Time–frequency characteristics and trend feature of the ENPEMF signal before Lushan $${{\varvec{M}}}_{{\varvec{w}}}$$ M w 6.6 earthquake via DE-DDTFA method

The Earth's natural pulse electromagnetic field (ENPEMF) signal, is generally considered to be a nonlinear or nonstationary signal received from our instrument, placed on the surface near the source area. To obtain latent information on the ENPEMF signal, this paper employs the time–frequency analysis (TFA) method to get the instantaneous frequency (IF) of the signal. The traditional Data-driven time–frequency analysis (DDTFA) requires to know the initial phase function (IPF) set of the signal, to accomplish the signal decomposition and its IFA. However, it's difficult to observe directly the IPF set of the ENPEMF signal. To acquire accurate time–frequency distribution, an improved DDTFA method was proposed, which adopts differential evolution (DE) to calculate multiple IPF of multi-component non-stationary signals. In this paper, the ENPEMF signal received from the Lushan $${M}_{w}$$ M w 6.6 earthquake on April 20, 2013, was taken as an example, and the improved method, DE-DDTFA, was used to decompose the signal into multiple intrinsic mode function (IMF) components, and obtained the IF of each IMF. It is demonstrated by the experimental that the number of IMF is 200% more than usual time, and the energy of the signal had grown by approximately 10–20 times, or even more compared with usual in just one week before the earthquake. The experimental result illustrates that the amount of IMF and the energy of the ENPEMF signal show an overall upward trend, which is a distinct trait before the earthquake, and DDTFA is a good reference for studying the time–frequency distribution and energy spectrum variation characteristics of electromagnetic signals before earthquakes.