Date of Award
Open Access Dissertation
Based upon a framework of time-frequency analysis, we outline condition based maintenance (CBM), or maintenance only upon evidence of need, for both electrical and mechanical systems. We apply novel time-frequency cross-correlation metrics to helicopter drivetrain systems and electrical cables as remaining useful life assessments by non-destructive and non-invasive tests. In both cases, these metrics for health assessment provide a basis for diagnostic and prognostic analysis of underlying systems and components by performing accelerated condition tests on actual mechanical and electrical systems. We present novel time-frequency domain vibration analysis of a gearbox failure in an AH-64 Apache drivetrain testbed and quantify transient precursors of failure where previous diagnostic methods rely on stationary power spectrum analysis to analyze nonstationary signals. Using time-frequency representations of the vibration signals, a shift in energy is seen from the first harmonic of the gearmesh frequency to the third and fourth harmonics in intermittent patterns indiscernible by the standard power spectrum over the course of 4 days leading to gearbox failure due to grease lubrication drought. We demonstrate a new form of Rényi entropy-based mutual information measure based upon Shannon and Hartley entropy and derived from a cross-time-frequency distribution of separate accelerometer vibration signals for comparing rotational harmonics from multiple bearings to create new condition indicators of damage in rotorcraft drivetrains. Baseline, unbalanced, and misaligned experimental settings of helicopter drivetrain bearings and shafts are quantitatively distinguished by the proposed techniques. With unbalance quantifiable by variance in the in-phase mutual information and misalignment quantifiable by variance in the quadrature mutual information, machine health classification is accomplished by use of statistical bounding regions. Utilizing similar methods to form a time-frequency cross-correlation derived metric, a process for non-invasively assessing the health of low voltage instrumentation cables and medium to high voltage feeder and underground transmission cables is proposed by way of Joint Time-frequency Domain Reflectometry (JTFDR). We introduce a new standardized method for determination of the optimal reference signal for reflectometry to allow implementation of a stand-alone reflectometer device for cable testing and provide theoretical background for a more generalized time-frequency enveloping function of this reference. Fault location and life estimation methods are verified in networks of instrumentation cable, cable tray and conduit systems, multiple localized fault scenarios, simulations of faults on endless lines, and three separate thermal accelerated aging tests of low to high voltage cables. A 24 hour thermal aging test of underground 15kV, tree-resistant cross-linked polyethylene (TR-XLPE) cable simulates 90 years of service life and shows a monotonic increase in the measured JTFDR metric. This is compared to aging of similar duration for other cables utilizing silicon rubber (SIR), cross-linked polyethylene (XLPE), and ethylene propylene rubber (EPR) insulation types. Expanding on this preliminary aging, we present a 916 hour extended accelerated aging test of XLPE insulated RG-58 instrumentation cable at reduced and more realistic temperatures to simulate 30 years of service. The time-frequency optimal reference signal is updated with a separated spectrum as enveloped by a Gaussian derivative function. Lastly, we utilize a single broadband monopole surface wave launcher and receiver in combination with the JTFDR algorithm to obtain fault location and health assessment metrics non-invasively and provide fault assessment in unshielded concentric neutral cables.
Coats, D. L.(2014). Comprehensive Joint Time-Frequency Analysis Toward Condition Based Maintenance Regimes for Electrical and Mechanical Components. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/2964