This is the theoretical basis for using TRIZ with the function taxonomy for configuration design. The design expert system is adequate for controlling the design knowledge, which is expressed as knowledge of parts, functional modeling, mapping rules between functions and parts, and TRIZ. The functional model of a new brake system in the bogie is introduced with the design expert system.