Abstract: Based on signal detection theory,the real-time leakage detection problem of submarine pipeline was discussed,by means of the observed series of mass flow difference of pipe.The adaptive filter algorithm based on innovation theory was introduced.The color noise of the observed series resulted from data gathering and long-distance transfer was whitened.A binary composite hypothesis test model with unknown and stochastic parameters was constructed.The sequential probability ratio test (SPRT) with the uniformly most powerful effect was provided for white Gaussian process.This novel method was used to detect the distant series of pipe mass flow difference of leakages produced in stationary and non-stationary operation situations, respectively.The uniformly most powerful test (UMPT) can exactly judge leakage signals of above both situations and obtain the detection probability of 95 percent.It is indicated that UMPT has a better accuracy,and the least detection time can keep more robust performance in varying pipe operating situation,compared with SPRT applied in some cases of industrial pipelines.

Key words: submarine pipeline, leakage, color noise, innovation theory, adaptive filter, hypothesis test

摘要： 针对海底管道泄漏监测问题,基于信号检测原理分析了受噪声干扰的流量差观测序列.引入了新息理论和自适应滤波算法,使管道运行的数据采集和远程传输过程中产生的非白噪声转化为白噪声.建立了以泄漏信号的幅值为随机参量的二元复合假设检验模型,提出了具有一致最大功效的序列似然比检验方法,并进行了平稳观测序列强泄漏信号和非平稳观测序列弱泄漏信号的检测.对以上两种情况无须预先假设可能发生的有关泄漏量统计知识,利用该方法即能做出准确判断,检测概率均达到95%以上.与现有的序列似然比检漏方法比较,一致最大功效检验探测泄漏信号的准确性较高,发现泄漏信号的时间短,更容易适应不同运行工况及泄漏规模.

关键词: 海底管道, 泄漏监测, 信号检测, 噪声, 随机参量, 自适应滤波算法, 检验模型