基于在线字典学习的管道微弱泄漏检测方法

Method of pipeline weak leak detection based on online dictionary learning

针对离线字典学习方法存在对管道泄漏检测长时期运行信号的适应性不足、计算量大的缺点,进行了基于在线字典学习的检测方法研究。借鉴参数化字典训练方法的思想,对管道泄漏动态压力信号进行多分辨率分解,对分解的子频带信号进行稀疏编码,并进行快速的在线字典训练与更新,根据稀疏编码结果进行微弱泄漏检测。现场实验数据的测试结果表明,提出的方法可检测出泄漏低频响应为0.2 Hz以上,流量变化量大于0.4%的微弱泄漏,有效提高了微弱泄漏的检出率,降低了虚警率。

The traditional algorithm of offline dictionary learning has the disadvantage of massive computation and lack of adaptability to long term run signal. This paper put forward a new method of oil and gas pipeline weak leak detection based on online dictionary learning. Firstly, it decomposed the signal of pipeline dynamic pressure by multi-resolution wavelet, and sparsely coded each sub-band signals. Secondly, it designed the algorithm of over-complete dictionary training. At last, it performed weak leak detection by sparse coding results. Simulation results show that the proposed method can detect low frequency response of the weak leak greater than 0.2 Hz, flow changes greater than 0.4%, the proposed method can improve the weak leak detection rate and reduce the false alarm rate.