MODEL-BASED DIAGNOSTICS AND PROGNOSTICS OF MACHINERY

摘要

A model-based diagnostic technique to detect, isolate and assess machine faults is presented and applied to motors. Signals measured from the real machine are compared to simulations derived from physics-based models of the machine. To make simulations match measured signals, parameters of the models are adjusted or "tuned". A direct physical correspondence between the model's parameters and real system elements permits the tuned parameters to locate faults and simplifies classification of faults. The degree to which faults in a degraded machine alters power and signal flows through the machine, indicates machine health. An analogy is formed between a machine and a communications channel, to apply powerful theorems from communications theory to assess machine functional health. For prognosis of machine health and functionality, parameter variations observed over time can be extrapolated into the future, and simulations of the model with extrapolated parameter values can be performed, enabling model simulations to forecast future machine behavior. A Graphical User Interface (GUI) permits a machine operator to select specific sensor outputs, desired parameters, and specific faults to tailor the diagnostics system to a specific machine or design. The aforementioned model based diagnostics techniques are computed behind the interface. In this article, these techniques are applied to motors. Motor voltages, currents, and speeds are measured, and parameters including stator resistance, rotor resistance and magnetic field features are tuned to detect and isolate shorted turns and broken rotor bars. Then functional health of degraded motors is assessed.