Exploiting Phase Fluctuations to Improve Machine Performance Monitoring

Machines are an integral and important part of modern life. Machine condition monitoring is of vital importance to modern industry in its quest for higher reliability, quality, and efficiency. A new signal-processing technique for machine performance monitoring is presented. This new technique exploits fluctuations in phase angles of machine rotational frequency signals to determine their dynamic temporal coherence. Temporal coherence is the key to automatically identify a fault condition and assess its severity. The exploitation of temporal coherence also provides increased spectral resolution and signal-to-noise ratio (SNR). Tests were conducted on an edger trimmer ball bearing assembly that was subjected to different levels of fault conditions, such as a hairline crack on the outer race and sand contamination in the bearing. An electric fan motor with bearing faults was also tested. The fault identification capabilities of incoherent power averaging are compared with those of the new coherence processors. Some rotational signals could not be identified in the average power spectra and, therefore, the average power could not be used for monitoring these signals. However, they were easily identified in the spectra of the new processors, with their increased SNR gain, and were used successfully for machine performance monitoring and diagnostics. This new processing capability for quantifying and exploiting the dynamic temporal coherences of a machine's signals provides a valuable capability for detecting existing and developing faults, and for monitoring their progress. This is also true during the startup and shutdown phases of machines, when their speeds and corresponding rotational frequencies are changing. Note to Practitioners - General guidance for monitoring a machine's performance is included below. Calculate the coherence parameter phi and establish the base-line temporal coherences of a machine operating in good condition for future reference, or for a similar machine that is known to be in good operating condition. Continuously monitor the machine's rotational signals via accelerometer or microphone, measuring the temporal coherences using the FFT-derived coherence parameter phi. Calculate/display the temporal coherence time histories of phi for the rotational signals. Analyze the phi coherence patterns. A constant thickness, continuously nonrandom, phi temporal coherence time history pattern, indicates fault-free machine operation. A decreasing spread in the phi time history pattern with time indicates increasing temporal coherence of a rotational frequency component due to a mechanical fault. The frequency indicates which machine component is faulty. Possible problem: a fault associated with a bearing (e.g., a crack in a bearing race). A broadening phi time history pattern of a rotational frequency component indicates decreasing temporal coherence. Possible problems: need for maintenance, worn out lubrication, or foreign matter (dust, grit, sand) in the bearing