I have a lot of data results, How do I find a parameter of an equation which can predict these results?

I have many data of dissipated hysteretic energy of a system and for each ground motion, each height of the system, each soil condition, and each period of the system, has different dissipated hysteretic energy. Is this called "multi-input-multi-output system"?, I am not sure because I don't have good knowledge about statistics. I want to create a equation which take account of those variables above. Which MATLAB app or function should I use for this analysis? Thank you.

[Walter Roberson]

I am not clear as to which of those are considered inputs and which are considered outputs?

[Sukrisna Gautama]

thank you for the response, the output will be the dissipated energy (which I already have the empirical data), and the input will be the soil condition, height of the building, and the period of the building.

[Walter Roberson]

That sounds like multiple inputs and single output.

I suspect that you should probably not be looking at statistics for this. Instead I would direct you to possibilities in the neural networks, or maybe the curve fitting toolbox (but I doubt it is suitable), and I am thinking that perhaps what you really need is System Identification toolbox

[Sukrisna Gautama]

Thank you, Mr. Roberson, I will take a look at those things you mention first.

Answer 1

[Star Strider]

There are several articles on the Internet. One recent (2015) open-source discussion is Estimation of Earthquake Input Energy, Hysteretic Energy and its Distribution in MDOF Structures (link), the Ph.D. dissertation of Mebrahtom Gebrekirstos Mezgebo, Syracuse University.

I would first look through an open-source publication by the same author, Hysteresis and Soil Site Dependent Input and Hysteretic Energy Spectra for Far-Source Ground Motions (link) to see if it has information relevant to you.

Answer 2

[Greg Heath]

Neural Networks seems like the most reasonable approach UNLESS you have plots that indicate the individual I-O plots are relatively easy to model via a low order polynomial

Hope this helps

Thank you for formally accepting my answer

Greg

[Walter Roberson]

Neural Networks are not easy to turn into a coherent "model", so if the poster is trying to figure out what the model is, I think they are going to need to use a different toolbox. Curvefitting might help eliminate some potential models but curve fitting is not good at saying that a model definitely is something or other. System Identification might be the best bet for coming out with a definite model.

Answer 3

[Ali Hammouche]

this data measurment for any material ,

can you provide it ?