neural network for modeling an aerodynamic database (multiple variables)

[rotw]

Hello,

I want to use a neural network to create a correlation model (sort of black box) of a database of aerodynamic data.
The database can be though off as a matrix. The columns contain the variables. It can be represented as follows:

Columns (say 7): X1, X2, X3, X4, Y1, Y2, Y3
X being inputs
Y being outputs

The matrix has about a thousand of sequential rows or lines. Each row contain a set of numerical values which are assigned to the variables as given above. It is basically a big array of data.

Using multiple regression techniques is an option (although I am not sure of the outcome) but I would like to try a neural network-based approach to hopefully short cut the tedious task of doing multiple regression.

My questions:
Can you advise some reference for a neural network system that ideally you have experience with or simply feel would be appropriate for modeling my problem? I also would like to implement this in an excel spreadsheet.
Any comments such as limitations to this approach, etc. are also very much appreciated.

Thanks

 

[3DDave]

This is a better question for the StackExchange website, though there are others that deal with software development.

 

[LeoChem]

rotw

Even though there's not much information about exactly what you are trying to do, I guess your approach is just fine. About the software you are looking for there are many, but I saw a review last week telling that for scientific uses the most common one is MATLAB ANN Toolbox. I'll send the review.

There are different regression functions that you can choose. Its not difficult to use (has a manual) and gives good result, so you can achieve your correlation.

Implementation on a spreadsheet I guess is more difficult and you will need some optimization algorithm that would be even more difficult in my opinion.

 

[broadtechengineering]

In this hybrid-engineering simulation model, the net wind speed and direction time series predicted by ANN in the met-mast position have
been used as input to forecast the power production of the wind turbines through CFD (Computational Fluid Dynamics) simulations.
The numerical modeling calculations can be implemented utilizing the Windsim CFD model: a numerical code based on PHOENICS, which easily solves the Reynolds Averaged Navier-Stokes (RANS) equations coupled with a multigrid-coupled numerical solver(MIGAL).