Using bond graphs to synthesize tree-structured transfer functions: Improved frequency-domain analysis of uncertain systems (Q5959144)

Various control applications require the value-set (in the complex plane) of a transfer function, which depends on a relatively large number of parameters. The latter are known to vary in intervals. Discretization is computationally inefficient due to combinatoric explosion. An alternative is the transformation of the given transfer function into a tree-structured (TS) one, where each term contains only one or few parameters. Based on bond graphs, the authors develop a simple rule to determine whether or not a TS transfer function with disjoint parameters exists for a given input-output pair. Further, algorithms for using TS transfer functions to synthesize value sets are presented that use complex-set operations (addition, multiplication and inversion) and work with intermediate value set boundary points only. This results in considerable reduction of computations (in comparison to discretization) and provides also insight into the influence of certain parameters on the system's behaviour. An example illustrating the proposed algorithms is provided.