Tebmer / BearingFaultDetection

DC competition of bearing fault detection: rank 40/1388.

BearingFaultDetection

This code is for Bearing Faul Detection competition in DC platform. Rank 40/1388.

Please see offical website: https://challenge.datacastle.cn/v3/cmptDetail.html?id=248

Code Overview

1. data/ directory stores the train data, test data and feature data containing time/frequency/time-frequency feature.
2. analyze/ directory is used to analyze the correlatoins between features by a correlation matrix.
3. preprocessing/ directory is main part to preprocess data: 1) Use wavelet filter to denoise data; 2) Use wavelet decoposing tool to extract time-frequency domain feature of dat. 3) And finally, apply PCA to reduce dimension of feature.
4. RF.py uses random forest as our classifier to train model.

Visualization

Framework

Comparison of signal processed by wavelet filter

Heatmap of correlation matrix of features

Report

Please refer to the report.pdf for details.