Discussion

In this part we want to discuss how we would like to unite and enhance the benefits of bond graphs and port-Hamiltonian systems.

Bond graphs

• subsystems are primitive elements

  • it would be nice to allow arbitrarily complex subsystems

• the composed system (network of elements) is itself closed

  • it would be nice to allow open bond graphs (outer interface)

• composition (i.e. hierarchical nesting) of bond graphs should then ...

  • exist whenever interfaces of systems match
  • be unique, given no further data than the bond graphs to be composed
  • be associative (unique flattened hierarchy of bond graphs)

Port-Hamiltonian systems

• port-Hamiltonian systems compose according to extra data ("interconnecting Dirac structure")

  • it would be nice to have them compose according to graphical syntax similar to bond graphs, that is composable just like the systems themselves

• subsystems and their (inner) ports are not explicit in representations of port-Hamiltonian systems

  • it would be nice to have subsystems and system interfaces as explicit parts of the framework

Conclusion

• bond graphs provide a network representation of systems, but the representation is not composable as such, at least not without specifying extra data

• port-Hamiltonian systems are composable, but composite systems are not defined using a graphical syntax similar to bond graphs

Can we combine these great inventions/discoveries to get something even more powerful?

Our answer is sketched out on the next page.