Model and Information Theory Based Diagnostics for Machinery and Gears

Abstract

Machines focus power. We view a machine as a “machine communications channel,” wherein components along the channel organize power and information flow. Errors from broken or degraded components disrupt this organization and the implicit information processing. Our model based diagnostics approach constructs detailed physics models of the machine having direct physical correspondence to components and faults, measures states off the real in-service machine, simulates states and sensor outputs of the machine under same service loads, compares simulated sensor outputs to real sensor outputs, and adjusts (tunes) the model’s parameters until simulated outputs closely mimic real outputs. The tuned model now contains information on the real system’s health condition. By comparing the numerical values of parameters to those of an ideal model—an exemplar of perfect health—faults are detected and located. To assess machine functional condition, Shannon’s theorems of information theory are applied as a health metric to the machine communications channel. For prognosis of future health, plots of the model’s parameter values extrapolated forward in time predicts future parameter values. Simulation of the machine channel model predicts “future” machine behavior, and future machine functional condition, using the aforementioned methods. This article applies these methods to a gearbox with tooth root cracking.