An autonomous and intelligent system for rotating machinery diagnostics

Teay, Siew Hon (2015) An autonomous and intelligent system for rotating machinery diagnostics. Doctoral thesis (PhD), Manchester Metropolitan University.

Preview

Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (4MB) | Preview

Abstract

Rotating machinery diagnostics (RMD) is a process of evaluating the condition of their components by acquiring a number of measurements and extracting condition related information using signal processing algorithms. A reliable RMD system is fundamental for condition based maintenance programmes to reduce maintenance cost and risk. It must be able to detect any abnormalities at early stages to allow preventing severe performance degradation, avoid economic losses and/or catastrophic failures. A conventional RMD system consists of sensing elements (transducers) and data acquisition system with a compliant software package. Such system is bulky and costly in practical deployment. The recent advancement in micro-scaled electronics have enabled wide spectrum of system design and capabilities at embedded scale. Micro electromechanical system (MEMS) based sensing technologies offer significant savings in terms of system’s price and size. Microcontroller units with embedded computation and sensing interface have enabled system-on-chip design of RMD system within a single sensing node. This research aims at exploiting this growth of microelectronics science to develop a remote and intelligent system to aid maintenance procedures. System’s operation is independent from central processing platform or operator’s analysis. Features include on-board time domain based statistical parameters calculations, frequency domain analysis techniques and a time controlled monitoring tasks within the limitations of its energy budget. A working prototype is developed to test the concept of the research. Two experimental testbeds are used to validate the performance of developed system: DC motor with rotor unbalance and 1.1kW induction motor with phase imbalance. By establishing a classification model with several training samples, the developed system achieved an accuracy of 93% in detecting quantified seeded faults while consumes minimum power at 16.8mW. The performance of developed system demonstrates its strong potential for full industry deployment and compliance.