人工免疫系统在信号异常检测方面有很好的应用前景。该文对人工免疫系统中的负向选择算法进行探讨。通过沿时间轴对采样信号加窗的方式构造向量集合,利用余弦相似度进行向量间亲合度的匹配计算,实现在实数域进行匹配计算的实数值负向选择算法。结合对仿真信号及实际转子系统振动信号的异常检测,对该算法的加窗方式及匹配阈值进行了分析。计算结果表明:这种算法可有效地检测出振动信号中的异常值,且仅需要对信号进行去均值的预处理,可望用于机械振动信号的在线监测。 Artificial immune system has good application prospect in signal anomaly detection. This paper discusses the negative selection algorithm in artificial immune system. The vector set is constructed by windowing the sampling signal along the time axis and the matching of vector affinity is calculated by cosine similarity to realize the real-valued negative selection algorithm for matching calculation in real number domain. Combined with the anomaly detection of the simulated signal and the actual rotor system vibration signal, the windowing method and matching threshold of the algorithm are analyzed. The results show that this algorithm can effectively detect the outliers in the vibration signal and only need to preprocess the signal to the mean value, which is expected to be used for on-line monitoring of mechanical vibration signals.