Deep Convolution Neural Network for Remaining Useful Life Prediction

This is an example repo that illustrates the concept of predictive maintainance through remaining useful life estimation. A deep convolution neural network is trained using a subset of the C-MAPSS dataset. This implementation utilized a threshold approach (called piecewise linear RUL) to define the engine pre-degradation cycle (R_early) when the engine is assumed to operate in a normal state. This implementation also uses selected sensor readings, as opposed to other implementation that used the whole dataset for model training.

Dataset

The dataset (Train_data.csv) is a cleaned version of train_FD001 subset in the C-MAPSS dataset. The test dataset (Test_engine_21.csv) is a subset of the C-MAPSS dataset containing engine number 21 sensor readings. Data cleaning is done by removing features that do not provide useful (and non-i.i.d) information from the dataset. This is done to move the implementation closer to real-world scenario.

Model

The RUL estimator has 4 convolutional hidden layers with varying filters and a kernel size of 3. It also uses a Globalmaxpooling layer and dropout layers stacked as shown in the notebook. The output is a single-unit Dense layer. Relu activation is used on all convolutional layers while the fully connected layer and the output activation is linear.

Usage

You may load the model checkpoint and predict the RUL directly or preprocess and train a new model using the functions in the CMAPSS data cleaner and the steps shown in the notebook.

If you like the example in this notebook, you will also like a new architecture proposed in this paper. It gives a state-of-the-art prediction, and can be used for other time-series prediction tasks. An adaptation of the code used in the paper is presented here.

Links:

1. A. Ayodeji et al, "Causal augmented ConvNet: A temporal memory dilated convolution model for long-sequence time series prediction", ISA Transactions, 2021.
2. O. Heimes, "Recurrent neural networks for remaining useful life estimation", International Conference on Prognostics and Health Management, pp. 1-6, 2008.
3. Li X, et al., "Remaining useful life estimation in prognostics using deep convolution neural networks" Reliability Engineering and System Safety pp.1-11, 2018.