基于改进的Levinsion算法对电机转速特性的研究

传统的Levinsion算法忽略了信号自相关序列值代替理论值时的误差，因而降低了算法性能。为了减小该误差的影响，采用改进Levinsion算法、AIC阶数判定准则，将现代谱估计中的AR模型应用于包含广泛频率成分的典型的电机转速振动信号分析，求得AR模型的功率谱。通过AR现代谱与Welch经典谱的实际对比分析，验证了所建AR模型的正确性和对电机转速振动信号分析的适用性，为电机转速振动特性的研究提供了可靠的分析技术。

Research of Motor Rotate Speed Characteristic Based on Improved Levinsion Algorithm

Traditional levinsion algorithm omits errors that are produced when estimated values of autocorrelation sequence take the place of their ideal values. In order to lower errors as much as possible,an improved levinsion algorithm and Akaike information criterion(AIC) order selection criteria are presented to improve its performance of power spectrum. Auto regressive(AR) model was used to process typical motor rotate speed vibration signals. AR spectrum based on the proposed algorithm was compared with the Welch method and the results of simulation show some remarkable advantages for motor rotate speed vibration signals.