I am trying to obtain an empirical equation/model having 2 inputs (frequency and S11) and one output (a physical parameter)
I have done various simulations using CST of the System and now I am trying to relate S11 and frequency to a physical parameter. Any suggestions of how I should go about it?
I have tried using system identification tool in Matlab but the results I obtained are rather confusing and I am not sure whether the approach is correct.
try metamodeling techniques such as Neural Networks, Support Vector Machines or response surface modeling and so on
if there is a relation between your inputs and your output (heuristic, probabilistic or statistical) Neural networks or, their generalization, Support Vector Machines they should be able to find that empirical model.
try metamodeling techniques such as Neural Networks, Support Vector Machines or response surface modeling and so on
if there is a relation between your inputs and your output (heuristic, probabilistic or statistical) Neural networks or, their generalization, Support Vector Machines they should be able to find that empirical model.
if you have any prior knowledge about the physics that could relates your inputs to your output, it would be helpful in designing your Neural network.
Actually, even if Neural Networks are considered among Black Box modeling techiques, the choice of the optimal architecture depends tightly on the nature of the relation between inputs and outputs.
For example, you should take care to verify if the relation is static or dynamic, i.e, if Output(i) depends only on Inputs(i) <=> static relation, or if Output(i) depends on Inputs(i) and Inputs(i-1) ... <=> dynamic relation.
It is a wrong idea to not consider available knowledge about the physics of your model even if it is basic.