Time-Frequency Bandpass Filter with Nonstationary Signal Decomposition Application

Abstract

Abstract In various applications such as speech processing, underwater sound, radar application and mechanical fault diagnosis, there are large number of nonstationary signals. In order to achieve nonstationary signal decomposition, the instantaneous frequency ridge is extracted through interactive mode, and a time-frequency bandpass filter (TFBPF) method is proposed. This method is the modified version of Intrinsic Chirp Component Decomposition (ICCD) and has the physical essence of time-frequency domain bandpass filter. The superiority and application potential of TFBPF in complex engineering signal decomposition are demonstrated through two numerical examples. In the first one, we prove that TFBPF can decompose the weak signal component directly through analyzing the vibration signal of hydraulic turbine; While in the second one, we prove that TFBPF can effectively overcome the drawbacks of ICCD through analyzing the vibration signal during the speed-up process of a complex equipment. TFBPF is still very effective even for complex signal decomposition in a strong noise environment.