Exploratory Data Analysis of the N-CMAPSS Dataset for Prognostics

Abstract

In the recent years, industries such as aeronautical, railway, and petroleum has transitioned from corrective/preventive maintenance to condition based maintenance (CBM). One of the enablers of CBM is Prognostics which primarily deals with prediction of remaining useful life of an engineering asset. Besides physics-based approaches, data driven methods are widely used for prognostics purposes, however the latter technique requires availability of run to failure datasets. In this manuscript authors have aimed at performing exploratory data analysis (EDA) on the New Commercial Modular Aero-Propulsion System Simulation (N-CMAPSS) dataset published by NASA. Although 8 datasets are publicly available, authors have chosen dataset 3 (DS03) for EDA in this paper which consists of 9.8 million instances and 47 features. The main aim of doing EDA is to gain better understanding of the dataset as it would facilitate in building a deep learning model that can be used for predicting RUL of the aircraft engines.