中文摘要：

针对转子碰摩声发射信号在传播过程中因衰减和噪声影响而难以定位的问题，引入最小方差自适应时延估计（LMSTDE）算法．对传统的LMSTDE算法进行了改进，提出了将衰减系数与时间延迟同步迭代估计的LMSTDE算法．该算法增加了一个衰减系数估计器，利用估计值与参考值的误差将时延与衰减系数同时进行迭代估计．为了加快算法的收敛速度，对该算法还进行了变步长处理，提出了当估计值与参考值的误差较大时利用误差的梯度来调整步长，当误差较小时引入遗忘因子对误差信号进行加权从而确定新步长的方法．理论分析及实验结果表明，改进的LMSTDE算法能够收敛到真实值处，与传统LMSTDE算法相比，能显著提高算法在低信噪比时的性能，具有更快的收敛速度．

英文摘要：

Due to difficulties of source localization of rub-impact acoustic emission influenced by at- tenuation and strong noise in propagation process, the least mean square adaptive time delay estimate （LMSTDE） algorithm is introduced. The traditional algorithm is improved, and an improved LM- STDE algorithm which estimates the attenuation coefficient and time delay simultaneously and itera- tively is proposed. The improved algorithm adds an attenuation coefficient estimator which estimates the time delay and attenuation coefficient simultaneously and iteratively according to the error be- tween estimated value and reference value. In order to accelerate the convergence speed, variable step size processing is used in the improved LMSTDE algorithm. When the error between the esti- mated value and the reference value is relatively large, the gradient of the error is used to adjust the step size; when the error is relatively small, the error signal is weighted by the forgetting factor to adjust the step size. Theoretical analysis and experimental results show that the improved LMSTDE algorithm can converge to the true value. Compared with traditional LMSTDE algorithm, the im- proved algorithm has better oerformance when the SNR is low and faster convergence speed.