System Modelling with Mathematica

*To*: mathgroup at smc.vnet.net*Subject*: [mg18784] System Modelling with Mathematica*From*: com3 at ix.netcom.com (Steve Comstock)*Date*: Tue, 20 Jul 1999 01:33:36 -0400*Organization*: Netcom*Sender*: owner-wri-mathgroup at wolfram.com

Does anyone know where one could find sample notebooks that can aid in modelling 2 input , 1 output systems ? We have time-domain test measurements (ASCII files) of the inputs and the outputs at roughly equal time points that capture the steady state as well as transient behaivor of a complex system. The "output" from this system is actually the mathematical average of several "true" outputs from the hardware under test and thus cannot be represented by the first principles of the underlying physics to the degree of accuracy needed. We were thinking along the lines of some sort of least squares minimization of the error approach which would solve for the coefficients of the underlying function. The problem is that we aren't sure what this function should look like. Perhaps Mathematica could be setup to search for this function ? Another system has a "System Identification Toolbox" that tries to solve this kind of problem (not sure of how or if it works). Has something similar in Mathematica been attempted ? A more specific newsgroup or website on this topic would be welcomed as well. Not interested at this time in neural networks, general order polynomials or other "purely mathematical" approaches but an approach (if possible) that is more "systems based" that has knowledge of how "well-behaved" systems respond to changing inputs. We have Mathematica version 3.0. Thanks in advance . . . Steve.