To solve the problem of weak echo signal and low signal-to-noise ratio caused by large lift-off distance of electromagnetic ultrasonic thickness measurement transducer protection, which makes it difficult to directly and accurately extract transit time to obtain accurate thickness value in time domain, this article proposes an O-VMD transit time extraction method based on particle swarm optimization of variational modal decomposition parameters in frequency domain. The fixed parameters are selected for decomposition layers and penalty factors respectively, and the particle swarm optimization algorithm based on the joint fitness function of kurtosis and power spectrum entropy is used to obtain the ergodic optimization parameters of the variational mode decomposition. The VMD processing is performed twice to filter high-frequency and low-frequency noise. The maximum energy mode is selected for signal reconstruction, and the transit time of echo signal is obtained by applying Hilbert transform. Under different lift off distances, O-VMD, EMD and other methods are used for signal contrast processing of aluminum plate detection data with different thicknesses. The results show that the maximum error of O-VMD method is 0. 67% when the lift off distance is within 0 ~ 2. 1 mm, and the error is proportional to the lift off distance, providing a basis for accurately obtaining thickness measurement data with high lift off distance.