This project looks into the development and operations of machine learning pipelines to demonstrate accurate, reliable and maintainable systems. Predictive maintenance can be used to forecast maintenance schedules. Data-driven decision support systems (DSS) can increase confidence in predictions by ensuring the high performance of machine learning algorithms. In this study, we look at the application of machine learning towards predictive maintenance of aircraft engines. By utilising the NASA CMAPSS turbofan engine simulated data, we apply random forest and test various feature selection methods to demonstrate high accuracy of RUL predictions, supported by an interactive dashboard displaying key metrics.
base_rf_pipe.joblib is the current base ML pipeline and model. It uses random forest.
data_ingestion.py is used to initialise the training database
settings.py contains extra functions used by the program