This module contain the code used in our survey paper: Deep Learning and Its Applications to Machine Health Monitoring. It has been accepted by Mechanical Systems and Signal Processing. Pls find the newest version there.

• Deep Learning and Its Applications to Machine Health Monitoring: A Survey
  • Table of Contents
  • Data
  • Code
    • Feature Extraction
    • Deep Learning Models
    • Machine Learning Models
    • Main Test

This folder contains two pickle files, which are extracted features and labels for tool wear sensing experiments. Each pickle file contain x_train, y_train, x_test, y_test. The task is defined as a regression problem.

• data_normal: each data sample is a vector. The features are extracted from the whole time sequences.
• data_seq: each data sample is a tensor. The features are extracted from windows of the time time sequences.

Especially, data_seq can be used by LSTM and CNN models. data_normal can be utilized by conventional ML models.

This folder contains codes for feature extraction, traditional machine learning models, deep learning models and test modules.

RMS, VAR, MAX, Peak, Skew, Kurt, Wavelet, Spectral Kurt, Spectral Skewness, Spectral Powder features are extracted from the input time series.

Based on Keras, autoencoder and its variants, implementations of DBN, LSTM, Bi-directional LSTM and CNN models are provided

SVR with two kernels (linear and rbf), Random Forest and Neural Network are provided.

To replicate the results reported in paper (python 2.7)

pip install -r requirement 
python main_test.py

The results will be stored in output.log. In addition, a python notebook file is provided to parse the raw log file for mean and std accuracies computing. And due to randomness, we run all of these models five times.