detenction of anomalies using autoencoders

Job ID: 33655465

Budget: €30 – €250 EUR

Hello, I have to send this task for my DL class, hope you can help me:

It is call the prognostic challenge 2012, please find attached the documentation about it.
My main task is to predict the RUL of the bearings using Deep learning technicques.
My teacher asked me to do the following:
1.Apply autoencoders to differentiate between anomaly and normal condition taking a time window of size W.
2.Apply autoencoders using a pre-trained network as an encoder to differentiate between anomaly and normal condition by taking a time window of size W.

I have already done part of it but I am blocked. I already made a datagenerator and have fed it to an autoencoder but the results seems really bad. I will also attached the .ipynb file so you can have a look at it.
I would need not only the answer but also an explanation of how you have done it in case you accept it. THANKS.
please also find the datasets here:

training: https://hdvirtual.us.es/discovirt/index.php/s/z75ekAqY3tt3aKy
Test: https://hdvirtual.us.es/discovirt/index.php/s/kZtaCMbgbAsdpNq
Validation: https://hdvirtual.us.es/discovirt/index.php/s/DMXxpQDgQMJ6L34
Related categories: Python Machine Learning (ML) Deep Learning