基于多维特征的序贯概率比振源检测算法

A Vibration Source Detection Algorithm Using Sequential Probability Ratio Test Based on Multidimensional Features

光纤振动传感器现已应用于管道安全预警系统中,被用来获取振动信号.已有的信号识别检测方法多是采用一维特征,并主要应用于高信噪比的情况.由于实际振源产生的振动信号经常是复杂多变的非平稳信号,本文基于此情况提出采用多维特征作为振源检测模型,根据振源数据冗余量一般较大的特点,选取了序贯概率比算法完成振源检测,并利用光纤传感器采集的现场振动数据对该方法进行验证.结果表明,本文所给方法可以有效地进行振源检测,极大地提高了系统的检测性能.

The fiber-optic vibration sensor is applied to obtain vibration signal in the pipeline security pre-warning system. The pattern recognition, based on one-dimensional feature, is utilized to detect vibration source in most of the existing methods. These methods are applied mainly in high signal to noise power ratio regions （SNR）. But the signals, generated by vibration source, are usually complex and non-stationary. Based on the above reasons, multidimensional features are applied to establish the vibration source detection model. As the data is redundant, the sequential probability ratio test（SPRT） is applied to detect vibration source. It is proved that the proposed algorithm is effective in vibration source detection and largely improves the detection performance of the system.