neural network, using simulink, then nn toolbox

so i have a set of date, in matrix or state space form (A,B,C,D), and i am trying to use nn to detect anomalies in this system. please how do I achieve the following.

1. get target data to use in the fitting tool
2. get the training data to use for this system,

[Olalekan Babatunde]

I have the inputs U outputs Y and target/tOUT FOR the systems and the data I obtained had order 601*2.I did some analysis usng nnet toolbox and i think i have some expected results. now I have created 2 models for this systems. but I'm confused as to how to train the system manually using Simulink/ matlab commands and how to identify the outliers. please help

AND PLEASE, how do I plot the step response from the neural network fitting tool. I have used it to analyse the data but I need the step response to complete the method

Answer 1

[Abel Babu]

Hi Olalekan,

This MATLAB example is a good place to start for anomaly detection : https://in.mathworks.com/help/ident/examples/fault-detection-using-data-based-models.html

It describes various approaches to fault detection of systems. All the methods start by first generating the system model as a state-space and then using various methods to account for faults or anomalies, those include - residual analysis, Normal-Deteriorated States etc.

After setting up the system, anomaly detection can be done using either of the methods:

• Fuzzy C-Means Clustering. See fcm() in Fuzzy Logic Toolbox.
• Support Vector Machine Classifier. See fitcsvm () in Statistics and Machine Learning Toolbox.
• Self-organizing Maps. See selforgmap() in Neural Network Toolbox.

This has been documented in the article as well.

Abel

[Olalekan Babatunde]

thanks a lot abel, i have been looking through your response and the link you provided, and it has helped me proceed from where i was too. thanks a lot.... i have update the question with the challenges am having now, please look through. thank you for your continued support