Learning-Based Methods for Vibration-Based Condition Monitoring

Abstract

In this chapter, the fundamentals of learning-based techniques are introduced in a condition monitoring context. The objective is to provide a general overview of learning-based techniques and detail how these techniques are used in vibration-based condition monitoring. This chapter introduces learning-based techniques and introduces the notable distinctions between the supervised and unsupervised learning frameworks. The distinctions between the learning-based techniques are highlighted as the implications of these distinctions are crucial for industrial applicability. Specifically, the assumption of a learning-based methodology of access to a labelled historical fault dataset is often infeasible in industrial applications. Popular methodologies such as supervised learning, semi-supervised learning and transfer learning all require access to a labelled fault dataset. Unsupervised learning techniques offer the opportunity to circumvent the labelled fault data requirements and provide useful condition monitoring insights for condition inference. Under suitable data evaluation strategies, where the preservation of the temporal structure present in time-series data is key, unsupervised learning offers health indicators that can be investigated for fault detection and isolation purposes. Finally, the complementary nature of signal processing and learning-based approaches in an unsupervised learning setting is discussed.