Predictive Maintenance on Aircraft Engine Using Degradation Simulation Dataset for Use by Maintenance, Repair, and Overhaul Organizations




Peña, George Patrick

Journal Title

Journal ISSN

Volume Title



Machine learning has impacted the manufacturing industry through the use of predictive maintenance (PdM) to reduce cost and downtime, thus attaining more research attention. Predictive maintenance strives to extend the useful life of machines and their parts, allowing maintenance to be conducted only when required to minimize unexpected downtime and unexploited equipment lifetime. An area of concern is the prediction accuracy in machine learning algorithms can be inadequate, which has resulted in continued research to find more effective models. This thesis uses machine learning techniques to apply predictive maintenance to predict the remaining useful life (RUL) on a degradation simulation turbofan dataset, which can be beneficial to maintenance, repair, and overhaul (MRO) organizations. MRO organizations in the aviation industry strive to restore all items on an aircraft to its working condition. Some challenges the industry is dealing with are inventory management of parts, moving parts, demanding customer base, and emerging market. This thesis proposes incorporating PdM with Lean Six-Sigma concept Just-in Time (JIT) through supply chain management within an MRO organization to develop a reliable production plan to improve on parts inventory management. Three machine learning algorithms were used to predict the RUL on an aircraft engine: Support Vector Regression (SVR), Regression Tree (RT), and Random Forest (RF). The dataset used was released in 2008 by the Prognostics Center of Excellence at NASA's Ames Research Center. The data consist of a separate training and test set, along with a true value of RUL for each engine's cycle. Upon pre-processing and analyzing the data, the machine learning algorithms were tested for accuracy through the use of a R-squared score. The SVR model obtained a score of 0.6904, the RT model obtained a score of 0.7325, and the RF model obtained a score of 0.9541. It was discovered through cross validation that the RT model had overfitted and the SVR and RF model did not. Since RF was able to achieve the best accuracy, it was the model of choice for applying predictive maintenance. Accurately predicting when an aircraft needs repairs can eliminate the need for preventative or unnecessary repairs, resulting in savings on labor, expedited shipping cost, and balancing inventory holds more efficiently.


This item is available only to currently enrolled UTSA students, faculty or staff. To download, navigate to Log In in the top right-hand corner of this screen, then select Log in with my UTSA ID.


Just-in-time, Machine learning, Maintenance, Repair, Overhaul, Predictive maintenance, Supply chain management



Mechanical Engineering