Condition-Based Monitoring Techniques for Concrete and Industrial Equipment

Abstract

This thesis presents hybrid techniques that combine signal processing with machine learning for the fault diagnosis and prognosis of concrete structures and industrial equipment. Based on the information collected during the monitoring process, abnormalities in the working conditions can be detected at an early stage. This ensures maintenance can be planned and performed accordingly, thus avoiding irreversible damage to the asset and the surrounding environment at minimal cost. In the third chapter, a scheme for leak localization on a cylinder tank bottom using acoustic emission (AE) is proposed, which utilizes similarity scores to group acoustic emission hits by their sources and analyzes the source locations through a Voronoi Diagram. The results under a one-failed-sensor scenario show a high level of accuracy across multiple test locations. Furthermore, in Chapter 4, an approach to perform leak state detection and size identification for industrial fluid pipelines with an acoustic emission activity intensity index curve (AIIC) using b-value and a random forest (RF) is proposed. This chapter shows that the AIIC outperforms traditional AE features in portraying the pipeline’s working states, and along with the classification power of RF, the proposed method also consistently surpasses two state-of-the-art reference methods in the size identification task. Afterwards, Chapter 5 proposes a new technique for the construction of a health indicator for remaining useful life (RUL) prediction of concrete structures based on the Kullback–Leibler Divergence (KLD) and deep learning. The KLD-based indicator is capable of descriptively portraying the concrete structure’s fracturing process, and its adaptation to the RUL task shows a high level of accuracy.