Data-driven nonlinear technique for condition monitoring

This paper describes a sensitive technique for distinguishing changes in a nonlinear process. The method obtains a phase-space (PS) representation of the process, which in turn is converted into a probability density function (PDF). Condition change is monitored by comparing two PS-PDFs via a {chi}{sup 2} statistical measure. One example application involves monitoring of brain waves to distinguish various states in an epileptic patient. A second example distinguishes different drilling conditions from spindle motor current data. A third example distinguishes balanced and unbalanced pumping conditions from power data.