Simulation-based methods for fault estimation and parameter identification of rotating machines

Abstract

The vibration signature of a rotating machinery often indicates the condition of the machine. Faults and parameters of the rotating shaft, the bearings, and the support have a significant influence on the dynamics of the system. To understand the root cause of unwanted vibrations and predict failure, it is important to identify the system faults and parameters through vibration and dynamic analysis. While the vibration signature of such machines can be measured using sensors, their dynamic behavior can be modeled using simulation methods. The combination of computationally efficient simulation models and measured vibration data creates the possibility for predictive maintenance through parameter identification, system diagnosis and condition monitoring.

The objective of this study is to develop simulation-based methods for accurate parameter estimation and computationally efficient modeling of rotor-bearing-support systems. In this regard, this dissertation introduces a method for the accurate identification and estimation of rotor unbalance which is one of the most common occurring faults in rotating machinery. Furthermore, the study focuses on developing computationally efficient methods for modeling the effect of bearing waviness and clearance on the system vibrations. Finally, simulation models are developed to capture the effect of support parameters on the dynamic behavior of rotating machines and to train an intelligent tool to identify the support stiffness. The modeling methods developed in the study are experimentally validated using the case study of a large flexible rotor. The results and analysis provide a summary of the accuracy and computational efficiency of the methods along with insight into the challenges and shortcomings. The concluding remarks include possible application of these methods in the industrial sector and future research topics.