A MapReduce-Based Ensemble Learning Method with Multiple Classifier Types and Diversity for Condition-based Maintenance with Concept Drifts

Abstract

Condition-based maintenance (CBM) is to collect a huge amount of production data stream continuously from sensors or IoT devices attached to machines to forecast the time when to maintain machines or replace components. However, as conditions of machines change dynamically with time owing to machine aging or malfunction, the concept of capturing the foresting pattern from the data stream could drift unpredictably so that it is hard to find a robust forecasting method with high precision. Therefore, this work proposes an ensemble learning method with multiple classifier types and diversity for CBM in manufacturing industries. Aside from manipulating data diversity from previous works, the proposed method includes the features of multiple classifier types, dynamic weight adjusting, and data-based adaption to concept drifts for offline learning models, to promote the precision of the forecasting model and precisely detect and adapt to concept drifts. With these features, the proposed method requires powerful computing resources to efficiently respond to practical CBM applications in manufacturing industries. Therefore, furthermore, the implementation of the proposed method based on the MapReduce framework is proposed to increase computational efficiency. Simulation shows that the proposed method can detect and adapt to all concept drifts with a high precision rate.