煤矿通风机振动信号是一种非平稳多分量信号。传统的非平稳信号特征信号提取方法存在自适应性差、对通风机早期故障的微弱特征辨识能力有限等问题，基于广义变分模式分解的特征提取方法的信号处理速度难以满足通风机振动信号特征快速提取的要求。针对上述问题，提出了一种基于改进变分模式分解的煤矿通风机振动信号特征快速提取方法。在广义变分模式分解算法的基础上，采用乘子交替方向法迭代求解，将约束优化问题转换为无约束优化问题。应用改进变分模式分解算法对信号进行等效分解，得到匹配目标信号特征的等效滤波器，通过内积变换原理快速提取通风机振动信号特征分量。仿真和实验结果表明，改进变分模式分解算法对不同强度的特征分量提取效果均较好，准确性和抗噪性良好，处理通风机实测振动信号的耗时为0.008 165 s，与广义变分模式分解算法相比，特征提取速度大幅提升。

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.