I am working on an experimental research that introduces the need for non-linear grey-box system identification.
The physical-system that we have in the lab is characterized by a well-defined mathematical model with:
* 15 input-signals (degrees of freedom);
* 30 fixed, unknown, parameters (estimation target);
* 4 output signals (observations).
I am interested in estimating these 30 unknown parameters, while the values of the 15 input signals may be set arbitrarily, and while monitoring the values of the 4 output signals. A known mathematical model, gives the 4 output signals as a function of: the 15 input signals and the 30 parameters, using highly non-linear functions. I wonder if the JMP SW package has a module that can perform this identification task.
The nonlinear platform in JMP is used for fitting nonlinear regression models. There is also a nonlinear experimental design platform, although you might also consider a space-filling design if you have little information about the likely values of the parameters. Space-filling designs are also used to create surrogate models (gaussian process or neural net) of complex nonlinear systems.
So, do you know the exact form of the "well-defined mathematical model" but just not the parameters?
