Abstract: The vibration signal of coal mine ventilator is a non-stationary multicomponent signal. Traditional methods for feature extraction of non-stationary signals suffer from poor adaptability and limited ability to identify weak characteristics of early faults in the ventilators. Additionally, the signal processing speed of feature extraction methods based on generalized Variational Mode Decomposition (VMD) is unable to meet the requirements of rapid feature extraction of ventilator vibration signals. To address these issues, a rapid feature extraction method for coal mine ventilator vibration signals based on improved VMD is proposed. On the basis of the generalized VMD algorithm, the multiplier alternating direction method was used for iterative solving, converting the constrained optimization problem into an unconstrained optimization problem. The improved VMD algorithm was applied to perform equivalent decomposition of the signals, obtaining an equivalent filter that matched the target signal features. The feature components of the ventilator vibration signals were quickly extracted based on the inner product transform principle. Simulation and experimental results showed that the improved VMD algorithm performed well in extracting feature components of different intensities, with good accuracy and noise resistance. The processing time for the measured ventilator vibration signals was 0.008 165 seconds. The feature extraction speed was significantly improved compared to the generalized VMD algorithm.