Project_NNDL

This repository is created for completing capstone project for NNDL

Predictive Maintenance — LSTM Classifier (NASA C-MAPSS)

Anomaly detection for industrial equipment: an LSTM classifier that flags turbofan engine failures 24 cycles ahead on the NASA C-MAPSS (FD001) dataset.

Capstone project · HSE · Business Analytics & Big Data Systems.

How to run (Kaggle)

The notebook is built to run on Kaggle with GPU.

1. Create a new Kaggle Notebook and upload notebook.ipynb (or copy the cells in).
2. In the right sidebar click Add Input and add the NASA C-MAPSS dataset:

   https://www.kaggle.com/datasets/behrad3d/nasa-cmaps

3. (Recommended) Turn on a GPU accelerator: Settings → Accelerator → GPU.
4. Run all cells.

KPI targets

Metric	Target	Result (test)
Recall	≥ 85%	88.9% ✓
FPR	< 5%	1.5% ✓

Engine level: 19/19 failing engines caught ≥24 cycles ahead.

Team

Anastasia Zaporozhets · Vladimir Kuznetsov · Pavel Manuilov