Fault detection and diagnosis of rotating machineries

Failure of rotating machineries is a frequent and common incident in process industries, leading to catastrophic outcomes. There is therefore need to continuously evaluate the condition of a monitored machine without interruption to its operation, and thereby successfully identify impending faults long before disastrous breakdowns occur. Such an approach towards monitoring and maintenance can provide for properly planned service schedules and the replacement of the failing components at the most appropriate time.;Time Synchronous Averaging exploits the natural periodicity of the signals by extracting a synchronous average of the signal over one period, removing the stochastic part efficiently and producing a single period of a deterministic signal. In this thesis, Wavelet and Cyclostationary Analysis has been combined together to generate Time Domain Averaging across all scales (TDAS). It captures the vibration generated by a rotating machinery over one complete revolution of the shaft, and extracts the periodic components from the noisy signal keeping the different scale representation of the wavelet analysis intact.;Failure of a mechanical system is always preceded with changes from linear or weakly non-linear to strong non-linear dynamics. A measure of non-linearity in the vibration signal gives a measure of deviation of the process from normal operation to the emergence of a fault. Bicoherence Analysis has been proposed as a technique that detects the increase in non-linearity due to generation of faults in the system.;This thesis also proposes the use of Inductive Monitoring System (IMS) to monitor rotating machines or plant units. It automatically classifies data into different clusters to specify different modes of operation of the system, and detects anomalous behavior of the system by classifying abnormal data.;Analysis of vibration signals is widely used to detect early faults in rotating machineries. Vibration signals from a rotating machine carry the signature of its internal fault, and as such early fault detection is possible by analyzing the vibration signals using different signal processing techniques. Wavelet Analysis is such a tool that has the ability to characterize the local features of the signal at different scales. The ability of Wavelet Analysis to separate specific frequency components of a signal has been utilized in this thesis to detect rub-impact in rotating machineries, and to quantify it by the proposed Rub Index.;Finally, all proposed signal processing techniques in this thesis have been demonstrated successfully on numerous simulations, pilot plant case studies, and industrial case studies.