By using newgrnn function in MATLAB.

@S. Anbazhagan ..newgrnn....can you share some sample source code for this?which demonstrated ANN training for EMTP-ATP generated Signals?

@Babak Mohammadi...How?........is there any reference source for more understanding?

Refer to Matlab/ANN toolbox user guide

As far as I have worked , Muhammad classification is the first step. Since you have the training data set and then use the test data for clustering and cross validation .  NPRTOOL is a GUI . Scripting is latter...

https://www.mathworks.com/help/nnet/gs/classify-patterns-with-a-neural-network.html

@Aparna Murthy....i know basic classification and training of ANN i-e Iris dataset.But my dataset is in the form of signals as file attached.How can i train this data?What inputs should i feed to ANN?

This is a waveform , it should be stored  as (Y, t(ms) ) format that is in numerical !!!

Get that file and edit it say time axis interpolation (For the phase say) .  Give me #s I can play with it and  give you in the some manipulations. You can plot and check for current responses.

@Aparna Murthy .......Thank you for such  useful answers.I am highly grateful.

My data is attached in file.(i have obtained this data from EMTP ATP Simulations).

i am trying this from previous 30 days but i didn't come with any solution yet.

I want to train ANN using this data to detect starting and ending point o saturation as shown in figures

Ok ... shall start give me some time , will update

I was thinking of NARX net.

I am not sure but i think so it will work

I looked at nprtool , non-linear AR that has time series and the delta variation (from current amplitudes)  for say few steps (in the slope) if compared , should tell us .

One more question , I shall try NARX too . I was wonder are the  signals considered  endo (that is part of the incoming signal) or exo ?

Thank you Aparna for your positive feedback.

yes i also agree with you delta variation (from current amplitudes) for say few steps (in the slope) if compared will giventhe required distorted region of signal.

Let me attach my file,that will show you what i want to do.I have highlighted some imp portions in pdf file.Esp a

ANN for detection(.pdf) is my main file

Rest are some reference papers.

Well Muhammad, here is the story !!

The GUI gives an exception ... (attached) . So I switched to command line (Atleast data is under my control) . Used narxnet () , although I am not convinced . That can be replaced by narnet() .

Now with this result a LOT of analysis has to done to may to the CT delta value...

I haven't read the docs you shared , if you can tell me in a nut shell then I can do some reverse engineering.

PS- l have used the default values for say Hidden Layers =10 , delays =2 and so on.

Play with the code ...