Abstract: Early rolling bearing faults are challenging to diagnosed due to the difficulty in accurately selecting the input parameters of feature mode decomposition (FMD) filter length L and the number of modal components n. To address this problem, an early fault diagnosis of rolling bearings based on pelican optimization algorithm (POA) optimised FMD parameters is proposed. The method utilizes the ensemble kurtosiss index as the fitness function to achieve optimal parameter combinations of FMD, and integrates with envelope spectrum analysis for fault diagnosis. Applying this method to the rolling bearing early fault simulation signals and experimental signals, the results show that it can accurately extract fault characteristic frequency and its multiplicative amplitude from the envelope spectrum by optimizing the FMD parameters, identifying the early fault classification in rolling bearing. Compared with the methods based on intrinsic time scale decomposition (ITD) and minimum entropy deconvolution (MED), the proposed approach demonstrates superior performance in extracting fault characteristic frequencies and multiplicative amplitudes, showing its certain application prospects and value in the early diagnosis of rolling bearing faults.