A signal filtering method for lubricating oil wear particle sensor based on dragonfly algorithm(DA)，variational mode decomposition(VMD)and Savitzky-Golay(SG) filtering is proposed to improve signal quality.Firstly

according to the generation mechanism of the sensor signal

the VMD technique is used to decompose the signal of the lubricating oil wear particle sensor

and extract its components and features.In this process

the DA is used to optimize the number of modes

K

and the penalty factor

α

by minimizing the envelope entropy as the fitness function.Then

each decomposed mode is processed using an SG filter

which f

urther reduces the noise level of the signal.Finally

according to the kurtosis principle

the modes containing abundant signal components are reconstructed to recover the effective components of the original signal.The experimental results show that this method significantly improves the clarity of the signal from the lubricating oil wear particle sensor.Compared with the wavelet denoising model

the recognition rate of wear particle is increased by 6%.Compared with the directly applying SG filtering to the signal

the recognition rate of wear particle is increased by 9%.