Detection of Machine Failure by Using Information on Higher-Order Correlations Between Sound and Vibration

摘要

In this study, a method for stochastic detection of the failure of machines based on changes in information not only on linear correlations but also on higher-order nonlinear correlations is proposed in a form suitable for on-line signal processing in the time domain on a personal computer, especially in order to find in detail the mutual relationship between sound and vibration emitted from rotational machines. More specifically, a conditional probability hierarchically reflecting various types of correlation information is theoretically derived by introducing an expression for the multidimensional probability distribution in orthogonal expansion series form. The effectiveness of the proposed theory is experimentally confirmed by applying it to observed data emitted from a rotational machine driven by an electric motor.