Fault detection and diagnostics for non-intrusive monitoring using motor harmonics

Abstract

Abstract-Harmonic analysis of motor current has been used to track the speed of motors for sensorless control. Algorithms exist that track the speed of a motor given a dedicated stator current measurement, for example [1][2][3][4][5]. Harmonic analysis has also been applied for diagnostic detection of electro-mechanical faults such as damaged bearings and rotor eccentricity [6][7][8][9][10][11][12][13][14][15][16][17]. This paper demonstrates the utility of harmonic analysis for fault detection and diagnostics in non-intrusive monitoring applications, where multiple loads are tracked by a sensor monitoring only the aggregate utility service. An optimization routine is implemented to maintain accuracy of speed estimation while using shorter lengths of data. I. SMART MONITORINGAt any point in the life of a system, mechanical and electrical equipment may be poorly operated. For example, as buildings age, both the electro-mechanical actuators and associated mechanical components wear, cease to function properly, and eventually fail, via myriad processes that are often undetected. Valves do not close fully, filters clog, airconditioning system dampers stick, refrigerant leaks, heating and cooling coils -from the smallest refrigerator to the largest building air-conditioning system -become fouled with dirt and debris, and belts slip. Energy waste and excessive plant wear are often exacerbated by closed-loop control. Under active control, damaged but still functioning equipment will operate by extending run times or operating points to meet user commands, leaving few obvious signs of compromised operation.For example, a number of surveys of airflow faults in buildings hint at the range and extent of these problems. One compendium of fault surveys [18], which examined 503 rooftop air-conditioning units in 181 buildings in five states in the Western U.S. from [2001][2002][2003][2004], found that the airflow was out of the specified range in approximately 42% of the units surveyed. A separate study [19] of 4, 168 commercial airconditioners in California reported that 44% of the surveyed units had airflow that was out of specifications. Studies of 29 new homes in Washington State [20] found that average duct leakage rates to the exterior ranged from 687 to 140 cubic feet per minute (CFM). Extrapolating from such fault surveys, one estimate for the total energy consumed by duct leakage is $5 billion/year [21].When "failure is not an option," the performance of important electro-mechanical loads on mission-critical systems like warships or power plants is often tracked by dedicated monitoring equipment [22]. An extensive sensing network can provide obvious advantages for fault detection, diagnosis, and prognosis. However, a large sensing network can be expensive and difficult to maintain.Smart Grid and Smart Meter initiatives hope to allow energy providers and consumers to intelligently manage their energy needs through real-time monitoring, analysis, and control of electrical power usage. The U.S. Department of Energy has i...

Non‐intrusive induction motor speed detection

“…Extrapolating from such fault surveys, one estimate for the total energy consumed by duct leakage is $5 billion/year [12]. If the speed of critical motors can be tracked for an HVAC system, the operating and diagnostic condition of fans and compressors can be inferred from motor shaft speed estimates [8].…”

“…Even motor design features like skewed rotor bars tend to leave some harmonic variation in the permeance of a machine as a function of rotation [13], and this variation leads to harmonic currents that can be tracked by the methods introduced here. The algorithm demonstrated in this paper builds on the harmonic estimation algorithm presented in [8]. Here, we demonstrate that a non‐intrusive load monitor (NILM) with transient event recognition [14] can serve as a platform for identifying induction motor speed in an aggregate data stream.…”

NilmDB: The Non-Intrusive Load Monitor Database

“…Many such diagnostics have been developed within the context of the NILM [8], [17], [19]- [21], and can be applied within the NilmDB framework.…”

An in depth study into using EMI signatures for appliance identification

“…Patel et al used low amplitude transients (up to a few kHz) for detecting electrical events [24,20], and showed that these events, combined with smart meter data, can help in appliance disaggregation. Leeb et al used low frequency harmonics up to a few kHz to disaggregate appliances and diagnose faults in electrical systems [22,28]. Other studies have utilized low frequency power signatures, either harmonics of 50 Hz sinusoid or transient noise generated by appliance switching, to distinguish household appliances [20,24].…”