LAUSR

Abstract The complexity–entropy causality plane based on permutation entropy is a powerful tool to discriminate signals from different systems. In this paper, we combine traditional statistical complexity measure and power spectral entropy and construct complexity–entropy causality plane in frequency domain. The power spectral entropy is derived from Fourier transformation, so some features that are obscure in time domain can be extracted in frequency domain. Comparing to permutation entropy, this method is free of parameters. Several time series generated from different classes of systems are analyzed to demonstrate the measure. Results show that these signals can be clearly distinguished in our plane. Then by adding sinusoidal abnormal signal into original one, the abnormal information can be efficiently detected. Finally, we apply it to bearing vibration signals. Empirical consequences illustrate that the start–stop time and classification of fault signal can be clearly determined.