RUL Prediction – Tesseract Technolabs

Case Study

project description

In this project, we have created a Recurrent Neural Network for the prediction of remaining useful life.
This can be used by the industrialist to the prediction of the remaining time for the hydrogen gas to burn off.
So we trained our AI model on the dataset with a time stamp of around 1000 hours, we took almost 70% of the data to train and the remaining 30% was used to test the model’s accuracy.
Moreover here rather than selecting regular models available for time series prediction we have selected ESN(Echo State Neural Network), which is best suited for such a task.

Technologies used

We have implemented different models like RNN, Bi-RNN, CNN-RNN, CNN-Bi-RNN, ESN, etc. in the PyTorch library of python specifically used for deep learning.
Along with this, we have also used matplotlib library of Python for plotting the data and for visualization of the output.

Difficulties we faced

The major problem with this project was, as other time-series data can be trained on regular RNN or LSTM models, but for this data, specifically for RUL prediction, we cannot use regular models so we need to find a proper model for this task.

Solutions

After research, we found that Echo State Neural Network (ESN) is more adaptive to the trends of RUL, and they provide good accuracy with such data.

Time Series Forecasting

12 months

Share now

gallery

project images

ready to get started?

Receive news, announcement and reports