This repository contains the results of 104 experiments to estimate the degradation process for two different case studies each, one from manufacturing and one from dry-bulk shipping.
The data set used for the case study is the well-known bearing data set provided by FEMTO-ST institute within PRONOSTIA, an experimental platform dedicated to the testing and validation of bearing failure detection, diagnostic, and prognostic approaches. The FEMTO bearing data set contains run-to-failure tests of 17 bearings, 6 for training and 11 for testing that can be found here: https://github.com/wkzs111/phm-ieee-2012-data-challenge-dataset.
The second data set consists of sensor data for 15 vessels, 12 for training and 3 for testing, with data ranging from beginning of 2016 to end of 2020. The data set is provided by an anonymous shipping company.
We choose to be an experiment, where
denotes a combination of preprocessing steps listed in the table below,
denotes if an health stage classifier is used, and
denotes the selected regression model for prediction, where
denotes Multiple Linear Regression,
denotes Gaussian Process Regression,
denotes Artificial Neural Network and
denotes Support Vector Machine.
For example, the notation represents the application of health stage classifier, followed by frequency analysis and statistical feature extraction, with a final regression of degradation using ANN.