The invention discloses an adaptive electromechanical equipment weak fault characteristic extraction method. The method comprises steps that firstly, vibration signals of electromechanical equipment are acquired; secondly, two-scale similarity transformation of selected initial multi-wavelets is carried out to improve the approximation order of a multi-wavelet scale function; thirdly, a multi-wavelet lifting matrix satisfying symmetry and translation constraint conditions is designed, and a multi-wavelet integration construction method is realized; and lastly, based on the vibration signals, agenetic algorithm is utilized to optimize free parameters, an adaptive multi-wavelet base function matching with fault characteristics is acquired, and an advantageous method is provided for weak fault characteristic extraction. The method is advantaged in that the dynamic matching bottleneck in the weak fault characteristic extraction process in the actual monitoring and diagnostic application of the electromechanical equipment is solved.