Reduced Mode Decomposition: A New Signal Decomposition Method
Jian Cheng, Haiyang Pan, Jinyu Tong, Jinde Zheng
Abstract
Although traditional signal processing methods have good decomposition performance in multimodal signals, they lack theoretical research on periodic pulse signals, resulting in insufficient decomposition. Based on this, a reduced mode decomposition (RMD) method is proposed in this paper, which can decompose reduced components (RCs) iteratively through the designed finite impulse response (FIR) filter bank. On the one hand, an adaptive index called reweighted kurtosis (RK) is defined as the objective function of filter banks, so as to fully consider the impulsivity and periodicity of signals and make the filter components contain rich state information. On the other hand, FIR is used to constrain various modal signals, so that RMD is not limited by filter bandwidth and center frequency, improves the decomposition ability and ensures the robustness of noise. The verification results of two types of roller bearing fault signals indicate that RMD is a novel multi-mode signal analysis method.