Cross-Method Overview of Fleet-Based Machine Health Estimation and Prediction: A Practical Guide for Industrial Applications

A reliable assessment of industrial machine health is crucial for economical and safe operation. To this end, data-driven approaches have gained prominence owing to the advancements in data acquisition and machine learning techniques. However, practical applications of these approaches often confront the challenge of data scarcity, due to heterogeneity among machines. To address the data scarcity problem, this study delves into health estimation and prediction methods that utilize fleet data. Unlike existing review papers that mainly focus on one specific fleet-based method, this work offers a cross-method overview. The methods are classified into six categories. All share three steps: data selection, model development, and model adjustment. This work also provides a step-by-step guide for industry practitioners to incorporate fleet knowledge, which emphasizes business requirements and highlights an iterative method development process. It helps industrial practitioners navigate through the complexities of various approaches to utilize fleet knowledge, paving the way to bring advanced methods to industrial implementations.