On the Predictability of Time Series by Metric Entropy

Abstract

The computation of the metric entropy, a measure of the loss of information along the attractor, from experimental time series is the main objective of this study. In this study, replacing the current warning systems (simple threshold based, on/off circuits), a new and promising prognosis system is tried to be achieved by the metric entropy, i.e. Kolmogorov . Sinai entropy, from chaotic time series. Additional to metric entropy, correlation dimension and time series statistical parameters were investigated.Condition monitoring of ball bearings and drill bits was achieved in the light of practical considerations of time series applications. Two different accelerated bearing run-to-failure test rigs were constructed and the prediction tests were performed.However, as a reason of shaft failure in both structures during the experiments, none of them is completed. Finally, drill bit breakage experiments were carried out. In the experiments, 10 small drill bits (1 mm ) were tested until they broke down, while vibration data were consecutively taken in equal time intervals. After the analysis, a consistent decrement in variation of metric entropy just before the breakage was observed. As a result of the experiment results, metric entropy variation could be proposed as an early warning system.