Hybrid monitoring, diagnostics and prognostics on machines and structures

点击翻译按钮获取中文摘要

Vibrations are the most critical source of structural and mechanical damage and failure. In modern engineering, while static loading has been thoroughly investigated and resolved by advanced computing and finite element methods, dynamic problems with nonlinearities, uncertainty and non-stationarity have not yet been solved explicitly, especially for existing mechanisms and structures. The best approach to understand and predict the dynamic behavior of such real-life systems is to conduct experimental analysis, diagnostics and prognostics. The core of this experimental investigation is modal analysis. Over the past two decades, knowledge of modal analysis has developed to the “in operation” level, taking into account actual structural and mechanical conditions of systems and their operating loads. Consequently, the challenges faced in operational modal analysis (OMA) are the variations in geometry, mass and stiffness during operation, the existing mechanical and structural damage, local modes, unknown harmonic and random excitations, and high noise levels. The research program proposed is designed to study and elucidate the problems above using the finite element method, time series vector autoregressive (VAR) model, Kalman filter and artificial intelligence. The objective of the program is to develop projects to resolve research issues and spawn innovations in OMA for the analysis and monitoring of existing mechanical systems subject to non-stationary vibrations. Through initial researches on stationary and short-time non-stationary vibrations, this method has demonstrated extraordinary performance and led to efficient applications in terms of precision, stability and robustness. Moreover, this methodology supports the handling of a large number of sensors and has huge potential for the development of advanced functionalities. Under this program, various research topics will be explored on non-stationary vibrations: optimization of sensor number and location, localization and estimation of damage; identification of damping; updating the finite element model; development of a time-frequency-damping technique for the detection of damage initiation and propagation; and adaptive model updating for signal processing. A new technique is proposed called “hybrid operational monitoring” (HOM), which combines operational modal monitoring and finite model updating. On this basis, a diagnostic-prognostic-health monitoring technology is developed for machine and structure management, prediction and maintenance. The program will focus on applications for different machines, e.g., rotating machines (hydraulic turbines and wind turbines with harmonic excitations) and processing machines (grinding robots with variations in geometry and stiffness), and will consider specific operating conditions and the various types of damage.