基于遗忘加权的变步长自适应管网泄漏定位方法  

Pipeline leak detection method based on variable step size LMS adaptive filter with forgetting weight

作　　者：陈建宇[1] 

机构地区：[1]大同煤矿集团有限责任公司供水分公司,山西大同037003

出　　处：《辽宁工程技术大学学报（自然科学版）》2017年第5期484-487,共4页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金项目(70971059)

摘　　要：供水管道发生泄漏时会产生泄漏声信号,为克服声信号传播速度不确定对泄漏定位的影响,提出基于三传感器泄漏检测与定位模型,解决声速依赖经验值的问题;对变步长的LMS自适应滤波算法做了改进,在步长因子迭代过程中引入历史误差的遗忘加权和进行补偿,同时加入滑动窗,提高收敛速度的同时降低稳态失调.研究结果表明:该方法能够实现准确的时延估计,在保持较小稳态失调的基础上提高算法的收敛速度,提高管网泄漏的定位性能.Acoustic signal of leakage will be generated when water pipes leak. A three sensor-based leak detection and location model is proposed, which solves the problem of sound velocity dependent on experience. Improvements on variable step size LMS adaptive filtering algorithm have been made, and the sum of historical errors with forgetting weight are introduced in the iteration process of the step size factor to compensate it, and a sliding window is added in the same time to improve the convergence speed while reducing homeostasis. Experimental results show that the method can achieve accurate delay estimation, improve convergence speed with small steady state offset and improve the performance pipeline leak positioning.

关 键 词：供水管网 泄漏 定位 时延估计 自适应 变步长 遗忘加权 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]