An AdaBoost Approach to Predict Engine Failures

Abstract

This study investigates the application of the adaptive boosting (AdaBoost) algorithm for predicting engine failures, a critical task in industries where equipment downtime incurs significant costs and safety risks. The choice of AdaBoost was motivated by its interpretability and low computational requirements, making it well-suited for real-time or embedded system implementations. The analysis utilized a dataset comprising operational data from 100 engines. During preprocessing, binary labels were created based on the median of the engines’ Time to Failure (TTF) to indicate the urgency of maintenance needs. The implemented model achieved an overall accuracy of 80%. The results demonstrate the model’s promising predictive capabilities, though the number of false negatives suggests the need for further optimization. The findings of this research open new avenues for enhancing predictive maintenance systems.